in the name of allah, most merciful, most gracious

IN THE NAME OF ALLAH, MOST MERCIFUL, MOST GRACIOUS

Proceedings of the Seventh Saudi Engineering Conference Riyadh 2-5, December 2007

  Volume IV Research and development to serve the industry and upgrade its services Chemical Engineering Electrical Engineering

PREFACE The Seventh Saudi Engineering Conference comes to complement the series of Saudi engineering conferences which started in 1402H and have been hosted successively by different colleges of engineering of the Saudi universities. The College of Engineering at King Saud University is honored to host the conference for the second time. These conferences have greatly contributed to the resettlement of technology, the dissemination and exchange of experiences between engineering professionals, and have helped to promote the scientific research besides advancing innovation and excellence. At a time of the advanced technology and the availability of information in various ways, nations have become closer and the world is turning into a small village, the economy has become the prime engine of the world. It is necessary for all nations to work hard to cope with this technical progress and benefit from it, and moreover create appropriate conditions to deal with this tremendous development and competition as much as possible. It is incumbent upon all professionals in general and engineers in particular to work hard to provide the proper environment in such circumstances. As a consequence, The Seventh Saudi Engineering Conference discusses an important and vital theme for researchers, engineers and industrialists. The theme is to provide an Engineering Environment to merge in a Competitive Global Economy in an open and boundary-less economy and profession. This conference is trying to answer this question through well-formulated seven topics. Conference topics discuss multiple issues related to engineering profession and engineering firm, engineering environment through education and labor market requirements, engineering rehabilitation, preservation of the environment, rationalization of resource consumption, Saudi construction code, development of the engineering sector to diversify sources of national income, and research and development to service the industry and upgrade its services.

PREFACE

The conference proceedings contain 168 refereed scientific research papers which are distributed into a number of volumes, and each volume contains one or more topic. A separate volume for paper abstract is also published in addition to electronic proceedings that includes all papers accepted in the conference. These proceedings will be a scientific reference for engineers in the Kingdom and the worldwide. Finally, thanks to Almighty God for his help in completing of this work and deep thanks for all members of the Conference Committees for their efforts, and special thanks to members of the Scientific Committee for their efforts to have this documentation of the huge scientific research, which is an important reference for researchers and engineers. Thanks also for authors and experts who have contributed their ideas, their research to the success of the conference.

Thanks Chair of organizing committee Prof. Abdulaziz A. Alhamid

Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007

VI

INTRODUCTION Under the high patronage of his Royal Highness the Prince Sultan Ibn Abdulaziz the crown prince and minister of defense, aviation and inspector general, the College of engineering at the King Saud university hosted the Seventh Saudi Engineering Conference during the period 22 to 25 Dhu Alqeeda 1428 corresponding to 2-5 December 2007. The theme issue of the conference is “Towards An Engineering Environment Competitive to the Economics of Globalization”. The response to contribute in the conference has been most encouraging. A large number of abstracts were received. After a thorough peer-review process for evaluating the submitted papers, the scientific committee has selected a total of 168 papers, presented by 300 researchers. The conference has drawn participants from the different kingdom universities, colleges, institutes and technical education establishments as well as governmental and national companies. The conference has also attracted international participation from universities and institutes of United Arab Emirates, Egypt, Sudan, Algeria, Tunisia, Malaysia, India, Great Britain, Germany, France, Deutschland, Canada, Japan and United States of America. One of the main objectives of the conference was to contribute to the review and development of important aspects of the engineering sector both public and private. The topics of the conference were chosen to tackle the challenges that engineering education and its outputs are facing. In addition, the themes also emphasized on the contribution of the engineers to the development of the country. The conference themes were as follows: • • • •

Engineering qualification and its role in the strategy of Saudization Engineering specialties as viewed from the educational establishments and the job market requirements Engineering sector contribution to resources conservation Engineering and environmental protection

VII Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007

INTRODUCTION

• • •

The Saudi building code Development of the engineering sector for diversification of income resources Research and development in the service of industry and for the improvement of services

In addition to the specialized scientific papers that covered the above mentioned themes the conference also hosted a number of plenary lectures and discussion forums that attracted the participation of key policy makers as well as academics and economic parties. The selected abstracts and papers have been documented in the proceedings which comprise of six volumes in accordance with the conference themes. The papers are also documented in CDs. Before concluding I would like to express my gratitude to all members of the Scientific Committee for their efforts and active participation to the success of the conference. Thanks are also due to the referees who have been of great help in selecting high quality papers for the conference. The support provided by the secretarial and technical staff of the college of engineering is also thankfully acknowledged. Finally on my own behalf and behalf of the Scientific Committee I would like to record our appreciation and sincere thanks to His Excellency the rector of the King Saud University and the Dean of College of engineering, the chairman of the organizing committee for their continued support and valuable guidance, We are all hopeful that this scientific conference will be of a support for recruiting engineering specialties on a larger scale and contribute to the growth and prosperity of our country. May Allah Almighty accept our sincere efforts.

Chairman of the Scientific Committee Prof. Khalid Ibrahem Alhumaizi


VIII

Contents

Page No.

Topic 7 Research and development to serve the industry and upgrade its services • Chemical Engineering TWO-PHASE NON-EQUILIBRIUM MASS TRANSFER OF VOLATILE CHEMICALS IN FIXED VOLUME

1

3

James Warner , Osama Al-Gahtani

EFFECT OF POTASSIUM ADDITION ON THE ACTIVITY OF NICKEL BASED CATALYST USED FOR DRY REFORMING OF METHANE

15

Fakeeha ,Anis H.; Al- Fatish , Ahemed S. ; Soliman M.A.Ibrahim A.A

PREDICTION OF MINIMUM FLUIDIZATION VELOCITIES OF SOLID MIXTURES

23

Mohammad Asif

A SHORTCUT METHOD FOR BINARY DISTILLATION COLUMN DESIGN

35

M. A. Soliman

ECONOMICS OF ETHYLENE AND ACETIC ACID PRODUCTION BY PARTIAL OXIDATION OF ETHANE

47

Anis H. Fakeeha, Moustafa A. Soliman

OXIDEHYDROGENATION OF PROPANE OVER NI-MO SUPPORTED METAL OXIDES CATALYSTS

63

M. K. Al-Mesfer, S. M. Al-Zahrani and A. E. Abasaeed

DRY REFORMING OF NATURAL GAS USING NICKEL BASED CATALYST

73

A.S. Al-Fatish and A.A. Ibrahim

KINETICS OF PARTICLE FORMATION SUPERCRITICAL GAS PROCESS

IN

Yousef Bakhbakhi , Paul Charpentier , Sohrab Rohani


THE

87

Contents
3), which is not suitable for Fischer–Tropsch and methanol syntheses [1-5]. Thus, research efforts have been directed towards Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007

DRY REFORMING OF NATURAL GAS

obtaining syngas with a more suitable H2/CO ratio of 2 or lower via methane partial oxidation or by CH4/CO2 reforming or autothermal oxidation. Partial oxidation of methane and CO2 reforming of methane have the potential to reduce the cost of syngas [6] and/or CO2 utilization technologies [7, 8]. Therefore many important attributes are given to dry reforming [9]. Nevertheless, catalysts deactivation by coke is one of the most serious problems in CO2 methane reforming. Since the reaction is endothermic, it proceeds at high temperature, thermodynamically favour coke formation. Al-Fatish [10] found that one of the main problems in carbon dioxide dry reforming of methane is the instability and the deactivation of the catalyst. Initially the catalyst activity was high, but it deactivated very fast. This phenomenon is observed by other investigators [11-13]. It has been pointed out in the literature that the main source of the problem is carbon deposition on the catalyst surface in the production of syngas with low H2/CO, which results in the catalyst deactivation, plugging of the reactor, and breakdown of the catalyst granules. Attempts to overcome this limitation of carbon deposition have focused on the development of improved catalysts. Nickel is considered as most promising catalyst for CO2/CH4 reforming due to low price; however, nickel-based catalysts easily induced formation of carbon, causing catalyst deactivation and plugging of a reactor. The catalyst deactivation is the main hindrance for a catalyst to be considered for an industrial application, the two most known causes are: coke deposition and sintering of the metallic active phase [14]. High temperature calcinations of Ni/γ-Al2O3 catalyst suppressed coke formation and the catalysts did not lose activities up to 120h However, the Ni/γ-Al2O3 catalyst maintained high activity and good stability [15]. On the other hand, the stability of Ni/ Al2O3 catalyst by adding Cu for CO2 reforming of methane was studied and found that 1 wt% Cu addition suppresses the carbon deposition; enhances the stability and the activity of the catalyst [16]. Pompeo et al. studied Ni and Pt catalysts supported on α-Al2O3, α-Al2O3-ZrO2 and ZrO2 in the dry reforming of methane to produce synthesis gas and obtained that all catalytic systems presented well activity levels with turn of frequency (TOF) (s-1) values between 1 and 3, with Ni based catalysts more active than Pt based catalysts [17]. In this paper, we present the results of experimental investigation of dry reforming reaction using Nickel based catalyst and the effects of two supports alumina and silica, calcination temperatures of the range from 500 to 800˚C, different reaction temperatures 500, 550 and 575˚C. Investigations of catalyst stability in terms of methane and carbon dioxide conversions and hydrogen yield have been performed.

EXPERIMENTAL The experimental equipment used in this study is a micro-reactor system which is shown schematically in Figure.1. The system is composed of the following sections:

74 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


1. Feed Section: The feed section contains three gas cylinders for CO2, CH4 and N2. Gases coming from the regulators pass through in-line filters are then introduced to Mass Flow Controllers (MFC), obtained from Brooks. The gases are mixed and passed to the reaction section. On line samples from the feed gas mixture are directed to gas chromatograph for analysis. 2. Reaction Section : The micro-reactor overall length is 17.2" (43 cm) with inside diameter of 1/4" i.e. (6.35 mm) made of stainless steel and surrounded by three zones heater. Each zone temperature can be controlled separately. The temperature in the reactor is measured by a thermocouple located in the catalyst bed. The outlet from the reactor (bottom end) is passed through a back pressure regulator (BPR) to control the pressure in the reactor and the product gases from the BPR were sent to analysis section. 3. Analysis Section: Reaction products as well as feed mixture are analyzed on-line using 3400 CX Varian gas chromatograph. Thermal conductivity detector (TCD) is used for analysis using Haysepe A columns. CATALYST PREPARATION The used supports used are γ-alumina (SA-6175), α-alumina (SA-5239) and silica (S151-10), while their surface area are (230 – 290 m2/gm), (0.15 – 0.45 m2/gm) and (< 1 m2/gm) respectively. Impregnation of the supports was carried out with aqueous solutions of nickel-nitrate. After impregnation, the catalysts were dried for 13 hours. at 110 °C and calcined for 5 hours at the required temperature in air atmosphere. Activation Procedure: The catalyst must be activated once before it is used for the first time. This is done by introducing hydrogen to the reactor at 20 ml/min for 2.5 hours. Then hydrogen is stopped while N2 is introduced at 40 ml/min for an hour. Catalyst pretreatment process was found to be essential for the reaction to take place. Comparing activated and un-activated catalysts, it was found that the conversion of both CH4 and CO2 as well as H2 yield for the activated catalyst are much higher than the un-activated catalyst as shown in table 1.



Table 1: Comparing activated and un-activated of Ni / SA-617515 catalyst at 500 °C Activated catalyst

Un-activated catalyst

Conversion of CH4

34.7

2.51

Conversion of CO2

35.1

5.11

H2 yield

31.4

1.1

RESULTS AND DISCUSSION The results for experimental investigation for dry reforming reaction using Nickel based catalyst prepared by impregnation method are presented. Effects of using different supports, calcination temperatures and operating conditions, stability aspects are elaborated. Effect of Using Different Supports Various supports have been used in the preparation of nickel catalysts: two types of alumina (SA-6175, SA-5239) and silica (S151-10). Impregnation of supports was carried out with aqueous solution of nickel (Ni). After impregnation, the catalyst was dried at 110 °C for 13 hours and calcined at (500°C) or (800°C) for 5 hours. The catalytic activity was evaluated at 500, 550, 575°C using a flow rate of 10 ml/min and atmospheric pressure. A group of four catalysts were prepared using Alumina SA-6175 (high surface area) as support. Two of them were prepared by using a loading of 12% Ni while pure support was used with the other two catalysts. One catalyst prepared with pure support and one with 12% Ni loading were calcined at 800°C while the other two were calcined at 500°C. Figures (2-5) depict the effect of support and calcinations. Figure 2 gives the trends of methane conversions using catalyst calcined at 500°C. It is obvious from the figure that catalyst supported with alumina, loaded with 12% Ni and calcined at 500°C gives higher conversion in comparison with catalyst supported with silica. On the other hand, catalyst supported with lower surface area alumina SA-5239 gives better methane conversion than catalyst supported with higher surface area alumina SA-6175.While in the CO2 conversion of figure 3, catalyst supported with alumina SA-6175 gives the highest values for all temperatures. In figure 4, catalyst supported with alumina SA-6175 and calcined at 800°C presents the lowest methane conversion. This is due to more solidification aspect of the catalyst at higher temperature. Therefore, lower calcined catalyst seemed to be better and showed less agglomerate. The higher calcinations temperature enhanced methane conversion on silica supported catalyst, while the higher surface alumina SA-6175 supported catalyst reduced considerably 76 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


the methane conversion. This stipulates the knowledge of appropriate calcination condition for each catalyst depending on its constituting matrix. Again figure 4 underscores the preference of lower surface area alumina SA-5239. Figure 5 depicts the CO2 conversion corresponding to condition of figure 4. It is clear from graph that comparable values of CO2 conversion are given by silica and low surface area alumina. Alternatively, catalysts prepared by using pure alumina only showed no activity at all calcination temperatures used. Stability Study To investigate catalyst durability, endurance tests of CH4/CO2 reforming reactions were carried out at 500 °C, 1atm, CH4/CO2 =1/1, flow 10 ml/min, F/W =20 ml/min. gm cat. The results of these tests are presented in the form of plots of CO2, CH4 conversion and H2 yield as a function of time for the Ni-supported catalysts. The prepared catalysts with 12% Ni loading were prepared by impregnation. The catalysts were dried at 110°C for 13 hours and calcined at (500°C) for 5 hours. On alumina (SA-5239) support, the initial conversion of CH4 is higher than CO2 conversion then decreases with time until the CH4 conversion become lower than that of CO2. The conversion of both CO2 and CH4 decreased rapidly but the decrease of the later was significantly faster. The initial conversion of CH4 was 24.6 % and decreased to 6.5 %. in 10 hours time on stream. The H2:CO product ratio which was initially 1.42 decreased to approximately 0.45 at the end of the run. The catalyst was found to agglomerate at the end of the experiment and black carbon appeared. The results of testing the above catalysts are presented in figure 6. Therefore, alumina SA-5239 is not a suitable support for the dry reforming reaction because the support is not stable, agglomeration and deposition of carbon occurs. On using alumina (SA6175) support, figure 7 shows CH4 and CO2 conversion during the stability study that lasted for almost 86 hours. The catalyst performance is almost stable. The H2:CO product ratio is stable at 0.6 all the time and did not deactivate during that time. The initial conversion of CH4, CO2 and H2 yield are 16.9%, 26.2%, 15.3% and at the end time are 16%, 25.8% 15.3% respectively. Thus, alumina support (SA6175) catalyst is a suitable support for dry reforming reaction because the support was stable during the stability study which lasted for 86 hours. While on silica (S151-10) support, the CO2 conversion is higher than the CH4 Conversion. The conversion of both CO2 and CH4 decreased rapidly with time as shown in figure 8. The initial conversion of CH4 and CO2 were 7%, 13% and decreased to 1.8%, 5.3% in 13 hours respectively. The H2:CO product ratio was initially 0.51 and decreased to 0.28 by the end of the run. The catalyst was found to agglomerate at the end of the run and black carbon appeared. From this result we infer that silica (S151-10) is not a suitable support for dry reforming reaction because the support was not stable, agglomeration and deposition of carbon takes place. Finally summarizing the stability study, the obtained results indicate that catalyst supported with SA-6175 illustrate good stability during testing and not generating agglomerate catalyst. For that sake, catalysts supported with SA-6175 is chosen for further studies. 77 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


Effect of calcination and Ni loading on catalytic activity of supported Ni/ SA6175 Figures 9 shows the calcination result. It is obvious that CH4 conversions decreased with increasing calcination temperatures. The conversion of CH4 and CO2 obtained for the catalysts calcined at 500, 600, 650, 700 and 800°C temperatures and tested at 550°C are (17.1%, 16.4%, 14.0%, 1.3%, 0.61% and 23.7%, 23.7 %, 23.4%, 2.8%, 1.1%) respectively. The results showed that increasing calcination temperature above 600°C is not desirable, so 600°C is the recommended calcination temperature for this type of catalyst. Alternatively, the catalytic activity was investigated using the optimum calcinations temperature of 600 ºC at temperatures 500, 550, 575 ºC. The variation of CH4 conversion at different nickel loading and temperatures are shown in Figure 10. It is obvious that the conversion increased with increasing reaction temperature and nickel content. When the nickel loading was greater than 12 wt%, a slight change in conversion occurred which indicated that equilibrium was almost reached (thermodynamic equilibrium). When loading was 22% the catalyst was found to agglomerate at the end of the experiment. It was found that CO2 conversions were always higher than CH4 conversion which indicated that the reverse water- gas shift reaction took place.

CONCLUSIONS The supports used for preparation of Ni catalyst of dry reforming reaction are of two types of alumina (SA-6175, SA-5239) and silica (S151-10). Each support was investigated using calcination temperatures in the ranged from 500°C to 800°C. • The catalysts must be activated once before it is used for the first time. • Catalyst prepared by using pure support showed no activity at all calcination temperature used. • Alumina SA-6175 with high surface area showed good stability during testing without solidification. This made it the best support tested. • If the ratio H2/CO is higher than 1.0, solidification is observed and methane conversion is higher than CO2 conversion due to decomposition of methane. • It was found that 600°C is the optimum calcination temperatures for alumina support (Ni /SA-6175). The results also showed that increasing calcination temperature above 600°C decreases the performance and hence not desirable. • Optimum Ni loading is found to be at about 12%. Further increase (above 20%) of Ni loading leads to agglomeration catalyst. •



REFERENCES [1]

Rezaei, M., S.M., Alavi, S,. Sahebdelfar and Zi-Feng Yan, 2006, "Syngas production by methane reforming with carbon dioxide on noble metal Catalysts", Journal of Natural Gas Chemistry, Vol.15, pp. 327-334.

[2]

Hotz, N., M. J., Stutz, S. Loher, W. J., Stark and D., Poulikakos,2007, "Syngas production from butane using a flame-made Rh/Ce0.5Zr0.5O2 catalyst", Applied Catalysis B: Environmental, Vol. 73 , pp. 336-344.

[3]

Barelli, L., G., Bidini, A., Corradetti and U., Desideri, 2007, "Study of the carbonation–calcination reaction applied to the hydrogen production from syngas", Energy, Vol. 32, pp.697-710.

[4]

Zhong-Wen L., Xiaohong Li, K., Asami and K., Fujimoto, 2007, "Syngas to iso-paraffins over Co/SiO2 combined with metal/zeolite catalysts", Fuel Processing Technology, Vol.88, pp. 165-170.

[5]

Choudhary, V. R,. and A. S. Mamman,1999, " Oxidative conversion of methane to syngas over NiO/MgO solid solution supported on low surface area catalyst carrier", Fuel Processing Technology, Vol.60, pp.203-211.

[6]

Bhattacharyya, A., V. W., Chang and D. J,. Schumacher, 1998, " CO2 reforming of methane to syngas: I: evaluation of hydrotalcite clay-derived catalysts", Applied Clay Science, Vol.13, pp.317-328.

[7]

Cao, W., and D,. Zheng, 2006, " Exergy regeneration in an O2/CO2 gas turbine cycle with chemical recuperation by CO2 reforming of methane ", Energy Conversion and Management, Vol.47, pp.3019-3030.

[8]

Song, C., 2006, " Global challenges and strategies for control, conversion and utilization of CO2 for sustainable development involving energy, catalysis, adsorption and chemical processing", Catalysis Today, Vol. 115, pp.2-32.

[9]

Edwards, J. H., and A. M., Maitra, 1995, " The chemistry of methane reforming with carbon dioxide and its current and potential applications", Fuel Processing Technology, Vol. 42 , pp. 269-289

[10]

Al-Fatish, A.A, 2003,"Dry reforming of methane over supported nickel catalyst", M.Sc. Thesis, Chemical Engineering department, King Saud University, Saudi Arabia .



[11]

Guo, J., H., Lou and X., Zheng, 2007, " The deposition of coke from methane on a Ni/MgAl2O4 catalyst", Carbon, Vol. 45, pp. 1314-1321.

[12]

Boukha, Z., M,. Kacimi, M. F. R., Pereira, J. L,. Faria, J. L,. Figueiredo and M., Ziyad, 2007," Methane dry reforming on Ni loaded hydroxyapatite and fluoroapatite", Applied Catalysis A: General, Vol. 317, pp. 299-309.

[13]

Liu, B. S., and C. T., Au, 2003, " Carbon deposition and catalyst stability over La2NiO4/γ-Al2O3 during CO2 reforming of methane to syngas", Applied Catalysis A: General, Vol. 244 , pp. 181-195.

[14]

Goldwasser, M.R., M.E., Rivas, M.L., Lugo, E. Pietri, J. Pérez-Zurita, M.L. Cubeiro, A. Griboval-Constant and G. Leclercq,2005," Combined methane reforming in presence of CO2 and O2 over LaFe1−xCoxO3 mixed-oxide perovskites as catalysts precursors", Catalysis Today, Vol.107-108, pp. 106-113.

[15]

Ren J and Y .G , Chen., 1994, "Intermediate hydrocarbon species for the CO2-CH4 reaction on supported nickel catalysts", Catal . Lett., Vol.29, pp.33-37.

[16]

Lee, J-H, E-Gu, Lee, Oh-Sh., Joo and K-D, Jung, 2004, " Stabilization of Ni/Al2O3 catalyst by Cu addition for CO2 reforming of methane", Applied Catalysis A: General Vol. 269 , pp. 1–6.

[17]

Pompeo, F., N.N., Nichio, M.M.V.M., Souza, D.V., Cesar, O.A., Ferretti, M., Schmal,2007," Study of Ni and Pt catalysts supported on α-Al2O3 and ZrO2 applied in methane reforming with CO2", Applied Catalysis A: General, Vol. 316, pp. 175-183.



Figure 1: Schematic of experimental setup. (HV1, HV2 and HV3 Filters , MFC's Mass Flow Controller, PG Pressure Gauge CV1 , CV2 and CV3 Shut-off Valves, SV1, Sampling Valve, F Furnace, R Reactor, T1, T2, T3 Temperature Measurement (by Thermocouple), GC, Gas Chromatograph, PCV,Pressure Control Valve.)



SA-6175

S151-10

SA-5239

70 Conversion CH4 %

60 50 40 30 20 10 0 500

525 550 Reaction temperature ºC

Figure 2. Effect of different types of supports on CH4 conversion at 500 ºC calcination temperature.

70

SA-6175

S151-10

SA-5239

Conversion CO2 %

60 50 40 30 20 10 0 500

525 550 575 Reaction temperature ºC

Figure 3. Effect of different types of supports on CO2 conversion at 500 ºC calcination temperature.


575


SA-6175

S151-10

SA-5239

70

60

50

Conversion CH4 %

40

30

20

10

0

500

525

550

575

Reaction temperature ºC

Figure 4. Effect of different types of supports on CH4 conversion at 800 ºC calcination temperature . 45

40

SA-6175

S151-10

SA-5239

% C on version of C O 2

35

30

25

20

15

10

5

0 490

500

510

520

530

540

550

560

570

580

0

Reaction temperature, C

Figure 5. Effect of different types of support on CO2 conversion at 800°C calcinations temperature



40

35

CH4 30

CO2

F=10 ml/min , P=15 PSI

SA-5239

CH4:CO2=1 T=500 ºC

Conversion %

25

20

15

10

5

0 0

1

2

3

4

5

6

7

8

9

10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31

Time (h)

Figure 6 Variation of CH4 and CO2 conversion with time for Ni/SA-5239

40

35 CH4 30

CO2


SA-6175

CH4:CO2=1 T=500 º C

Conversion

25

20

15

10

5

0 -2

2

6

10

14

18

22

26

30

34

38

42

46

50

54

58

62

66

70

74

78

82

86

90

Time (h)

Figure 7 Variation of CH4 and CO2 conversion with time for Ni/SA-6175



20 CH4

18

CO2

16


S151-10

CH4:CO2=1 T=500 º C

14

Conversion

12

10

8

6

4

2

0 0

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

Time (h)

Figure 8 Variation of CH4 and CO2 conversion with time for Ni/S-151-10

T=500 ºC

T=550 ºC

T=575 ºC

70

60

Conversion CH4 %

50

40

30

20

10

0

500

550

600 650 700 calcination temperature ºC

750

800

Figure 9 Effect of calcination temperature on CH4 conversion



70

T=500 ºC

T=550 ºC

T=575 ºC

60

Conversion CH4%

50

F=10 ml/min , P=15 PSI 6175

SA-

40

30

20

10

0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 Loading Ni %

Figure10 Effect of Ni Loading on CH4 Conversion


KINETICS OF PARTICLE FORMATION IN THE SUPERCRITICAL GAS PROCESS Yousef Bakhbakhi1, Paul Charpentier2,, Sohrab Rohani2, Department of Chemical Engineering, King Saud University, Riyadh, Saudi Arabia 2 Department of Chemical and Biochemical Engineering, The University of Western Ontario, London, Ontario, Canada.

1

ABSTARCT The kinetics of the supercritical gas antisolvent crystallization (GAS) process using phenanthrene-toluene-carbon dioxide as a model system was investigated applying a rigorous mathematical model. This model accounts for the governing physical phenomena, i.e., the thermodynamics of near- or supercritical solutions, and the particle formation process controlled by primary and secondary nucleation, and crystal growth. Simulations were performed for changes in the main operating parameters, i.e., the antisolvent addition rate and saturation level. The simulations were performed at a process temperature of 25oC, while the antisolvent addition rate, Q A , was varied between 1 and 100 ml/min, and the initial solute concentration was varied between 25% and 100% of the concentration ratio. The model was successfully able to predict/represent the experimental observations phenomenologically. It was shown that the simulation findings were consistent with the experimental results, and good quantitative agreement was achieved.

KEYWORDS: supercritical fluids, crystallization kinetics, modeling, GAS.

1. INTRODUCTION The manufacturing of particles with controlled size and size distributions has attracted significant interest in the scientific and industrial communities with applications for pharmaceuticals, food, nutraceuticals, chemical, paint/coating, and polymer industries1-5. The important properties of these products are narrow particle size distribution, uniform morphology, and enantiomeric purity7. The employment of supercritical fluid techniques has attracted considerable interest as an emerging “green” technology for the formation of particles in these size ranges9. Particle formation using supercritical fluids (SCFs) can be carried out according to several different techniques, including antisolvent techniques such as the gasantisolvent (GAS) process. Antisolvent techniques exploit the low solubility of most compounds in the antisolvent, in particular CO2, which has to be miscible with


KINETICS OF PARTICLE FORMATION IN THE SUPERCRITICAL GAS PROCESS

the organic solvent10. In the GAS process, high pressure CO2 is injected into the liquid phase solution, which causes a sharp reduction of the solute solubility in the expanded liquid phase. As a result, precipitation of the dissolved compound occurs. The potential advantages of the GAS recrystallization process lies in the possibility of obtaining solvent free, micron and submicron particles with a narrow size distribution11. By varying the process parameters, the particle size, size distribution and morphology can be “tuned” to produce a product with desirable qualities. This makes the GAS technique attractive for the micronization of high-valued products, such as pharmaceuticals5. The theoretical studies that investigated the antisolvent processes primarily consisted of phase equilibria calculations. Muhrer et al12., to rationalize the precipitation kinetics in the GAS process, presented a model which accounts for solution thermodynamics and particle formation and growth. The developed model was constructed under the assumption of instantaneous phase equilibrium of the vapor and liquid phases upon antisolvent addition (i.e., there is no mass-transfer resistance). Their results showed that the model predicts correctly the variation of particle size and particle size distribution with the main operating parameter, the antisolvent addition rate. Their findings also demonstrated the possibility of adjusting the antisolvent addition rate in accordance with the final product specifications. The objective of this work was to provide a theoretical framework for the interpretation of the experimentally reported results. For a better understanding of the particle formation dynamics of the GAS process, the effect of process parameters, namely, antisolvent addition rate, and solute concentration on the particle size and size distribution of phenanthrene was examined. Furthermore, a mathematical model was developed to describe the elementary phenomena involved in the gas antisolvent crystallization process, i.e., the thermodynamics and the kinetics of particle formation that govern the process. 2.

THEORY

The aim of the mathematical modeling of the GAS process is to acquire a fundamental understanding of the crystallization mechanisms governing this unconventional crystallization technique, and how nucleation (primary and secondary), and growth occur during the expansion process, and how this affects particle size and size distribution. To determine the nature of the crystallization kinetics in the GAS process, the developed model has to relate the volumetric expansion of the liquid phase to the dynamics of particle formation. A proposed model by Muhrer12 was employed to fill the gap between the experimental results obtained from the GAS crystallization of phenanthrene from toluene using CO2 as antisolvent, and the theoretical 88 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007

Yousef Bakhbakhi , et al

understanding of the particles formation mechanisms and the influence of the process parameters in the GAS process.

CRYSTALLIZATION KINETICS In the GAS process, the supersaturation of the solute is created by dissolving a pressurized gas as an antisolvent into the liquid phase solvent. One of the most important process stages is the volumetric expansion of the liquid phase, which can be simulated using the Peng-Robinson equation of state (PR-EOS). The nucleation and growth of the particles in the GAS process is described using the population balance equation, which describes the evolution of the particle size distribution with time. The population balance approach to the analysis of crystallizers was formalized and presented by Randolph and Larson14. This technique accounts for both the size and number of particles. A number balance over a size range L and L + dL over an increment of time dt is given as:

∂n ∂ (nG ) n (1) + + = B (L ) − D (L ) ∂L ∂t τ where 'n' represents the population density of the crystals at time 't ' of a given size ' L' per unit mass of solvent per unit size, G represents the growth rate of particles, B(L ) represents the birth function of particles of size L at any instant by means of agglomeration and breakage, D(L ) is the death function representing the death of particles of size L by means of breakage at any instant. The equation for a semibatch crystallizer, in which particle agglomeration and breakage are not included, and particle shapes are uniform, and size-independent growth rate is given as:

∂n ∂ (n ) n d ( N L vL ) +G + =0 ∂t ∂L N L vL dt

(2)

The material balance on the antisolvent in the crystallizer is given by:

d ( N L x A + NV y A ) (3) = QA dt where Q A is the molar flow rate of antisolvent, and N L , NV , are the molar holdups of the liquid and the gas phase in the crystallizer, and x A , y A , are the mole fractions of the antisolvent in the liquid and the gas phase, respectively. The material balance on the solvent in the crystallizer is given by: 89 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


(

)

d N x +N y L S V S =0 dt

(4)

where xS , y S , are the mole fractions of the solvent in the liquid and the gas phase, respectively. The material balance on the solute, which relates the change in solute concentration to changes in magma density, is given by the following equations:

d ( N L xP + N P )

(5) =0 dt N v k m (6) NP = L L V 3 vP where N P , xP , kv,, vP and m3 are the molar hold-up of the solute in the solid phase, the mole fractions of the solute in the liquid phase, the volume shape factor, the molar volume of the solid solute, and the third moment of the population density function, respectively. The third moment of the population density is related to the total weight of crystals in the crystallizer, and is given as:

mi =

Lmax

∫ L n(L )dL ;

i=3

i

(7)

0

The boundary condition for the population balance equation, is the ratio of nucleation to growth rate at size zero.

n (t, 0) =

B G

(t > 0)

(8)

where B is the nucleation rate and G is the growth rate, which are given by the following equations:

B = B′ + B′′

( S > 1) ,

B′ = 1.5 D ( cP N A )

cP =

7

3

else B = 0

⎡ −16π ⎛ γ ⎞3 ⎛ vP ⎞ ⎛ 1 ⎞ 2 ⎤ exp ⎢ ⎟⎜ ⎜ ⎟ ⎜ ⎟ ⎥ kT N A ⎢⎣ 3 ⎝ kT ⎠ ⎝ N A ⎠ ⎝ ln S ⎠ ⎥⎦

γ vP

xP vL


(9)

(10)

(11)


B′′ =

⎡ ⎛ γ d 2 ⎞2 1 ⎤ exp ⎢ −π ⎜ M ⎟ ⎥ ⎢⎣ ⎝ kT ⎠ ln S ⎥⎦

α ′′av D d M4

(12)

av = ka m2 kT D= 2πηd M dM =

3

(13) (14)

vP NA

(15)

G = k g ( S − 1) if g

( S > 1) ,

else G = 0

(16)

where B′ is the primary nucleation rate, B′′ is the secondary nucleation rate, cP is the solute concentration, k tension, N A

is the Boltzman constant,

is Avogadro’s number,

α ′′ is

γ

is the interfacial

the secondary nucleation rate

effectiveness factor, av is the specific surface area, k a is the surface shape factor,

m2 the second moment of the population density function, D is the solute diffusion coefficient, η is the dynamic viscosity of the liquid phase, d M is the molecular diameter, and S is the supersaturation. The supersaturation is given as: f liq (17) S = Psol fP

⎡ v (P − P0 )⎤ f Psol = f Psol exp ⎢ P ⎥ RT ⎣ ⎦ 0

f Psol = f Pliq (P0 , T , x0 )

(18)

0

3.

(19)

RESULTS AND DISCUSSION

The population balance model was implemented in a dynamic simulation program that solves complex sets of partial integro-differential equations using a numerical algorithm of a finite-element type known as ‘Galerkin h–p method’ with a time discretization of Rothe’s type. The simulations were performed for the GAS crystallization of phenanthrene from toluene, as a model system, using carbon dioxide as the antisolvent. The process 91



temperature was kept constant at 25oC for all of the simulations. The antisolvent addition rate, Q A , was varied between 1 and 100 ml/min. The initial solute concentration was varied between 25% and 100% of the concentration ratio (the concentration ratio is defined as the ratio between the actual concentration of the liquid solution and the saturation concentration). The final pressure, where the antisolvent addition was terminated, is 60 bar. Particles are assumed to have spherical shape (i.e., the volume shape factor, kv = π 6 and the surface shape factor , k a =

π

). The physical properties of solid phenanthrene employed in the

implemented model are given in Table 1. The interaction parameters,

δ ij and η ij

k , Avogadro’s number, N A , and the values of interfacial tension, γ , the solute diffusion coefficient, D , and the dynamic viscosity, η , are taken from the literature and provided in Table 1, along with the (PR-EOS), Boltzman constant,

corresponding references. In the employed population balance model, the secondary nucleation rate effectiveness factor, α ′′ , was shown to be the only parameter controlling the qualitative effect of the process variables, i.e., the antisolvent addition rate, on the particle size, and size distribution of the final precipitate14 . The secondary nucleation rate parameter, α ′′ , is defined as a measure of the effectiveness of the physical events leading to secondary nucleation and ultimately determines which of the two phenomena between primary and secondary nucleation is predominant. Moreover, α ′′ is the only parameter in the relationships describing nucleation mechanisms that cannot be estimated based on physical properties of the model system. With the only exception of α ′′ , k g and g , the same set of parameters reported in Table 1, has been used in the accomplished simulations. The optimization algorithm varies the secondary nucleation rate effectiveness parameter, α ′′ , the growth rate parameter, k g , and the growth exponent, g , in the model; until the error between simulated and experimental variables is minimized. It is worth noting that the final time particle size distribution is the only variable included in the optimization objective function. Therefore, the objective is to find the value of the mentioned parameters, α ′′ , k g and g , that minimizes the difference between the measured and simulated particle size distribution. A correct match between the experimental and simulated particle size distribution is an indication of an accurate estimation of the primary and secondary nucleation rates. The estimated secondary nucleation rate parameter, α ′′ , within the investigated level of antisolvent addition rate of 1 ml/min , as presented in Table 2, is in the range of 8.11*10-17. An important means of testing any model is to use it for prediction of data not used to estimate the unknown model parameters, e.g. by 92 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


changing certain experimental conditions such as the antisolvent addition rate (50 and 100 ml/min), and initial solute concentration (25% solute concentration ratio). Failure of the model (within experimental error) to predict the full particle size distribution, indicates an incorrect/incomplete kinetic scheme. Figure 2 shows the highest value attained for both primary and secondary nucleation rates at the antisolvent addition rates of 1, 50 and 100 ml/min. At 100 ml/min addition rate, the maximum primary nucleation rate, B ′ , attained a higher order of magnitude of 9.9*1012 than 1.5*109 #/m3s for the secondary nucleation rate, B′′ . Thus, at this level of addition rate, primary nucleation is much faster than secondary nucleation, whose contribution to the final unimodal particle size distribution is far smaller. It is evident that the primary nucleation rate is more responsive to the antisolvent addition rate than the secondary nucleation rate. At the very beginning of the process, when particles do not yet exist, the dynamics of the GAS process is dominated by primary nucleation and antisolvent addition competition. The faster the addition rate, the higher the volumetric expansion rate, i.e., the supersaturation magnitude, where nucleation is initiated. Therefore, the higher the primary nucleation rate, the larger the number of nuclei and particles produced. The contrary phenomena happened at a low antisolvent addition rate of 1 ml/min. At this level of addition rate, primary nucleation had a lesser effect. In this case, secondary nucleation reaches higher rates than primary nucleation and thus, has a larger role in determining the final shape of the particle size distribution. At this level of the antisolvent addition rate, the maximum secondary nucleation rate, B′′ , attained a higher order of magnitude of 5.8*10 6 than 8.9*10 5 #/m3s for the primary nucleation rate, B ′ . Therefore, the primary nucleation burst forms enough particles and enough surface area to trigger secondary nucleation, whose rate under these conditions is large enough that the secondary nucleation burst forms much more particles than the primary one. However, the two nucleation bursts produce a closer number of particles. Nevertheless, the two sets of particles (i.e., those formed during the first and the second burst of nucleation) are born at different times, these grow for a longer and for a shorter time, respectively. As a consequence, the final particle size distribution is distinctively bimodal, as shown in Figure 1. At the intermediate level of antisolvent addition rate of 50 ml/min, the difference between the maximum values attained for both primary and secondary nucleation rates, is far less in comparison to the higher addition rate, (i.e., 100 ml/min). Moreover, the particles formed during the second burst of nucleation have a shorter time to grow than the primary nucleation generated particles (i.e., to a smaller and to a larger size, respectively). Therefore, the produced final particle size distribution is more broad than the 100 ml/min addition rate, as illustrated in Figure 1. However, the maximum primary nucleation rate, B ′ , attained a higher order of magnitude of 6.1*10 10 than 0.2*10 9 #/m3s for the secondary nucleation rate, B′′ , whose role into the shaping of the final unimodal particle size distribution is moderately smaller. 93 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


Comparison between the measured and simulated volume percent particle size distributions at 25% solute concentration ratio (50 ml/min antisolvent addition rate) is displayed in Figure 3. It is apparent that the predicted and experimental particle size distributions are in a good agreement. However, small discrepancies can be observed. Figure 4 shows the highest value attained for both the primary and secondary nucleation rates at the initial solute concentration ratio of 25%. At this solute concentration ratio, the maximum primary nucleation rate, B ′ , attained a higher order of magnitude of 0.95*1012 than 0.89*109 #/m3s for the secondary nucleation rate, B′′ , whose contribution to the final unimodal particle size distribution is far smaller. It is apparent that the primary nucleation rate is less sensitive to the initial solute concentration than the secondary nucleation rate, as the maximum primary nucleation rate attained a lower order of magnitude of 6.1*1010 #/m3s at 100% solute concentration ratio. At higher solute concentrations, the supersaturation profile tends to get quickly closer to the saturation line initiating a primary nucleation burst, and thus, longer time for the particles formed during the first burst of nucleation to grow, i.e., the growth mode dominates and superimposes to secondary nucleation, and thus, larger size particles with broad particle size distribution are produced. As shown in Table 3 of the simulation results, the particle size distribution of the produced particles were analyzed in terms of the moments of the distribution given by equation (7). The mean of the distribution , L , providing the average particle size:

L = m1 m0

(20)

Comparisons between the measured and simulated volume percent based average particle size for the tackled cases of the antisolvent addition rate, 1, 50, 100 ml/min (100% solute concentration ratio) and 25% solute concentration ratio (50 ml/min antisolvent addition rate) are displayed in Figure 5. It is apparent that the predicted and experimental results match reasonably well.

4.

CONCLUSIONS

In this work, a theoretical model was implemented to describe the elementary phenomena involved in the GAS crystallization process, through the thermodynamics and kinetics of particle formation that govern the process. It was demonstrated that the size and size distribution of the precipitated particles can be strongly influenced in the GAS process through the manipulation of the process parameters, antisolvent addition rate, and concentration. It was shown that the simulation results were consistent with the experimental results, and good quantitative agreement was achieved. The theoretical findings showed that the observed pattern of behaviour could be explained by invoking differences in the relative weight of the primary and of the secondary nucleation 94 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


rate. Furthermore, the obtained simulation results demonstrated the importance of the antisolvent addition rate and the initial saturation level to control the final particle size distribution and to tune it in accordance with the product specifications. In fact, if possible, i.e., in the case when primary nucleation is dominant or when primary and secondary nucleation are of similar importance, the average particle size of the product may be adjusted over a relatively wide range by selecting both the initial saturation level, and the antisolvent addition rate, accordingly.

ACKNOWLEDGMENTS We wish to acknowledge Dr. Gerhard Muhrer at Novartis Pharma AG for fruitful correspondence on the GAS modeling. The authors acknowledge the financial support of the Natural Sciences and Engineering Council (NSERC) of Canada, the Canadian Foundation for Innovation (CFI).

REFERENCES 1.

Cansell, F.; Chevalier, B.; Demourgues, A.; Etourneau, J.; Even, C.; Garrabos, Y.; Pessey, V.; Petit, S.; Tressaud, A.; Weill, F., 1999, "Supercritical Fluid Processing: a new route for materials synthesis". J. Mater. Chem. 9, 67-75.

2.

Subramaniam, B.; Rajewski, R. A.; Snavely, W. K., 1997, "Pharmaceutical Processing with Supercritical Carbon Dioxide". J. Pharm. Sci. 86, (8), 885890.

3.

Woods, H. M.; Silva, M. C.; Nouvel, C.; Shakesheff, K. M.; Howdle, S. M., 2004, "Materials processing in supercritical carbon dioxide: surfactants, polymers and biomaterials". J. Mater. Chem. 14, 1663-1678.

4.

Ye, X.; Wai, C. M., 2003, "Making Nanomaterials in Supercritical Fluids: A Review". J. Chem. Ed. 80, (2), 198-203.

5.

Tan, H. S.; Borsadia, S., 2001, "Particle Formation Using Supercritical Fluids: Pharmaceutical Applicatons". Exp. Opin. Ther. Patents 11, (5), 861872.



6.

Dixon, D. J.; Johnston, K. P., 1991, "Molecular thermodynamics of solubilities in gas antisolvent crystallization". AIChE J. 37, 1441 1449.

7.

York, P., 1999, "Strategies for particle design using supercritical fluid technologies". PSIT 2, (11), 430-440.

8.

Perrut, M., 2000, "Supercritical Fluid Applications: Industrial Developments and Economic Issues". Ind. Eng. Chem. Res. 39, 4532-4535.

9.

Gallagher, P. M. C., M. P.; Krukonis, V. J.; Klasutis, N., 1989, "Gas antisolvent recrystallization: new process to recrystallize compounds insoluble in supercritical fluids". In Supercrit. Fluid Sci. Technol., ACS Symposium Series 406: pp 334-354.

10.

Muller, M.; Meier, U.; Kessler, A.; Mazzotti, M., 2000, "Experimental study of the effect of process parameters in the recrystallization of an organic compound using compressed carbon dioxide as anti-solvent". Ind. Eng. Chem. Res. 39, 2260-2268.

11.

Muhrer, G.; Lin, C.; Mazzotti, M., 2002, "Modeling the gas antisolvent recrystallization process". Ind Eng Chem Res. 41, 3566-3579.

12.

Elvassore, N.; Parton, T.; Bertucco, A., 2003, "Kinetics of Particle Formation in the Gas Antisolvent Precipitation Process". AIChE Journal 49, 859-868.

13.

Randolph, A.; Larson, M. A., 1988, Theory of particulate processes. 2nd edition. Academic Press, Inc.: San Diego, CA.

14.

Bakhbakhi, Y.; Rohani, P.; Charpentier, P. A., 2006, "Micronization of Beclomethasone-17,21- dipropionate for Pulmonary delivery: Experimental Study Using The GAS Process". J.Pharm. Sci. 309 (1-2); 71.



Table 1. Thermodynamic and Kinetic Parameters for Model Simulations.

Parameter

Value

Units

Referenc e

vP

1.512x10-4 *

m3/mol

6

δ12 a

0.09*

dimensionles s

6

δ 23 b

0.12*

dimensionles s

6

δ13 c

0*

dimensionles s

6

ηij

0*

dimensionles s

6

k

1.38x10-23 **

J/K

11

NA

6.022x1023 **

mol-1

11

γ

1.75x10-2 **

J/m2

11

D

3.46x10-9

m2/s

η

3x10-4 **

Pa·s

11

a

Interaction parameter CO2-toluene (Peng-Robinson equation of state, PR-EOS). Interaction parameter CO2-phenanthrene (PR-EOS). c Interaction parameter toluene-phenanthrene (PR-EOS). *Dixon & Johnston, 1991. **Muhrer et al., 2002. b



Table 2. Simulation Results (parameter estimates).

α′

8.11*10-17

dimensionless

kg

4.49*10-5

m/s

g

1.87

dimensionless

Table 3. Simulation Results (moments and mean of the particle size distribution)*.

Simulation run

m0

m1 3

L

3

( µm )

(1 m )

(m m )

A1

9.52*109

1.466*106

153.992

A2

5.95*109

2.52*107

42.353

A3

6.52*1010

1.29*108

19.785

C1

4.78*1010

1.16*108

24.428

*Each simulation run is characterized by a label, indicates the type of the experiment according to the operating condition. The lable letters A1, A2, A3, and C1 points to the 1ml/min, 50 ml/min, 100 ml/min, and 25% concentration ratio experimental run.



12

10

Volume(%)

8

6

4

Experimental Simulation

2

0 1

10

100

1000

10000

Particle Size (microns)

Figure 1. Comparison between the experimental and simulated volume percent particle size distributions at the antisolvent addition rate of 100, 50 and 1 ml/min.

1 .0 E+ 1 4

P r im a r y N u c le a t io n 1 .0 E+ 1 2

S e c o n d a r y N u c le a t io n

Nucleation rate, #/m

s

1 .0 E+ 1 0

1 .0 E+ 0 8

1 .0 E+ 0 6

1 .0 E+ 0 4

1 .0 E+ 0 2

1 .0 E+ 0 0

A1

A2

A3

S im u la t io n r u n

Figure 2. The highest value attained for both primary and secondary nucleation rates at different antisolvent addition rates. The lable letters A1, A2, and A3, points to the 1 ml/min, 50 ml/min, and 100 ml/min.



10

Volume

8

6

Experimental 4

Simulation 2

0 1

10

100

1000

10000

P a r t i c l e S i z e ( m i c r o n s)

Figure 3. Comparison between the experimental and simulated volume percent particle size distributions at 25% solute concentration ratio.

1 .00 E+14

Pr ima ry Nuc leation 1 .00 E+12

Se c o ndary Nu c lea tion

Nucleation rate, #/m

s

1 .00 E+10

1 .00 E+08

1 .00 E+06

1 .00 E+04

1 .00 E+02

1 .00 E+00 C1

Sim u latio n r u n

Figure 4. The highest value attained for both primary and secondary nucleation rates at the initial solute concentration ratio of 25%.



180

Ξm

140

Predicted mean particle size,

160

120 100 80 60 40 20 0 0

20

40

60

80

10 0

120

140

160

180

200

Ex p e r im e n t a l m e a n p a r t ic le s iz e , Ξ m

Figure 5. Comparisons between experimental and simulated volume percent based average particle size.


PERFORMANCE STUDY OF MO-V-NB CATALYST FOR ETHANE OXIDATIVE DEHYDROGENATION TO ETHYLENE

Y.S. Al-zeghayera, S.I. Al- Maymanb, T.C. Al-Smaric, and A.A. Ibrahima a CHE, College of Eng., KSU, Box 800 , Riyadh 11421, Saudi Arabia, ,, [email protected] b Petroleum & petrochemicals Research Institue, KACST, Riyadh, Saudi Arabia c Sabic R& T, Riyadh, Saudi Arabia

ABSTRACT A suitably active and selective Mo-V-Nb catalyst for ethane oxidative dehydrogenation was developed and implemented. The sought-after study was carried out in a micro-reactor system. The catalysts were prepared by impregnation method. The study established the effect of some parameters that include: calcinations and drying temperatures, supports, and catalyst shapes, on the performance of Mo-V-Nb catalyst promoted with different elements. Catalyst samples were examined over reaction temperatures of 250, 280, 300 and 325 °C, 14.7 psi of pressure, 30 ml/min.gm (F/W) and the oxygen/ethane feed ratio of 1/3. Three different catalyst shapes were used for this study. Two catalysts, cylindrical and cylindrical with channel, were shaped by a German company for SABIC R&T. The third catalyst, spherical shape, was prepared at KSU lab by filing/rounding off cylindrical shape catalyst. Different supports, namely, alumina (SA-5239), silica (S151-10) and activated carbon were used. Palladium and other promoters were employed. Sensitivity analysis for the catalysts was carried out at different reaction temperatures. It was found that Spherical shaped catalyst gave better effect of performance than other catalyst shapes (cylindrical and cylindrical with channel).While Catalyst support with silica S151-10 gave better performance than catalyst supported with alumina SA-5239 and activated carbon. On the other hand it was noted that ethylene selectivity decreases by increasing the reaction temperature, while the selectivity of COx products increases with the increase of reaction temperature. Moreover, higher calcination temperatures were not favorable not only due to low C2H6 and O2 conversion, but also low C2H4 selectivity and yield and high COX selectivity and yield. On the other hand with respect of drying temperature on catalysts, it was found as following: For catalyst promoted with Pd, the results showed that the increase of drying temperature leads to lower C2H6 and O2 conversions, lower COX selectivity/yield and increase of C2H4 selectivity/yield. While catalyst promoted with Ni did not show significant effect of drying Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007

PERFORMANCE STUDY OF MO-V-NB CATALYST FOR ETHANE OXIDATIVE

temperature on catalyst activity. On the other aspect catalyst promoted with Ag and dried out at low temperature, 25 °C gave much better results compared to other catalyst dried out at 70 and 100 °C. While catalyst promoted with Pt and dried out at higher temperature, 70 and 100 °C gave much better results compared to other catalyst. Key words: Ethane, Mo-V-Nb catalyst, Oxidative-Dehydrogenation, Promoters

INTRODUCTION Saudi Arabia is endowed with a large production and reserve of natural gas Ethylene is produced by thermal steam cracking of hydrocarbons which may be ethane contained in natural gas, ethane-propane mixtures originating from refinery processes and naphtha [1]. The highly endothermic thermal cracking processes consumes a large amount of energy and involve significant formation of coke which requires frequent process shut-downs for its removing from the reactor. To prevail over the energy and coke problems associated with thermal cracking, recent technologies employ direct oxidative-dehydrogenation of ethane. This requires the development and instigation of a suitably active and selective catalyst [2,3]. Current catalytic system characteristically encounters an apparent selectivity/conversion barrier that limit single pass process yield to less than about 35%[4]. The decrease in selectivity with conversion is primarily due to secondary combustion reactions of the primary product, ethylene. High reaction temperature is generally used in the selective oxidation of short chain alkanes, especially ethane, as a consequence of their low reactivity [4]. V-containing catalysts have been widely used in the oxidative dehydrogenation (ODH) of alkanes [5]. Mo–V–Nb mixed oxides have been proposed to be the most active and selective catalysts in the ODH of ethane at relatively low reaction temperatures (300–400 °C) [6]. The influence of both the composition and the calcination conditions of Mo–V–Nb mixed oxides catalysts on their catalytic behavior in the ODH of ethane have been studied in the last years [7] Supported vanadyl phosphates catalysts were found more active than iron modified samples and catalysts calcined at 6500C gave better catalytic performances than those calcined at 5500C [8] Several investigators proposed Mo-V-Nb mixed oxides to be the most active and selective catalysts for ODH of ethane at relatively low reaction temperatures (3000C – 4000C) [9]. On the other hand, the study of oxidative dehydrogenation of ethane into ethylene by CO2 over a series of Silica – supported chromium oxide catalysts showed that the catalysts were effective for the reaction and CO2 in the feed promoted the catalytic activity [10] In this work, different catalysts were prepared using specific promoter/supports. These catalysts are tested under various conditions of temperatures, flow rates. The study covers the effect promoter type, drying temperature and calcination temperature and supports on catalyst performance. The ultimate goal is to shed light 104 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007

Y.S. Al-zeghayer , et al

on to Mo-V-Nb catalysts, for oxidative dehydrogenation of ethane in terms of activity and selectivity to ethylene. Experimental Impregnation method was used to prepare the catalysts in this study. 1.43 gm ammonium m-vanadate is dissolved in 75 ml distilled water while stirring and heating at 870C. Similarly, 2.394 gm niobium oxalate (21.5% Nb2O5) is dissolved in 75 ml distilled water while stirring and heating at 630C. The two solutions are mixed together while stirring and heating at 870C. After that, 3.69 gm oxalic acid powder is added to the final solution while stirring and heating at 880C. At this point, 5.38 gm ammonium p-molybdate is dissolved in to 75 ml distilled water stirring and heating at 600C. Then it is added to the former solution while stirring and heating at 850C. Now 0.1 gm of promoter is added to the resulting solution. After this, 10 gm of support (activated carbon, silica, or alumina) is added to the Mo-solution while stirring and heating at 70° C for 45 min until water evaporate, then calcinations take place for 4 hr at 700° C. During the calcination furnace temperature is raise to 3250C (10C/min) for 1.5 hr, different temperatures are tested (325,400 °C). The experimental equipment used in this study is micro-reactor system as shown schematically in Figure. 1. The feed section contains two gas cylinders for oxygen, nitrogen and ethane. Gases coming from regulators pass through in-line filters (molecular sieves, 5A) are then introduced to the Mass Flow Controllers (MFC), obtained from Bronkhorst. The gases are mixed and passed to the reaction section. On line samples (reactor By-Pass) from the feed gas mixture are directed to gas chromatograph for analysis. The micro-reactor overall length is 300 mm with inside diameter of made of 8 mm stainless steel and surrounded by heater. Surrounding temperature can be controlled. The reactor temperatures are measured at three locations by means of thermocouple located in the catalyst bed. The outlet from the reactor (bottom end) is passed through a back pressure regulator (BPR) to control the pressure in the reactor and the product gases from the BPR were sent to analysis section. Reaction products as well as feed mixture are analyzed on-line using Varian system model CP-3800 RGA gas chromatograph. Thermal conductivity detector (TCD) is used for analysis using Haysepe A Columns.



RESULTS & DISCUSSION The results for preparation and testing of promoted catalyst for ethane oxidative dehydrogenation to ethylene are presented. The study establishes the effect of drying temperature and calcination temperature, promoter type, and supports on catalyst performance. Catalyst samples were tested under reaction condition at temperatures of 250, 280, 300 and 325 °C, 14.7 psi of pressure, 30 ml/min.gm (F/W) and the oxygen/ethane feed ratio of 1/3.



Effect of Drying Temperature Different groups of catalysts (MoV Nb) were prepared with different promoters (Ni, Pt, Pd, Ag) and dried at different temperatures (25,70 and 100 °C) for 16 hrs and calcined at 350 °C for 1.5 hr. Testing the prepared catalysts at four different temperatures (250,280,300 and 325 °C), atmospheric pressure and feed flow rate of 15 ml/min, some of the results of testing these catalysts are shown in Figures 2-5 and table.1. Figures 2, of 1st group of catalyst Mo16V6.26Nb2.01Pd0.1(Pd-cat), shows that catalysts dried at 25°C provide higher 7% C2H6 conversion at 280°C, compared with other catalysts dried at higher temperatures. It was also obtained as shown in table1, lower C2H4 selectivity (14.7%) and lower C2H4 yield (0.23%) at 250°C. However, Pd-cat dried at 100°C gives higher selectivity (45%) when 325°C was used. In general Pd-cat dried at 25°C gives higher C2H4 selectivity compared with other catalysts this is due to reaction between ethane and O2. The decrease in C2H4 selectivity is accounted by an increase in CO2 selectivity and yield. Therefore, for the reaction temperature ranges of 250-325°C, .it can be summarized that increasing pd-cat drying temperature decreases C2H6 and O2 conversion, increases C2H4 selectivity/yield, and lowers CO2 selectivity / yield. Figure 3, of 2nd group of catalysts Mo16V6.26Nb2.01Ni0.1 (Ni-cat), depicts that there is no significant effect of drying temperature on C2H6 conversion. Therefore drying temperature provides no significant effect on Ni-cat performance. Figures 4, of 3rd group of catalysts Mo16V6.26Nb2.01Ag0.1 (Ag-cat), illustrates that catalyst dried at 25°C gives higher C2H6 conversion (5% at 250°C) compared to other catalysts dried at higher temperatures. However, at 300°C and 325°C, Ag-cats dried out at 70 and 100°C provide ethane conversion of (35 and 40%) respectively. On the other hand as table1 shows Ag-cat dried at 25°C gives higher C2H4 selectivity / C2H4 yield at all reaction temperatures compared to Ag-cats dried at higher temperatures. In contrast Ag-cats dried at 70 and 100°C give higher CO2 selectivity/yield compared to Ag-cat dried at 25°C. In conclusion, Ag-cats dried at higher temperatures did not improve performance. Figures 5 of 4th group of catalysts Mo16V6.26Nb2.01Pt0.1 (Pt-cat), shows that Pt-cats dried at 70 and 100°C provide higher conversion of C2H6 (10.5 and 11% respectively at 280°C) compared to Pt-cat dried at 25°C. But Pt-cat dried at 25°C gives higher O2 conversion (100% at all reaction temperatures) compared to Pt-cats dried at higher temperatures, which provide maximum 90% at 325 °C. In table 1, Pt-cats dried at higher temperatures give higher C2H4 selectivity/ yield (8.5% at 250°C) at all reaction temperatures compared to Pt-cat dried at 25°C. In summary Pt-cats dried at higher temperatures give much better results compared to Pt-cat dried at 25°C.



9 8

(Td=25 C)"

7

(Td=70 C)"

C2H6 Conv.%

6 5

(Td=100 C)"

4 3 2 1 0 220

240

260

280

300

320

340

Temperature,C

Figure 2. Effect of Drying Temperature on %C2H6 Conversion (Promoter=Pd) 20

(Td=70 C)

18

(Td=100 C)"

16

C2H6 Conv.%

14

(Td=25 C)

12 10 8 6 4 2 0 220

240

260

280

300

320

Temperature, C Figure3. Effect of Drying Temperature on % C2H6 Conversion (Promoter=Ni)


340


45 40

Td=25 C

%C2H6 Conv.

35 30

Td=70 C

25 20

Td=100 C

15 10 5 0 220

240

260

280

300

320

340

0

Temperature, C Figure4: Effect of Drying Temperature on %C2H6 Conv .(Promoter=Ag) 18

16

C2H6 Conv.%

14

12

10

8

(Td=70 C)

6

(Td=25 C)

4

(Td=100 C)

2

0 220

240

260

280

300

320

340

Temperature,C Figure5. Effect of Drying Temperature on C2H6 Conversion(Promoter=Pt)



Table 1: Effect of Drying Temperature:(Calcination Temperature: 350 C for 1.5 hr F/W: 30 ml/min.gm. Cat., P = 1 atm) Drying

Catalyst Composition

Tempe

Rxn Temp.

rature,

C

C

O2 Conv. %

C2H4

C2H4

Selec.%

Yield

CO2 Selec. %

CO2 Yield

Mo16V6.26Nb2.01Ni0.1

70

325

89.63

81.26

14.71

18.74

3.39

Mo16V6.26Nb2.01Ni0.1

70

300

53.67

88.24

10.47

11.76

1.39

Mo16V6.26Nb2.01Ni0.1

70

280

36.02

91.94

6.74

8.06

0.59

Mo16V6.26Nb2.01Ni0.1

70

250

25.16

93.86

3.3

6.14

0.22

Mo16V6.26Nb2.01Ni0.1

100

325

85.38

82.84

15.31

17.16

3.17

Mo16V6.26Nb2.01Ni0.1

100

300

52.91

88.8

10.7

11.2

1.35

Mo16V6.26Nb2.01Ni0.1

100

280

36.88

92.03

7.2

7.97

0.62

Mo16V6.26Nb2.01Ni0.1

100

250

25.53

94.53

3.43

5.47

0.2

Mo16V6.26Nb2.01Pd0.1

25

325

99.68

5.9

0.46

94.1

7.26

Mo16V6.26Nb2.01Pd0.1

25

300

99.69

2.44

0.18

97.56

7.18

Effect of Calcination Temperatures Different catalysts (Mo-V-Nb) prepared with Pd promoter were dried at 100 °C for 16 hrs and calcined at two different temperatures 325 and 400 °C for 1.5 hr. Testing the prepared catalysts, at four different temperatures (250,280,300 and 325 °C), atmospheric pressure and feed flow rate of 15 ml/min, provided the results given in the following figures 6 and 7 and. The purpose of calcination is to eliminate extraneous materials such as binders and un-wanted hydrocarbon as well as volatile 110 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


and unstable anions and cations that have been previously introduced, but is not desired in the final catalysts. Furthermore, the degrees of removal of un-wanted materials also affect the performance of the catalyst. Removal of these un-wanted hydrocarbons depends on the procedure of the calcination such as temperature, duration and atmosphere [11]. Increasing calcination temperature was found to: decreases (i) both C2H6 and O2 conversions particularly at lower reaction temperatures. (ii) Decreases C2H4 selectivity and yield. (iii) Increases COx selectivity and yield. Therefore increasing the calcination temperature is not favorable as the results displays that increasing the calcinations temperature decrease the catalyst activity. Figure 6 exhibits clearly that yield of C2H4 reduces with the increase of calcination temperature. While figure 7 depicts that selectivity of undesired product CO increase with the increase of the calcination temperature. Effect of Supports ON Mo-V-Nb-Pd Catalyst Different supports have been used in the preparation of Mo16V6.26Nb2.01Pd0.1 catalysts (alumina SA-5239, silica S151-10, and activated carbon). The catalysts were dried out at 100 °C for 16 hrs and calcined at 400 °C for 1.5 hrs. The catalytic activity was evaluated at 250, 280, 300, 325 °C using flow rate of 15 ml/min and atmospheric pressure. The results obtained show that un supported catalyst provides low C2H6 conversion except at 325 °C where the conversion shoots up. On the other hand the unsupported catalyst gives also higher O2 conversion compared to all another catalysts as shown in figure 8. While it also gives very low C2H4 selectivity therefore the C2H4 yield drops down. This decrease in C2H4 selectivity is accounted by an increase in CO selectivity which rends it quite undesirable at 325 °C. At lower temperature such as at 250°C, unsupported catalyst has a very high selectivity towards C2H4 however this is accompanied by a very low C2H6 conversion which brings again very low yield in ethylene production. The alumina SA-5239 catalyst shows very low conversion of C2H6 and O2 compared to other catalyst. While it gives high C2H4 selectivity. The low C2H4 yield (maximum 0.3% at 325 °C) of the catalyst is due to low C2H6 conversion. In addition it provides low CO2 selectivity as well as CO2 yield compared with other catalysts. Alternatively, the silica S151-10 catalyst provides low C2H6 conversion and higher O2 conversion compared to other supported catalysts. Moreover it shows higher C2H4 selectivity compared to other catalysts except at 325 °C, where alumina SA-5239 catalyst gives higher C2H4 selectivity. The silica S151-10 catalyst also manifests very low CO2 selectivity and yield. On the other hand, activated carbon support catalyst shows low C2H6 and O2 conversions, low C2H4 selectivity/yield and high CO2 selectivity which make it undesirable for performance selection. The silica S151-10 catalyst improves the reaction selectivity to ethylene compared with unsupported catalyst that gives higher COx selectivity. Therefore, it offers much better results compared to other catalysts.



1.6

Tc=325 C 1.4

C2H4 Yield

1.2

Tc=400 C

1 0.8 0.6 0.4 0.2 0 220

240

260

280

300

320

340

0

Temperature, C

Figure 6 Effect of Calcination Temperature on C2H4 Yield 100 90

Tc=325 C

80

CO Select.%

70

Tc=400 C

60 50 40 30 20 10 0 220

240

260

280

300

320

340

0

Temperature, C

Figure7. Effect of Calcination Temperature on CO Selectivity



100 90

No Support Alumina Sa-5239 Silica S151-10 Activated carbon

% O2 Conversion

80 70 60 50 40 30 20 10 0 220

240

260

280

300

320

340

0

Temperature, C Figure 8 Effect of Support on O2 Conversion



C2H6 Conv.%

Effect of Catalyst Shape Three catalyst shapes of Mo-V-Nb-Pd catalyst have been tested: (Cylindrical, Spherical, and Cylindrical with channel). The catalytic activity was evaluated at temperatures 240, 260, 280 °C using flow rate of 40 ml/min and atmospheric pressure. Figure 9 shows the effect of catalyst shape on ethane conversion. The graph underscores that spherical shaped catalyst has clear advantage at all three temperatures over other shaped catalysts.. For a rich hydrocarbon feed process certainly this will be very beneficial as it will reduce the recycle ratio. At temperature range between 240 to 280 °C, spherical catalyst selectivity ranges between 66.18-79.34% while cylindrical catalyst 60-79.28%. This fact places the spherical shape ahead of cylindrical because conversion of spherical is higher than cylindrical. Consequently, ethylene yield for spherical catalyst is higher On the other hand; cylindrical with channel catalyst has selectivity range of 51.5-62.7% which is much lower than other two catalysts. Moreover, it adds to the selection of spherical shaped catalyst. Hence, considering all the evaluating parameters, spherical shaped catalyst has shown the best performance over other shapes, cylindrical shaped catalyst comes next in performance while cylindrical with channel catalyst has shown the least performance. This can be explained with the fact that for same dimension surface area for sphere is largest and this implies that feed gas molecules will have more chance to get in contact with active sites of catalyst and therefore better results.

4.5 4 3.5 3 2.5 2 1.5 1 0.5 0 230

Cylindrical Spherical Cylindrical with channel

240

250 260 270 Temperature,C

280

290

Figure9. Effect of Catalyst Shape on C2 H6 Conversion



%C2H6 Conv.

Effect of Promoters Four promoters (Pd, Ni, Ag and Pt), have been used in preparation of Mo16V6.26Nb2.01 catalysts. A group of 8 catalysts were prepared while using promoters of 0.1% loading. Catalysts were dried out at 100°C for 16 hr and calcined at 325°C and 400°C. The prepared catalysts were tested at four different temperatures 250, 280, 300 and 325°C at atmospheric pressure and feed flow rate of 15 ml/min. In figure 10 catalyst promoted with nickel and calcined at 400°C showed the highest C2H6 conversion at all reaction temperatures. It gave also high C2H4 selectivity and yield. While Pt-promoted catalyst showed a very low C2H6 conversion and the C2H4 selectivity at all temperatures is almost negligible. Moreover CO and CH4 selectivity for Pt-promoted catalyst is also higher compared to other catalysts, thus it is not suitable for ethane ODH to ethylene reaction. On the other hand, Ag.promoted catalyst behaves better than Pd-promoted catalyst in terms of C2H6 conversion except at 325°C, C2H4 selectivity except at 250°C. However, C2H4 yield for Ag-promoted catalyst is always higher than Pd-promoted catalyst. In figure 11 Pd-promoted catalyst and calcined at 325°C showed low C2H6 conversion. Similarly, Pt-promoted catalyst showed low C2H6 conversion as well as low C2H4 selectivity/yield and high CO and CH4 selectivity making it undesirable candidate. On the other hand, Ag-promoted catalyst gave better C2H6 conversion than Pdpromoted catalyst. Again at this calcination temperature, Ni-promoted catalyst exhibited better results when compared to all other catalysts in terms of C2H6 conversion. In consequence, Ni-promoted catalyst with 0.1% loading and calcined at 325°C gave much better activity compared to other catalysts. 25

Pd

20

Ni

15

Pt

10

Ag

5

0

220

240

260

280

300

320

340

0

Temperature, C

Figure10 :Effect of Promoters on % C2H6 Conversion (Tc=4000 C)



25

%C2H6 Conv.

20

Pd

Ni

Pt

Ag

15

10

5

0 220

240

260

280

300

320

340

0

Temperature, C 0

Figure11: Effect of Promoters on %C2H6 Conversion(Tc=325 C)

CONCLUSIONS The following conclusions are drawn from the results of this work: • It was found that, ethylene selectivity decreases by increasing the reaction temperature, while the selectivity of COx products increases with the increase of reaction temperature • No significant effect of drying temperature upon the activity of Ni-promoted catalyst. • Ag-promoted catalyst dried out at low temperature, 25 °C gave much better results compared to those dried out at 70 and 100 °C • Pt-promoted catalyst dried out at higher temperature, 70 and 100 °C gave much better results. • Increasing drying temperature decreases the C2H6 and CO2 conversions and increases C2H4 selectivity/yield and decrease of COX selectivity/yield for Pdpromoted catalyst. • It was found that higher calcination temperatures were not favorable not only due to low C2H6 and O2 conversion, but also low C2H4 selectivity and yield and high COX selectivity and yield. • Catalyst support with silica S151-10 gave better performance than catalyst supported with alumina SA-5239 and activated carbon. • Spherical shaped catalyst gave better effect of performance than other catalyst shapes (cylindrical and cylindrical with channel). • Ni is the best element tested for promoting prepared Mo V Nb catalyst. 116 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


REFERENCES: 1. Kniel, L., O., Winter, K, Stork,.1980,"Ethylene: Keystone to the Petrochemical Industry", Macel Dekker, New York. 2. Donsì, F., S. Cimino, R. Pirone, G. Russo and D. Sanfilippo. " Crossing the breakthrough line of ethylene production by short contact time catalytic partial oxidation",2005, Catalysis Today, Vol. 106, pp.72-76 3. James, E., M., Mary, E., Lindesey, G., Allen, Z., Conrad, R., Rajeev, W., George, M., Amitesh, K., Dominic, K.," Oxidative dehydrogenation of ethane over iron phosphate catalysts",2002., Applied Catalysis A: General , Vol. 231, pp.281-292. 4. Concepción, P., M.T. Navarro, T. Blasco, J.M. López Nieto, B. Panzacchi and F. Rey., " Vanadium oxide supported on mesoporous Al2O3: Preparation, characterization and reactivity ", 2004, Catalysis Today, Vol. 96, pp.179-186. 5. Klose, F., T. Wolff, H., Lorenz, A.S., Morgenstern, Y., Suchorski, M., Piórkowska and H., Weiss , "Active species on γ-alumina-supported vanadia catalysts: Nature and reducibility", 2007, Journal of Catalysis, Vol. 247, pp176-193. 6. Botella, P., E. García-González, A. Dejoz, J. M. López Nieto, M. I. Vázquez and J. González-Calbet, " Selective oxidative dehydrogenation of ethane on MoVTeNbO mixed metal oxide catalysts",2004, Journal of Catalysis, Vol. 225, pp. 428-438 7. Botella, P., J. M, López Nieto,. A, Dejoz, M. I ,Vázquez,. and A., MartínezArias, " Mo–V–Nb mixed oxides as catalysts in the selective oxidation of ethane", 2003, Catalysis Today, Vol. 78, pp.507-512. 8. Casaletto, M. P., L. Lisi, G. Mattogno, P. Patrono, G. Ruoppolo and G. Russo, " Oxidative dehydrogenation of ethane on γ-Al2O3 supported vanadyl and iron vanadyl phosphates: Physico-chemical characterisation and catalytic activity",2002, Applied Catalysis A: General, Vol. 226 pp.41-48. 9. Botella, P., A. Dejoz, J.M. López Nieto, P. Concepción and M.I. Vázquez, " Selective oxidative dehydrogenation of ethane over MoVSbO mixed oxide catalysts", 2006, Applied Catalysis A: General, Vol. 298, pp.16-23. 10. Mimura, N., I. Takahara, M. Inaba, M. Okamoto and K. Murata., "Highperformance Cr/H-ZSM-5 catalysts for oxidative dehydrogenation of ethane to ethylene with CO2 as an oxidant", 2002, Catalysis Communications, Vol. 3, pp. 257-262. 11. Desponds, O., R.L Keiskie,. and G.A Somorjai,." Kinetics of the water-gas shift reaction over several alkane activation and water-gas shift catalysts", 1993, Applied Catalysis A: General, Vol., 101, pp. 317-338.


EFFECT OF ZEOLITE TYPE AND METAL CONCENTRATION ON THE CATALYTIC ACTIVITY OF NANOPOROUS CATALYSTS FOR CONVERSION OF HEAVY AROMATICS INTO XYLENES M. A. Ali1, S. Al-Khattaf, S. A. Ali and K. Al-Nawad, Research Institute, Research Institute, King Fahd University of Petroleum & Minerals, Dhahran, Saudi Arabia, and T. Okamoto and K. Ishikawa, Nippon Oil Research Institute Co., Ltd. Tokyo, Japan. Email: [email protected] 1 Correspondance Author, Email: [email protected]

ABSTRACT Four catalysts evaluated in a fixed-bed reaction system were zeolites-based and are as described below: Molybdenum loaded mordenite based catalyst (Mo-MOR), Rhenium loaded mordenite based catalyst (Re-MOR), High Molybdenum-Cerium loaded mordenite based catalyst (5Mo-5Ce-MOR), and Low Molybdenum-Cerium loaded USY based catalyst (3Mo-3Ce-USY). The feed containing 95% C9 aromatics was used, and the reaction conditions were as follows: reaction temperatures; 340, 360, 380, 400, 420 and 440 ºC, liquid hour space velocity; 1.0, 1.5 and 3.3, hydrogen to feed ratio; 2.0 and hydrogen carrier gas pressure; 3 Mpa. The reaction products were separated into gas and liquid portions and were analyzed using gas chromatography based Analyzers. All four catalysts exhibited increase in C9 conversion with increase in temperature and decrease in space velocity. Re-MOR catalysts produced more gaseous products as compared to Mo-MOR catalyst. Mordenite-based catalysts offered a number of advantages over the USY-based catalysts which include higher C9 conversion, higher xylenes selectivity, and reduced C10+ and ethylbenzene production, whereas the disadvantages are higher yield of benzene and toluene. The results revealed that high loading of metals decreased C9 conversion. It was determined that for the conversion of C9 aromatics to xylenes, mordenite zeolite are better than USY. Keywords: Transalkylation, Nanoporous, Mordenite, USY, Xylenes, Heavy aromatics.


EFFECT OF ZEOLITE TYPE AND METAL CONCENTRATION ON THE CATALYTIC ACTIVITY

1.0 INTRODUCTION A joint research program has been started between the Nippon Oil Research Institute Co., Japan and King Fahd University of Petroleum & Minerals (KFUPM), Dhahran, Saudi Arabia. The main objective of this program is to develop novel catalytic systems for efficient conversion of heavy aromatics, improved xylene selectivity, and reduced higher aromatics (C10+) formation and lower gas formation. Transalkylation is the process that converts low-value toluene and heavy aromatics into value-added mixed xylenes. Transalkylation processes use silica-alumina and zeolites such as dealuminated mordenite, ultra stable Y-zeolite (USY) and ZSM-12 [1]. Nippon Oil Corporation has demonstrated a new technology for the production of mixed xylenes via the transalkylation of benzene and C9 aromatics [2]. A highactive catalyst based on nickel supported high silica mordenite was used in the production of toluene and mixed xylenes. The catalyst showed higher C9 aromatics conversion activity with less deactivation and higher selectivity to mixed xylenes. Transalkylation technologies have been developed by a number of companies including UOP, ExxonMobil, Toray, and SK Corporation. The Advanced Transalkylation (ATA) process was developed by SK Corporation to produce high xylenes from the transalkylation of C9+ aromatics and utilizing a catalyst based on platinum-promoted mordenite. Platinum was used for its hydrogenation function along with tin or lead to control the hydrogenating activity [3]. The catalyst exhibited strong dealkylation activity, the ethyl and propyl groups were easily dealkylated to provide high BTX yield and a low concentration of ethylbenzene and propylbenzene in the product. Toluene disproportionation also occurred as evidenced by the high conversion of toluene. As the concentration of C9+ aromatics increased in the feed, the TDP reaction was suppressed and hence the toluene conversion and benzene yield decreased while mixed xylenes yield increased. Other processes also utilized zeolite based catalysts mainly mordenite loaded with metals such as molybdenum, rhenium and nickel. Mixed xylenes fraction consists of three isomers namely: para-xylene, ortho-xylene, and meta-xylene (Figure 1). The ortho-xylene and meta-xylene are processed further by isomerization reaction to produce para-xylene, which is used for the production of polyester fibers, resins, and films. Because of the growth in demand for para-xylene, new technologies are needed to convert low value aromatics (C9A+) into mixed xylenes and specifically para-xylene.

120


M. A. Ali , et al.

Figure 1. Isomers in mixed xylenes fraction [2]. During the process of xylene production, a number of reactions such as transalkylation, disproportionation, and dealkylation take place, The methyl groups are shifted from one benzene ring to the other via transalkylation and disproportionation to produce mixed xylenes. During dealkylation, the ethyl, propyl, and butyl groups attached to benzene and methylbenzenes are removed to produce benzene and methylbenzene. Undesirable reactions that may occur include ring saturation (lead to substituted cyclohexanes), hydrocracking (lead to ring opening) and coking and condensation (many benzene rings condensed to produce polynuclear aromatics). Since large aromatic molecules such as trimethylbenzenes, methylethylbenzenes and heavier alkylsubstituted aromatics (C10+) are involved in the transalkylation reaction; relatively large-pore zeolites are able to catalyze this reaction towards maximum xylenes formation. The literature reported a variety of supports and metals used to prepare the catalysts for these reactions. The supports utilized include zeolites including mordenite, Y, beta, L, omega, and MCM and SAPO while the metals that have been used include molybdenum, bismuth, chromium, cobalt, copper, iron, nickel, lead, lanthanum, palladium, platinum, silver and zirconium. Catalysts having strong acid sites are less selective for producing mixed xylenes and are easily deactivated by coke. Zeolites with small and medium pores such as ZSM-5 act only by the external active sites. However, ZSM-5 alone can be applied for toluene disproportionation reaction, but is not suitable for disproportionation and transalkylation of C9 or higher aromatic hydrocarbons because of its pore size limitation. Mordenite and zeolite Y have better performance in the conversion of C9+ aromatics to BTX.

121



1.1 Literature Review Transalkylation and disproportionation are the two main processes for the conversion of aromatics by alkyl group transfer reactions, especially for the production of dialkylbenzenes. Transalkylation process provides conversion of toluene and heavy aromatics, for example C9 alkylbenzenes, to xylenes, while disproportionation reaction yields benzene and xylenes using toluene as a feed. The mixed xylene yields produced by conventional transalkylation and disproportionation processes are strictly controlled by thermodynamic equilibrium isomer distribution. Since the thermodynamic equilibrium xylene yield is known to increase with increasing feed methyl-to-ring ratio, transalkylation reaction is more favorable than toluene disproportionation especially in terms of the demand of xylenes over benzene and the availability of excessive heavy aromatics [4]. Developments of cost effective transalkylation process tend to aim at the utilization of heavy aromatics which possess high methyl-to-ring ratio. This in turn, promotes the thermodynamic equilibrium xylene yields and also the feed cost is lowered. Zeolites that possess 12 membered ring pore aperture, including beta [5,6], mordenite [7] and USY [8] are the potential solid acid catalysts for conversion of alkylbenzenes. The major problem associated with heavy aromatics feed is that it contains catalyst poisoning components, which deactivate the catalyst due to coke formation and hence its efficiency and cycle length are reduced. It has been reported that the coke formed during toluene disproportionation over USY is mostly polyaromatic in nature, and its structures depend on the reaction temperature; the higher the temperature the more condense rings are formed during coking [9]. A probable solution to resolve the aging problem during heavy aromatics processing is to incorporate metals into the zeolite catalyst to hydrogenate polyaromatics coke precursors. Different zeolites incorporated with various metals (Cu, Ni, Pt, etc.) have been studied for transalkylation [6,7,10-12]. Roger et al. [13,14] studied the conversion of 1,2,4-trimethylbenzene over amorphous silica-alumina and HZSM-5 in gas phase. It was concluded that paring reaction played a decisive role during the conversion over H-ZSM-5 at elevated temperature (450 °C) and in the reaction sequence; xylenes and TeMBs were produced as intermediates. Over both silica-alumina and H-ZSM-5, isomerization of 1,2,4-TMB to 1,2,3-TMB and 1,3,5-TMB were found to be the fastest reaction having 90% selectivity. This reaction was believed to occur on the external surface of the zeolite crystals. They also reported that over silica-alumina, the reaction almost terminated at the disproportionation step, whereas over H-ZSM-5, the disproportionation of 1,2,4-TMB was followed by rapid paring dealkylation of the TeMBs. Cejka and others [15] used beta, Y, L and mordenite type zeolites which exhibited significant difference in conversion. They found higher conversions of TMBs and high selectivity to xylenes and TMBs at 400 °C with Y and beta compared to other large pore zeolites (mordenite and zeolites L). Collins and co122


M. A. Ali , et al.

workers [16] have reported transformation of trimethylbenzenes over Y zeolitebased catalysts. They have concluded that disproportionation reaction seems to be directly proportional to the total conversion over most of the conversion range studied for both 1,2,4- and 1,3,5-trimethylbenzene (TMB), whereas for 1,2,3-TMB, isomerization was much more favored than disproportionation (Figure 2). In the disproportionation reaction, the transfer of one methyl group from TMB led to the formation of equal amount of xylene and tetramethylbenzene (TeMB) isomers. It has been concluded by Matsuda and coworkers [17] that isomerization reaction of TMB were found to occur only over Bronsted acid sites, whereas both Bronsted and Lewis sites are responsible for disproportionation reaction.

Figure 2. Reactions scheme for conversion of trimethylbenzene [15]. This paper presents the results of four catalysts utilized for the conversion of heavy aromatics into xylenes. Effect of zeolite type, metal type and their concentrations on the catalytic activity is discussed and highlighted.

2.0 EXPERIMENTAL 2.1 Catalysts Preparation Four catalysts were used in this study and are described as follows: Molybdenum loaded mordenite based catalyst (Mo-MOR), Rhenium loaded mordenite based catalyst (Re-MOR), 3%Molybdenum-3%Cerium loaded USY based catalyst (3Mo3Ce-USY), and 5%Molybdenum-5%Cerium loaded USY based catalyst (5Mo-5Ce123



USY). The zeolite mordenitre and USY were acquired from Tosoh Chemicals, Japan. Alumina binder AP# was obtained from Chemicals and Catalysts Industries Company, Japan. All catalysts were prepared in extrudate form and the metal loading was performed by impregnation method. Mo-MOR was prepared by the following method: Alumina binder was dispersed in deionized water and added with Mordenite zeolite (HSZ-690HOA), mixed and kneaded well. The kneaded material was converted into 1.5 mm extrudates. The ratio of mordenite zeolite to AP3 was 2:1. The extrudates were dried and calcined. The extrudates were impregnated with Mo to achieve 3 wt% loading. The extrudates were again dried and calcined. ReMOR was prepared in the similar way using the same type of materials as Mo-MOR catalysts except the metal loaded was 0.5 wt% rhenium. 3Mo-3Ce-USY was prepared by the following method. The alumina binder was dispersed in deionized water, added with USY (HSZ-370HUA), mixed, kneaded well, converted into 1.5 mm extrudates, dried at 120 ºC and calcined at 500 ºC for 2 hours. The extrudates were impregnated sequentially with 3 wt% Mo and 3 wt% Ce using hexaammonium heptamolybdate tetrahydrate and cerium nitrate hexahydrate salts. The resulting catalyst was dried and calcined. 5Mo-5Ce-USY was prepared in the similar way using the same type of materials as 3Mo-3Ce-USY catalysts except the Mo and Ce metal loading was 5 wt% each., , 2.2 Characterization These catalysts were characterized for surface area and porosity characteristics using NOVA 1100 system. The degassing temperature for each catalyst was 250 ºC maintained for 10 min under helium atmosphere. After cooling, nitrogen gas was adsorbed and then desorption process was started. At the end of the run, the system performed the necessary calculations and results were obtained. 2.3 Catalytic Evaluation All four catalysts were evaluated for transalkylation reactions. These catalysts were reduced under hydrogen gas flow prior to catalytic reaction. The catalysts were evaluated in fixed-bed reaction system in the temperature range 340 to 440 ºC at 1.0, 1.5 and 3.3 LHSV using the aromatic feedstock containing 95% C9 aromatics. The liquid reaction products were analyzed using a gas chromatography based PIONA Analyzer. 3.0 RESULTS AND DISCUSSION 3.1 Characterization The surface area and porosity characteristics values for all four catalysts are given in Table 1. The results indicated higher surface area for USY-based catalysts as compared to mordenite-based catalysts. This is because the USY zeolite obtained from Tosoh Chemicals in powder form has a surface area of 640 m2/g while modenite has 420 m2/g. Following the similar trend, USY-based catalysts have higher total pore volumes and average pore radii. 124


M. A. Ali , et al.

3.2 Catalytic Evaluation The catalytic results obtained for converting C9+ into C6, C7 and C8 were evaluated. Figure 3 shows C9 conversion obtained as a function of temperature at 1.5 LHSV. C9 conversion was observed increasing with increase in reaction temperature. Mordenite-based catalysts exhibited higher C9 conversion as compared to USYbased catalysts. The plot revealed that at 1.5 LHSV and 30 bar hydrogen pressure and reaction temperature of 400 ºC, Mo-MOR exhibited 69% C9 conversion and 58% xylene selectivity. Table 1. Surface area and Porosity of Catalysts. Catalyst Designation Mo-MOR Re-MOR 3Mo-3Ce-USY 5Mo-5Ce-USY

Surface Area (m2/g) 389 384 503 437

Total Pore Volume (cc/g) 0.36 0.31 0.59 0.49

Average Pore Radius (Angstrom) 16.6 16.4 23.4 22.5

Figure 4 shows C9 conversion plots obtained using four catalysts as a function of reaction temperature in the range 340 to 440 ºC at 3.3 and 1.0 LHSV. C9 conversion was observed increasing with reaction temperature and decreasing space velocity. Both plots indicated that Mo-MOR exhibited higher C9 conversion as compared to Re-MOR catalyst at 3.3 and 1.0 LHSV.. The plot exhibited that Mo-MOR exhibited 70% C9 conversion at 3.3 LHSV and 440 ºC, which is similar to C9 conversion at 1.0 LHSV and 400 ºC.

C9 Conversion (wt%)

80

340

360

380

400

40

0 Mo-MOR Re-MOR

3Mo5Mo3Ce-USY 5Ce-USY

Figure 3. C9 conversion as a function of reaction temperature at 30 bar hydrogen pressure and 1.5 LHSV. 125


80

340 400

360 420

380 440

40

80 340

C9 Conversion@ 1.0 LHSV (wt%)

C9 Conversion@ 3.3 LHSV (wt%)


360

380

40

40

0 Mo-MOR

Re-MOR

0 Mo-MOR

Re-MOR

Figure 4. C9 conversion as a function of reaction temperature at 30 bar hydrogen pressure and at 3.3 and 1.0 LHSV.

80 340

360

380

400

40

0 Mo- Re- 3Mo- 5MoMOR MOR 3Ce- 5CeUSY USY

p-Xylene Selectivity @ 1.5 LHSV (wt%)

Xylenes Selectivity@ 1.5 LHSV (wt%)

Figure 5 shows two plots of xylenes selectivity and p-xylene selectivity as a function of temperature for all four catalysts at 1.5 LHSV. Xylenes selectivity was found increasing with reaction temperature at all space velocity. At 1.5 LHSV, xylenes selectivity and p-xylene selectivity of Re-MOR is slightly better than MoMOR catalyst. Xylenes selectivity and p-xylene selectivity of 3Mo-3Ce-USY and found better than 5Mo-5Ce-USY catalysts probably due to less metal loading. This trend is also reflected in Figure 6 that shows higher xylenes selectivity of Mo-MOR compared to Re-MOR at 3.3 LHSV in the temperature range 340 to 440 ºC. 10

340 380

360 400

5 0 Mo- Re- 3Mo- 5MoMOR MOR 3Ce- 5CeUSY USY

Figure 5. Xylenes selectivity and p-xylene selectivity as a function of reaction temperature for four catalysts at 30 bar hydrogen pressure and at 1.5 LHSV. Figure 7 shows that the concentration of C10 and C11+ in the reaction products obtained at 1.5 LHSV and were found lower for mordenite-based catalysts compared to USY-based catalysts. This is because mordenite does not facilitate formation of higher aromatics due to its unique pore size and structure as compared to USY. Figure 8 shows the benzene and toluene concentration in the reaction products at 1.5 LHSV and 30 bar hydrogen pressure. Benzene and toluene concentrations were 126


M. A. Ali , et al.

observed higher for mordenite-based catalysts as compared to USY-based catalysts. This is because of the higher pore size and three-dimensional pore structure of USY zeolite.

Xylenes Selectivity@ 3.3 LHSV (wt%)

Figure 8 exhibited ethylbenzene concentration obtained at 30 bar hydrogen pressure and 1.5 LHSV in the temperature range 340 to 400 C. An interesting trend was observed for EB concentration associated with the type of zeolite used in the catalyst preparation. The EB concentration increased with rise in temperature for USY-based catalysts while decreased for mordenite-based catalysts with increase in temperature. EB concentration was observed lower for mordenite-based catalysts compared to USY-based catalysts. 80

340 420

360 440

380

400

40

0 Mo-MOR

Re-MOR

25 340

360

380

400

0 MoMOR

Re- 3MoMOR 3CeUSY

5Mo5CeUSY

C11+ Concentration (wt%)

C10 Concentration (wt%)

Figure 6. Xylenes selectivity as a function of reaction temperature for two catalysts at 30 bar hydrogen pressure and at 3.3 LHSV. 340 5

360

380

400

0 MoMOR

ReMOR

3Mo3CeUSY

5Mo5CeUSY

Figure 7. C10 and C11+ concentration obtained at 30 bar hydrogen pressure and 1.5 LHSV in the temperature range 340 to 400 C.

127


3340

360

380

400

2 1 0 MoMOR

ReMOR

3Mo3CeUSY

5Mo5CeUSY

Tol. Concentration (wt%)

B Concentration (wt%)


20

340 380

10

360 400

0 MoMOR

ReMOR

3Mo3CeUSY

5Mo5CeUSY

EB Concentration (wt%)

Figure 8. Benzene and toluene concentration at 30 bar hydrogen pressure and 1.5 LHSV in the temperature range 340 to 400 C. 3

340

360

380

400

2 1 0 Mo-MOR

Re-MOR 3Mo-3Ce- 5Mo-5CeUSY USY

Figure 9. Ethylbenzene concentration obtained at 30 bar hydrogen pressure and 1.5 LHSV in the temperature range 340 to 400 C. CONCLUSIONS Following conclusions were drawn from this study: 1. 2.

3. 4. 5.

All four catalysts exhibited increase in C9 conversion and xylene selectivity with increase in temperature and decrease in space velocity. Mordenite-based catalysts offered a number of advantages over the USY-based catalysts which include higher C9 conversion, higher xylenes selectivity, and reduced C10+ and ethylbenzene production, whereas the disadvantages are higher yield of benzene and toluene. The results revealed that high loadings of metals on the catalysts decreased C9 conversion. It was determined that for the conversion of C9 aromatics to xylenes, mordenite zeolite based catalysts performed better than USY. Mo loaded mordenite based catalysts performed better in terms of higher C9 128


M. A. Ali , et al.

6.

7. 8.

conversion as compared to Re loaded catalyst. The EB concentration increased with rise in temperature for USY-based catalysts while decreased for mordenite-based catalysts with increase in temperature. EB concentration was observed lower for mordenite-based catalysts compared to USY-based catalysts. Benzene and toluene concentrations were observed higher for mordenite-based catalysts as compared to USY-based catalysts. The concentrations of C10 and C11+ were found lower for mordenite-based catalysts as compared to USY-based catalysts.

ACKNOWLEDGEMENTS The authors appreciate and acknowledge Japan Cooperation Center, Petroleum (JCCP), with the subsidy of the Ministry of Economy, Trade and Industry, Japan, and King Fahd University of Petroleum & Minerals (KFUPM), Dhahran, Saudi Arabia, for supporting this work under the Joint Saudi-Japanese Joint Research Program. REFERENCES 1. Tsai, T., W. Chen, S. Liu, C. Tsai, I. Wang, 2002, Catal. Today. Vol. 73, pp. 3948. 2.

Stachelczyk, D., 2000, Presented at the National Petrochemical & Refiners Association, San Antonio, Texas.

3.

Boon, A.; S. Oh, S. Lee, K. Seong, Y. Kim, J. Lee, 2002, Presented at the 2nd Asian Petrochemicals Technology Conference, Seoul 2002.

4.

Tsai, T.C., S.B. Liu, I. Wang, Appl. Catal. A: Gen. 181 (1999) 355.

5.

Wang, I., T.C. Tsai, S.T. Huang, Ind. Eng. Chem. Res. 29 (1990) 2005.

6.

Das, J., Y.S. Bhat, A.B. Halgeri, Catal. Lett. 23 (1994) 161.

7.

Wu, J.C., L.J. Leu, Appl. Catal. A: Gen. 7 (1983) 283.

8.

Dumitriu, E., V. Hulea, S. Kaliaguine, M.M. Huang, Appl. Catal. A: Gen. 135 (1996) 57.

9.

Magnoux, P., C. Canaf, F. Machado, M. Guisnet, J. Catal. 134 (1992) 286.

10. Meshram, N.R., S.B. Kulkarni, P. Ratnasamy, J. Chem. Technol. Biotechnol. 34A (1984) 119.

129



11. Halgeri, A.B., J. Chem. Technol. Biotechnol. 31 (1981) 541. 12. Ribeiro, M.F., F.R. Ribeiro, P. Dufresne, C. Marcilly, J. Mol. Catal. 35 (1986) 227. 13. Roger, H.P. K.P. Moller, C.T. O’Connor, 1998, J. Catal. Vol. 176 pp. 68–75. 14. Roger, H.P., W. Bohringer, K.P. Moller, C.T. O’Connor, 2000, Stud. Surf. Sci. Catal. Vol. 130, pp. 281–286. 15. Cejka, J., J. Kotrla, A. Krejci, 2004, Appl. Catal. A: Gen. Vol. 277 pp. 191–199 16. Collins,D.J., C.B.Quirey, J.E.Fertig, B.Davis, 1986, Appl. Catal. A: Gen. Vol. 28, pp. 35–55. 17. Matsuda, T. M. Asanuma, E. Kikuch1, 1988, Appl. Catal. A: Gen. Vol. 38, pp. 289–299.

130


INVESTIGATION OF THE INFLUENCE OF MICROPOROUS LAYER ON THE WATER TRANSPORT AND PERFORMANCE OF A PEM FUEL CELL

Hasan K. Atiyeh1,*, Kunal Karan2, Brant Peppley2, Jon Pharoah2, Ela Halliop2 and Aaron Phoenix3 * 1 : Current Address: Department of Chemical Engineering, King Saud University, P.O. Box 800, Riyadh 11421, Saudi Arabia, [email protected] 2: Queen’s-RMC Fuel Cell Research Centre, Kingston, ON, Canada K7L 5L9 3: Permanent address: Department of Chemical Engineering, University of Saskatchewan, Saskatoon, SK Canada S7N 5A9

ABSTRACT An automated fuel cell test station was modified in house to collect water in both the anode and cathode exhaust streams of a standard 100 cm2 active area PEM fuel cell. Various combinations of cells were built and tested with porous transport layers (PTL) at the electrodes using either carbon fiber paper with a microporous layer (MPL) (SGL 10BB) or carbon fiber paper without a MPL (SGL 10BA). The net water drag coefficient at three current densities (0.3, 0.5 and 0.7 A cm-2) for two combinations of anode/cathode relative humidity (60%/100% and 100%/60%) was determined from water balance measurements. The overall water balance consistently had an error of less than 5%, which provided statistical confidence in the determination of the net drag coefficient. The addition of a MPL to the carbon fiber paper PTL at the cathode did not cause a statistically significant change to the overall drag coefficient although there was a significant improvement to the fuel cell performance and durability when a MPL was used at the cathode. The results indicate that the function of the MPL in improving fuel cell performance is not associated with overall water drag as has been proposed by some researchers. KEY WORDS Water transport, microporous layer, PEM fuel cell, net drag coefficient

INTRODUCTION A polymer electrolyte membrane (PEM) fuel cell is an electrochemical device that converts the chemical energy of a reaction directly into electrical energy. The Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007

INVESTIGATION OF THE INFLUENCE OF MICROPOROUS LAYER ON THE WATER TRANSPORT

electrochemical reaction starts with the oxidation of hydrogen at the anode to provide protons and freeing electrons in the process that pass through an external circuit to reach the cathode. The protons diffuse through the PEM to the cathode to react with oxygen and the returning electron. Water is then produced at the cathode. Water management in a PEM fuel cell remains a challenging design and operational problem (Berg et al., 2004; Eikerling, 2006). Sufficient hydration of the polymer electrolyte membrane is essential to maintain high proton conductivity and minimize ohmic losses. On the other hand, removal of water generated at the cathode is important to avoid flooding of pores in the catalyst layer as well as the pores in the porous transport layer (PTL), which connects the catalyst layer to the flow-field plates (FFP) in a PEM fuel cell. Excessive flooding of the cathode catalyst layer impedes oxygen transport to the active sites on the electrodes resulting in an increased mass transport limitation, which is manifested via a lowering of the limiting current densities. Further, depending on the humidification level of the reactant feed streams, localized drying of the membrane can occur leading to excessive ohmic heating and, ultimately, to pinhole formation in the membrane. Therefore, improper water management may not only compromise fuel cell performance but also contribute to a rapid degradation of the membrane electrode assembly of a PEM fuel cell. To improve water transport in PEM fuel cells, the wettability of the PTL is altered by treating it with a hydrophobic material such as Teflon. Such a treatment leads to pockets of hydrophilic and hydrophobic pores in the PTL (Mathias et al., 2003). It is thought that hydrophobic regions allow a pathway for gas transport whereas the hydrophilic regions facilitate liquid transport. In addition to the hydrophobic treatment of the PTL, the use of a surface layer or microporous layer (MPL) has become a common practice. The MPL is made up of a mixture of carbon particles and a hydrophobic agent. The size of open pores in a MPL ranges between 100-500 nm in contrast to the PTL pores which range between 10-30 microns. It is thought that micro-sized pores help water removal by acting like a wick (Mathias et al., 2003). Several studies have investigated the modeling of water transport in PEM fuel cells (Nam and Kaviany, 2003; Pasaogullari and Wang, 2004; Weber and Newman, 2005; Qi and Kaufman, 2002; Lin and Nguyen, 2006). Nam and Kaviany (2003) presented computational results for a system in which a MPL was placed between the cathode catalyst layer and the PTL. The placement of the MPL helped in reducing water saturation in the adjacent catalyst layer. Passaogullari and Wang (2004) reported similar findings. However, both studies were based on half-cell models and considered water transport in the PTL/MPL only but did not include the membrane in the model (Nam and Kaviany, 2003; Passaogullari and Wang, 2004).


Hasan K. Atiyeh, et al.

Weber and Newman (2005) employed a two-phase, 2-D, fuel cell model to study the influence of the MPL on water transport. They fitted key model parameters (PTL porosity, MPL porosity and MPL fraction of hydrophobic pores) to single-cell polarization curves reported Qi and Kaufman (2002), who had reported PEM fuel cell performance for single-cells with and without a MPL. Based on the predictions from the tuned model, they claimed that the MPL acts as a valve that pushes water away from the cathode PTL through the membrane to the anode. Lin and Nguyen (2006) have investigated the effect of MPLs on PEM fuel cell performance for several PTLs. They observed that PEM fuel cells with a MPL exhibited better performance than the fuel cells without a MPL. They hypothesized that the MPL helps keep liquid water in the cathode catalyst layer and minimizes liquid water transport to the cathode PTL. This means that the MPL increases the back-diffusion rate of water from the cathode through the membrane to the anode. Although this hypothesis is in agreement with the modeling result of Weber and Newman (2005), it is in contrast with the simulation results of Nam and Kaviany (2003) and Passaogullari and Wang (2004), both of whom showed that the MPL enhances water removal rate from the catalyst layer to the cathode PTL. These studies (Nam and Kaviany, 2003; Pasaogullari and Wang, 2004; Weber and Newman, 2005; Kaufman, 2002; Lin and Nguyen, 2006) clearly present contradicting explanation on the role of MPL on the water transport and neither one has reported nor used experimental data from water balances to support their hypothesis. The role of a microporous layer (MPL) in a PEM fuel cell remains unclear. Few researchers such as Jannsen and Overvelde (2001) and Murahashi et al. (2006) have conducted experimental studies comparing the water transport in fuel cells with a MPL to those without. Experimental data on water transport are usually discussed in terms of net water drag coefficient, i.e. the moles of water dragged from anode through the membrane to the cathode per mole of proton transported. Janssen and Overvelde (2001) presented net drag coefficients for a wide range of operating conditions such as current density, temperature, pressure, stoichiometry and humidity of the inlet gases, which were either fully saturated or dry. They found that the humidity and the stoichiometry of the inlet gases had a much larger effect on the drag than did the different fuel cell components. They also examined the effect of the MPL on PEM fuel cell performance and water transport to a varying extent. However, the PTL type and/or the catalyst loading for any two sets of their experiments were different. Murahashi et al. (2006) studied the effect of humidification level of reactants and current density on the net water drag. A lower net water drag coefficient was observed when the cathode was at a higher humidification level than the anode. Unfortunately, none of their data reported would allow a direct investigation of the effects of MPL solely on water drag coefficient and none used catalyst coated membranes (CCM). 133 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


In summary, the role of a MPL on water transport in PEMFCs remains unresolved. An absence of experimental data in open literature on the water balance across the PEM fuel cell has perpetuated the confusion on whether the role of a MPL is to enhance back-diffusion of water from the cathode through the membrane to the anode or to improve water removal from the cathode catalyst layer through the MPL to the PTL. Therefore, the primary objective of this study was to obtain reliable experimental data that can help resolve the question: whether, how and to what extent does the MPL on either electrode or on both electrodes affect water transport in PEM fuel cells? A secondary but essential objective was to develop an experimental system and protocol such that reliable water balance data could be generated. Recently, we reported the experimental results on the effect of MPL on the net drag coefficient of water from the anode through the membrane to the cathode and on fuel cell performance when a MPL was present only on the cathode (Karan et al., 2007). In this paper, we present some of our results on the influence of a MPL, when it is used at either electrode or at both electrodes, on the net drag coefficient at various operating conditions. The full study can be seen in the recently published article (Atiyeh et al., 2007). EXPERIMENTAL Fuel cell components The flow field plate (FFP) used for the anode and cathode sides of the fuel cell were similar, each having seven serpentine parallel channels. Catalyst coated membranes (CCMs) with Nafion 112 and 0.3 mg/cm2 platinum catalyst loading on each electrode were used in all experiments. The total active electrode area of the cell was 100 cm2. Two types of porous transport layer (PTL), SGL 10BA and SGL 10BB carbon papers (SGL Carbon Group, USA), were investigated. The SGL 10BB carbon paper has a microporous layer (MPL) on one face. Fuel cell test system Tests were conducted on an automated fuel cell test station (Hydrogenics FCATSS800). This test station is equipped with a primary water collection system for both anode and cathode effluent streams. Each primary water collection system consists of a knock-out drum and a collection vessel connected to a differential pressure transducer (DPT) and solenoid valve (SV) as shown for the cathode effluent in Figure 1. The DPT activates the SV to open when the condensed water level in the collection vessel reaches a preset height allowing water to flow out into a collection flask. The time to fill the collection vessel to the level needed to actuate the SV to open varies based on the operating conditions. 134 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


Preliminary water balance measurements showed that the standard humidification configuration and primary water collection systems did not provide sufficiently accurate results for a meaningful calculation of the net water drag coefficient. Therefore, a secondary water collection system was built in-house to obtain reliable measurements Figure 1. The secondary water collection systems condensed the water from the gases leaving the knock-out drums using a condenser and a cold trap. This resulted in a more accurate water balance that was verified experimentally. Following these modifications, the overall water balance consistently closed to within 5%. Experimental conditions The net drag coefficient was determined from water balance measurements for each fuel cell build (Table 1) at three current densities (300, 500 and 700 mA cm-2) and under two different anode/cathode relative humidities (60%/100% and 100%/60%). Pure H2 and air were supplied to the anode and cathode, respectively, in a current based flow control mode at a rate such that the stoichiometric ratio of H2 was 1.4 and the stoichiometric ratio of air was 3. The supply pressures to the anode and cathode were maintained at 35 kPa.

Electronic Load

Dew Point gas Tdew

mΩ

V

TI

Humidified Ai r In Over-Saturated Air Moisture Separator

Level Sensor

Air Out

TI

TI

TI

TI

H 2 Out to water collection system, same as cathode (not shown) Humidif ied H 2 In

TCW Fuel Cell

TI

Air + Steam

TI

SV

Cooling Water Knock Loop Out Drum

Condensate to Sink

Exhaust Air T TI

Condens er Air Exhaust

Humidification System DPT

Collection Vessel

Water Chiller

SV

Cold Trap Cat hode Primary Water Collection

Automated Fuel Cell Test Station

Cat hode Secondary Water Collection

In-House Addition

Figure 1. Schematic diagram of experimental setup used.



The inlet gas temperatures and cell temperature were controlled at a nominal value of 60°C. Lines between humidifiers and the fuel cell were heated to avoid condensation. The inlet dew point gas temperatures, inlet and outlet gas temperatures and pressures of the anode and cathode streams along with the exhaust gas temperatures from the primary and secondary water collection systems were recorded using the data acquisition software supplied with the fuel cell test station. At the end of each run, the water collected from the primary and secondary collection systems of the anode and cathode was weighed and recorded.

Table 1. Experimental fuel cell builds. Cell builds Cell builds rejecteda c A1 to A3 ---B1c to B8 B3, B5, B7, B8 C1 and C2 C1 D1 and D2 ----

Anode PTLb

Cathode PTLb

SGL 10BA SGL 10BA SGL 10BB SGL 10BB

SGL 10BB SGL 10BA SGL 10BA SGL 10BB

a

Runs discarded because beginning of life cell voltage was less than 0.4 V at 0.7 A cm-2 b 10BA without a MPL; 10BB with a MPL c Cell builds A1 and B1 were used for commissioning of the water balance system Fuel cell performance monitoring The fuel cell performance was monitored using standard polarization measurements at a nominal operating temperature of 60°C, inlet pressure of 35 kPa, and at 100% relative humidities for both inlet gases. The stoichiometric ratios of the H2 and air were 1.4 and 3, respectively. Polarization curves were generated for each cell built after initial conditioning and at the end of the water transport experiments.

CALCULATIONS An overall water balance was performed around the fuel cell, accounting for the water generated through the electrochemical reaction, to assess the reliability of the experimental results. An overall water balance was considered “closed” when the difference between the “water in + generation” and the “water out” was less than the 95% confidence limits of the calculation. The net water drag coefficient, α (mol H2O/mol H+), was calculated solely from a water balance performed on the cathode side of the fuel cell:



α=

F ⋅ (Wout − Win ) − i ⋅ A 2 i⋅ A

[1]

where F is Faraday’s constant, Win and Wout are, respectively, the cathode side inlet and outlet water flow rates (mol/s), i is the current density (A cm-2) and A is the total active electrode area (cm2). More details on the experimental work and the calculations done can be found elsewhere (Atiyeh, et al., 2007). RESULTS AND DISCUSSIONS Fuel cell performance Standard polarization curves for all configurations tested (Table 1) are shown in Figure 2. Individual plots correspond to the polarization curves for each build after initial conditioning and at the end of the water transport experiments. It can be readily seen that the polarization curves for the cells with a MPL on either electrode (builds A2, A3 and C2) or on both electrodes (builds D1 and D2) in Figure 2a show a more consistent and in general better performance than those without a MPL (builds B2, B4 and B6) in Figure 2b. In addition, cells with a MPL showed very little deterioration in performance between the initial and final polarization curves (an average of 570 hours of operation). However, cells without a MPL on the cathode exhibited more variability in post conditioning performance and an obvious degradation between the post conditioning and final polarization curves (an average of 250 hours of operation). All of the cell builds with a MPL present at the cathode by itself or in combination with a MPL at the anode met the beginning of life critical performance of a cell voltage higher than 0.4 V at 0.7 A cm-2 (Table 1). However, several builds without a MPL and only one cell build with a MPL on the anode were discarded because they did not meet the beginning-of-life critical performance.



1.0 RH (A/C) = 100% / 100% Stoic (A/C) = 1.4 / 3 TCell = 60 °C

a 0.9

Voltage (V)

0.8

0.7 Build A2 MPL on cathode Build A3 Build C2 MPL on anode Build D1 MPL on anode and on cathode Build D2 Solid symbols: start of testing Open symbols: end of testing

0.6

0.5

0.4 0

0.2

0.4 0.6 -2 Current Density (A cm )

0.8

1

1.0 RH (A/C) = 100% / 100% Stoic (A/C) = 1.4 / 3 TCell = 60 °C

b

Voltage (V)

0.8

0.6

0.4 Build B2 Build B4

no MPL Build B6 Solid symbols: start of testing Open symbols: end of testing

0.2

0.0 0

0.2

0.4 0.6 Current Density (A cm-2)

0.8

1

Figure 2. Cell performance curves for builds with various combinations of porous transport layers at stoichiometric ratio of H2/air of 1.4/3 for a) anode/cathode relative humidity of 100/60% and b) anode/cathode relative humidity of 60/100%. Error bars not visible are smaller than the symbols.



Influence on net water drag The net drag coefficient of water from the anode to the cathode, α, (mol H2O/mol H+), for all cases considered are seen in Figure 3. The data is grouped according to inlet humidification conditions: A/C RH = 100/60% (Figure 3a) and A/C RH = 60/100% (Figure 3b). The scales on the two sub-figures are different. Only error bars that show typical experimental uncertainty from representative builds A2, B2, B6, C2 and D1 are included in the figure for ease of comparison. The net drag coefficient was between 0.01 and 0.11 mol H2O/mol H+ for cell builds at A/C RH of 100/60% (Figure 3a). However, it was lower when the cathode was at a higher humidification level than the anode (A/C RH of 60/100%) in Figure 3b. This corresponding change in the drag coefficient as a result of change in the RH difference between the anode and the cathode provided confidence in the reliability of the measurements. This is also consistent with results presented by Jannsen and Overvelde (2001) and Murahashi et al. (2006) who reported a lower net drag coefficient when the cathode was at a higher humidification level than the anode. In the present study, as water is removed from the anode PTL into the unsaturated (60% RH) anode feed, it would be expected that there would be an increase in the back-diffusion of water from the cathode to the anode near the inlet of the flow filed. The difference in the net drag coefficient for the two combinations of RH is larger at higher current densities (Figure 3). This would also be expected as the nitrogen in the 60% RH air feed at 700 mA cm-2 would have 2.33 times the capacity to remove water and increase the net water transport from anode to cathode compared to the flow of nitrogen in the feed at 300 mA cm-2. Conversely, the nitrogen in the 100% RH air feed would not pick up additional water while the 60% RH hydrogen would tend to decrease the net water transport from the anode to the cathode.



+

Net Water Drag (mol H 2O/mol H )

0.12

Stoic (A/C) = 1.4 / 3 RH (A/C) = 100% / 60% TCell = 60 °C

a

0.10 0.08 0.06 0.04

Build Build Build Build Build Build

0.02 0.00

A2 A3 B2 C2 D1 D2

MPL on cathode no MPL MPL on anode MPL on anode and on cathode

-0.02 0.2

0.3

0.4 0.5 0.6 -2 Current Density (A cm )

0.7

0.8

+

Net Water Drag (mol H 2O/mol H )

0.12 Stoic (A/C) = 1.4 / 3 RH (A/C) = 60% / 100% TCell = 60 °C

b

0.08

0.04

0.00 Build Build Build Build Build Build Build

-0.04

-0.08

A2 A3 B4 B6 C2 D1 D2

MPL on cathode no MPL MPL on anode MPL on anode and on cathode

-0.12 0.2

0.3

0.4 0.5 0.6 -2 Current Density (A cm )

0.7

0.8

Figure 3. Net water drag coefficients for builds with various combinations of porous transport layers at stoichiometric ratio of H2/air of 1.4/3 for a) anode/cathode relative humidity of 100/60% and b) anode/cathode relative humidity of 60/100%. Error bars for builds A2, B2, B6, C2 and D1 are only shown for ease of comparison.



The results of Figure 3 also demonstrates that there is inherent variability in the performance of seemingly similar fuel cells, and accounting for measurement errors and the inherent performance variability, the difference in net drag is statistically insignificant. The results of the net drag coefficient of one fuel cell build fall within the experimental uncertainty of the other as seen by the overlapping error bars. This indicates that the addition of a MPL to the carbon fiber paper PTL at the cathode did not cause a statistically significant change to the overall drag coefficient although there was a significant improvement to the fuel cell performance and durability when a MPL was used. CONCLUSIONS An experimental program to study the net water transport in a standard 100 cm2 active area PEM fuel cell was undertaken. Cells with a MPL on either electrode or on both electrodes exhibited better overall performance and durability compared to cells without a MPL. The net drag coefficient was determined from water balance measurements that were consistently accurate to within 5%. As would be expected, the net drag coefficient was lower when the inlet RH was 60% for the anode and 100% for the cathode, compared to when the RH was 100% for the anode and 60% for the cathode for the three current densities studied. However, there was no statistically significant difference in the net drag coefficient for cells with a MPL at the cathode and no MPL on the anode compared to cells without a MPL at the conditions studied. The results indicate that the function of the MPL in improving fuel cell performance is not associated with overall water drag and further work is required to reveal the mechanism by which the presence of the MPL affects PEM fuel cell performance. ACKNOWLEDGEMENT This work was partially funded through a Natural Science and Engineering Council, Collaborative Research Grant in partnership with E.I. Du Pont of Canada Research and Development. REFERENCES Atiyeh, H.K., K. Karan, B. Peppley, A. Phoenix, E. Halliop, J. Pharoah, 2007, “Experimental Investigation of the Role of a Microporous Layer on the Water Transport and Performance of a PEM Fuel Cell”, Journal of Power Sources, Vol. 170, pp. 111–121. Berg, P., K. Promislow, J. St. Pierre, J. Stumper, and B. Wetton, 2004, “Water Management in PEM Fuel Cells”, Journal of Electrochemical Society, Vol. 151, pp. A341-A353. 141 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


Eikerling, M., 2006, “Water Management in Cathode Catalyst Layers of PEM Fuel Cells: A Structure-Based Model”, Journal of Electrochemical Society, Vol. 153, pp. E58-E70. Janssen, G.J.M. and M.L.J. Overvelde, 2001, “Water Transport in the ProtonExchange-Membrane Fuel Cell: Measurements of the Effective Drag Coefficient”. Journal of Power Sources, Vol. 101, pp. 117–125. Karan, K. , H. Atiyeh, A. Phoenix, E. Halliop, J. Pharoah, B. Peppley, 2007, “An Experimental Investigation of Water Transport in PEMFCs: The Role of Microporous layer” Electrochemical Solid-State Letters, Vol. 10, pp. B34-B38. Lin, G. and T.V. Nguyen, 2006, “A Two-Dimensional Two-Phase Model of a PEM Fuel Cell”, Journal of Electrochemical Society, Vol. 153, pp. A372-A382. Mathias, M., J. Roth, J. Fleming and W. Lehnert, 2003, Diffusion media materials and characterization. In: W. Vielstich, A. Lamm and H. Gasteiger, (Eds) Handbook of Fuel Cells—Fundamentals, Technology and Applications, John Wiley & Sons, New York, pp 1-21. Murahashi, T. M. Naiki, E. Nishiyama, 2006, “Water Transport in the Proton Exchange-Membrane Fuel Cell: Comparison of Model Computation and Measurements of Effective Drag”, Journal of Power Sources, 162, pp. 1130–1136. Nam, J.H. and M. Kaviany, 2003 “Effective Diffusivity and Water-Saturation Distribution in Single- and Two-Layer PEMFC Diffusion Medium”, International Journal of Heat Mass Transfer, Vol. 46, pp. 4595-4611. Pasaogullari, U. and C.-Y. Wang, 2004, “Two-Phase Transport and the Role of Microporous Layer in Polymer Electrolyte Fuel Cells”, Electrochimca Acta, Vol. 49, pp. 4359-4369. Qi, Z. and A. Kaufman, 2002, “Improvement of Water Management by a Microporous Sublayer for PEM Fuel Cells”, Journal of Power Sources, Vol. 109, pp. 38-46. Weber, A.Z. and J. Newman, 2005, “Effects of Microporous Layers in Polymer Electrolyte Fuel Cells”, Journal of Electrochemical Society, Vol. 152, pp. A677A688.


THE ESTIMATION OF LIQUID PROPYLENE POLYMERIZATION RATE USING PRESSURE-DROP DILATOMETRY

Mohammad Al-haj Ali1, Ben Betlem2, Brain Roffel2 and Günter Weickert2 1: King Saud University, Faculty of Engineering, Department of Chemical Engineering, E-mail: [email protected] 2: University of Twente, P.O.Box 217, 7500 AE Enschede, The Netherlands

ABSTRACT A new technique to measure the polymerization rate of liquid propylene in filled batch reactors was examined. This technique is based on dilatometry principle. Pressure-drop dilatometry utilizes pressure and temperature changes that occur on polymerization as well as the compressibility of the reactor contents. The estimated polymerization rates compare well to those obtained using the well-developed isoperibolic calorimetry technique. It was also found that pressure-drop dilatometry gives more kinetic information during the initial stages of the polymerization reaction than the other technique. In addition, pressure-drop dilatometry can be used to estimate the polymerization rate online. However, similar to the isoperibolic calorimetry method, pressuredrop dilatometry depends on a few parameters which are not accurately known.

KEYWORDS Liquid propylene, Polymerization, Calorimetry, Dilatometry, Ziegler-Natta

INTRODUCTION In polymerization reactions, the rate of reaction has a major impact on polymer morphology and polymer molecular properties, such as molecular weight distribution (MWD). In addition, the rate of reaction is a critical issue in safety studies, since it is directly related to the amount of heat produced during polymerization reactions. In general, the polymerization rate can be experimentally determined by measuring the change in any property that differs for the monomer(s) and the polymer such as solubility, refractive index and density[1] . However, the measurement of such properties is not a simple task during the polymerization reaction because sensors in the reaction mixture are readily fouled by the


THE ESTIMATION OF LIQUID PROPYLENE POLYMERIZATION RATE USING PRESSUREDROP DILATOMETRY

sticky polymer. Furthermore, sensor technology for on-line monitoring is still in its infancy[2] . Fortunately, the situation is better when dealing with sensors that monitor reactor operation, by measuring temperature, pressure and different flows. Thus, different techniques, which based on these measurements, were developed to estimate the polymerization rate: (i) the flow rate technique, (ii) the calorimetric method, and (iii) the dilatometric technique.

Flow rate technique This technique is based on continuous compensation of monomer consumption such that the pressure in the system is kept constant. This technique is usually used in gas phase or slurry polymerization reactions. However, this method is not applicable in case of isothermal liquid-pool polymerization, since reactor pressure is constant with conversion, as long as there is liquid monomer in the reactor.

Calorimeter technique The principles of the calorimetric method were developed a long-time ago. However, since only 30 years intensive developments and applications of calorimeters are reported in the literature. Without exception, the basis of calorimetric measurement is a heat balance of the stirred tank reactor. For this purpose the heat flux produced by the chemical reaction and power added by the stirrer are compared to the heat removal by accumulation, convection, conduction, and heat loss to the surrounding in a macroscopic heat balance. The reaction mass is regarded as perfectly mixed, and changes in both kinetic and potential energies as well as heats of mixing and solution are neglected. By means of isoperibolic calorimetry, the jacket temperature is kept constant during the experiment. As shown in Figure 1[3], after catalyst injection, the reactor temperature increases slightly reaching a quasi-steady state value after few minutes. As in isothermal operation, the temperature difference between the cooling jacket and the reaction mass corresponds to the polymerization rate. Assuming isothermal conditions, which is reasonable because about 0.5 K is sufficient to measure the entire rate profile accurately[4] , the reaction rate is calculated by[5]:

R p ≈ C ⋅ ( ∆T − ∆TBL )

(1)

where ∆TBL, the so-called baseline temperature difference, is the temperature difference between the reactor temperature and the average temperature of the cooling jacket of the reactor when no reaction takes place. C is defined as:


Mohammad Al-haj Ali, et al.

C=

U.A wall mc .∆H r

(2)

C is assumed to be constant during the polymerization reaction. However, it should be emphasized that this technique only works as long as the increase in reaction temperature is small (~1.0 K), since the kinetic constants should be related to a constant reaction temperature. To overcome this problem, the amount of catalyst added should be reduced; unfortunately, this increases the probability of catalyst poisoning.

Figure 1. Temperature profile for isoperibolic calorimetry during propylene polymerization.

Dilatometer technique In industrial applications, polypropylene is usually produced in loop reactors. These reactors consist essentially of a main tubular closed loop system. The reaction mixture flows through the pipe impelled by an axial pump, the reactor is operated fully-filled with liquid monomer[6]. The production of polypropylene results in a shrinkage of the reactor’s content because of the large difference in polymer and monomer density. This shrinkage phenomenon is a kinetic signal and the experimental technique is known as Dilatometry. Dilatometry is an accurate method for polymerization reactions with a large difference in density between monomer and polymer[1] as long as the volume shrinkage can be related to the reaction kinetics. Since the reactor is operated filled with liquid monomer, volume shrinkage will lead to a sensitive decrease in the reactor pressure[6]. This can be used to develop a new strategy to measure reaction rate profiles in liquid propylene polymerization[4]. This strategy can be implemented in two ways: 145 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


1.

Pressure-drop dilatometry: the reaction rate is calculated using the pressure drop recorded during the polymerization reaction. 2. Compensation dilatometry: in this approach the reactor pressure is kept constant through the continuous compensation of the monomer consumption. The reaction rate is calculated using the amount of monomer fed to the reactor. This article focuses on pressure-drop dilatometry. Detailed discussion of compensation dilatometry can be found in [3].

Pressure-drop dilatometry This method is based on the compressibility behavior of the monomer inside the reactor. The polymerization reactor usually contains not only monomer but also amorphous and crystalline polymer and hexane that is used to inject the catalyst. The presence of such a mixture affects the compressibility of the monomer; consequently an average compressibility should be used instead of the value for pure monomer. The reaction rate can be calculated by the pressure drop recorded during the polymerization reaction. Because of the change in reactor temperature during the polymerization reaction, the change in monomer volume depends not only on the reaction pressure, P, but also on the reaction temperature, T. A relation connecting P, T and the specific volume of the monomer, ν, can be expressed by the functional equation:

ν = ν (T, P )

(3)

The total differential of ν is defined as:

⎛ ∂ν ⎞ ⎛ ∂ν ⎞ dν = ⎜ ⎟ ⋅ dT + ⎜ ⎟ ⋅ dP ⎝ ∂P ⎠ T ⎝ ∂T ⎠ P

(4)

Or in a more formal form this equation can be rewritten as:

dν = β ⋅ dT − κ ⋅ dP νo

(5)

Where β and κ are defined as follows[7]:

Volume expansivity:

β≡

1 ⎛ ∂ν ⎞ ⋅⎜ ⎟ ν o ⎝ ∂T ⎠ P


(5)


Isothermal compressibility: κ ≡ −

1 νo

⎛ ∂ν ⎞ ⋅⎜ ⎟ ⎝ ∂P ⎠T

(6)

here νo is the original specific volume [8]. Because of the presence of different components in the system during the polymerization reaction, Equation (5) will be rewritten as: n

n

n

i =1

i =1

i =1

∑ dν i = ∑ βi ⋅ ν o,i ⋅ dT − ∑ κi ⋅ ν o,i ⋅ dP where n is the number of components in the system. The average volume expansivity and isothermal compressibility can be defined, respectively, as

β= κ=

β m ⋅ ν o ,m + β p ⋅ ν o , p ν o ,m + ν o ,p κ m ⋅ ν o,m + κ p ⋅ ν o,p ν o,m + ν o,p

(7)

(8)

here νo,m and νo,p are the original specific volumes for the monomer and polymer respectively. Note that the effect of other components in the reaction system, such as hydrogen gas, cocatalyst and hexane, is assumed negligible since their concentration is low. Using these new terms, and dividing both sides by the time increment, dt, Equation (5) will have the following final form:

1 dν dT dP ⋅ =β⋅ −κ⋅ ν o dt dt dt

(9)

It can be proved that the differential of the specific volume and the density for any material, ρ, are related by the following equation:

dν dρ =− ν ρ

(10)

Combining this equation with Equation (10) gives:

−(

1 ρ m ,o

)⋅

dρ dP dT = −κ ⋅ +β⋅ dt dt dt

(11)



where ρm,o is the initial monomer density. In batch reactors, the relationship between reaction rate and density profile can be described by Equation (13) below, assuming constant volume system1

−

dρ m ⎡m ⎤ = Rp ⋅ ⎢ c ⎥ dt ⎣V⎦

(12)

where Rp, in kg/gcat. hr, is the polymerization rate, mc is the catalyst mass, and V is the volume of the reactor. Substituting Equation (13) into Equation (12), the reaction rate can be expressed as a function of both temperature and pressure profiles:

R p = M1 ⋅

dP dT + M2 ⋅ dt dt

(13)

with

⎡ κ ⋅ ρo ⋅ V ⎤ M1 = − ⎢ ⎥ ⎣ mc ⎦

(14)

⎡ β ⋅ ρo ⋅ V ⎤ M2 = ⎢ ⎥ ⎣ mc ⎦

(15)

It is well-known that the volume expansivity and isothermal compressibility are functions of system temperature and pressure. Variations in κ and β can be neglected for small changes in reaction temperature and pressure. The yield can be calculated by integrating the rate, Equation (14), over the reaction time: t

Yield calc = ∫ R p ⋅ dt

(16)

0

The measured yield can be determined by weighing the dry product from the polymerization. The calculated yield should be equal to the measured yield. This fact is used to determine the values of M1 and M2 using fitting techniques.

1

If reactor contents’ volume becomes smaller than reactor volume, gas phase will start to form and system pressure will fall to propylene vapor pressure, and this approach will not be applicable anymore. 148



EXPERIMENTAL PART Chemicals Propylene used in the experiments was “polymerization grade” and obtained from Indugas. The purity was more than 99.5%, with propane as a main impurity. Hydrogen used had a purity higher than 99.999%; it was further purified by passing it over a reduced BTS copper catalyst, obtained from BASF, and subsequently passing through three different beds of molecular sieves, with pore sizes of 13, 4 and 3 Å, respectively. Propylene was purified in the same way; additionally it was passed over a bed of oxidized BTS catalyst to remove CO. TiCl4 supported on MgCl2 with phthalate as internal donor and an external silane donor was used as a catalyst with TEA as a cocatalyst and scavenger.

Reactor system A 5-liter stainless steel jacketed batch reactor (Büchi BEP 280) with a separately heated cover plate was used; it is described elsewhere [3, 9-11]. For intensive mixing, the reactor was equipped with a turbine stirrer operated at 2000 rpm. The pneumatic injection system allows the introduction of liquids and slurries into the reactor, even at high reactor pressures. The cooling medium temperature is kept constant within ± 0.01 K during isoperibolic experiments.

Experimental procedure The reactor was flushed with nitrogen gas five times at 90°C and purged with propylene gas at the beginning of the experiment, then filled with liquid propylene and heated up to the reaction temperature. When the temperature reached the set-point, hydrogen was injected. The reactor temperature and pressure were monitored as a function of time. As soon as both became stable for an interval of three minutes, the reaction was started by injecting the prepared catalyst into the reactor. The experiments were executed under isoperibolic conditions. Thus, just after the catalyst injection, the temperature control system becomes active to keep the jacket temperature constant, the reaction temperature increases slightly reaching a quasi-steady-state after about 1.5 minutes in case of using a fully pre-activated catalyst. The heat of polymerization was measured under quasisteady state conditions. Data was collected every three seconds. The polymerization reaction was finally terminated by rapidly flushing the unreacted propylene. After each experiment, the resulting polymer was dried under vacuum at 50°C for 4 hours.



RESULTS AND DISCUSSION A number of polymerization tests was carried out; the recipes and the polymerization yields are shown in Table 1. The reactions were carried out at constant temperature, cocatalyst concentration and donor concentration. The main difference between these experiments is the hydrogen concentration. Reproducibility The reproducibility of the experiments has been tested by repeating a standard experiment at 70 °C and 43 bar with 150 mg hydrogen at fixed concentrations of the catalyst, cocatalyst and external donor. Figure 2 shows the reaction rate profiles for three standard experiments. The figure shows that the three rate profiles are almost the same, which is a signof excellent reproducibility.

Table 1 Recipes used in the polymerization experimentsα. Run H2 added, mg Yield, kg/gcat. hr 1 0 14.5 2 50 57.1 3 150 82.5 4 250 83 5β 1000 78.2 α Other polymerization conditions: T = 70 °C, catalyst = 3.78 mg, TEA = 1040 mg, and donor = 50 mg. β Other polymerization conditions: T = 70 °C, catalyst = 1.58 mg, TEA = 1040 mg, and donor = 50 mg.



120

R p , kg/g

cat .

hr

100 80 60 Exp. 17

40

Exp. 28

20

Exp. 38

0 0

5

10

15

20

25

30

35

Time, min Figure 2. Polymerization rate-time profiles of repeated standard experiments. Note that the numbers of experiments in this figure differ from those in Table 1.

Temperature and pressure profiles in a fully-filled reactor Before discussing the results of pressure-drop dilatometry, it is worth to take a closer look at the pressure, reactor temperature, and inlet and outlet jacket temperature profiles. Figure 3 shows these profiles for Run 3. Before catalyst injection, the reactor temperature and pressure are 70 °C and 48 bar respectively. When the catalyst is injected, point A, system pressure and temperature increase immediately. Since the reactor is operated in isoperibolic mode, the outlet jacket temperature increases due to the heat production during the polymerization reactions. After a short time, 2.5 minutes in this experiment, at point B, the pressure reaches its maximum value (49 bar) and starts decreasing, while the reaction temperature still increases. During this period, two factors affect the pressure: (i) temperature increase, and (ii) polymer production. The increase in temperature has a more pronounced effect which leads to a pressure increase. When the system reaches point B, the second factor becomes predominant and the pressure starts decreasing. At point C, the polymerization temperature reaches its maximum value (72.8 °C), and starts decreasing. By the end of the experiment, the decrease in temperature does not exceed 1.0 °C, compared to 16 bar decrease in reaction pressure.


A B C

73.5 73 72.5 72 71.5 71 70.5 70 69.5

50

Tr

45

P

40

Tj,in

P, bar

T, C


35

Tj,ou

30 -5

0

5

10

15

20

25

30

35

Time, min. Figure 3. Temperature and pressure profiles during a typical propylene polymerization experiment in a filled batch reactor.

Pressure-drop dilatometry In Figures 4a, 4b and 4c the time versus polymerization rate plots are shown for Runs 1, 2 and 4. From these plots, we can state different facts. First, the curves resulting from dilatometric and the corresponding calorimetric data are almost similar; only in the initial stages there is a discrepancy. In isoperibolic calorimetry, the quasi-steady state heat balance is used to estimate the polymerization rate; thus, calculating the polymerization rate during the first few minutes is not possible. The dilatometric approach includes the dynamics of both reaction pressure and temperature which are measured fast with a limited lag (~ 3 seconds) from the early beginning of the polymerization reaction; this results in more accurate estimations of polymerization rate particularly at the beginning. Note that the oscillations, in the initial period, are attributed to the differential approach used to estimate the time derivates of pressure and temperature.



Second, since the difference between the estimated polymerization rates using both techniques is small, the assumption of constant fitting parameters, M1 and M2 in Equation (14), seems acceptable; consequently β and κ could be assumed constant since other parameters in Equations (15) and (16) are independent of the specific experiment. Table 2 gives the M1 and M2 values for the experiments. The results show that M2 does not change with changing hydrogen concentration in the system, for hydrogen additions up to 250 mg. The situation is different for M1; the compressibility changes when adding small amounts of hydrogen; however, this parameter is constant at high hydrogen concentrations. Table 2. Fitting parameters for Equation (14). Run

1

2

4

5

M1

1.85

1.57

1.62

3.10

M2

2.01

1.99

2.05

4.8

M2/M1

1.09

1.27

1.27

1.55


Rp, kg/gcat.hr


70 60 50 40 30 20 10 0

a Dilatometry

Calorimetry 0

10

20

30

40

50

60

Rp, kg/gcat.hr

Time, min. 120 100 80 60

b Calorimetr y

40 20 0

Dilatometry 0

10

20

30

40

50

60

Rp, kg/gcat.hr

Time, min. 90 80 70 60 50 40 30 20 10 0

c

Calorimetry Dilatometry

0

10

20

30

40

50

60

Rp, kg/gcat.hr

Time, min

140 120 100 80 60 40 20 0

d

Calorimetry

Dilatometry

0

10

20

30

40

50

Time, min.

Figure 4. Pressure-drop dilatometry versus isoperibolic calorimetry for propylene polymerization with (a) 0.0 mg H2 (b) 50 mg H2 (c) 250 mg H2 (d) 150 mg H2.



The predictability of this approach is examined by fixing the values M1 and M2 to be 1.6 and 2.0 respectively and estimating the polymerization rate when 150 mg H2 is added (Run 3). The results, Figure 4d, show satisfactory agreement with the calorimetric method. The predictability of this technique is examined further by considering the polymerization experiment using an extremely high hydrogen concentration (Run 5). For this experiment, catalyst mass and ρmo are 1.58 mg and 355 kg/m3, respectively, compared to 3.78 mg and 425 kg/m3 for the others. M1 and M2 for Run 5 are estimated by two approaches: (i) using the fixed values for these parameters that are obtained for the previous experiments, this approach gives 3.2 and 4 respectively, and (ii) fitting these parameters, this method gives 3.1 and 4.8 respectively. Figure 5 illustrates the results for both approaches; the former approach shows a good agreement with the calorimetric method with a maximum error of 20 %. The latter approach gives better agreement with 5 % maximum error. The obtained polymerization profiles are further used to estimate the reaction yield. The second method gives a yield of 85 g compared to 90.4 g obtained experimentally. Interestingly, the difference between the two methods is not high, even though the fitting parameters, for the first approach, are obtained at hydrogen concentrations far below that for the considered experiment. The comparison between the fitted and the estimated model parameters (Run 5) gives a clear indication that the volume expansivity term, M2, is the reason for the difference in the obtained fits. A possible reason for this difference in M2 values could be that increasing hydrogen concentration has a pronounced effect on β ; however, this would need experimental verification. Based on these results, it can be stated that for a specific reactor system, pressuredrop parameters can be taken as constants within certain operating conditions and this method can be used to predict the polymerization rate even when the polymerization reaction takes place, i.e. the method can be used on-line. Pressure-drop dilatometry is based on the fact that the changes in reactor pressure (and temperature if the reaction does not take place under isothermal conditions) affect monomer density because it is a compressible fluid. These changes in density can be related to the polymerization rate. The applicability of these methods to calculate the polymerization rates and the verification of the resulted kinetic data were demonstrated using several polymerization runs. The simultaneous use of isoperibolic calorimetry to measure the reaction rate showed that the polymerization rate profiles which are obtained by both techniques are comparable. Contrary to the isoperibolic calorimetry method, the pressure-drop dilatometry method allows to estimate the polymerization rate at the initial stages of the polymerization reaction. Similar to the isoperibolic calorimetry method, the dilatometric method depends on a few parameters which are not known accurately. Yet, the estimated parameters can be used to calculate the polymerization rate at other reaction conditions; subsequently, the dilatometric techniques can be used to estimate the polymerization



rate on-line. This is applicable within a predefined range of operating conditions relatively close to the operating point at which the parameters were found.

cat .

200 150

Rp , kg/g

hr

250

100

b a

50

c

0 0

10

20

30

40

50

Time, min

Figure 5. Dilatometry versus isoperibolic calorimetry for propylene polymerization with 1000 mg H2. (a) isoperibolic calorimetry, (b) dilatometry model with optimal fitted parameters, (c) dilatometry model with fitted parameters which is estimated using polymerization experiments with different H2 concentration. CONCLUSIONS The calculation of polymerization profile for liquid-pool polymerization of propylene in a filled batch reactor was discussed. The use of a fully-filled reactor provided us with a unique pressure behavior that can be used to estimate the polymerization profile. This pressure-drop, which resulted from the shrinkage of the reactor content, is a kinetic signal and this measurement technique is known as Dilatometry. Dilatometry technique can be implemented in two ways: (i) pressuredrop method, and (ii) constant pressure approach.



ACKNOWLDEGEMENT This work has been funded by Dutch Polymer Institute (DPI), project DPI #114. We greatly acknowledge R. Emonds and K. van Bree for the technical assistance. NOTATIONS m2

∆Hr

average wall area inside the reactor available for heat transfer heat of polymerization reaction

mc

mass of catalyst injected

mg

M1

fitting parameter

kg/gcat. bar

M2

fitting parameter

kg/gcat. °C

P

reaction pressure

bar

Rp

rate of polymerization

kg/gcat. hr

T

reaction temperature

°C °C

t

difference between reactor temperature and average jacket temperature before starting reaction time

U

average heat transfer coefficient

Watt/m2. °C

V

reactor volume

m3

Aw

∆TBL

kJ/kg

hr

Greek letters β

isothermal compressibility

1/°C

κ

volume expansivity

1/bar

ρm

monomer density

kg/m3

ρp

polymer density

kg/m3

νm

Specific volume of the monomer

mol/m3

νp

Specific volume of the polymer

mol/m3



REFRENCES 1.Odian, G.,2004, Principles of Polymerization, 3rd ed, John Wiley & Sons, New York. 2.Kammona, O., E.G. Chatzi, and C. Kiparssides, 1999,"Recent Developments in Hardware Sensors for the On-Line Monitoring of Polymerization". Journal of Macromolecular Science-Reviews in Macromolecular Chemistry and Physics, Vol. C39, pp. 57-134. 3.Al-haj Ali, M., Modeling and Control of Molecular Weight Distribution in a Liquid-phase Polypropylene Reactor. 2006, Twente University: Enschede. 4.Weickert, G. High Precision Polymerization Rate profiles. in 3rd International Workshop on Heterogeneous Ziegler-Natta Catalysts. 2003. Japan. 5.Pater, J.T.M., G. Weickert, and V.S. W.P.M., 2003,"Propene Bulk Polymerization Kinetics: Role of Prepolymerization and Hydrogen". AIChE Journal, Vol. 49, pp. 180-193. 6.Zacca, J.J. and W.H. Ray, 1993,"Modeling of the Liquid Phase Polymerization of Olefins in Loop Reactors". Chemical Engineering Science, Vol. 48, pp. 3743-3765. 7.Smith, J.M., H.C. Van Ness, and M.M. Abbott,1996, Introduction to Chemical Engineering Thermodynamics, 5 ed, McGraw-Hill, Singapore. 8.Bernard, S.M.,1989, Mechanics of fluids, 6th edition ed, Van Nostrand Reinhold, London. 9.Samson, J.J.C., P.J. Bosman, G. Weickert, and K.R. Westerterp, 1999,"LiquidPhase Polymerization of Propylene with a Highly Active Ziegler-Natta Catalyst. Influence of Hydrogen, Cocatalyst, and Electron Donor on Reaction Kinetics." Journal of Polymer Science: Part A: Polymer Chemistry, Vol. 37, pp. 219-232. 10.Pater, J.T.M., G. Weickert, and W.P.M. Van Swaaij, 2002,"Polymerization of Liquid Propylene with A Fourth Generation Ziegler-Natta Catalyst: Influence of Temperature, Hydrogen, Monomer Concentration, and Prepolymerization Method on Polymerization Kinetics". Chemical Engineering Science, Vol. 57, pp. 3461-3477. 11.Shimizu, F., J.T.M. Pater, W.P.M. Van Swaaij, and G. Weickert, 2002,"Kinetic study of a highly active MgCl2-supported Ziegler-Natta catalyst in liquid pool propylene polymerization. II. The influence of alkyl aluminum and alkoxysilane on catalyst activation and deactivation". Journal of Applied Polymer Science, Vol. 83, pp. 2669-2679.


USING ULTRASONIC EMULSIFICATION TO ENHANCE THE PRODUCTION OF FREE FATTY ACIDS FROM ENZYMATIC HYDROLYSIS OF PALM OIL

Sulaiman Al-Zuhair1 and K.B. Ramachandran2 1: Chemical and Petroleum Engineering Department, United Arab Emirates University, P.O. Box 17555, Al-Ain, United Arab Emirates, Email: [email protected] 2: Department of Biotechnology, Indian Institute of Technology Madras, Chennai 600036, Email: [email protected] ABSTRACT The hydrolysis of oil by lipase takes place at the interface between the oil and the aqueous solution containing the enzyme. For such systems, interfacial area between the oil phase and the aqueous phase influences the rate of hydrolysis. In this study, to enhance the hydrolysis rates of lipids, ultrasonication instead of mechanical agitation was used for interfacial area generation. By ultrasonication, larger interfacial areas could be produced as compared to mechanical agitation. Experiments carried out to assess the hydrolysis rate of lipids showed that it was significantly enhanced by ultrasonic emulsification and the interfacial area saturation with enzyme was not observed due to the large interfacial area generated. KEY WORDS Lipase; Oil hydrolysis, Ultrasonication; Mechanical agitation, Interfacial area Introduction The applications, importance and significance of lipase in oleochemical industry have been thoroughly demonstrated in literature [1]. The most important among these applications is the use of lipase for the production of fatty acids from oils. It is attempted as an energy-saving method, especially for producing high value-added products from heat sensitive fatty acids [2]. Lipase catalysed reactions take place at the interface between the aqueous phase containing the enzyme and the oil phase, where the enzyme has to penetrate the interface as a first step in the reaction [3–8]. In a mechanically stirred bioreactor, at any particular operating condition, the total free interfacial area is limited. However, increasing the agitation speed can increase the interfacial area [9]. However, high speeds result in increased shear rate on the Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007

USING ULTRASONIC EMULSIFICATION TO ENHANCE THE PRODUCTION OF FREE FATTY

lipase molecules and lead to denaturation of the enzyme, which results in reduced activity [10]. On the other hand, the applicability of power ultrasound for emulsification purposes has been known for long although in most cases application is still restricted to laboratory scale [11]. Currently, ultrasonic emulsification is gaining interest in many industrial fields, as a better alternative to conventional emulsification processes. These fields include water processing, powder production, combustion of fuels, cleaning, surface hindering, impregnation and coating, complex vibration testing, post-thermal treatment of hardened steels, founding and casting, and so many more [12–14]. Ultrasonic influence in such situations is much more effective than mechanical stirrers. Further, it is almost immediate and the solubility of agents, which are difficult to dissolve, improved significantly [11]. Abismail et al. [15] compared the drop sizes of emulsions produced by mechanical agitation with that of power ultrasound using the model system water/kerosine/polyethoxylatedsorbitan-mono-stearate. They found that with ultrasound emulsification, smaller average drop sizes can be obtained and also were less polydipersed and more stable compared to emulsification by mechanical agitation. Li et al. [13] studied the enzymatic hydrolysis of a variety of pulps for its enhancement with continuous ultrasonic irradiation. They proposed a model for the enhancement effect of the ultrasonic irradiation on the enzymatic hydrolysis of cellulose. Recently, Freitas et al. [16] have demonstrated a continuous contact and contamination free ultrasonic emulsification system. Using this system, they could generate vegetable-oil-in water emulsions with droplet size of 0.5 µm and below with good reproducibility. Though many applications of ultrasound for generating fine dispersions of oil in water have been demonstrated, very few have used it to study its effect on the enzymatic hydrolysis of oils. In this paper, ultrasonic emulsification was used to generate oil-water emulsions for use in hydrolysis by lipase. Ultrasonically emulsified palm oil was used to measure the improvement in hydrolysis rate and compared with that obtained with emulsification by mechanical agitation. MATERIALS AND METHODS Materials Two types of lipase were used in this study, namely, solid lipase from Candida rugosa obtained from Sigma Chemical Co., Japan and liquid lipase from Mucor miehei obtained from Novo Nordisk, Denmark. For convenience, in rest of the paper, lipases A and B will refer to the first and the second lipases used, respectively. Refined, bleached and deodorised (RBD) palm oil used in this study was obtained from Lam Soon (M) Berhad, Malaysia. Analytical grade tributyrin was 160 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007

Sulaiman Al-Zuhair and K.B. Ramachandran

obtained from Sigma Chemicals Co., Germany. All other chemicals used were of analytical grade. Determination of enzyme activity The method is based on the hydrolysis of tributyrin by the enzyme and titrating the butyric acids produced with 0.05M NaOH in isopropanol [17]. The alkali consumption is registered as a function of time under standard conditions of 30oC and pH 7.0, using an auto-titrator (Metrohm 702 SM Titrino, Metrohm UK Ltd., UK). From the amount of alkali consumed, the equivalent amount of butyric acid in the samples was calculated and the enzyme activity determined. The enzyme activity is expressed in lipase unit (LU), where 1 LU is defined as the amount of enzyme that liberates 1µmol of titrable butyric acid per minute at 30 oC. Both lipases A and B, were examined in this experiment. Lipase A solution was prepared by dissolving different weights of solid lipase powder in 100 ml of distilled water and lipase B solution was prepared by dissolving different volumes of lipase in 100 ml of distilled water. Ten milliliters of enzyme solution were added to the reaction mixture to initiate the reaction. For both lipases, five repetitive experiments were carried out to determine the original enzyme activity. It was found that the activity of lipase A was 9,520 LU g-1 of the enzyme with a standard deviation of ±300 LU g-1. For lipase B, the activity was found to be 93,200 LU ml-1 with a standard deviation of ±880 LU ml-1. Hydrolysis experiments For hydrolysis experiments with ultrasonicated emulsions, the reaction solution made of 400 ml acetate buffer, the required amount of oil, distilled water and the emulsifier Gum Arabic were dispersed using ultrasonicator (Labsonic 2000 U, B. Braun Biotech International, Germany). Enzyme was not added in the above mixture. The ultrasonication power used was 175 W and the probe used was of standard type (19 mm). Ultrasonication was carried out in for 1 min. The emulsion was found to be stable up to 2 h and no significant coalescence was observed in this period. The reactor which consisted of a standard 600 ml beaker immersed in a temperature controlled water bath (Techne-Tempette TE 8D, Protocol Instruments Limited, West Byfleet, UK) set at 40 oC. The solution was gently mixed using a mechanical stirrer throughout the progress of the hydrolysis experiment. The hydrolysis reaction was initiated by adding the required amount of enzyme solution. Two milliliters samples were withdrawn at regular intervals for determining the rate of hydrolysis. Similar hydrolysis experiments were carried out using emulsions generated by mechanical agitation [3–6]. A four-flat pitched-blade turbine immersed in the solution at one-third-depth was used for agitation. The impeller was connected to a motor (Model IKA Labortechnik-RW 20, IKA, Germany). The reactor was small enough to get adequate mixing in the absence of baffle plates. 161 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


0.4

Ultrasonic Mixing 1300 rpm 1000 rpm 800 rpm

0.35

-3

-1

Initial rate of reaction, υ (mol m min )

For estimation of fatty acids produced, the samples were evaporated under a vacuum of 300 mbar and at 90 oC using vacuum evaporator (Model Buchi Rota Vapor R144, Buchi Laborotechnik AG, Switzerland). The fatty acids were then extracted in 50 ml of isopropanol and titrated using the auto-titrator, with 0.05 N NaOH in isopropanol to determine its concentration. A plot of product concentration against time was drawn and the slope of the plot at the origin gave the initial rate of reaction. Experiments were run at different substrate concentrations, enzymeconcentrations and stirrer speeds to determine their effects on the initial rate of hydrolysis of palm oil.

0.3 0.25 0.2 0.15 0.1 0.05 0 0

500

1000

1500

2000

2500

-3

Enzyme concentration, (kLU m )

Fig. 1. Comparison between ultrasonic mixing and mechanical agitation on the initial rate of reaction at different lipase A concentration. (S = 660.7 mol m-3 and T = 45 oC)



Initial rate of reaction, ν (mol m-3 min-1)

2.5 Ultrasonic mixing Mechanical agitation

2 1.5 1 0.5 0 0

100000

200000

300000

400000

Enzyme concentration (kLU m-3)

Fig.2. Comparison between ultrasonic mixing and mechanical agitation on the initial rate of reaction at different lipase B concentration. (S = 660.7 mol m-3 and T = 45 oC).

RESULTS AND DISCUSSION The experimental results and discussion presented here are found in our previous work [18]. Experiments were run at different lipase A concentrations and stirrer speeds to determine their effect on the initial rate of palm oil hydrolysis. The results are shown in Fig. 1. The same figure also shows the hydrolysis rates measured at different enzyme concentrations with ulatrasonically emulsified solution. The solid lines shown in the figure are connection between the experimental data, shown to highlight the trend. It can be seen that at low enzyme concentrations, the initial rate of reaction increased linearly. At high enzyme concentrations, this linear relationship tends to flatten, especially for low agitation speeds. This is because at these agitation speeds, when the enzyme concentration is high, the interfacial area could be saturated with the enzyme molecules [3,6]. This limits the substrate available for hydrolysis reaction and hence, any further increase in enzyme concentration in the bulk could not enhance the reaction rate. The figure also shows that the point where the effect of enzyme concentration tended to fade shifts to the right (higher enzyme concentration) as the agitation speed is increased. This point is not pronounced when ultrasonication was used for emulsification. This is due to the high interfacial area available for the enzyme to occupy in case of ultrasonication 163 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


and hence even higher enzyme concentration is needed to saturate it. The same trend was found when lipase B was used for the hydrolysis of palm oil at the same substrate concentration and temperature employed earlier (i.e., S = 661.0 mol m−3 and T = 45 oC), using ultrasonicator and mechanical agitation at 1000 rpm. Fig. 2 shows the results for lipase B activity in the range of 0-345 LU ml-1. The solid lines shown in the figure are connection between the experimental data, shown to highlight the trend. Here again, higher initial rates were obtained with ultrasonication and the enzyme saturation takes place at a higher concentration compared to stirred system. For the hydrolysis of palm oils by lipase A, at low enzyme concentrations, the following model equation was derived to predict the rate of reaction for a wider range of substrate concentrations [4-6]:

υ=

1.8 ×10 −3 (Et )m S ⎛ 7.7 ×107 ⎞ + 1⎟⎟ + S 5.65⎜⎜ 2 a t ⎝ ⎠

(1)

± 0.018

Where, at is the specific total interfacial area, (Et)m is the total enzyme mass concentration, S is the bulk substrate concentration and υ is the initial reaction rate. However, when dispersion is done using mechanical agitation, the model equation (Eq.1) deviates largely from the experimental data, at high enzyme concentrations [4,6]. On the other hand, as noted earlier, when ultrasonicator was used, the interfacial area of palm oil-water system is much higher than that formed when mixing was by mechanical agitation. Therefore, for hydrolysis using ultrasonicator, the apparent Michaelis Menten constant, K M = 5.65 7.7 × 10 7 a t2 + 1 , would be much less

[(

) ]

than substrate concentration, S and accordingly, Eq. (1) can be simplified to:

υ = 1.8 × 10 −3 (Et )m

(2)

Fig. 3 shows a comparison between the experimental results and the simplified model (Eq. 2) for the initial rate of reaction for dispersion using ultasonicator at different lipase A concentrations. The figure clearly shows that the experimental data are represented fairly well by the straight line given in the simplified model equation.



Fig. 3. Comparison between the experimental results and the simplified model on the initial reaction rate using ultasonication at different lipase A concentrations (S = 660.7 mol m-3 and T = o45 C)

The effect of increasing the substrate concentration on the initial rate of reaction is shown in Fig. 4. At a lipase B activity of 11.5 LU ml-1, it can be seen that, when ultrasonicator was used, the maximum rate of reaction was reached at a substrate concentration less than 300 mol m−3, indicating that the entire added enzyme is fully used at this substrate concentration due to high interfacial area generated. Hence, further increase in interfacial area due to increased substrate concentration had no effect on the initial rate. However, when mechanical agitation at 1000 rpm was used, the maximum rate of reaction was not reached even at the substrate concentration of 990 mol m−3. This indicates that the added enzyme is not fully utilized due to inadequate interfacial area generated even at this high substrate concentration. It can also be seen from the figure that the apparent Michaelis Menten constant, KM, value for the reaction with ulatrsonicated emulsion is much smaller compared to the reaction carried out with mechanically agitated system, which is due to larger interfacial area.


Initial rate of reaction, ΰ (mol m-3 min )


0.7 Ultrasonication Mechanical agitation (1000 rpm)

0.6 0.5 0.4 0.3 0.2 0.1 0 0

300

600

900

1200 -3

Substrate concentration,S (mol m )

Fig. 4. Comparison between the effect substrate concentration on the initial rate of reaction using ultrasonicator and mechanical agitation (1000 rpm at 11,500 kLU m-3 and 45 oC)

CONCLUSIONS The use of ultrasonicator is investigated for generating the oil-water emulsion for hydrolysis by lipase. It was found that by ultrasonic emulsification, the rate palm oil hydrolysis could be enhanced significantly and interfacial area saturation with enzyme was not observed. The maximum rate of reaction was also reached at a much lower substrate concentration due to increased substrate availability. The results confirm the applicability of ultrasonicator as a better alternative to mechanical agitation for increased rate of hydrolysis of lipids by lipase. SYMBOLS at

:

Specific total interfacial area (m2 m-3)

(Et)m

:

Total enzyme mass concentration (g m-3)

S

: Bulk substrate concentration (mol m-3)

Greek Letters

υ

: Reaction rate (mol m-3 min-1) 166



REFERENCES [1]

Hasan, F., A.A. Shah, A., Hameed 2005, “Industrial application of microbial lipases”, Enz Microbiol Technol, Vol. 49, pp. 235-251

[2]

Rooney, D., L.R. Weartherley, 2000, “The effect of reaction conditions upon lipase catalysed hydrolysis of high oleate sunflower oil in stirred liquid-liquid reactor” Process Biochem, Vol. 36, pp. 947-953.

[3]

Noor, I.M., M. Hasan, K.B. Ramachandran, 2003, “Effect of operating variables on the hydrolysis of palm oil by lipase”, Proc Biochem, Vol. 39, pp. 13-20.

[4]

Al-Zuhair, S., M. Hasan and K.B. Ramachandran, 2003, “Kinetic hydrolysis of palm oil using lipase”, Proc Biochem, Vol. 38, pp.1155–1163

[5]

Al-Zuhair, S., M., Hasan and K.B., Ramachandran, 2004, “Unsteady-state kinetics of lipolytic hydrolysis of palm oil in a stirred bioreactor”, Biochem Eng J, Vol. 19, pp. 81–86

[6]

Al-Zuhair, S., M. Hasan and K.B. Ramachandran, 2004, “High enzyme concentration model for the kinetics of hydrolysis of oils by lipase”, Chem Eng J, Vol. 103, pp. 7-11.

[7]

Tsai, S.W. and C.S. Chang, 1993, “Kinetics of Lipase-Catalysed Hydrolysis of Lipids in Biphasic Organic-Aqueous Systems”, J Chem Tech Biotechnol, Vol. 57, pp. 147-154.

[8]

Verger, R., C.E.M. Maria and H.D. Gerard, 1973, “Action of Phospholipase A at Interfaces”, J Biological Chem, Vol. 248, pp. 4023-4034.

[9]

Al-Zuhair, S., K.B. Ramachandran and M. Hasan, 2004, “Investigation of the specific interfacial area of a palm oil-water system”, J Chem Technol Biotecnol, Vol. 79, pp. 706-710.

[10]

Shamel, M.M., K.B. Ramachandran and M. Hasan, 2005, “Operational stability of lipase Enzyme: Effect of temperature and shear”, Dev Chem Eng Min Proc, Vol. 13, pp. 599-604.

[11]

Behrend, O., K. Ax and H. Schubert, 2000, “Influence of continuous phase viscosity on Emulsification by Ultrasound”, Ultrasonics Sonochem, Vol. 7, pp. 77-85. 167



[12]

Behrend, O. and H. Schubert, 2001, “Influence of hydrostatic pressure and gas content on continuous ultrasound emulsification”, Ultrasonics Sonochem, Vol. 8, pp. 271-276.

[13]

Li, C., M. Yoshimoto, N. Tsukuda, K. Fukunaga and K. Nakao, 2004, “A kinetic study on enzymatic hydrolysis of a variety of pulps for its enhancement with continuous ultrasonic Irradiation”, Biochem Eng J, Vol. 16, pp. 155-164.

[14]

Chanamai, R., J.N. Coupland and D.J. McClements, 1998, “Effect of temperature on the ultrasonic properties of oil-in-water emulsions”, Colloids Surfaces. A: Physiochem Eng Aspects, Vol. 139, pp. 241-250

[15]

Abismail, B., J.P. Canselier, A.M. Wilhelm, H. Delmas and C. Gourdan, 1999, “Emulsification by ultrasound: drop size distribution and stability”, Ultrasonics Sonochem, Vol. 6, pp. 75-83.

[16]

Freitus, S., G. Hielscher, and H.P. Merkle, 2005, “Continuous contact-and contamination-free ultrasonic emulsification-a useful tool for pharmaceutical development and production”, Ultrasonics Sonochem, Vol. 13, pp. 76-85.

[17]

NOVO Industrials. Analytical methods handout, Enzyme Process Division, NOVO industrials, Denmark (1995).

[18]

Ramachandran, K.B., S. Al-Zuhair, C.S. Fong and C.W. Gak, 2006, “Kinetic study on hydrolysis of oils by lipase with ultrasonic emulsification”, J Biochem Eng, Vol. 32, pp.19–24


FED-BATCH PRODUCTION OF HIGH FRUCTOSE SYRUP AND ETHANOL FROM SUCROSE BY Saccharomyces cerevisiae ATCC 36858

1

Hasan K. Atiyeh 1: Department of Chemical Engineering, King Saud University, P.O. Box 800, Riyadh 11421, Saudi Arabia, [email protected]

ABSTRACT Fed-batch production of enriched fructose syrups and ethanol from synthetic media with sucrose using the mutant Saccharomyces cerevisiae ATCC 36858 was studied. The effect of initial concentrations of ethanol and KH2PO4 on the production process was investigated. Results indicate that more than 90% of the theoretical amount of fructose was recovered in all tests. The theoretical ethanol yield was over 72%, in media containing a total (initial plus added) sucrose concentration of 364 g/L and with no ethanol added. The fructose and ethanol concentrations in these media were about 180 and 74 g/L, respectively. The fructose content in the produced syrup was above 92% of its carbohydrate amount. The ethanol yield was less than 69% of the theoretical value when the total sucrose and initial ethanol concentrations in the media were about 361 and 32 g/L, respectively. The fructose and ethanol concentrations in these media were about 173 and 85 g/L, respectively. The fedbatch process offered a method for the production of enriched fructose syrups with high fructose concentrations. Considering the above characteristics, this mutant has the potential for industrial utilization in the production of fructose and ethanol from media that contain sucrose such as molasses or from media with glucose/fructose mixture such as date syrups.

KEY WORDS Fed-batch, fructose, sucrose, ethanol, Saccharomyces cerevisiae

INTRODUCTION Fructose is naturally found in fruits, vegetables and their juices, as well as in honey. It is the sweetest natural sugar that gives these foods their sweet taste. Due to its sweetness, smaller quantities of fructose are required to produce the same sweetness


FED-BATCH PRODUCTION OF HIGH FRUCTOSE SYRUP AND ETHANOL FROM SUCROSE

as sucrose, resulting in lower calorie intake. Fructose is of high commercial value and is widely used in beverages, baking, canning and other food products. Fructose is available as a component of invert sugar and of high fructose corn syrups (HFCS) in which fructose and glucose are present in nearly equimolar concentrations. Three HFCS containing 42, 55 and 90% fructose are commercially produced. In sweetness, the 55 HFCS is about equal to sucrose, while the 90 HFCS is 20-60% sweeter than sucrose (Long, 1991). The production of pure, solid fructose is carried out by crystallization of syrups containing 90-95% fructose. Existing industrial methods use expensive chromatographic techniques to produce the 90 HFCS from the 42 HFCS (Coker and Venkatasubramanian, 1987). Novel technologies, which involve the selective conversion of glucose to ethanol, have been considered for the production of high fructose syrups (HFS) from sucrose media using mutants of Tricholoma nudum NRRL 2371 (Reusser et al., 1960), Fusarium sp. F5 (Ueng et al., 1982), Pullularia pullulans ATCC 9348 (Fan, 1988), Zymomonas mobilis ATCC 39676 (Doelle and Greenfield, 1985), Z. mobilis ATCC 53432 (Suntinanalert et al., 1986), and Z. mobilis ATCC 53431 (Doelle and Doelle, 1991; Edye et al., 1989; Kirk and Doelle, 1994). Studies with the majority of the above microorganisms showed that a significant fructose consumption and/or production of by-products such as sorbitol were noticed. S. cerevisiae ATCC 36859 demonstrated high ethanol and fructose yields in batch fermentation with glucose/fructose mixtures (Koren and Duvnjak, 1993). However, this strain was not able to utilize sucrose, which limited its use to media with glucose/fructose mixtures. The production of fructose and ethanol in batch fermentation using S. cerevisiae ATCC 36858, which possesses a capability to selectively ferment glucose and galactose to ethanol from synthetic media with sucrose (Atiyeh and Duvnjak, 2001a,b) or raffinose (Atiyeh and Duvnjak, 2003a) as well as beet (Atiyeh and Duvnjak, 2002) and cane (Atiyeh and Duvnjak, 2003b) molasses media was studied. The production of high concentrations of fructose and ethanol is advantageous in reducing the distillation and evaporation costs when producing syrups. However, this requires carrying out fermentation in media with high sucrose concentrations. The use of media with high sucrose concentrations in batch process resulted in the inhibition of growth and ethanol production by S. cerevisiae ATCC 36858 (Atiyeh and Duvnjak, 2001a). The fed-batch process offers a method for obtaining high concentrations of fructose and ethanol from sucrose media. The objective of the present study is to examine fed-batch production of fructose and ethanol from synthetic sucrose media using S. cerevisiae ATCC 36858.


Hasan K. Atiyeh

MATERIALS AND METHODS Microorganism, media, growth and production conditions The yeast Saccharomyces cerevisiae ATCC 36858 (ATCC, USA), obtained by mutation (Lobo and Maitra, 1977) was maintained on malt extract agar slants. The medium for inoculum was composed of (g) glucose, 10.0; yeast extract, 30.0; peptone, 3.5; KH2PO4, 2.0; MgSO4⋅7H2O, 1.0; (NH4)2SO4, 1.0; deionized water up to 1.0 L. The synthetic medium used for the production of fructose and ethanol differed from the medium for inoculum in the concentrations of yeast extract (10.0 g/L), KH2PO4 (2.0 and 6.0 g/L) and the type of sugar used. In this medium, glucose was substituted by sucrose. The initial concentration of sucrose was about 240 g/L. In addition, about 30 g/L of ethanol was initially added for two tests to check the effect of initial ethanol concentration on the growth and fermentation capability of the mutant. In a fed batch mode, about 10 mL of concentrated sucrose solution was sterilized and then added to the fermentation media when the concentration of glucose in the media was below 30 g/L. The inoculum and production tests were carried out in 500 mL Erlenmeyer flasks containing 100 mL of medium. About 10 mL of a concentrated inoculum was added to each 100 mL of the tested production media. Experiments were carried out in a rotary shaker at 160 rpm and 33οC. All media were sterilized at 115οC for 15 min. Samples were taken aseptically at various times to determine biomass, ethanol and sugar concentrations. Analytical methods Ethanol concentration was determined enzymatically using alcohol dehydrogenase (Bernt and Gutmann, 1974). Sucrose, glucose, fructose, fructo-oligosaccharides and glycerol were measured using a 600E system controller Water’s high performance liquid chromatography (HPLC). A Sugar-Pak I column (Waters, Massachusetts) operated at 75οC, with deionized water as the mobile phase flowing at 0.5 mL/min, was used. The biomass concentration was measured by drying the samples at 105οC for 24 h in pre-weighed aluminum dishes.

RESULTS AND DISCUSSION Fermentation pattern of S. cerevisiae ATCC 36858 in a sucrose medium with no ethanol added into the medium The kinetics of growth of S. cerevisiae ATCC 36858 and fructose and ethanol production in fed-batch fermentation in a medium containing 364.3 g/L total sucrose is displayed in Figure 1. The initial concentrations of sucrose, biomass, ethanol and KH2PO4 were 241.5, 8.4, 0.8 and 2.0 g/L, respectively. A fast, and complete hydrolysis, of sucrose was observed in less than 9 h from the beginning of 171 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


fermentation. This resulted in the accumulation of fructose and ethanol. In the beginning of the process, glucose started also to accumulate since its production by hydrolysis of sucrose was faster than its consumption. Glucose concentration then started to decrease. When the concentration of glucose was about 10 g/L, a fresh concentrated sucrose solution was quickly added into the medium. This was done 12 h after inoculation. The fermentation was then continued in a batch mode. The added sucrose plus the initial sucrose at time zero combined together correspond to a total sucrose concentration of 364.3 g/L. The addition of fresh sucrose was done to supply the yeast with more sucrose in order to produce a syrup with a high fructose and ethanol concentrations. As a result of adding fresh sucrose, the fructose and ethanol concentrations in the medium increased from 131 to 181 g/L and from 53 to 74 g/L, respectively, by the end of fermentation (Figure 1). A syrup the carbohydrate content of which was more than 92% fructose was produced.

300

5 0 0

25 50 Time (h)

100

4.4 50

1.0

0.1

3.4

0

10

350

20 30 Time (h)

40

0 50

200

100

0

Figure 1. Fed-batch production of fructose and ethanol using S. cerevisiae ATCC 36858 in a medium containing 364.3 g/L sucrose (initial plus added). A fresh concentrated sucrose solution was added after 12 hours of inoculation. Initial ethanol and KH2PO4 concentrations were 0.8 and 2.0 g/L, respectively: (▲) biomass; (♦) pH; (●) ethanol; (■) sucrose; (□) glucose; (▼) fructose; (○) oligosaccharides; (+) glycerol.


Sucrose; glucose; Fructose (g/L)

10.0

pH

Biomass (g/L)

100.0

150

10

Ethanol (g/L)

Fructooligosaccharides; Glycerol (g/L)

5.4

500.0

Hasan K. Atiyeh

The specific growth rate and biomass yield were 0.071 h-1 and 0.022 g/g of monosaccharides consumed, respectively. However at the end of fermentation, the fructose and ethanol yields were 95% and 76% of the theoretical values, respectively. A small amount of fructo-oligosaccharides was produced in the beginning of the process and simultaneously consumed by the end of fermentation. In addition, some glycerol (7.7 g/L) was also produced. In this study, the concentration of fructose in the syrup produced using S. cerevisiae ATCC 36858 was 27% higher than when Z. mobilis ATCC 53431 was used in fedbatch process with sucrose media at a concentration of 350 g/L (Edye et al., 1989). In their study, the fructose and ethanol concentrations were 142 and 77 g/L, respectively, at the end of fermentation. Edye et al. (1989) found that about 5% of sucrose accumulated in the medium and the fructose and ethanol yields were 81% and 72%, respectively. There was also a production of about 16 g/L of sorbitol that lowered the fructose yield. Fermentation pattern of S. cerevisiae ATCC 36858 in a sucrose medium with ethanol initially added into the medium The kinetics of growth of S. cerevisiae ATCC 36858 and the fructose and ethanol production in fed-batch fermentation in a medium containing initial ethanol and total sucrose concentrations of 33.0 g/L and 371.7 g/L, respectively, is shown in Figure 2. Similar to the sucrose medium with no ethanol added, glucose started to accumulate in the beginning, and then it was consumed by the yeast. When the glucose concentration in the medium decreased to 28 g/L after 15 h after inoculation, a fresh concentrated sucrose solution was quickly added into the medium and the fermentation proceeded in a batch mode. The fructose concentration in the medium increased from 113 to 174 g/L between 15 and 49 h of the fermentation process. A maximum ethanol concentration of 87 g/L (added plus produced) was also obtained. The fructose and ethanol yields were 90% and 64% of the theoretical values, respectively. The biomass yield was 0.012 g/g of glucose and fructose consumed. It was noticed that about 6 g/L of fructo-oligosaccharides were produced in the beginning of the process but they were not completely consumed. In addition, about 5 g/L of glycerol accumulated in the fermentation medium.


300

2

120 0

25 50 Time (h)

4.4

80

1.0

0.1

40

3.4

0

10

20 30 Time (h)

40

0 50

200

100

0

Figure 2. Fed-batch production of fructose and ethanol using S. cerevisiae ATCC 36858 in a medium containing 371.7 g/L sucrose (initial plus added). A fresh concentrated sucrose solution was added after 15 hours of inoculation. Initial ethanol and KH2PO4 concentrations were 33.0 and 2.0 g/L, respectively: (▲) biomass; (♦) pH; (●) ethanol; (■) sucrose; (□) glucose; (▼) fructose; (○) oligosaccharides; (+) glycerol.

Effect of initial ethanol concentration The effect of ethanol and KH2PO4 on growth and fermentation capabilities of S. cerevisiae ATCC 36858 was also investigated. In sucrose medium with 2.0 g/L of KH2PO4, the results showed that the specific growth rate of the mutant and the biomass yield decreased by 38% and 45%, respectively, when the initial ethanol concentration was 33.0 g/L compared to no ethanol added (Table 1). Similar decrease in specific growth rate and biomass yield was noticed when ethanol was added into the medium and for KH2PO4 concentration of 6.0 g/L. The specific initial rate of sucrose hydrolysis decreased by more than two fold with an increase in the initial ethanol concentration from 0.7 to 33.0 g/L. In addition, the ethanol yield and productivity decreased as a result of adding ethanol to the fermentation medium (Table 2). However, the fructose yield did not significantly change.


Sucrose; glucose; Fructose (g/L)

10.0

350

4

0

pH

Biomass (g/L)

100.0

160

6

Ethanol (g/L)

Fructooligosaccharides;

5.4

500.0

Glycerol (g/L)


Hasan K. Atiyeh

Table 1. Biomass yield, specific growth rates and specific initial rates of sucrose hydrolysis for media with about 363 g/L sucrose and various initial ethanol and KH2PO4 concentrations using S. cerevisiae ATCC 36858. Initial Specific Specific initial rate Biomass Total Initial Initial yieldc of sucrose sucros ethano KH2P biomas growth rate, µ O4 s (g/g) hydrolysisd e l (h-1) (g/L) (g/L) (g/L) (g/L) (g/g⋅h) 364.3a 0.8 2.0 8.4 0.071 0.022 8.38 364.5a 0.7 6.0 8.8 0.074 0.020 7.72 371.7b 33.0 2.0 8.9 0.044 0.012 3.63 31.4 6.0 8.6 0.048 0.014 3.45 351.1b a Based on initial sucrose and fresh sucrose added after 12 hours from beginning of fermentation b Based on initial sucrose and fresh sucrose added after 15 hours from beginning of fermentation c Values are calculated at the end of fermentation (49 hours) d Values are calculated after 2.6 hours of hydrolysis

The low biomass and ethanol yields as well as ethanol productivity are due to the inhibition effect of the added ethanol into the media. This is consistent with results of other researchers who showed that ethanol inhibits growth, viability and fermentation capability of many organisms (Jones et al., 1981; Kosaric et al., 1983; Stewart et al., 1983; D’Amore et al., 1990). It has been reported that ethanol has inhibitory effects on key glycolytic enzymes such as hexokinase and alcohol dehydrogenase as well as on cell constituents and membrane, which affects the cells properties for substrate uptake and ethanol excretion (Jones et al., 1981; Kosaric et al., 1983; Nagashima, 1990). Table 2. Fructose and ethanol yields, and ethanol productivities for media with about 363 g/L sucrose and various initial ethanol and KH2PO4 concentrations using S. cerevisiae ATCC 36858. Initial Ethanol Ethanol Fructose Total Initial Initial biomas productivityc yieldc yieldc sucrose ethanol KH2PO4 (g/L) s (%) (%) (g/L) (g/L) (g/L⋅h) (g/L) 364.3a 0.8 2.0 8.4 1.56 94.6 76.2 364.5a 0.7 6.0 8.8 1.62 92.6 72.2 371.7b 33.0 2.0 8.9 1.26 90.0 64.4 351.1b 31.4 6.0 8.6 1.24 93.4 68.6 a Based on initial sucrose and fresh sucrose added after 12 hours from beginning of fermentation b Based on initial sucrose and fresh sucrose added after 15 hours from beginning of fermentation c Values are calculated at the end of fermentation (49 hours)



A complete hydrolysis of sucrose by the end of fermentation was noticed in all media. Syrups the carbohydrate content of which was more than 92% fructose were produced by S. cerevisiae ATCC 36858 using sucrose media with no ethanol added (Table 3). However when ethanol was initially added into the sucrose media, the fructose fraction in the produced syrups decreased to about 82% due to the slow uptake of glucose and its accumulation. Glycerol was produced in low concentrations by S. cerevisiae ATCC 36858 (Table 3). In addition, small amounts of fructo-oligosaccharides accumulated in the media when ethanol was initially added into the media. It is appropriate to mention that, at the end of the process, neither sorbitol nor residual sucrose were detected in the tested sucrose media when S. cerevisiae ATCC 36858 was used. However in fedbatch process with Z. mobilis ATCC 53431, sorbitol (16 g/L) was produced as a byproduct and about 5% of the total sucrose (350 g/L) added to the medium was not utilized by the bacterium (Edye et al., 1989). Table 3 Maximum and residual fructo-oligosaccharides, final glycerol concentrations in the fermentation broth and fructose contents in the produced syrups using S. cerevisiae ATCC 36858 in media with about 363 g/L sucrose and various initial ethanol and KH2PO4 concentrations. Total Maximum Residual Fructose Glycero Initial Initial sucros ethan fructofructocontents KH2P l e O4 oligosaccharides oligosaccharid in syrups ol es produce dc (g/L) (g/L) (g/L) (g/L) % (g/L) (g/L) 364.3a 0.8 2.0 8.6 0.0 92 7.7 364.5a 0.7 6.0 8.0 0.0 98 8.0 33.0 2.0 5.8 1.9 83 4.7 371.7b 351.1b 31.4 6.0 5.9 1.8 80 4.4 a Based on initial sucrose and fresh sucrose added after 12 h from beginning of fermentation b Based on initial sucrose and fresh sucrose added after 15 h from beginning of fermentation c Values are calculated based on the final total sugar concentrations


Hasan K. Atiyeh

Effect of initial KH2PO4 concentration Phosphorus, provided in the form of phosphate, plays a role in determining the rate of fermentation (Jones et al., 1981). Phosphorus (as H2PO4-) is also essential for cell growth, controls synthesis of lipids and carbohydrates, and maintains the integrity of the cell wall. In the present study, the increase in KH2PO4 concentration from 2.0 to 6.0 g/L has an insignificant effect on the rate of fermentation or on the fructose and ethanol yields (Tables 1 to 3). This shows that the increase in the concentration of KH2PO4 from 2.0 to 6.0 g/L has no effect on the overall fermentation kinetics. The results of the present study showed that S. cerevisiae ATCC 36858 has more advantages over Z. mobilis in the production of fructose and ethanol. Furthermore, S. cerevisiae ATCC 36858 has the ability to produce syrups with high fructose and ethanol concentrations in fed-batch process.

CONCLUSIONS The mutant S. cerevisiae ATCC 36858 was able to produce enriched fructose syrups in fed-batch fermentation. High fructose yield (above 90%) was obtained in all media tested. Ethanol yield was above 72% of theoretical value when ethanol was not initially present in the media with a total sucrose (initial plus added) concentration of 364 g/L. However, lower ethanol yields were observed in media in which about 32 g/L ethanol was added at the beginning of the fermentation process. Complete hydrolysis of sucrose was noticed in the media tested. The overall fermentation kinetics were not affected by changing the initial concentration of KH2PO4 in the medium from 2.0 to 6.0 g/L. Syrups the carbohydrate content of which contains more than 92% fructose can be produced from media with a total sucrose concentration of 364 g/L and no added ethanol. The fructose and ethanol concentrations in the produced syrup at the end of fermentation were about 180 and 74 g/L, respectively. The fed-batch process offered a method for obtaining high concentrations of fructose and ethanol from sucrose media using S. cerevisiae ATCC 36858.

REFERENCES Atiyeh, H. and Z. Duvnjak, 2001a, “Production of Fructose and Ethanol from Media with High Sucrose Concentration by a Mutant of Saccharomyces cerevisiae”. Journal of Chemical Technology and Biotechnology, Vol. 76, pp. 1017-1022. Atiyeh, H. and Z. Duvnjak, 2001b, “Study of the Production of Fructose and Ethanol from Sucrose Media by Saccharomyces cerevisiae”. Applied Microbiology and Biotechnology, Vol. 57, pp. 407–411. 177 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


Atiyeh, H. and Z. Duvnjak, 2002, “Production of Fructose and Ethanol from Sugar Beet Molasses Using Saccharomyces cerevisiae ATCC 36858”. Biotechnology Progress, Vol. 18, pp. 234-239. Atiyeh, H. and Z. Duvnjak, 2003a, “Utilization of Raffinose and Melibiose by a Mutant of Saccharomyces cerevisiae”. Journal of Chemical Technology and Biotechnology, Vol. 78, pp. 1068–1074. Atiyeh, H. and Z. Duvnjak, 2003b, “Production of Fructose and Ethanol from Cane Molasses Using Saccharomyces cerevisiae ATCC 36858”. Acta Biotechnologica, Vol. 23, pp.37-48. Bernt, E. and I. Gutmann, 1974, “Ethanol Determination with Alcohol Dehydrogenase and NAD”. In: Methods of Enzymatic Analysis, Bergermeyer, H. U. (ed.), Academic Press, New York, pp. 1499-1502. Coker, L. E. and K. Venkatasubramanian, 1987, “Corn Sweeteners”. In: Food Biotechnology, Knorr, D. (ed.), Marcel Dekker Inc., New York, pp. 443-460. D’Amore, T., C. J. Panchal, I. Russell and G. G. Stewart, 1990, “A study of Ethanol Tolerance in Yeast”. Critical Reviews in Biotechnology, Vol. 9, pp. 287-304. Doelle, M. B. and H.W. Doelle, 1991, “High Fructose Formation from Sugarcane Syrup and Molasses Using Zymomonas mobilis mutants”. Biotechnology Letters, Vol. 13, pp. 875-878. Doelle, H. W. and P.E. Greenfield, 1985, “Fermentation Pattern of Zymomonas mobilis at High Sucrose Concentrations”. Applied Microbiology and Biotechnology. Vol. 22, pp. 411-415. Edye, L. A., Johns, M. R. and K.N. Ewings, 1989, “Fructose Production by Zymomonas mobilis in Fed-Batch Culture with Minimal Sorbitol Formation”. Applied Microbiology and Biotechnology, Vol. 31, pp. 129-133. Fan, L., 1988, “Process for Producing Fructose”. U.S. Patent 4774183. Jones, R. P., N. Pamment and P. F. Greenfield, 1981, “Alcohol Fermentation by Yeasts: The Effect of Environmental and Other Variable”. Process Biochemistry, Vol. 16, pp. 42-49. Kirk, L. A. and H.W. Doelle, 1994, “Simultaneous Fructose and Ethanol Production from Sucrose Using Zymomonas mobilis 2864 Co-Immobilized with Invertase”. Biotechnology Letters, Vol. 16, pp. 533-538. 178 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007

Hasan K. Atiyeh

Koren, D. W. and Z. Duvnjak, 1993, “Kinetics of the Selective Fermentation of Glucose from Glucose/Fructose Mixtures Using Saccharomyces cerevisiae ATCC 36859”. Acta Biotechnologica, Vol. 35, pp. 311-322. Kosaric, N., A. Wieczorek, G. P. Cosentino, R. J. Magee and J. E. Prenosil, 1983, “Ethanol Fermentation”. In Biotechnology, Vol. 3, H. J. Rehm, G. Reed and H. Dellweg, (eds.), VCH, Weinheim, Germany, pp. 257-385. Lobo, Z. and P.K. Maitra, 1977, “Genetics of Yeast Hexokinase”. Genetics, Vol. 86, pp. 727-744. Long, J. E., 1991, “High fructose corn syrup”. In: Alternative sweeteners, Nabors, L. O. and Gelardi, R. C. (eds.), 2nd ed., Marcel Dekker Inc., New York, pp. 247-258. Nagashima, M., 1990, “Progress in Ethanol Production with Yeasts”. In: Yeast Biotechnology and Biocatalysis, H. Verachtert and R. De Mot (eds.), Marcel Dekker Inc., New York, pp. 57-84. Reusser, F., Gorin, P. A. J. and J.F.T. Spenser, 1960, “The Production of Fructose as a Residue of Sucrose Fermentation by Tricholoma nudum”. Canadian Journal of Microbiology, Vol. 6 pp. 17-20. Stewart, G. G., C. J. Panchal, I. Russell and A. M. Sills, 1983, “Biology of Ethanol Producing Microorganisms”. CRC Critical Reviews in Biotechnology, Vol. 1, pp. 161-188. Suntinanalert, P., Pemberton, J. P. and H.W. Doelle, 1986, “The Production of Ethanol Plus Fructose Sweetener Using Fructose Utilization Negative Mutant of Zymomonas mobilis”. Biotechnology Letters, Vol. 8, pp. 351-356. Ueng, P. P., McCracken, L. D., Gong, C. S. and G.T. Tsao, 1982, “Fructose Production from Sucrose and High Fructose Syrup: A Mycelial Fungal System”. Biotechnology Letters, Vol. 4, pp. 353-358.


ON THE DEVELOPMENT OF AN INTEGRATED KNOWLEDGE BASED SYSTEM FOR CHEMICAL PROCESS FLOWSHEET SYNTHESIS

Abdullah Alqahtani 1,Klaus Hellgardt 2,Richard Holdich 1, Iain Cumming 1 1: Department of Chemical Engineering, Loughborough University, Loughborough, Leicestershire, LE11 3TU, UK. [email protected], [email protected], [email protected] 2: Department of Chemical Engineering and Chemical Technology, Imperial College London, London, SW7 2AZ, UK. [email protected]

ABSTRACT Since the last four decade, very few systematic procedures have been proposed for the synthesis of a complete chemical process flowsheet. Mathematical design and heuristics are the two main methods usually used in process synthesis. Most approaches use heuristics based on studying reaction and separation systems in isolation. This paper discusses the development of process synthesis software that integrates knowledge based system with Aspen HYSYS process simulator, HYSYS optimizer and Aspen Icarus economic evaluator utilising knowledge from existing industrial processes to obtain heuristic rules. The prototype IKBS has been applied for the selection of reactor systems for the ethylene oxide and ethylene glycol manufacturing processes. A wide range of chemical reactors are considered during the selection process, and then elimination of reactors takes place at different steps until better alternatives are selected and justified. Analysis by the software suggests the use of two reactor systems and a list of suitable reactors. The list contained new and currently used reactor types in addition to the recommended reactors by industrial research. Modular simulation of reactors has been conducted to account for the non-ideal behaviour.

Keywords: Chemical process synthesis, chemical reactor system selection, knowledge based system, modular process simulation.

1. INTRODUCTION Chemical process synthesis is one of the most important areas of chemical process design as it deals with the problem of how to develop and integrate flowsheets for a chemical product manufacturing processes. Since Rudd1, proposed the first method for process synthesis, several works have been published based on the systematic Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007

ON THE DEVELOPMENT OF AN INTEGRATED KNOWLEDGE BASED SYSTEM FOR CHEMICAL

generation of a flowsheet, evolutionary modification, and superstructure optimization. Due to the fact that process synthesis problems are by nature combinatorial and open ended, a number of different approaches have been proposed. The two main approaches for process synthesis are heuristic methods, which consist of a series of heuristic rules to screen process alternatives, and the mathematical methods which rely on optimization techniques. When only heuristics are used, optimal design is not guaranteed and the method is limited to the state of knowledge. The mathematical programming methods restrict design considerations to the proposed superstructure and only limited size problems can be handled.2 Based on the previous research efforts in process synthesis, existing approaches mostly use heuristics based on the study of reactors and separation systems in isolation. Therefore, the synthesis of a total process flowsheet using a practical method has not yet been fully investigated. Douglas3,4 has produced a design methodology based on the hierarchical decomposition approach to process synthesis. Douglas’ process synthesis hierarchical method relies on sets of rules at different stages during process development. An overview of MINLP optimization techniques and applications for process synthesis can be found in5, 6. Linke and Kokossis7 have also developed a general framework for selecting process designs through simultaneous exploitation of reaction separation options. The objective of this research is to develop an Integrated Knowledge Based System (IKBS) for the synthesis of a complete chemical process flowsheet. This work has been structured around two themes, the structure of a hierarchical knowledge base, and the development of software that can provide an automation of the synthesis procedure to exploit interactions between reaction and separation utilising thirdparty software. This research differs from the previous research work in the following aspects:

Proposed approach integrates knowledge based systems with third-party process simulators, flowsheet optimizer and economic evaluators,

The developed software can synthesis multiple and novel reactor-separatorrecycle systems for petrochemical processes,

Databases have been successfully incorporated to obtain physical properties, prices, and safety and environmental impacts,

Modular simulation of technical reactors has been conducted to account for the non-ideal behaviour,

Software results have been successfully validated using existing commercial processes and industrial research. 182


A Abdullah Alqahtani, et al.

2. A GENERIC REPRESENTATION FRAMEWORK FOR CHEMICAL PROCESS SYNTHESIS Knowledge Based System (KBS) is essentially a computer program that has a specialised knowledge about a specific area and solves a specific class of problems using that knowledge.8 The structure of the IKBS is illustrated in Figure 1. Excel Visual Basic for Application (VBA) is being used as a tool to provide the basic elements of the integrated knowledge based system. In IKBS, the user communicates with the system using the user interface in Excel. The database of facts invoked by the rules includes the fact list, which contains the data from which inferences are derived. The knowledge-base contains all the rules used by the expert system. The inference engine infers by deciding which rules are satisfied by facts, prioritises the satisfied rules, and executes the rule with the highest priority. In each level of the synthesis, the user is guided in a step by step manner to generate process alternatives.

Knowledge Base (Rules)

User

User Interface (Excel)

Database (Facts)

Aspen Icarus Process Evaluator (Aspen IPE)

Inference Engine

Process Simulator (Aspen HYSYS)

Aspen Simulation Workbook (ASW)

Process Optimization (HYSYS Optimizer)

Figure1: Structure of IKBS for chemical process synthesis

Aspen HYSYS chemical process simulator is being used as a design tool in the development of chemical flowsheet and is being considered as a means to evaluate different design options. Aspen Simulation Workbook (ASW) is being used as a tool for interfacing Aspen HYSYS with Excel worksheets. HYSYS Optimizer is being used as flowsheet optimizer to determine the near optimal operating conditions. Aspen Icarus Process Evaluator (IPE) is a tool to extend the results of HYSYS to generate rigorous size and estimates the capital and operating costs.9 This structured method allows systematic identification of the most economical process flowsheet.



2.1. Systematic Procedure for Chemical Process Flowsheet Synthesis The systematic procedure integrates heuristics with process simulation and economic evaluation in a set of synthesis levels as illustrated in Figure 2. This work addresses the automated synthesis of process configurations that exploit interactions between reaction and separation to maximise the process performance. The user starts to specify some process relevant data required by IKBS to build a knowledge base for the process in a form of input/output information. There are often a number of alternative reaction paths to manufacture a certain product. Economic potential is used to eliminate some of the alternatives that are not economically viable.

Figure 2: Flowchart of the proposed systematic procedure The main task of the separation system synthesis consists of selecting the type, location, sequences and operating conditions of the separation system. The design of recycle from separators to reactor involves recycle component classifications, number of recycle streams determination, and the specifications and locations of liquid/vapour recycle and purge. These alternative flowsheets can be simulated using Aspen HYSYS to solve the mass and energy balances given by the user, calculate the thermodynamic properties of process streams, and determine operating conditions. A generic flowsheet is used to simulate different alternative reactor-separator-recycle systems. Figure 3, illustrates the simplified generic flowsheet for a reactor system. A matrix of splitter ratios is used in Excel to specify the direction and magnitude of streams between the reactors. Internal recycle around the reactors and the distribution of reactants between reactors are used to examine different configurations. 184 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


Figure 3: Simplified generic flowsheet for reactor system synthesis and optimization HYSYS Optimizer provides a tool for chemical process flowsheet optimization. It uses an advanced algorithm for optimization based on sequential quadratic program (SQP) technology. Optimized flowsheets are subsequently loaded into Aspen IPE for sizing and economic evaluation. A limited number of process flowsheets are proposed based on meeting the design requirements at low investment cost and high profit. 2.2. Reactor System Synthesis Strategy The synthesis of reactor system is an important part of an overall chemical process flowsheet development. In a chemical process, feed preparation, product recovery and recycles steps are directly influenced by the reactor system. The two main previous works in chemical reactor system synthesis using heuristics method were carried out by Schembecker and co-workers 10,11 and, Jacobs and co-workers 12- 14. READPERT (reactor development, selection and design expert system) which was developed by Schembecker and co-workers10,11 as heuristic-numerical consulting system contains five different modules as illustrated in Figure 4. READPERT does not consider the separation system and does not use precise kinetic information. It also does not use databases to import the facts required during the synthesis such as physical properties. Furthermore, READPERT does not size the reactor nor perform economic evaluation.

Figure 4: The main modules of READPERT



The other work in reactor synthesis is KBS for reactor selection which was developed by Jacobs and co-workers12-14. As illustrated in Figure 5, KBS for reactor selection contains seven steps of synthesis. KBS reactor selection method works in a choice of matrix which represents the reactors that are left and the reactor properties that can be used for the selection. A new choice matrix represents the selection process progress after the rejection of some reactors. This selection process is described as a sequential construction of new reduced choice matrices. These new reduced choice matrices are constructed by matching reactor properties suitable for the desired chemical process.

Figure 5: KBS for reactor selection synthesis decomposition levels

KBS synthesises reactor systems in isolation from separation system. It also relies heavily on the user to provide many facts during the synthesis steps instead of utilising a database for this purpose. KBS does not include reactor sizing or economic evaluation. Recently, Montolio-Rodriguez and his co-workers15 have proposed a systematic identification of optimal design for Acetic acid process. Their case study involves only conventional reactors and separators such as CSTR, series of fixed bed reactors, simple two phase separator and single distillation column. Their work accounts for the internal recycle around the reactor and the distribution of the reactants between reactor zones to improve the process performance and maximise the economic potential. Figure 6, illustrates the proposed strategy for reactor system synthesis as a part of the total chemical flowsheet development. It starts with providing process chemistry information. Chemical equations are provided by selecting reactants and products chemical names from a database where chemical prices and other physical properties are imported automatically. If there is any safety and environmental concern about the economically viable paths, warning will be given to the user based on the database information. Conversion, selectivity and recycle of unreacted material are considered.



Figure 6: Reactor system synthesis strategy

General input information about the process such as reaction phase, temperature and pressure, the use of catalyst and its lifetime etc. are provided to start the general reactor selection process. Further details on the reaction exotherm, residence time and kinetics lead to suggested list of single and multiple technical reactors. The scoring system used in the selection process starts from “not suitable” then four levels of ranking of selection criteria from 0 to 3 where 0 is “not recommended”, 1 “acceptable”, 2 “recommended” and 3 “highly recommended”. The highly recommended rank is given to the criteria which are implemented in an existing commercial process using the same reactor type. Multiple reactor systems can be synthesised and decisions are explained16. Proposed reactor systems are simulated using a generic flowsheet in HYSYS. In the reactor system synthesis, non-ideal behaviour of reactors are considered, such as the modular simulation of fluidized bed reactor. The performance of fluidized bed reactor has been studied based on the use of a series of CSTRs and PFR. In this work, the fluidized bed reactor is divided into several segments in series. In each stage, the flow of gas is considered a plug flow through the bubbles and perfectly mixed through the emulsion phase17. Results from reactor and separator systems are linked for total flowsheet synthesis.



3. ETHYLENE GLYCOL REACTOR SYSTEMS SYNTHESIS CASE STUDY Ethylene glycol is an important industrial petrochemical in many countries. It is a feedstock for the production of polyester fibres and resins including polyethylene terephthalate (PET) which is used to produce film and packaging such as bottles. Other formulation of ethylene glycol is used as antifreeze and deicing solutions. Ethylene glycol is also used as a general purpose solvent in paints and plastic industries. There are many different reaction routes to synthesis ethylene glycol such as (1) hydration of ethylene oxide, (2) ethylene oxide via ethylene carbonate, (3) synthesis gas, (4) formaldehyde via glycolaldehyde, (5) directly from ethylene etc..18 Some of these reaction routes are being evaluated and this case study illustrates the reactor system synthesis for the ethylene oxidation to ethylene oxide and the subsequent hydration to ethylene glycol. 3.1. Results and Discussions Analysis by the software of different alternative reaction route to the synthesis of ethylene glycol shows that the route via ethylene oxidation and ethylene oxide hydration is economically viable as long as the other production costs are sufficiently lower than the gross profit. By-product of a reaction is not taken into account but can have a significant impact on the economics of the process. Figure 7, is a screenshot of the user interface showing input process chemistry information for ethylene oxidation primary and secondary reactions.

Figure 7: Process chemistry input screen

Selection results in Table 1, illustrate that two reactor systems are required. The user will not be required to specify the number of reaction systems required. The IKBS will make this decision based on the information provided, such as reaction conditions and phase, and the use of catalyst. 188 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


For the ethylene oxide reactor system, the multi-tubular fixed bed reactor has the highest scores among the alternative reactors. This reactor is currently used in commercial processes because of the special requirements on temperature control throughout the catalyst bed.

Table 1: List of proposed reactors for ethylene oxide and ethylene glycol processes Reactors Type Continuous Stirred Tank Reactor (CSTR) with Jacket CSTR with Jacket and internal coil CSTR with external heat exchanger on circulation loop Sparged CSTR Simple tubular reactor Simple tubular reactor with circulation of heat transfer fluid Simple tubular reactor placed in a furnace Adiabatic fixed bed reactor Fixed bed with intermediate cooling/heating Fixed bed with cold/hot shot Multitubular fixed bed reactor with indirect cooling/heating Trickle-bed reactor Fluidized bed reactor Moving bed reactor Riser reactor Bubble column Spray column reactor Falling thin-film reactor Agitated thin-film reactor Monolith reactor Gauze reactor Reactive distillation

Ethylene oxide reactor system scores Not Suitable Not Suitable Not Suitable Not Suitable Not Suitable Not Suitable Not Suitable Not Suitable 11 11 13 Not Suitable 13 13 13 Not Suitable Not Suitable Not Suitable Not Suitable 12 12 Not Suitable

Ethylene glycol reactor system scores 11 12 13 Not Suitable 12 13 Not Suitable Not Suitable Not Suitable Not Suitable Not Suitable Not Suitable Not Suitable Not Suitable Not Suitable Not Suitable Not Suitable Not Suitable Not Suitable Not Suitable Not Suitable 10

The next highest scores were given to fluidized bed, riser and moving bed reactors. The use of these reactors can improve the heat removal from such highly exothermic reactions. These three reactors may have two drawbacks, possible catalyst attrition and the back mixing of ethylene oxide may result in a long residence time; hence more oxidation of ethylene oxide. Fixed bed reactors with intermediate cooling or cold shot are alternatives currently under consideration in industrial research19. Monolith and gauze reactors are low pressure drop alternatives that can be recommended for this highly exothermic and fast gas catalytic reaction. The software carried out a heat balance, which suggests that the reactors should be operated isothermally. Heat carrier such as methane can be used to increase the rate of heat transfer. This temperature control will reduce the loss of selectivity and catalyst performance. Results for the ethylene glycol reactor system show CSTR and tubular reactors can be used. As the reactions are mixed parallel and series excess of one reactant can be 189 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


used to improve the selectivity and plug flow reactors are preferred to back mixed reactors (CSTR) to minimize the formation of higher glycols. Based on the heat balance carried out by the software, reaction can take place adiabatically. Therefore, adiabatic tubular reactor can be the best choice for such liquid phase reaction. This type of reactor is currently implemented. Another alternative is the use of reactive distillation column where reaction and separation are taking place simultaneously20, 21 . This can be an attractive option as it combines the reaction and separation in a single unit22, which reduces the capital cost and utilises the heat required for the reaction to separate the desired product for unreacted materials.

4. CONCLUSIONS AND FUTURE WORK This paper discusses the development of chemical process synthesis software and the proposed systematic procedure for the synthesis and optimization of a total chemical process flowsheet. The synthesis software integrates the knowledge base with third party software and databases. The prototype software has been successfully applied for the selection of chemical reactors for the manufacture of ethylene glycol via different reaction routes. The synthesis of ethylene oxide and ethylene glycol reactor systems proves that the developed software is able to suggest multiple and novel reactor systems for petrochemical processes which have been successfully validated using existing commercial processes and industrial research. The ongoing work on the synthesis of reactor-separator-recycle systems will lead to linking the developed alternative flowsheets to the simulator, equipment sizing, economic evaluator and flowsheet optimizer. The total chemical flowsheet synthesis software under development is expected to yield significant improvements in the petrochemical industries by providing a cost effective chemical process flowsheeting strategy.

REFERENCES (1) Rudd, D.F. 1968, "The synthesis of system designs: I. Elementary decomposition theory", AICHE Journal, vol. 14, no. 2, pp. 343-349. (2) Grossmann, I.E. 1985, "Mixed-integer programming approach for the synthesis of integrated process flowsheets", Computers & Chemical Engineering, vol. 9, no. 5, pp. 463-482. (3) Douglas, J.M. 1988, Conceptual design of chemical processes, McGraw-Hill, New York.



(4) Douglas, J.M. 1992, "Process Synthesis for Waste Minimization", Industrial & Engineering Chemistry Research, vol. 31, no. 1, pp. 238-243. (5) Grossmann, I.E. 1996, "Mixed-Integer Optimization Techniques for Algorithmic Process Synthesis" in Advances in Chemical Engineering. J.L. Anderson, Academic Press, London, pp. 171-246. (6) Linke, P. 2001, Reaction and Separation Process Integration, PhD thesis. UMIST, UK. (7) Linke, P. and Kokossis, A. 2003, "On the robust application of stochastic optimisation technology for the synthesis of reaction/separation systems", Computers & Chemical Engineering, vol. 27, no. 5, pp. 733-758. (8) Han, C., Stephanopoulos, G. and Liu, Y.A. 1996, "Knowledge-based approaches in process synthesis", International conference on intelligent system in Process Engineering, eds. J.F. Davis, A. Stephanopoulos and V. Venkatasubramanian, vol. 92 of AIChE Symposium Series no. 312. CACHE and AIChE, Snowmass, Colorado, pp. 148. (9) Seider, W.D., Seader, J.D. and Lewin, D.R. 2003, Product and process design principles: synthesis, analysis, and evaluation, 2nd ed, Wiley, New York. (10) Schembecker, G., Droege, T., Westhaus, U. and Simmrock, K.H. 1995a, "READPERT - development, selection and design of chemical reactors", Chemical Engineering and Processing, vol. 34, no. 3, pp. 317. (11) Schembecker, G., Droege, T., Westhaus, U. and Simmrock, K.H. 1995b, "A heuristic-numeric consulting system for the choice of chemical reactors", Fourth International Conference on Foundations of Computer-Aided Process Design. T. Biegler and M. Doherty, AIChE, Snowmass, Colorado, pp. 337. (12) Jacobs, R. 1998, A knowledge based system for reactor selection, PhD thesis. University of Amsterdam, The Netherlands. (13) Jacobs, R. and Jansweijer, W. 2000a, "A knowledge-based system for reactor selection", Computers & Chemical Engineering, vol. 24, no. 8, pp. 1781-1801. (14) Jacobs, R. and Jansweijer, W. 2000b, "A knowledge-based method for the automatic derivation of reactor strategies", Computers & Chemical Engineering, vol. 24, pp. 1803-1813. (15) Montolio-Rodriguez, D. Linke, D. and Linke P. 2007, "Systematic Identification of optimal process designs for the production of acetic acid via ethane oxidation", Chemical Engineering Science, in press. 191 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


(16) Alqahtani, A. Hellgardt, K. Holdich, R. and Cumming, I., 2007. "An Integrated Knowledge Based System for Chemical Process Flowsheet Synthesis". Saudi Innovation Conference, Newcastle, UK. May 2007, [CD ROM]. (17) Alqahtani, A. Hellgardt, K. Holdich, R. and Cumming, I., 2007. "Integrated Knowledge Based System for Chemical Process Synthesis". 17th European Symposium for Computer Aided Process Engineering, Bucharest, Romania. May 2007, pp. 437-442. (18) Kirk-Othmer. 2005. Encyclopaedia of chemical Technology, 5th ed, Vol 12, Wiley, New York. (19) Schwaar, R. 1997, Ethylene oxide and ethylene glycol. SRI report no. 2F. (20) Alqahtani, A., Teo, HTR. and Saha, B. 2005, Esterification of dilute acetic acid with iso-amyl alcohol: heterogeneous kinetics and measurement of residue curve map. The 7th World Congress of Chemical Engineering (WCCE 2005). (21) Teo, HTR., Alqahtani, A. and Saha, B. 2005. Reactive distillation for synthesis of iso-amyl acetate. The 7th World Congress of Chemical Engineering (WCCE 2005). (22) Alqahtani, A. 2004, Recovery of dilute acetic acid through esterification in a reactive distillation column. MSc. dissertation. Loughborough University, UK.


ELECTROCHEMICAL CORROSION BEHAVIOR OF STEEL, COPPER, AND ALUMINUM IN SELECTED IONIC LIQUIDS Mansour AlHazzaa* and Inas AlNashef * Department of Chemical Engineering, College of Engineering King Saud University, P. O. Box 800 Riyadh 11421, Kingdom of Saudi Arabia

ABSTRACT Nowadays, there is a great concern about the protection of environment against pollution. One of the main sources of pollution is the industrial activities, in specific the chemical industries. Therefore the usage of green solvents (safe to the environment) which ionic liquids belong to is very important from, environment, industrial, and economical point of views. In order to build an integrated characteristic for these liquids, the corrosion behavior and the corrosivity of these liquids towards different materials should be investigated. This work is an attempt in that direction. The corrosion behavior of three commonly used metals, steel, copper, and aluminum in three commercially available ionic liquids, Ethaline 197 (E197), Reline 203 (R 203), and Glyceline 200 (G 200), was investigated. The corrosion current densities were determined by extrapolation from Tafel plots and by polarization resistance measurements. It was found that E197 is the most aggressive IL towards steel, copper, and aluminum. The corrosivity sequence for steel is E 197> G200>R203. Steel showed passivity in R 203. The corrosivity sequence for aluminum is E 197> R 203 > G200. For copper there is interference in the corrosivity sequence around the free corrosion potential, but in the potential range -250 mv and up the sequence is E197> R 203 > G 200.

INTRODUCTION Cutting edge technology requires cutting edge products. Novel ionic liquids can provide green chemistry and solvent-free processing. Ionic liquids are salts that are liquid at ambient temperatures. Unlike traditional solvents that can be described as molecular liquids, ionic liquids are composed of ions. This creates the potential to behave quite differently from conventional solvents. Due to the unique chemical physical properties of ionic liquids, they have been called "Green Solvents". Ionic liquids (ILs) are organic salts whose cations, substituents, and anions can be varied virtually at will to change their chemical and physical properties. In principle, the IL solvent can be tailored for a given application. The most important feature of these compounds is that, while they are liquid in their pure state at or near


ELECTROCHEMICALCORROSION BEHAVIOR OF STEEL COPPER, AND ALUMINUM IN SELECTED LONIC LIQUIDS

room temperature, they have essentially no vapor pressure. ILs are non-flammable and have high thermal stability. ILs can be used in several applications including: Solvent and catalyst in reactions, gas separations, liquid separation, solvent for cleaning operation, electrolytes, in Fuel cells, lubrication, and as heat transfer media, to name some. Ionic liquids, which until fairly recently were mainly confined to research laboratories, have penetrated the chemical process industries, as new IL solvents are tailored to suit the needs of a specific chemical reaction across a wide range of applications. The first commercial process to use ILs, launched by BASF AG in 2002, is already considered a huge success, and several other processes are in the pilot stage.1 Some other processes which have not yet been realized include the DIFASOLprocess2 and the desulphurization of hydrocarbon feed.3 Also, a number of companies are offering literally hundreds of ILs, and say they are ready to produce them in ton quantity whenever the demand arises. However, for their commercial implementation, ionic liquids have to meet a number of requirements such as purity, the commercial availability, the thermal stability, toxicological issues and the corrosivity to materials typically used for vessels, piping and other technical equipment.4 Material corrosion can be a risk for plant integrity, reduce plant efficiency, cause plant shutdowns, yield loss or contamination of produced products, waste valuable resources, result in over-design and may require costly maintenance.4 Uerdingen et al.4 investigated the corrosion behavior of carbon steel, austenitic stainless steel, nickel based alloy C22 copper, brass, and aluminum in seven ILs with different chemical structure under flow conditions at temperature up to 90°C. They found that stainless steel proved resistant in all water free and water diluted systems tested. They also found that for carbon steel and aluminum the corrosivity of IL media strongly depends on the chemical structure of the cationic moiety and nature of the anion in IL molecule. They also showed that corrosion inhibition in ILs is possible. Reddy et al.5 investigated the thermal stability and corrosivity of various ILs They investigated the corrosivity of some ILs against 316 stainless steel and 1018 carbon steel by electrochemical techniques at room temperature. They obtained corrosion rates less than 13 m/year, thus indicating the outstanding resistance of the alloys to uniform corrosion in ILs. The observed localized corrosion using 1-butyl3-methylimidazolium chloride, they attributed this to the presence of deleterious Clions. Perissi et al.6 studied the corrosion behavior of several metals and metal alloys (copper, nickel, AISI 1018 steel, brass, Inconel 600) exposed to a typical ionic liquid, the 1-butyl-3-methyl-imidazolium bis-(tri-fluoromethanesulfonyl) imide, ([C4mim][Tf2N]) by electrochemical and weight-loss methods. They determined corrosion current densities by extrapolation from Tafel plots and by polarization resistance measurements. They also performed 48 h immersion tests at 150, 250, 275 and 325°C. They reported that room temperature results showed low corrosion 194 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007

MANSOUR ALHAZZAA AND INAS ALNASHEF

current densities (0.1–1.2 µA/cm2) for all the metals and alloys investigated. At 70°C, the corrosion current for copper dramatically increases showing a strongly dependence on temperature. At 150°C copper shows significant weight-loss while nickel, AISI 1018, brass and Inconel do not. At higher temperatures (>275°C), the copper sample crumbles and localized corrosion occurs for the other metals and alloys.

EXPERIMENTAL WORK The ILs used in this study, along with some physical properties, are listed in Table (1) below. The ILs were obtained from Sionix (UK).

Table 1 List of ionic liquids used in this study.

No. 1

Name Ethaline 197 (E197)

Mp (°C ) 9-11

Purity > 98%

Viscosity (cP) 1.2 at 22°C

2 3

Reline 203 (R203) Glyceline 200(G200)

21 < -35

> 98% > 98%

1072 at 21°C 154 at 25 °C

Ethaline 197 is consists of Ethanediol C2H6O2 and choline chloride C5H14NOCL. . Reline 203 is consists of Urea, CH4N2O and C5H14NOCL choline chloride. Glyceline 200 is consists of Glycerol, C3H8O3 and choline Chloride C5H14NOCL. All ILs are supplied by Sionix Company. All chemicals used were of >98% purity and were used without further purification. Three commonly used metals with a purity of >98% were used as test. These are: Steel, copper, and aluminum. For each metal to be tested several coupons were prepared. The surface area of the coupon was one square centimeter. One face of the coupon was ground using 1000 emery paper, polished to a mirror like surface, and finally washed with acetone. The electrical connection was accomplished by welding a copper wire on the second face of coupon. All of the second face including the welding point and the copper wire was coated by an epoxy resin. This assembly (working electrode), see Figure 1, was kept dry up to the testing time. The working electrode was the metal under investigation; the counter electrode a platinum one square centimeter sheet, while saturated calomel, SC, was used as reference electrode. A potentiostat / galvanostat (ACM) connected to a 195 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


computerized data acquisition were used for both linear polarization resistance and Tafel plot measurements.

RESULTS AND DISCUSSION Fig.2 shows the potential VS time for steel, copper, and aluminum in Glyceline 200 for one day. Similar behavior for both steel and copper where there was a drop in potential in the first hour, the potential became steady for the rest of the day. The steady potential of copper was more negative (-510 mv) compared for that for steel (-400 mv). For aluminum the potential increases to more positive from (-750 mv) to (-640 mv) within the first hour of measurement, then it fluctuated around (-640 mv) for the rest of the day. The fluctuation of Al potential( break down and repair of the of the surface film) may be due to that the condition for free immersion at room temperature is slightly alkaline due to the presence of hydroxyl group in all tested ILs. As shown in Fig.3 the linear region for steel and copper is difficult to be predicted, also the measurements gave negative current during polarization which means that the cathodic process in the vicinity of free corrosion potential is uncertain . For aluminum the measurement gave positive currents (corrosion) for all four trials at different potentials as seen in Fig.4.The alkaline behavior of Glyceline is the main reason for AL corrosion. In conclusion the corrosion events around the free corrosion potential for all metals are not well defined. This may be due to the complex electrochemical behavior of the IL itself. Based on the LPR findings the comparison between the electrochemical behavior of steel, copper, and aluminum will be focused on the anodic portion of the Tafel plot. As shown in Fig5. for aluminum there was an active dissolution (corrosion) between (-750 up to -500) mv at which the current increased 100 times (0.001-0.1) m A/cm2, but between (-500 to+250) mv Al showed a sort of resistance behavior (may be due to the formation of a protective film) at which the dissolution current is only increased five times (0.1-0.5) mA/cm2. For copper and steel the dissolution is very sharp in a shorter range of potential (-400 to -250) mv, at which the dissolution increases 1000 times (0.01-100) mA/cm2. Copper show passivity within a potential range between -250 mv and zero. The corrosion products formed on the steel surface slow down the dissolution rate as noticed in the range of (-350 mv up to zero). Finally in Glyceline 200, the alkalinity of C3H8O3(see the chemical structure below) is the reason for the active dissolution of AL on the beginning of exposure afterwards a protective film formed which slow down the dissolution rate. For steel the alkalinity is not enough to form any sort of a protective film of (Fe (OH) 2).For copper the alkalinity is sufficient to slow down the corrosion process. The dissolution rate of AL is much slower than that for both steel and copper. Fig. 6 196 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


shows the potential vs. time for all three metals in Ethaline 197 see the chemical structure below.

C3H8O3 chemical structure

Ethanediol chemical structure Different behavior in the first hour of recording was noticed for steel and copper in Ethaline197. Steel potential was decreased from -300mv to -500mv, while that for copper was increased from -600mv to -550 mv. The tow potentials of both coincide after 12 hours, and remain steady for the next 12hours. For Al it was steady for the first 12 hours (-650) mv, then drop to -750mv for tow hours (may be due to a break down in the surface film) afterwards the free corrosion potential increases for the rest of the day. As seen in Fig. 7 the events and reactions for all metals by scanning the potential around the free corrosion potential (LPR testing) are uncertain, amphoteric, and/ or complicated, therefore determining the linear polarization resistance is impossible .The complexity of Ethaline197 molecule may be the reason of the uncertain prediction of Rp. Fig 8 shows the anodic dissolution (corrosion) behavior of Al, Cu, and steel in Ethaline 197. There are three stages of the anodic dissolution of Al, first stage between (-700 and -600) mv at which the dissolution rate increased 20 times, while the second stage (-600 to -300)mv the rate increased 10 times, in the third stage (-300 up zero) mv the rate increased by 200 times. The anodic dissolution of steel started with a sharp rate between (-560 and -250) mv, then a narrow passive region between (-250 and -100) mv at which the dissolution rate is 250mA/cm2. A steady increase in the dissolution rate was noticed for positive potentials more than (-100) mv. The anodic dissolution of Cu started at -430 mv with a sharp increase up to -250 mv, then a passive region formed between (-250 and zero) mv at a dissolution rate of 270 mA/cm2. Similar behavior of Cu was noticed in both E197 and G200. The passivity region of Cu in both ILs are identical 197 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


between (-250 and zero) mv. The alkaline nature, due to presence of OH- group in both molecules of G200 & E197, may be the reason of the similarity of copper behavior in these ILs. From Fig 9 the corrosivity of R 203 towards the investigated metals is less than the other ILs. The presence of urea group will slow down the corrosivity of CL- towards the metals. The anodic dissolution of steel started at -400mv, and then it shows a sort of passivity at -150 mv with a dissolution rate of 0.2 mA/cm2. For aluminum the anodic dissolution started at -700 mv, and then it shows passivity at -400 mv with a rate of dissolution of 0.4 mA/cm2. 1000 times of anodic dissolution rate (0.1 to 100) mA/cm2 was noticed for copper within a potential range of (-720 to -200) mv. The dissolution rate of copper became slower with potentials more positive than -200 mv with out any sign of passivation in these ranges of potentials. It is obvious that the absence of ( OH-) group from R 203 molecule is the reason for the corrosivity towards copper. Due to the complexity of the molecules of the studied ILs the cathodic processes which may take place on the metals surfaces are very difficult to be predicted as shown by the fluctuation of potential vs. current curves. In conclusion E197 is the most aggressive IL towards steel, copper, and aluminum. The corrosivity of studied ILs are summarized in Figs 10, 11, and 12. The corrosivity sequence for steel is E197> G200>R203. Steel showed a tendency of passivation in R 203 (may be due to the presence of urea group in R 203 as shown in the chemical structure below.

Urea chemical structure The corrosivity sequence for aluminum is E197> R 203 > G200 see fig 12. For copper the behavior is quite similar in G 200 & E197. Both Ecorr (-400mv) and Icorr (1mA/cm2) can be determined as shown in fig (11).There is an interference in the corrosivity sequence around the free corrosion potential of copper, but in the potential range -250 mv and more positive the sequence is E197> R 203 > G 200.More hydroxyl groups present in the molecule, it will make it less aggressive towards copper. From the corrosivity sequence for all metals, it can be seen that the presence of Ethanedoil, C2H6O2 group as a constituent of the ILs is more harmful towards the tested metals compared to other groups.

Conclusions The corrosion behavior of three commonly used metals, steel, copper, and aluminum in three commercially available ionic liquids, Ethaline197, Reline 203, and Glyceline 200 was investigated .The anodic dissolution of metals according to reaction (1) will be affected by the presence of both OH- ,CL-.Those ions are present 198 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


in E197 and G 200.For AL both ions are harmful. For steel and copper CL- is aggressive while OH- is non corrosive. So there is a mutual competition between these tow ions against steel and copper surfaces. M → Mn+ + ne-

(1)

It was found that E 197 is the most aggressive IL towards steel, copper, and aluminum. The corrosivity sequence for steel is E 197> G200>R203. Steel showed passivity in R 203. The corrosivity sequence for aluminum is E197> R 203 > G200. For copper there was interference in the corrosivity sequence around the free corrosion potential, but in the potential range -250 mv and up the sequence was E197> R 203 > G 200. The fluctuation of potential current curves may be due to the complexity of the cathodic processes which may take place on the metals surfaces.

REFERENCES 1. M. Fremeantle, Chem. Eng. News, 81, 9 (2003). 2. F. Favre, A. Ferestiere, F. Hugues, H. Olivier-Bourbigou, and J. A. Chordoge, Pet. Tech., 44, 104 (2002). 3. J. Esser, P. Wasserscheid, and A. Jess, Green Chem. 6, 316 (2004). 4. M. Uerdingen, C. treber, M. Balser, G. Shmitt, and C. Werner, Green Chem. 7, 321 (2005). 5. R. G. Reddy, Z. Zhang, M. F. Arenas, and D. M. Blake, Appl. Them. Eng. 22, 357 (2002). 6. I. Perissi, U. Bardi, S. Caporali, A. Lavacchi, Corrosion Science 48, 2349 (2006). 7. http://www.khdesign.co.uk/TechIntro.htm. ACKNOWLEDGMENTS The authors would like to acknowledge KACST for the finical support of this study

. 199 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


Fig (1) cell assembly

Fig 2 potential vs. time plots for AL, Cu, and

Steel in G 200

Fig 3 LPR plots of Cu &Fe in G200 G200


Fig 4 LPR plots of AL in


Fig 5 Tafel plots of steel, Cu, and AL in G 200

Fig 6 potential versus time for steel, Cu, and AL in E197



Fig7 LPR plots of Cu, steel, and AL in E 197



Fig 9 Tafel plots of steel, AL in E 197

Fig 10 Tafel plots of steel in ILs

Fig 8 Tafel plots of steel, Cu, and Cu, and AL R 203

Fig 11 Tafel plots of copper in the studied studied ILs



Fig 12 Tafel plots for aluminum in the three studied ILs.


Direct Hydrogenation of α,ß-Unsaturated Aldehyde in A parallel Approach Using Ruthenium(II)-Functionalized Catalysts in Simple Low Pressure GlassDesigned Reactor Ismail K. Warad Department of Chemistry, King Saud University P.O Box 2455, Riyadh-11451, Saudi Arabia [email protected] ABSTRACT Several diamine-bis(phosphine)ruthenium(II) complexes using four types of diamines co-ligands (L1- L4) were made available (Scheme1). These complexes were characterized by NMR, IR, and mass spectroscopy in addition to the elemental analyses. The chemical behavior of the phosphine and diamine co-ligands during the complexation reactions was manipulated by 31P{1H} NMR spectroscopy at room temperature which confirms the right complexes formation. The main target of this work is to compare the hydrogenation catalytic activity of these complexes toward selective hydrogenation of α,ß-unsaturated aldehyde substrate (cyanamid aldehyde) under identical hydrogenation conditions. The roles of the solvent and the cocatalysts have been investigated in this study. These complexes revealed a high degree of catalytic activity and selectivity under mild hydrogenation condition. The hydrogenation processes were carried out in a new simple low-pressure reactor, this glass-designed reactor has facility us handling of the hydrogenation processes which empowered use to do farther more trails, in order to screen several reaction parameters.

KEYWORDS: Ruthenium(II) complexes, phosphine, diamine, co-catalyst and hydrogenation. Introduction Asymmetric hydrogenation is a core technology in fine chemicals synthesis particularly for pharmaceuticals, agrochemicals, flavors, and fragrances, which requires a high degree of stereochemical precision [1]. Asymmetric hydrogenations of C=C, C=O, and C=N functionalities have found important applications in organic synthesis which depend directly on catalysts structures [2-13]. Considerable effort has been made to establish an empirical relationship between catalysis potential and structural behavior following the 31P{1H} NMR chemical shift of the phosphine ligand during reaction processes [14-16]. Small changes in structure may lead to dramatic changes in selectivity and activity of the catalysts [14, 17]. Hydrogenative reduction of prochiral ketones to chiral alcohols is a powerful tool for precise stereocontrolled organic synthesis [2-13, 18-25]. A high turnover frequency (TOF) can be obtained by designing suitable molecular catalysts and reaction conditions Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007

Direct Hydrogenation of α,ß-Unsaturated Aldehyde in A parallel Approach Using Ruthenium(II)-

[18-23]. Preferential reduction of a C=O function over a coexisting C=C linkage is an important and difficult task. Its versatility is manifested by the asymmetric synthesis of some biologically significant chiral compounds [1-4]. Although there are many examples of highly efficient catalysts for olefin and ketone reduction, imine hydrogenation is still a challenge in terms of both the turnover frequency and the lifespan of the active catalyst [2]. One of the best transition-metal complexes for ketone hydrogenation that has been discovered is the chiral Ru(II)–diphosphine-1,2diamine complex, which was developed by Noyori and Ohkuma [1, 3, 21]. Moreover, many efforts have been made to achieve this target, after reporting by R. Noyori (Nobel prizzer 2001) that such systems in the presence of strong bases, as co-catalysts, and 2-propanol as solvent, proved to be an excellent catalyst for the hydrogenation of prochiral ketones under mild conditions. We have already reported an open procedure to synthesize neutral diamine(diphosphine)ruthenium(II) complexes by what so called later ligands exchanged technique [16, 26, 27]. In addition, several neutral and cationic diamine–bis(ether– phosphine)-ruthenium(II) complexes using hybrid ligand were made available. Compounds of this type can be easily supported, by the introduction of T-silyl functions into the ether−phosphine ligands, several polysiloxane sol-gel matrixes were prepared and tested as xerogels catalysts [6, 8-13]. Our current research portfolio is focused at the extension of methodology to related system, namely the reduction of the C=O function group in the presence of the C=C groups. In connection with our ongoing interest in this field of research, I represent in this paper the comparable catalytic reaction towered hydrogenation of Cyanamid aldehyde substrate using several ruthenium(II) complexes with three different types of mono-phosphine ligands as well as four different types of diamine co-ligands.

2. Experimental 2.1 General remarks, materials, and instrumentations All reactions were carried out in an inert atmosphere (argon) by using standard high vacuum and Schlenk-line techniques unless otherwise noted. Prior to use, CH2Cl2, n-hexane, and Et2O were distilled from CaH2, LiAlH4, and from sodium / benzophenone, respectively. The diamines were purchased from Acros, and Merck and were purified. Triphenylphosphine and 2-propanol from Fluka, were used without further purification. [RuCl2.xH2O] is available from Chempur. [Cl2Ru(PPh3)3] and etherphosphine ligands were prepared by the published method [16, 28]. Elemental analysis was carried out on an Elementar Varrio EL analyzer. High-resolution 1H, 13 C{1H}, DEPT 135, and 31P{1H} NMR spectra were recorded on a Bruker DRX 250 spectrometer at 298 K. Frequencies are as follows: 1H NMR 250.12 MHz, 13 C{1H} NMR 62.9 MHz, and 31P{1H} NMR 101.25 MHz. Chemical shifts in the 1 H and 13C{1H} NMR spectra were measured relative to partially deuterated solvent 206 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007

Ismail K. Warad

peaks which are reported relative to TMS. 31P chemical shifts were measured relative to 85% H3PO4 (δp = 0). Mass spectra: FAB-MS; Finnigan 711A (8kV), modified by AMD and reported as mass/charge (m/z). Infrared spectra of the solid complexes were recorded on a Bruker IFS 48 FT-IR spectrometer using KBr disk. The analyses of the hydrogenation experiments were performed on a GC 6000 Vega Gas 2 (Carlo Erba Instrument) with a FID and capillary column PS 255 [10 m, carrier gas, He (40 kPa), integrator 3390 A (Hewlett Packard)]. 2.2. General Procedure for the Catalytic Study The respective diamine-bis(lphosphine)ruthenium(II) complexes 4-6 [0.01 mmol, Ru(II)] was placed in a 100 ml Schlenk tube and solid KOH, tBuOK and Na2CO3 [0.10 mmol each] were added individually as co-catalysts. The entire apparatus was evacuated and filled back with argon three times to establish an inert atmosphere. The solid mixture was stirred and warmed during the evacuation process to remove traces of oxygen and water. Subsequently the Schlenk tube was filled with argon and 10 ml of 2-propanol. The mixture was vigorously stirred, degassed by two freeze-thaw cycles, and then sonicated for 30 min (to ensure complete dissolving of components and remove trace of remained gases). A solution of Cyanamid aldehyde compound (10.0 mmol) in 30 ml of 2-propanol was subjected to a freeze-thaw cycle in a different 100 ml Schlenk tube and was added to the catalyst solution. Finally the reaction mixture was transferred to 200 ml pressure Schlenk tube the entire apparatus was evacuated and filled back with dihydrogen gas three times to wash out the argon inert sphere then pressurized with 1 bar of dihydrogen. The reaction mixture was vigorously stirred at room temperature for about 1 h. During the hydrogenation process samples were taken from the reaction by a special glass syringe then directly inserted into a gas chromatography to control the conversion and selectivity. The kind of the reaction products was compared with authentic samples.

3. Results and Discussion 3.1. Catalysts Preparation Dichloro-bis(triphenylphosphine)ruthenium(II) complex was prepared from RuCl3. xH2O according to literature methods [9, 26, 27, 30]. The desired catalysts were prepared as in Scheme 1.



Cl Ph3P

NH2

PPh3

Ph2P

PPh3

O

Ru Cl

NH2

O

1 2 Ph2PCH2CH2OCH3

- 1 PPh3

- 3 PPh3

- 3 PPh3

Cl

Ph3P

Ph2 Cl P

NH2

Cl

O

O

NH2

O

O

O

Cl

Cl 3

4

Ph2 P

Ru

Ru

Ru Ph3P

Ph2 Cl P

Ph2 P

O

2

O Ph2 Cl NH2 P Ru P NH2 Ph2 Cl

O

O

Ph2 Cl P

O

NH2

Ru

O

P Ph2

NH2

Cl

O

5

6

Selected diamines H2N

H2N

L1

H2N

H2N

H2N

H2N

H2N L2

H2N L3

L4

Sceme1: Synthesis of the ruthenium(II) complexes. To establish the unrestrictive structural behavior of the incorporation active site during the reaction, 31P {1H} NMR as a power tool was investigated, together the multiplicity, the chemical shifts and the coupling constant of these complexes 208 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007

Ismail K. Warad

confirmed the expected trans-configured dichloro-RuCl2 B isomer formations [2630]. The stoichiometry of five and six-coordinated of these complexes, is supported by elemental analyses FAB-MS, IR, 31P{1H} 1H, 13C{1H}NMR spectroscopy (see experimental section for details), which are also consistent with a chelate coordination of diamine ligands at room temperature. In the 31P{1H} NMR spectrum of such complexes at room temperature is very simple, it exhibited a singlet with (δp = 40-46 ppm) revealing that the chemical equivalence of phosphine groups in solution due to the C2v symmetry of the RuCl2(P)2diamine complex are equivalent. 3.2 The Catalytic Activity 3.2.1 Selective Hydrogenation Ruthenium(II) Complexes

of

C=O

of

Cyanamid

Aldehyde

via

To study the catalytic activity of the ruthenium(II) complexes, trans-4phenyl-3-propene-2-al was selected, because three different rego-selective hydrogenation are expected (Scheme 2). O

S

H

A

OH

O

OH H Ruthenium(II) catalysts

H

+

B

H

+

C

Scheme 2. Different hydrogenation possibilities of substrate ِS: Selective carbonyl function group hydrogenation to produce A, selective C=C function group hydrogenation to produce B, full hydrogenation path with no selectivity to produce C. The selective hydrogenation of the carbonyl group affords the corresponding unsaturated alcohol (A). Unwanted and hence of minor interest both the hydrogenation of the C=C double bond, leading to the saturated aldehyde (B) and the full hydrogenation of C=O and C=C bonds resulting the saturated alcohol (C) formation.



3.2.2 Hydrogenation Apparatus: Low-Pressure Schlenk-Tube Reactor The next Figure represents the designed Hydrogenation Reactor, which was constructed of high pressure Pyrex glass.

Figure 1. Schematic representation of a low-pressure Schlenk-tube for hydrogenations. During the hydrogenation process samples were frequently taken and directly inserted into a gas chromatography, in order to confirm the degree of conversion and control selectivity. The hydrogenation reactions using complexes 4L1-4L4, 5L1-5L4 and 6L1-6L4 as catalysts were carried out at 25 oC with a molar substrate: catalyst (TON, S/C) ratio of 500: 1, under 1 bar of hydrogen pressure, in 40 ml of 2-propanol [Ru: Co-catalysts (KOH, tBuOK and Na2CO3): cyanamid aldehyde] [1:10:500]. The reactions under the above condition using these complexes were finished within one hour (in most cases) with no side products (B and C), the conversions as TOF of the hydrogenation processes as a function of ligand types was illustrated in Figure 2.


Ismail K. Warad

Figure 2: Hydrogenation catalytic activity of the desired ruthenium(II) complexes. The catalysts were only effective in the presence of excess hydrogen and a strong basic co-catalyst like KOH, NaOH, NaBH4 and tKOBu, while the weak bases like K2CO3 co-catalyst was totally inactive. 2-propanol served as a solvent. All these complexes (except4L4 - 6L4) are highly active under mild conditions and gave rise to a 100% selective hydrogenation toward the C=O group in the presence of a C=C function. Complexes 4L4 - 6L4 were fewer actives under the identical compared condition compared by the other, which resonated to the presence of the phenyldiamine co-ligands, which found to be inactive under such mild hydrogenation condition [3-5].



3.2.3 Role of Solvent in the Hydrogenation Process The hydrogenation reaction was studied using eight different solvents and complex 5L2 at identical represented above condition, 2-propanol was found to be the best choice, contaminate it with traces of water was poisoned the catalyst, the result are given in Figure 3.

Figure 3: Solvents effects on the hydrogenation process.


Ismail K. Warad

3.2.4 Role of the Co-catalyst Using eight types of well known co-catalysts the hydrogenation reaction was performed in the presence of 5L2 under above condition, Strong basic condition was the corner stone to activate such complexes, the result are given in Figure 4.

120

Converation after 2h

100 80 60 40 20 0 KOH

NaOH

t KOBu

NaBH4

BF3

K2CO3

(Pr)2 NH

NH4 Cl

Figure 4: Co-catalysts effects on the hydrogenation process

4. Conclusion The following conclusions can be drawn from the present study. 1- All the aliphatic diamines complexes of diamine bis-(phosphine)ruthenium(II) showed a high degree of hydrogenation catalytic activity under mild condition, while the phenylic diamine complexes are totally inactive under the same identical conditions. 2- Solvent such as 2-propanol played an important role in activating these complexes, exchanging 2-propanol to ethanol lead to strong decrease in the catalytic activity while carrying out hydrogenation with other solvents killed the process totally. 213 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


3- The hydrogenation process never take place without co-catalyst, strong basic cocatalysts are media such as KOH, NaOH, NaBH4 and tKOBu required essentially to active the Cl-Ru complexes to H-Ru hydride catalysts which are the most accepted method for such hydrogenation. 4- The presence of at least one N-H function group in the backbone of the used diamine co-ligand are important to activate with the hydride function H-Ru in the complex the whole process at mild hydrogenation condition.

References [1] Noyori, R.,1994, "Asymmetric Catalysis in Organic Synthesis", 1st , J. Wiley and Sons, New York, (1994) (Chp 2). [2] H. Bauer, C. Malan, B. Pugin, F. Spindler, H. Steiner and M. Studer, Adv. Synth. Catal. 345 103–151 (2003). [3] R. Noyori, and T. Ohkuma, Angew. Chem., Int. Ed., 40, 40-120 (2001) and references cite therein. [4] M. Kitamura, M. Tokunaga, T. Ohkuma, and R. Noyori, Tetrahedral Lett, 32, 4163-4168 (1991). [5] T. Ohkuma, M. Koizumi, K. Muniz, G. Hilt, C. Kabuta and R. Noyori, J. Am. Chem. Soc., 124, 6508-6509 (2002). [6] I. Warad, S. Al-Rsayes and K. Eichele, Z. Kristallogr. NCS, 221, 1779-1782 (2006). [7] N. Shan, H. Adams and J. A. Thomas, Inorg. Chim. Acta, 358, 3377- 3377 (2005). [8] E. Lindner, Z-L. Lu, H. A. Mayer, B. Speiser, C. Tittel and I. Warad, Electrochemistry Communication, 9, 1013-1020 (2005). [9] I. Warad, E. Lindner, K. Eichele and H. A. Mayer, Inorg. Chim. Acta, 357, 18471853 (2004). [10] E. Lindner, I. Warad, K. Eichele and H. A. Mayer, Inorg. Chim. Acta, 350, 4956 (2003). [11] E. Lindner, A. Ghanem, I. Warad, K. Eichele, H. A. Mayer and V. Schurig, Tetraheron :Asymmetry, 14,1045-1053 (2003). [12] Z. Lu, K. Eichele, I. Warad, H. A. Mayer, E. Lindner, Z. Jiang and V. Schurig, Z. Anorg. Allg. Chem., 629, 1308-1315 (2003). [13] E. Lindner, S. Al-Gharabli, I. Warad, H. A. Mayer S. Steinbrecher, E. Plies, M. Seiler, and H. Bertagnolli, Z. Anorg. Allg. Chem., 629, 161-171 (2003). [14] I. Warad, S. Al-Gharabli, A. Al-labadi, and A. Abu-rayyan, J. Saudi. Chem. Soc. 9, 507-518 (2005). [15] C. D. Gilheany, and M. C. Mitchell, In the Chemistry of Organophosphorus Compounds, Hartley, F. R., Ed.; J. Wiley and Sons: New York, (1990). 214 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007

Ismail K. Warad

[16] E. Lindner, I. Warad, K. Eichele and H. A. Mayer, J. Organomet. Chem. 665, 176-176 (2003). [17] A. C. Tolman, Chem. Rev., 77, 313-348 (1977). [18] R. Noyori, and T. Ohkuma, Angew. Chem., Int. Ed., 40, 40-120 (2001) and references cite therein. [19] M. Kitamura, M. Tokunaga, T. Ohkuma, and R. Noyori, Tetrahedral Lett, 32, 4163-4168 (1991). [20] T. Ohkuma, M. Koizumi, K. Muniz, G. Hilt, C. Kabuta and R. Noyori, J. Am. Chem. Soc., 124, 6508-6509 (2002). [21] J.-X. Gao, T. Ikariya and R. Noyori, Organometallics, 15, 1087-1089 (1996). [22] K. Abdur-Rashid, M. Faatz, J. A. Lough and R. H. Morris, J. Am. Chem. Soc., 123, 7473-7474 (2001). [23] Y. Jiang, Q. Jiang and X. Zhang, J. Am. Chem. Soc., 120, 3817-3818 (1998). [24] P. Gamez, F. Fache and M. Lemaire, Tetrahedron Asymmetry, 6, 705-718 (1995). [25] G. A. Grasa, A. Zanotti-Gerosa, J. A. Medlock and W. P. Hems, Org. Lett., 7, 1449-1451 (2005). [26] I. Warad, J. Saudi. Chem. Soc. 11, 15-24 (2007). [27] I. Warad, J. King. Saud. Uni. Science 2. (2007) in press. [28] T. A. Stephenson and G. Wilkinson, J. Inorg. Nucl. Chem., 28, 945-956 (1966). [29] A. Batista, M. Santiago, C. Donnici, I. Moreira, P. Healy, S. Berners-Price and S. Queiroz, Polyhedron, 20, 2123-2128 (2001). [30] I. Warad, G. Al-Sousi, M. Al-Nuri, S. Al-Gobari, S. Al-Reasyes, J. Saudi. Chem. Soc. (2007) in press.


Topic 7 Research and development to serve the industry and upgrade its services • Electrical Engineering

THREE-PHASE SHUNT ACTIVE POWER LINE CONDITIONER

A. A. Mansour1, A. M. Zaki1, O. A. Mahgoub2, and E. E. Abu-Elzahab2 1: Electronics Research Institute, Power Electronics Dept, El-Tahrir St., Dokki, Cairo, Egypt, Post Code: 12622, Email: [email protected], [email protected] 2: Faculty of Engineering, Electrical Power & Machines, Cairo University, Giza, [email protected], [email protected]

ABSTRACT The shunt active power line conditioner represents one of the advanced techniques used to overcome the harmonic currents drawn by the non-linear loads. This paper introduces the harmonics extraction applying the average control technique based on the d-q reference frame. The technique is studied through both simulation and experimental implementation.

KEY WORDS Active Power Filters (APF's), Active Power Line Conditioner, Active Filters, Power Line Conditioner.

I. INTRODUCTION Due to the enhancement in the power electronics technology, modern control systems have been established. These loads such as variable speed drives, converter circuits, personal computers, Static VAR Compensators (SVC), and many other loads based on power semiconductor circuits. The previous loads play an important role in increasing harmonics in the electric power networks. So, the power quality suffers and degrades due to the presence of these harmonics which represent an incessant distortion in both voltages and currents in the electric network. Increasing the harmonic levels means more troubles in the electric grid. The problems arising from harmonics can be epitomized in increasing the RMS current and excessive neutral currents. These drawbacks lead to incessant heating of the electrical equipment such as transformers and supply apparatus, resonance causes over stress of power factor capacitors, trip of circuit breakers (i.e. blinding their operation), malfunction of phase locked loop circuits, and communication interference. So, the term power quality and its related harmonic problems have Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


gained much attention. Solving the harmonics problems can be achieved using different techniques according to the harmonics compensation requirements. Different techniques can be used to compensate for the harmonics: Passive filters and Active Power Filters (APF's). APF's are now mature and can be classified as shunt, series, and universal APF's [1, 2]. The extraction of the harmonic current reference is the first step in the compensation process. This can be achieved by different techniques, in the frequency domain such as Fast Fourier Transform (FFT) and Wavelet Transform or on the time domain such as p-q theory and d-q reference frame [3-10]. This paper studies the shunt three-phase APF as shown in Fig. 1. The control strategy is based on the d-q reference frame using the Average Control Technique, ACT to extract the harmonic current reference, then applying the hystersis current controller for closed loop implementation. There is no need for a previous knowledge of the harmonic equations of the load current, since this technique splits only the fundamental component from the load current and the remainder represents the total harmonics content of the load.

I Supply

I Load Non-Linear Load

AC

I Compensator

APF

Fig.1 Three-Phase Shunt APF


A. A. Mansour , et al

II. THE AVERAGE CONTROL TECHNIQUE The transformation from three-phase three-wire system to the d-q reference frame is helpful in decoupling the variables used in the control action into two components: the direct and quadrature components as in (1).

⎡v d ⎤ ⎡v a ⎤ ⎢v ⎥ =C1⎢v ⎥ ⎢ q⎥ ⎢ b⎥ ⎢⎣ 0 ⎥⎦ ⎢⎣ vc ⎥⎦

(1)

Where, ⎡ ⎢ Cosθ 2 ⎢ C1= ⎢−Sinθ 3 ⎢ ⎢ 1 ⎢ 2 ⎣

2π ⎤ 2π ) Cos(θ + ) ⎥ 3 3 2π ⎥ 2π − Sin(θ − ) − Sin(θ + ) ⎥ 3 ⎥ 3 1 1 ⎥ ⎥ 2 2 ⎦ Cos(θ −

θ is the transformation angle. It is known that for any transformation to the d-q reference frame rotating with the synchronous rotating speed, the vector components appear as dc components in the d-q rotating frame. So a question arises: what about the other components that rotates with speeds of integer multiples of the fundamental? "i.e. the harmonic components". Fig. 2 shows the representation in the stationary frame for a voltage vector consisting of two components as follows: v(t ) = sin(ωs t ) + 0.2 sin(5ωs t ) . Since the 5th harmonic component is a negative sequence component, the relative speed ω relative between the two vectors (i.e. the fundamental and the 5th) will be th ωrelative = ωs − (−5ω s ) = 6ωs . So, the 5 harmonic rotates with 6ω s w.r.t the fundamental in the d-q reference frame. It is shown that the overall vector deviates from the locus of the circle that represents the fundamental component [3].

Applying the transformation to the d-q reference frame, the 5th harmonic appears as both cosine and sine waves in the d and q components as shown in Fig. 3, and having six complete cycles along one cycle of the fundamental. So, any order of harmonics appears as sinusoidal wave in the d-q reference frame. The average of complete cycles of sine wave is equal to zero. Hence, tacking the average in the d-q 221 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


reference frame splits the fundamental component in the d-q frame from the harmonic content of the original wave. Splitting the harmonics from the distorted wave in the d-q reference frame is achieved using the average algorithm called the ACT [9].

Fig.2 Three-phase fundamental signal plus 0.2 P.U. 5th harmonic in stationary frame.

Fig.3 Three-phase fundamental signal plus 0.2 p.u. 5th harmonic in rotating reference frame.

III. SIMULATION RESULTS The simulation program has been developed using MATLAB software package. The closed loop simulation program of the APF has been developed for only R-C load fed from three-phase rectifier bridge as individual load and a general non-linear load as shown in Fig. 4, which consists of three parallel loads, three-phase linear inductive loads, three-phase rectifier bridge feeding R-L load, and three-phase rectifier bridge feeding R-C load.



Fig.4 Power circuit for the general load For three-phase rectifier bridge feeding R-C load, C=200µF, R=150Ω, and the supply voltage is equal to 380volt. Fig. 5 shows the line load current drawn by the load ia-Load. The THD of this load current is equal to 104% based on the spectrum of the load current shown in Fig. 6. Applying the ACT to the previous wave in the d-q reference frame gives the harmonic content ih-Load as shown in Fig. 7. Using the extracted harmonic waveform, the APF can be controlled to inject equal and anti-phase waveform using the hystresis current controller. The supply current appears as a sinusoidal wave, as shown in Fig. 8, except for some spikes. These spikes appear because the supply inductance feeding the load is neglected as well as the APF is connected to AC mains through an inductor, so the compensation will be limited at the sharp transitions. The THD of the supply current is reduced to 13.2%.



Fig. 5 Line current drawn by the load.

Fig. 7 The harmonic current contents of the load.

Fig. 6 Spectrum analysis of the load current.

Fig.8 Supply current after compensation.

For general non-linear load, the drawn line current is shown in Fig. 9, with THD=46% based on the spectrum shown in Fig. 10. This load consists of two individual R-C and R-L are feeding from three-phase rectifier bridge and another linear R-L load shown in Fig. 4. Fig. 11 shows the extracted harmonic of this load current. After harmonic compensation using the APF, the supply current appears sinusoidal wave form as shown in Fig. 12. The THD of the supply current is reduced to be 4%.



Fig.9 Line load current.

Fig. 10 Spectrum analysis of the load current.

Fig.11 The harmonic current contents of the load

Fig.12 Supply current after compensation



IV. EXPERIMENTAL RESULTS Fig. 13 shows the implemented closed-loop system block diagram of the APF used to compensate the harmonics load current. The inductor of the APF has an inductance "L " equal to 116 (µH) and dc capacitor link "C" equal to 3250 (µF) to make the DC link voltage nearly constant. Since the scope signals are in voltage scale, and then every current reading referred to the voltage scale is calculated using Equation (2) which represents the calibrated current sensor equation as: i = 0.2278 +2.8678 x vsensor Amp (2) Where vsensor is the output voltage of the current sensor. For three-phase rectifier bridge feeding only the R-C load shown in Fig. 13 has load parameters as follows R=6.2Ω, and C=350µF. Fig. 14, shows the drawn line current by the load on Channels 1,and 2. Also, the extracted harmonic content of both line "A", and line "B" on Channels "3" and "4" are shown. The harmonics current waveforms have been extracted based on the Average Control Technique in the d-q reference frame which is written in C++ language. The control is based on microcomputer system. Fig. 15, shows the harmonic compensation of the load current appearing on channel "R1", the supply voltage on channel "1", and the supply current on last channel "3". It is seem that the supply current has the sine wave track with high frequencies switching noise. These noises will be decreased using a small size passive filter, and enhancing the system speed by using DSP based system. Since the control algorithm is executed with frequency reached to only 8 kHz. The APF can be used to correct the fundamental power factor by adding the fundamental reactive component to the control signal used to control the APF. The supply current after compensating both harmonics and reactive component appears in Fig. 16 on channel "3" where its ground coincides with the same ground of the supply voltage appearing on channel "1". The supply current has a good power factor than that appearing in Fig. 15.


350 uF

A C

6.2 Ohm


4 Ohm

152mH

L =116

Inverter

Data Acquisition

Gating

C = 3250

Fig.13 Experimental Circuit

Another general load introduces three parallel three-phase loads as shown in Fig. 13. Fig. 17, shows the line load currents on channel "1", and channel "2". The extracted harmonics for line "A" and line "B" appearing on channels "3, 4". Fig. 18 shows the load current on channel "R1", the supply voltage on channel "1", and the supply current on channel "3" after compensation. A sinusoidal supply current is achieved. Adding the fundamental reactive current signal component to achieve power factor correction, the supply current appearing in Fig. 19 on channel "3" has a good power factor than that appearing in Fig. 18.



CH1 ia-Load of line "A" CH3 iha-Load

CH2 ib-Load of line "B" CH4 ihb-Load

Fig.14 The ia-Load , iha-load and ihb-load in three-phase system

CH1 Va Phase "A" R1 ia-Load Line "A" CH3 isupply Line "A" Fig.15 Harmonic compensation for R-C load fed from rectifier bridge

R1 ia-Load CH1 Va Phase "A" line "A" CH3 isupply line "A"

CH1 ia-Load of line "A" CH2 ib-Load of line "B" CH3 iha-Load CH4 ihb-Load

Fig.16 Harmonic compensation and power factor correction

Fig.17 The ia-Load , iha-load and ihb-load



R1 ia-Load line CH1 Va Phase "A" "A" CH3 isupply line "A" Fig. 18 Harmonic compensation only

CH1 Va line "A" R1 ia-Load line "A" CH3 isupply line "A" Fig. 19 Harmonic compensation and power factor correction

V. CONCLUSIONS This paper introduces both simulation and experimental results for three-phase APF at two different non-linear load types. The harmonics extraction algorithm is based on the average control technique on the d-q reference frame. The closed loop system is implemented using the hysteresis current controller. It is clear from the results obtained that harmonic current extraction has been accomplished successfully. Good results have been achieved for both harmonics compensation and power factor correction.

VI. REFERENCES [1] Bhim Singh, Kamal Al_Haddad, and Ambrish Chandra, Oct. 1999, “A Review of Active Filters for Power Quality Improvement” IEEE Trans. Ind. Elect, Vol. 46, No. 5 pp 960-971. [2] M.EI-Habrouk, M.K.Darwish, and R. Mehta, Sept 2000,"Active power filters: A review", IEE Proc.–Electr. Power Appl., Vol. 147, No. 5, Pp. 403-413.



[3] A. A. Mansour, July 2004, "Three-Phase Shunt Active Power Filter for Industrial Loads", Ph.D. thesis, Cairo Univ. [4] Akagi H., Kanazawa Y., And A. Nabae, May/June 1984, “Instantaneous Reactive Power Compensators Comprising Switching Devices without Energy Storage Components”, IEE Trans. On Ind. Appl., Vol. IA-20, No. 3, pp. 625630. [5] Leszek S. Czarnecki, January 2006, "Instantaneous Reactive Power p-q Theory and Power Properties of Three-Phase Systems", IEEE Transactions On Power Delivery, Vol. 21, No. 1. [6] Patricio Salmerón and Reyes S. Herrera, July 2006, "Distorted and Unbalanced Systems Compensation Within Instantaneous Reactive Power Framework", IEEE Transactions On Power Delivery, Vol. 21, No. 3. [7] María Isabel Milanés Montero, Enrique Romero Cadaval, and Fermín Barrero González, , January 2007, "Comparison of Control Strategies for Shunt Active Power Filters in Three-Phase Four-Wire Systems", IEEE Transactions On Power Electronics, Vol. 22, No. 1. [8] Hongyu Li, Fang Zhuo, Zhaoan Wang, Wanjun Lei, and Longhui Wu, May 2005, "A Novel Time-Domain Current-Detection Algorithm for Shunt Active Power Filters", IEEE Transactions On Power Systems, Vol. 20, No. 2. [9] Seung-Gi Jeong and Myung-Ho Woo, June 1997, "DSP-Based Active Power Filter With Predictive Current Control ", IEEE Trans. Ind. Electron., Vol. 44, pp. 329-336. [10] B.N.Singh, A.Chandra And K.AI-Haddad, March 2000, "DSP-based indirectcurrent-control led STATCOM Part 1: Evaluation of current control techniques", IEE Proc. – Electr. Power Appl. Vol. 147, No. 2, pp. 107-112.


STRESS INDUCED DEGRADATION OF SILANE AND DCP CURED XLPE INSULATIONS PRODUCED FROM LOCAL RAW MATERIALS

M.I. Qureshi1, A.A. Al-Arainy2, N.H. Malik3 1: Research Center, College of Engineering, King Saud University, P.O. Box 800, Riyadh 11421, [email protected] 2: E.E. Dept., College of Engineering, King Saud University, P.O. Box 800, Riyadh 11421, [email protected] 3: E.E. Dept., College of Engineering, King Saud University, P.O. Box 800, Riyadh 11421, [email protected]

ABSTRACT In this experimental investigation, initiation and propagation of electrical trees was monitored using a PC-based on-line system. Silane and DCP cured XLPE sample moulds were prepared using locally produced PE pallets by imparting curing on them for different time intervals. The electrical stress was applied in needle-plane electrode configuration in two different ways, using voltage-time steps of 1 kV/15 minutes and 3 kV/15 minutes. The results indicate that branch type trees are produced under the former case, while the bush-type trees occur under the later. However, subjecting these samples to heat cycling results in branch-bush type trees. Curing time and temperature cycling were found to have profound effect on this degradation. Generally there exists an optimum curing time/cross-linking index at which the electrical tree's growth is the lowest. Moreover, the temperature cycling enhances the tree initiation onset voltage in DCP cured sample while it is opposite in silane cured samples. KEYWORDS Stress induced degradation, Electrical treeing, Curing techniques, Bush type trees, Branch type trees, Curing time, Heat cycling.

INTRODUCTION Saudi Arabia is one of the largest producers of polyolefinic compounds. Out of these, polyethylene (PE) has a huge consumption in the local market. The major share of this product is consumed by the local electric cable compound and power Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007

STRESS INDUCED DEGRADATION OF SILANE AND DCP CURED XLPE INSULATIONS

cable manufacturers. This material is produced in the form of granules which can be extruded with conductors to form cylindrical cable insulation which can be employed in the temperature range of −50 °C to 75 °C. A more advanced version that can be used up to 90 °C and that has outclassed all other types of high voltage cable insulations is the cross-linked polyethylene (XLPE) compound. It is obtained by cross-linking low density PE by different techniques. The most commonly used cross-linking methods employed by cable manufacturers are known as (i) peroxide, and (ii) silane curing techniques. In the former case, 1-2% dicumyle peroxide (DCP) is added as an initiator in the base PE compound and then subjected to heat (usually 180 °C) and pressure (~250 bar) in presence of tin catalyst extruder to cure and convert PE to XLPE insulation which is denoted as P-XLPE in this paper. In the later case the cable's compound is first grafted with silane (also known as Sioplas material, which is a mixture of LDPE with grafting components and master batch ready for cross-linking). After extrusion the finished product is immersed in "sauna" for hydrolysis and condensation to form Si-O-Si links between the PE molecules. The combination of hydrolysis and condensation reaction leads to cross-linking. The silane based XLPE, termed here-to-forth as S-XLPE, has silicone in its matrix which renders it abrasion-resistant while the silane process is considered more economical than the DCP cured XLPE [1,2]. The cable manufacturers in Saudi Arabia use both of these curing techniques in the production of medium voltage cables. However, the most serious degradation process of XLPE is due to 'electrical treeing' that has caused serious threat to utilities worldwide. It is a latent damage in the insulation that consists of microscopic channels formed by partial breakdown of defective or overstressed regions produced during the manufacture and/or during the handling, installation and heat cycling of the cable. They are also reported to get initiated at the sharp tips of water trees that are also produced due to impurity ionic migration at the interface of the cable. Once initiated at the cable's interface, or at the water tree tips, these electrical trees propagate and eventually cause cable's breakdown [1]. Since the cable manufacturers in Saudi Arabia produce medium voltage cables using both types of curing methods, it was felt necessary to compare the electrical properties of both types of insulations. This paper presents the comparison of the two by the dint of optical on-line scanning technique in which the initiation and propagation parameters of electrical trees were examined. The results thus compare the propensity of S-XLPE and D-XLPE toward electrical treeing and also address the role of curing time as well as degassing of cable's insulation.

PREPARATION OF SAMPLES Both P-XLPE and S-XLPE materials used in this study were supplied by Plascom Ltd., Riyadh, Saudi Arabia which produces pallets of several polymers for use in cable production. P-XLPE samples were prepared in the form of 145 mm × 145 mm 232 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007

M.I. Qureshi , et al

× 2 mm thick plates by melting PE pallets at 120 °C for 10 minutes and crosslinking at 180 °C under 180 bar pressure for 15, 30 and 60 minutes, respectively. These samples are denoted here as X15, X30, and X60. Three similar S-XLPE samples were prepared first by compression molding at 120 °C for 20 minutes and then were boiled at 90 °C in distilled water for 2 hours, 4 hours and 6 hours. These three samples are denoted here to forth as S2, S4 and S6, respectively. From each sample plate, a number of sample blocks were prepared and high carbon steel needle electrode having a tip radius of 10 µm was inserted in each of these insulation blocks as shown in Fig. (1). The bottoms of each sample block was coated with a conductive paint.

2

5

Fig. (1): Test sample.

PC BASED EXPERIMENTAL SYSTEM The details of the experimental set-up are elaborated elsewhere [4], however, in brevity Fig. (2) shows the block diagram of the experimental set up, which consists of electrical as well as optical circuits. The output of the camera is linked to the PC, where the image of the stressed electrode gap can be viewed on its monitor with a magnification of X120. With this arrangement, either a complete video of the events that occur from tree onset to breakdown of the sample can be prepared or slides can be frozen and saved at specific time intervals of tree growth. These can be retrieved at later time for analysis and record.



1000:1 Probe

Fig. (2):

Block diagram of complete experimental set-up for real time acquisition of electrical treeing characteristics.

Modes of Electrical Stress Application The effect of variation of tree length with time is a sensitive parameter which depends on the voltage step (∆V) and time step (∆t) used for the initiation and growth of electrical trees. Most of the tree initiation stress (Ei) values reported in literature vary in the range of ~250 kV/mm to ~500 kV/mm [3]. Beside the morphology of the insulation, the accuracy in the estimation of tree parameters, and the mode of the application of voltage are perhaps the dominating factors in the large scatter of Ei values. Recent studies on this aspect have shown that if the voltage step is kept to 1 kV and maintained for 15 minutes before escalating to next step, then with this strategy mostly branch-type electrical trees appear. However, if the voltage is increased in 3 kV steps maintained for 15 minutes then bush type trees are generated in XLPE cable's insulation [5], [6]. This strategy of voltage application was followed in these experiments. For each sample tested, the voltage was ramped first at 1 kV/s to 10 kV and then it was maintained there for 15 minutes. It was then increased in either 1 kV or 3 kV steps as explained above. This way the pattern of voltage application has been applied such that each type of sample was tested twice i.e. at first with (∆V/∆t) of 1 kV / 15 minutes and then with (∆V/∆t) of 3 kV/minutes. The samples tested with 1 kV/15 minutes stress are denoted with subscript 'a' such as X15a or S2a, whereas samples subjected to 3 kV/15 minutes stress are denoted as X15b or S2b etc. It has been reported in literature [2] that longer boiling time for curing in S-XLPE leads to higher cross-linking density, while heating also known as degassing of XLPE samples reduces the volatile by-products present therein. To evaluate this 234 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


effect, both types of insulation samples were additionally aged in an oven maintained at 90 °C for 72 hours. These "conditioned" samples are denoted as X15c, X30c, X60c, S2c, S4c and S6c. Table (1) summarizes briefly the salient experimental features of these tested samples. The following sub-sections present the results obtained for these 18 samples subjected under electrical stress degradation using alternating current power supply operating at 60 Hz in a voltage test range of (1030) kVrms. Table (1): Summary of tested samples and symbols used. Sample Symbols X15a X30a X60a X15b X30b X60b X15c X30c X60c S2a S4a S6a S2b S4b S6b S2c S4c S6c

Curing Applied on XLPE DCP DCP DCP DCP DCP DCP DCP DCP DCP Silane Silane Silane Silane Silane Silane Silane Silane Silane

Curing Time

15 minutes 30 minutes 60 minutes 15 minutes 30 minutes 60 minutes 15 minutes 30 minutes 60 minutes 2 hours 4 hours 6 hours 2 hours 4 hours 6 hours 2 hours 4 hours 6 hours

Degassing (Conditioning) Applied − − − − − − Yes Yes Yes − − − − − − Yes Yes Yes

Volt/Time Step Used 1 kV / 15 minutes 1 kV / 15 minutes 1 kV / 15 minutes 3 kV / 15 minutes 3 kV / 15 minutes 3 kV / 15 minutes 1 kV / 15 minutes 1 kV / 15 minutes 1 kV / 15 minutes 1 kV / 15 minutes 1 kV / 15 minutes 1 kV / 15 minutes 3 kV / 15 minutes 3 kV / 15 minutes 3 kV / 15 minutes 1 kV / 15 minutes 1 kV / 15 minutes 1 kV / 15 minutes

RESULTS AND DISCUSSION Stress Degradation of P-XLPE Based Insulations (a)

Samples Type Xa

Three samples of DCP cured XLPE insulation were subjected one by one under voltage-time (∆V/∆t) stress of 1 kV/15 minute step. In each case, a minuscule filament measuring in length of (3~5) µm appeared as a sudden outshoot at the tip of the needle electrode at a critical threshold of voltage and time. The value of voltage and tree initiating time (incubation time) varied in each sample. Once initiated, the 235 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


tree rapidly propagated toward the plane electrode with its branches increasing in time and length. Fig. (3 a–d) displays photo frames of growth of a branch-type tree produced in the sample X60a. Fig. (4) compares the growth-time curves of the three differently cured type Xa samples of XLPE insulation. It is clear from these results that, generally, the tree grows initially at a high velocity up to a certain length while it possess few branches but this growth rate retards with the birth of new branches and an increase in total length. The critical initial rapid growth in case of sample X15a is longer than in sample X30a which indicates that the increased curing time, which results in higher cross-linking index, arrests the growth of tree filaments that are produced due to mechanical fracture of molecular bonds of the polymer. In case of sample X60 the tree initiated at 13.0 kV and hence the growth of tree is swifter than in case of samples X15a and X30a.

Fig. (3):

(a)

(b)

(c)

(d)

Branch type growth in sample X60a, axial length L of tree (µm) and time t, (a): L=158.2, t=1.0s, (b): L=293.8, t= 3.0s, (c): L=524.4, t=9.0s, (d): L=745.8, t=21.0s.



X60a

X30a X15a

Time (minutes) Fig. (4):

(b)

Comparison of tree growth with time for three types of XLPE insulation materials subjected to (∆V/∆t) of 1 kV/15 minutes.

Samples Type Xb

Three samples namely X15b, X30b, and X60b were subjected in a similar manner as the samples type Xa, but using a voltage-time step (∆V/∆t) of 3 kV/15 minutes, initiatory filaments emerged at a higher voltage. However, the value of tree initiating voltage and time after the application of voltage varied in each sample. In this case too, the initial growth rate was higher and almost linear in length range of 500 ~ 800 µm. However, several branches appeared near the point electrode, which increased in number and time rendering the tree into a bush-type structure. When the initial filamentary structure is transformed into a bush-type tree, the growth rate also retards swiftly toward saturation. Fig. (5 a–d) displays selected photo-frames of a bush-type tree. Fig. (6) compares the growth-time curves of the three differently cured type Xb samples. It is clear from these results that the trees in each sample initially grow rapidly in a linear fashion but then the growth rate gets retarded as it becomes associated with the emergence of newer bunch of filaments which transforms it to a bush-type structure. In this case too, though the trees in samples X30b and X60b are initiated at the same voltage level (13 kV), but the two growth-time characteristics are different. Whereas in case of X30b the growth rate rapidly increases up to 770 µm before it tends toward saturation, it only grows up to ~ 500 µm in sample X60b. This is due to the longer curing time which corresponds with comparatively higher cross-linking 237 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


index of the polymer. That means that higher cross-linking index helps in retarding the growth rate of electrical trees. This is in line with similar behavior reported for the growth of water trees in XLPE insulation, as well [2].

Fig. (5):

(a)

(b)

(c)

(d)

Bush type growth in sample X30c, axial length L of tree (µm) and time t, (a): L=113, t=2.0s, (b): L=284.76, t=12.0s, (c): L=497.2, t=64.0s, (d): L=506.24, t=240.0s.

Nevertheless, there is another aspect of cross-linking as well. It is evident in the growth-time curve of sample X15b as it exhibits a different scenario. In this case the electrical tree emerged at 16 kV and after much longer incubation time than in samples X30b, and X60b. This shows that there exists an optimum value of crosslinking after which the polymer exhibits favorable propensity toward electrical treeing and beyond which the beneficial effect of cross-linking certainly gets weekend.



X15a

X30a

X60a

Time (minutes) Fig. (6):

(c)

Comparison of tree growth with time for three types of XLPE insulation materials type Xb subjected under (∆V/∆t) = 3 kV/15 minute. Samples Type Xc

Three degassed samples type X15c, X30c, and X60c were subjected under (∆V/∆t) of 1.0 kV/15 minutes. Here the branch-bush type trees emerged instead of branch type trees reported above for the same samples where such a conditioning treatment was not employed. The results show that the tree initiation stress is increased by 1~2 kV in samples X15c and X30c whereas it declined in sample X60c. This further substantiates the presence of an optimum of around 30 minutes of curing time of DCP cured XLPE. In addition, the growth rate in sample X30c after increasing to a length of ~500 µm has rapidly retarded to saturation thus arresting the further propagation of tree. Table (2) summarizes the salient features of electrical tree parameters for the three cases of X-type insulation studied. It is clear that the stress at micro-protrusion of 10µm tip radius in this type of cable insulation will lead to formation of branch type trees at electrical stress level (Ei) of ≤ 300 kV/mm, whereas bush-type trees will be produced at Ei ≥ 330 kV/mm. Table (3) illustrates impact of the curing time (and hence the cross-linking index) on the initial velocity of propagation of electrical trees (vi). In case of branch-type trees produced in insulation X15 and X30, although the trees have initiated at the same 239 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


stress level, but the average propagation velocity (vi) of tree growth is 10 µm/s in sample X15 while it is 5 µm/s in sample X30 and is thus much slower. It indicates that insulation matrix with higher cross-linking index poses arresting behavior on the rate of growth of electrical tree. A similar behavior can be noticed in the growth rate of bush-type trees, as well. Here, vi in case of sample X30 is 27 µm/s while it reduces to 15 µm/s in case of longer cured sample X60. This aspect is confirmed in case of branch-bush trees, as well. Although the tree initiating voltage in case of sample X30 is 13 kV, its vi = 13 µm/s as compared to that of sample X15 in which case, this tree is propagating under a voltage of 12 kV only and is much smaller. Table (2): Comparison of tree parameters in new and degassed insulation type PXLPE.

Sample #

New

X15a X30a X60a X15c

Conditioned X30c X60c

Table (3):

(∆V/∆t = 1 kV / 15 minutes) (∆V/∆t = 3 kV / 15 minutes) ti Vi Ei Tree ti Vi Ei Tree (min:sec) (kV) (kV/mm) Type (min:sec) (kV) (kV/mm) Type 25:00 11.0 283 Branch 37:59 16.0 411 Bush 18:00 11.0 283 Branch 16:59 13.0 334 Bush 47:59 12.0 310 Branch 22:29 13.0 334 Bush 39:29 12.0 310 BranchBush 56:59 13.0 334 Bush 00:19 10.0 357 BranchBush

Impact of curing time on the initial velocity of propagation (vi) of electrical trees and their initiating voltage (Vi) in P-XLPE insulation. Tree Type

Branch under (1 kV/15 min) Bush under (3 kV/15 min) Branch-Bush (conditioned sample) under (1 kV/15 min)

vi Vi vi Vi vi Vi

X15 10 µm/s 11 kV 53 µm/s 16 kV 15 µm/s 12 kV

Types of Sample X30 5 µm/s 11 kV 27 µm/s 13 kV 13 µm/s 13 kV

X60 50 µm/s 13 kV 15 µm/s 13 kV 7 µm/s 10 kV

Stress Degradation of S-XLPE Based Insulations Three samples each of Sa, Sb, and Sc type blocks of silane cured S-XLPE insulation namely S2a, S4a and S6a were subjected one by one under voltage-time (∆V/∆t) stress 240 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


step of 1 kV/15 minutes as well as 3 kV/15 min. protocol as under insulations type P-XLPE. Similar to in P-XLPE, treeing was seen in S-type insulation, also. Tables (4) and (5) summarize the salient features of electrical tree parameters for three cases of S type insulation. It is clear that stress at protrusions left as manufacturing defects or produced during service malfunctions and having a tip radius of 10 µm shall lead to formation of branch-bush type electrical trees at electrical stress levels of Ei ≤ 260 kV/mm and bush type trees at Ei ≥ 330 kV/mm. Both of these Ei values are smaller than their corresponding values in DCP cured XLPE. Table (4): Comparison of tree parameters in new and degassed S-XLPE insulation.

Sample #

New

S2a S4a S6a S2c

Conditioned S4c S6c

(∆V/∆t = 1 kV / 15 min) (∆V/∆t = 3 kV / 15 min) ti Vi Ei Tree ti Vi Ei Tree (min:sec) (kV) (kV/mm) Type (min:sec) (kV) (kV/mm) Type 31:13 12 311 Branch 15:12 13 337 Bush 34:10 12 311 Branch 36:38 16 411 Bush 13:00 10 257 Branch 1:57 10 257 Bush 7:17 10 257 BranchBush 14:50 10 257 BranchBush 12:25 10 257 BranchBush

Table (5): Impact of curing time on the vi and Vi for S-XLPE insulation. Tree Type Branch under (1 kV/15 min) Bush under (3 kV/15 min) Branch-Bush (conditioned samples) under (1 kV/15 min)

vi Vi vi Vi vi Vi

S2 40 µm/s 12 kV 33 µm/s 13 kV 43 µm/s 10 kV

Types of Cured Samples S4 S6 15.8 µm/s 11 µm/s 12 kV 10 kV 30 µm/s 9 µm/s 16 kV 10 kV 2.5 µm/s 7 µm/s 10 kV 10 kV

CONCLUSIONS The results indicate that branch-type electrical trees are produced under the 1 kV/15 minutes steps whereas bush-type trees are produced under the 3 kV / 15 minutes aging stress. However, when both types of insulations are subjected to degassing, branch-bush type trees are produced. Generally, there exists an optimum crosslinking index at which treeing growth is the slowest. Degassing removes volatiles 241 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


from DCP cured samples which enhances the treeing inception voltages. However, for silane cured samples the effect is opposite.

ACKNOWLEDGEMENT The financial and technical support provided by Research Center, College of Engineering, under grant # 12/426 is thankfully acknowledged. We are also grateful to Plascom Ltd., Riyadh for providing DCP and Silane curved samples of XLPE prepared using their polymer pallets. REFERENCES [1]

L.A. Dissado and J.C. Fothargill, "Electrical Degradation and Breakdown in Polymers", Peter Perigrinous Ltd., UK, 1992.

[2]

Y. Zhu, H.G. Yoon and K.S. Suh, IEEE Trans. on Dielectric and Electrical Insulation, Vol. 6, No. 2, pp. 162-168, 1999.

[3]

N.H. Malik, A.A. Abdullah, A.A. Al-Arainy, and M.I. Qureshi, European Transactions on Electrical Power (ETEP), Vol. 16, pp. 205 – 218, 2006.

[4]

A.A. Al-Arainy, N.H. Malik, and M.I. Qureshi, Research Report # 12/426, Research Center, College of Engineering, King Saud University, Riyadh, 2007.

[5]

A.A. Al-Arainy, M.I. Qureshi and N.H. Malik, 14th GCC CIGREE Seminar on Transmission Lines Design and Operation, Kuwait, pp. 199-205, Dec. 2004.

[6]

A.A. Al-Arainy, M.I. Qureshi, and N.H. Malik, Proceedings of ICPADM 2006, IEEE Publication 06C37773, Vol. 1, Paper D2, pp. 115-118.


ENVIRONMENTALLY INDUCED DEGRADATION OF LOCALLY PRODUCED INSULATING POLYMERS

A.A. Al-Arainy1, N.H. Malik2, M.I. Qureshi3 1: E.E. Dept., College of Engineering, King Saud University, P.O. Box 800, Riyadh 11421, [email protected] 2: E.E. Dept., College of Engineering, King Saud University, P.O. Box 800, Riyadh 11421, [email protected] 3: Research Center, College of Engineering, King Saud University, P.O. Box 800, Riyadh 11421, [email protected]

ABSTRACT Polymers are widely used for a variety of industrial and engineering applications which include in particular the high voltage apparatus. In the Kingdom, several manufacturers are producing polyethylene based raw materials as well as insulated wires and cables. When such insulations are subjected to electrical stress and environmental parameters such as ionic species, water and temperature etc., their dielectric properties degrade. This paper assesses the impact of such modes of degradation on locally produced polymers that are commonly used by electrical industry. The main mechanisms which are reported for such degradation are identified and results elucidated. It is shown that the aging time as well as the differently manufactured polymers are influenced significantly by environmentally induced stresses. Water tree population and their lengths were subjected to Weibull and log-normal models of statistical distributions and the data were found to fit better on the log-normal model. This suggests that the most likely mechanism for environmentally induced water tree degradation is electro-mechanical fatigue based. Results, analysis, discussion and conclusions are presented in the paper.

1.

INTRODUCTION

There are two phenomena that cause the degradation of polymeric insulation. One of them is the 'electrical treeing' which causes the degradation due to high electric stress and the other is the 'water treeing' which causes degradation in the presence of moisture and ionic species present in the surrounding environment [1]. Studies show that the reduction in the dielectric strength of XLPE insulation is a direct function of water tree length. When a given cable is aged under specified conditions, the resulting lengths of water trees exhibit variations. Therefore, statistical methods have been used to understand mechanisms that control initiation Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


and growth of water treeing phenomena in polymeric insulated cables [1-4]. Kuma et al. [6] and Sletback and Ildstad [7] found that the cumulative distribution of water tree length was exponential. However, later studies showed that exponential distribution generally gives erroneous conclusions on the physical phenomena. Therefore data that shows skewness in its distribution, like that of water trees, should be attempted on other distributions to get valid conclusions. The log-normal and Weibull distributions are the most popular choices for analyzing continuous stochastic variables that have skewed distributions. The log-normal distribution is applied commonly in engineering areas where mechanical fatigue mechanism is operative, whereas Weibull distribution is applicable in dielectric breakdown of cable insulation system [1-4]. This paper describes the studies carried out to determine the impact of the later on the degradation of polymeric insulation produced in Saudi Arabia and other GCC countries. Medium voltage, 15 kV rated, XLPE insulated power cables were used for carrying out such investigations. The main parameters investigated included length and number density of water trees for a selected ionic salt solution as a function of aging time under elevated but constant applied voltage to assess treeing parameters in four different cable types to obtain a comparative treeing index and subject them to statistical models to determine the most fitting one. 2.

EXPERIMENTAL SYSTEM AND PROCEDURES

2.1

Tested Cables

For carrying out the environmental induced degradation of polymers, XLPE insulated power cables that are commonly used by a local electrical power utility and supplied by three national and one regional XLPE cable manufacturer, were selected. All four types of cables were produced according to the same specifications and were acquired from the local market. All cables were rated 15 kV, and were 3 core, 185 mm2 copper conductor, XLPE insulation, copper screen, steel wire armourd, and had an outer sheath of PVC. Table (1) gives the symbols assigned to these cables.


A.A. Al-Arainy , et al

Table (1): Summary of the cables used. S/No. 1 2 4

3

2.2

Manufacturer GCC Manufacturer Saudi Manufacturer # 1 Saudi Manufacturer # 2 Saudi Manufacturer # 3

Cable Type embossed on sheath 3 x 185 mm2, 15 kV, Cu/XLPE/PVC Manufactured 2002 3 x 185 mm2 Cu/XLPE/SWA/PVC 15 kV, Manufactured 2004 3 x 185 mm2/35mm2 Cu/XLPE/PVC/15 kV, Manufactured 2004 3 x 185 mm2 8.7/15 kV/Electric Cable Cu/XLPE/PVC/ Manufactured 2004

Symbol assigned A B C

D

Experimental Set up

The above cable samples were desheathed and single core samples were subjected to similar multifactor stress accelerated aging. These samples had strippable outer semiconducting screens. For each cable type, four meter long, single core samples were cut and their copper screens were removed. Around 1.00 meter length of insulation semiconducting screen was stripped away from each cable at both ends and ‘Elastimold’ stress cones were inserted. Each sample under test was provided with a ground wire by wrapping around the cable sample copper wire of ∼ 2 mm2 cross-section and connecting it to stress cone’s ground point. Each sample was placed inside a stainless steel tank which was filled with a purposefully selected ionic solution. Fig. (1) illustrates the complete set up to impart accelerated environmental and electrical aging on these cable samples in the presence of ionic aqueous solution.



HV

Fig. (1):

Experimental set up used to evaluate the impact of environmentally induced degradation of a cable sample.

For each type of study, cable samples were connected in parallel and fed at 3Uo = 26.1 kV (Uo = rated phase to ground voltage = 8.66 kV) from a 50 kV, 30 kVA, 60 Hz power supply. The ionic solution was circulated in closed loop with the help of a microprocessor controlled thermal fluid circulator which has a working temperature range of 25 oC – 140oC, and temperature control accuracy of ± 0.02oC. In these samples, a temperature of 50oC was maintained for 6 h, while it was kept switched off for the subsequent 18 h period each day. With this arrangement the temperature of the insulation shield rises to 50oC within 20 minutes after the heaters of the circulator are switched on. When shut off, the insulation shield temperature gradually falls to ambient level in about 12 h. The cable conductor was not subjected to any current cycling in this study. The ionic solution studied was 0.01 mole copper sulphate (CuSO4) for all the investigations reported here. This salt solution was selected since it is formed around the copper conductor as well as copper screen of the cables that are buried in soil. Sub-surface soil in Saudi Arabia is reported to contain very high concentration of sulphates [5]. CuSO4 has been found to exert a very strong influence on the water treeing degradation of XLPE insulation [3]. This experimental set up remained energized uninterrupted for a predetermined time period for each tested case. 246 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


Two aging regimes were used: A.

Cable type B was aged for aging periods of 500 hours, 1000 hours, and 1500 hours to determine the effect of aging time on the environmental induced degradation.

B.

Cables A, C, and D were aged for 800 hours to determine the comparative rates of environmental induced degradations in these cables.

After the end of each predefined aging period, the cables were removed. Insulation wafers were prepared and dyed for microscopic investigations. This is essential, since the dry water trees are not distinguishable by naked eye, therefore their staining has to be carried out. Although several techniques have been used by various researchers, our experience indicates that the ASTM method gives the best results and thus it was followed [1,2,4]. The cable insulation wafers were arranged on the microscope glass discs after dyeing for inspection under a high resolution imaging system consisting of X120 magnification microscope, CCD camera, monitor and video printer available in the High Voltage Laboratory. This system is also equipped with type PAXCAM-2 digital microscope camera with 2 Megapixel memory and USB 2.0 interface with PC. Its software ‘PAXIT’ has the features to capture images on-line and also can perform analysis.

3.

RESULTS AND ANALYSIS

Each cable sample displayed a different behavior regarding the water tree parameters. The studied tree parameters included, type of water tree (bow-type or vented-type) and maximum tree length (Lmax). In addition, the number density i.e. the number of trees per cubic centimeter of insulation volume was also estimated and compared. 3.1

Water Trees Photos

Few vented and many bow-tie water trees were seen in different cable samples for various aging conditions. Photos (1) and (2) illustrate few selected examples of bow-tie water trees recorded in different cases. Photo (3) illustrates a vented water tree.



Photo (1):

A 329 µm long bow-tie water tree after 1500 hours of aging in cable type B.

Photo (3):

3.2

Photo (2):

Many small bow-tie water trees after 1500 hours of aging in cable type B.

A vented water tree that emerged from the insulation screen of the tested cable type A.

Effect of Aging Time on Water Treeing

Using cable sample B, and aging times of 500 hours, 1000 hours and 1500 hours, the total number of water trees as well as tree length distributions of these trees and water tree's number density were measured. Many bow-tie water trees of different lengths were seen in insulation slices. The length ≥ 10 µm of every tree detected was measured for various aging times. Fig. (2 a−c) shows statistical variations of tree lengths and tree numbers as histograms for the three aging times studied. These results clearly show that: (i) the aging time 248 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


strongly influences the treeing related degradation of the insulation since the number of water trees increases with the aging time, (ii) for each aging time, there is a large statistical scatter in both the lengths as well as the numbers of water trees, and (iii) the shape of statistical distributions are non-Gaussian, skewed and aging time dependent. These observations show that like all other engineering phenomena, water treeing is also controlled by a physical relationship which is related to a statistical model [5]. This aspect of these results is discussed in detail in section (3.3) below. In order to have a better understanding of the effect of aging time on the relative level of water treeing degradation in cable B, the tree number density, nt, (number of trees per mm3 volume of the insulation) was measured for the cable samples corresponding to different aging times. Fig. (3) shows the treeing density for aging times of 500 hours, 1000 hours and 1500 hours for the tested cable samples which were taken from the same cable drum. It can be seen from this figure that all data follow a straight line plot on a semi-log paper which mean that tree density is proportional to logarithm of aging time. This linear relation can be approximated by the following equation: nt = 10.773 log10 (t) – 29.015

(1)

where nt = water tree density (number of trees per mm3) t = aging time (hours). This figure also compares the 'nt' values of cables A,C, and D that were aged for 800 hours, under similar aging conditions. It is clear that cable type (D) exhibits the lowest water tree index, while cable A exhibits the highest water tree index among the four cables tested which could be attributed to parameters related to cable's extrusion and materials.



(a) after 500 hours of aging. (µm)

(b) after 1000 hours of aging.

(µm)

(c) after 1500 hours of aging Fig. (2):

Histogram of number of trees versus the tree length (µm)in cable B subjected to different aging times.



B

A

C D

Fig. (3):

3.3

Variation of tree density nt versus aging time for cable type B. The nt values obtained for cables A, C and D aged for 800 hours are also given for comparison.

Tree Length Statistical Distributions

It was shown in the previous section (3.2) that the histograms shown in Fig. (2) exhibit skewed behavior which is indicative of a certain engineering physical phenomena. To evaluate this aspect, two statistical models were applied as the tree length distributions. The Weibull distribution is of general nature that can fit in a wide variety of data, and in the present case the comprehensive count and length measurements of water trees when plotted on Weibull statistical model are expected to yield the stochastic behavior of these trees. The cumulative Weibull distribution function is given by:



F(X ) = 1 −

⎛ X−γ ⎞ ⎟⎟ − ⎜⎜ α ⎠ ⎝ e

β

(2)

Where α = scale parameter; β = shape parameter; γ = location parameter, and X = random variable, which in the present context is tree length or tree number variations. This is typically a 3 – parameter WD. It can be reduced to 2 – parameter distribution by letting γ = 0 Yet another statistical distribution that also has skewed shape is the log-normal (LN) distribution that has been successfully utilized in rotating machine insulation [4]. The LN distribution results when many random quantities cooperate multiplicatively, so that the effect of a random change is in every case proportional to the previous value of the quantity. Fatigue mechanism in materials is one of the engineering areas in which the log-normal distribution is frequently applied. The LN-distribution is a 2- parameter distribution and its probability density function of the distribution is given by:

F(X′) =

1 σ' 2 π

1 ⎡ X '− µ ' ⎤ − ⎢ ⎥ e 2 ⎣ σ' ⎦

2

(3)

where X´ = ln X, where X is the measured variable. In the present context, it is taken as tree length µ´ = mean of the natural log of X σ´ = Standard deviation of the natural log of X. For further statistical inference the results obtained including the characteristic parameters were evaluated for the two types of statistical distributions and are presented next. Here the main purpose of subjecting the tree lengths to these statistical models is to determine whether the tree growth is dominated by rapid propagation from weak sites (Weibull model) or whether it is the result of a mechanism in which several factors act cooperatively (LN-model). For all aging times, when the tree length data were subjected to a Weibull statistical distribution, a very poor fit was obtained which indicates that water trees lengths do not obey Weibull distribution. Fig. (4a) displays such a selected case. However, an excellent fit was obtained for log-normal distribution plots for all cases studied. Fig. (4b) exhibits data of Fig. (4a) plotted on L-N plot. LN-plots of all other aged cables (not shown here) display similar trend. Thus, it can be concluded that water tree 252 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


initiation and growth is dominated by a mechanism in which many factors cooperate, such as in fatigue failure of components. These studies substantiate this attribute which was forwarded earlier by Al-Arainy et al. [4] while reporting water treeing in cables aged under aqueous ionic solution of sodium chlorides. These results therefore confirm this mechanism related to water treeing, which is not yet fully understood despite the efforts being exerted in several research centers worldwide. This report is part of a continuing research at King Saud University laboratory and is thus a step forward toward the understanding of this puzzle.

(a) Fig. (4):

(b)

(a) Weibull distribution plot of bow-tie tree length data for aged cable B for 800 hours aged under 0.01 mole CuSO4 solution maintained under 3 Uo voltage, (b) Log-normal distribution plot of data given in Fig. (4-a).



4.

CONCLUSIONS

This experimental investigation leads to the following conclusions: 1)

Medium voltage XLPE insulated cables of different manufacturers display different degrees of propensity toward water treeing. Although these cables are manufactured according to the same specifications but the purity of raw materials used and manufacturing quality control play the major role in such a degradation deficiency.

2)

The statistical behavior of water trees length follow log-normal distribution.

3)

The water treeing growth in a cable is linearly related with its service life span while its initiation is controlled by a fatigue based mechanism.

5.

ACKNOWLEDGEMENT

Authors would like to thankfully acknowledge the financial and technical support provided by Research Center, College of Engineering, Riyadh, through their grant # 12/426. The help provided by Eng. N.R. Wani during the experimental part of this work is thankfully acknowledged.



6.

REFERENCES

[1]

L.A. Dissado and J.C. Fothargill, "Electrical Degradation and Breakdown in Polymers", Peter Perigrinous Ltd., UK, 1992.

[2]

I.F. Hodgson et al., Proc. of 2nd Symposium on Geotechnical Problems in Saudi Arabia, Riyadh, pp. 164-175, 1989.

[3]

Crine, J.-P., “Electrical, chemical and mechanical processes in water treeing”, IEEE Transactions on DEI, Vol. 5, No. 5, pp. 681 – 694, (Oct. 1998).

[4]

A.A. Al-Arainy, et al., "Statistical Evaluation of Water Tree Lengths in XLPE Cables at Different Temperatures", IEEE Trans. on Dielectric and Electrical Insulation, Vol. 11, No. 6, p. 995-1006, 2004.

[5]

Czaszejko, T., “Determination of statistical distribution of water tree lengths: Monte Carlo simulation”, 6th International Conference on Conduction and Breakdown in Solid Dielectrics (ICSD '98), pp. 333 – 336, (22-25 June 1998).

[6]

Kuma and K. Soma, "Statistical Analysis of Bow Tie Trees", Proc. 14th Sympos. Electr. Insul. Materials, Japan, pp. 491-495.

[7]

J. Sletbak and J. Ildstad, "The Effect of Service and Test Conditions on Water Tree Growth in XLPE Cables", IEEE Trans. Power App. Syst., Vol. 102, pp. 2069-2076, 1983.


MEASUREMENTS OF EARTH RESISTIVITY IN DIFFERENT PARTS OF SAUDI ARABIA FOR GROUNDING INSTALLATIONS

1: 2: 3: 4:

N.H. Malik1, A.A. Al-Arainy2, M.I. Qureshi3, M.S. Anam4 E.E. Dept., College of Engineering, King Saud University, P.O. Box 800, Riyadh 11421, [email protected] E.E. Dept., College of Engineering, King Saud University, P.O. Box 800, Riyadh 11421, [email protected] Research Center, College of Engineering, King Saud University, P.O. Box 800, Riyadh 11421, [email protected] Distribution Engineering Department (COA), Technical Support Division, Saudi Electric Company, P.O. Box 57, Riyadh 11411, [email protected]

ABSTRACT Electrical supply systems are grounded to improve safety of equipment and personnel and to enhance reliability of power supply. The nature of soil, specifically the soil resistivity, strongly influences the performance of a grounding system. This paper discusses the methods and tools used for measuring soil resistivity. Using such tools, soil resistivity measurements were carried out in more than 40 locations covering major areas of Kingdom of Saudi Arabia. The results of these measurements are analyzed and reported in this paper along with implications of measured parameters on the design and performance of grounding installations in transmission and distribution networks. Steps that must be taken to achieve acceptable grounding installations in high resistivity soils are listed also.

KEYWORDS Earth resistivity, grounding, earthing, grounding resistance, distribution networks.

INTRODUCTION The term "ground" is defined as a conducting connection by which a circuit or equipment is connected to the earth. The connection is used for establishing and maintaining as closely as possible the potential of the earth on the circuit or equipment connected to it [1-3]. Generally, three-phase, four wire power systems are grounded by connecting one or more selected neutral points to buried earth electrode systems. Such earths are referred to as system earths. At electrical Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007

MEASUREMENTS OF EARTH RESISTIVITY IN DIFFERENT PARTS OF SAUDI ARABIA

installations, all non-live conductive metallic parts are interconnected and also earthed to protect people against electric shock, and in this role the earth is referred to as a protective earth. Under normal conditions, there is only a residual current or no current at all in the earth path. However, very high magnitudes of current return to source via the earth path under fault conditions. The earth also conducts lightning currents and the current path may involve part of a power system either directly or by induction. The earthing system, or part of it, may therefore also be specifically designed to act as a lightning protective earth. The earth is a poor conductor and, therefore, when it carries high magnitude current, a large potential gradient will result and the earthing system will exhibit an earth potential rise (EPR) or ground potential rise (GPR) which is defined as the voltage between an earthing system and the reference earth [1-4]. The magnitude of power frequency earth fault currents can range from a few kA up to several tens of kA, and earth impedances of electrical installations may lie in the range from less than 0.1 Ω to over 100 Ω [5]. Consequently, there is a potential risk of electrocution to people in the vicinity of a power network during earth faults, and damage to equipment may also occur unless measures are taken to limit the earth potential rise and/or to control the potential differences in critical places. Lightning transients can also generate currents of several tens of kA in the earthing system, and this requires the power system to be protected against over-voltages. The discharge of transient current into earth may also present an electrocution hazard. Thus, earthing systems are designed to control potential differences within and around the electrical installation. These potential differences are specifically referred to as the step and the touch voltages. Furthermore, consideration is also given to the maximum earth potential rise of the grounding system. The extent of the rise of potential on the ground surface around a substation can be described in terms of a hot zone, and this is used to identify whether third party in the vicinity of the installation is affected. Transferred potential levels are another important risk factor which is associated with presence of metallic objects coming to the earthing installation under consideration from other earthing installations and entering with it in the same earth fault current loop. Thus, as mentioned above, although the purpose of earth connections in different parts of the power system differs, generally the following requirements must be fulfilled for a proper grounding [5]: a)

Proper Earth Fault Current Protection: the earth connection must provide a path of low impedance and adequate thermal capacity for satisfactory operation of the protective relays.


N.H. Malik. , et al

b)

Proper Equipment or Protective Grounding: in order to ensure the safety of the consumers and of the technical personnel as well as equipment, caused by sparking.

c)

Limitation of Earth Potential Differences: in order to avoid injury or death to persons or to animals and to limit the damage caused to the buried equipment, for instance cable sheaths which could be overstressed and fail during normal and fault conditions.

d)

In order to provide safe discharge path for energy of lightning and switching surges to dissipate to earth, proper lightning and protection grounding must be provided.

To achieve the above objectives, a proper grounding system is required. A grounding system generally consists of ground conductors, ground rods, items to be grounded and the earth itself. Soil and rock resistivity may vary considerably from region to region, and it is rarely constant either vertically or horizontally in the area of interest around an electrical installation. This variability makes the construction of earth models for the prediction of earth potential rise a very difficult task. The earthing resistance highly depends on soil resistivity, which in turn depends on a number of factors, such as: soil type, chemical composition of soil, salts dissolved in soil, moisture content of the soil, temperature's seasonal variation, grain size and distribution, as well as soil composition and pressure [6]. Earth resistivity plays an important role in the design and performance of an earthing system. Saudi Arabia has a great variety of earth geology and as a result different types of soils posses a large range of earth resistivity values. However, there is lack of comprehensive information about the typical earth resistivity values found in different types of Saudi Arabian soils. The investigators have measured the earth resistivity in different parts of the Kingdom with a view to develop a comprehensive data base that could be useful for the Saudi Electricity Company in the design of transmission and distribution systems grounding installations. This paper discusses the latest methods and equipment used for such measurements [6-7]. It also presents the results of earth resistivity measurements along with analysis and implications of the measured resistivity values for the design of transmission and distribution systems groundings.

EARTH RESISTIVITY MEASUREMENTS Earth resistivity measurements are useful for, (i) Estimating the ground resistance of a proposed substation or grounding installation, (ii) Estimating potential gradients including step and touch voltages, (iii) Computing the inductive coupling between neighboring power and communication circuits. 259 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


Earth resistivity varies not only with the type of soil but also with temperature, moisture, salt content, and compactness. The literature indicates that the values of earth resistivity vary from 0.01 to 1 Ω.m for sea water and up to 109 Ω.m for sandstone. The resistivity of the earth increases slowly with decreasing temperatures from 25°C to 0°C. Below 0°C the resistivity increases rapidly. In frozen soil, the resistivity may be exceptionally high in winter [6]. Several methods have been discussed in literature for measuring earth resistivity. The most accurate method of measuring the average resistivity of large volumes of undisturbed earth is the four-point method. In this method, four small electrodes spaced at intervals a are driven vertically into ground, all at depth b. A test current I is passed between the two outer electrodes and the potential V between the two inner electrodes is measured. Then V/I gives the resistance R in ohms. Earth resistivity is related to R and two different alternatives of the four-point method are often used: 1)

Equally Spaced or Wenner Arrangement. With this arrangement the electrodes are equally spaced as shown in Fig (1a). Let a be the distance between two adjacent electrodes and b is the driven length of electrode into the soil. When a ≥ 10b, ρ is given as; ρ = 2πaR

Here ρ gives approximately an average value for the resistivity of the surface layer of soil having a thickness a. A set of readings taken with various probe spacings gives a set of resistivities which, when plotted against spacing, indicates whether there are distinct layers of different soil or rock and gives an idea of their respective resistivities and depth.


N.H. Malik. , et al

Fig. (1):

2)

Four-point method of resistivity measurement ; (a) Equally spaced, (b) Unequally spaced electrodes. Unequally-spaced or Schlumberger-Palmer Arrangement. In order to be able to measure earth resistivity with large spacings between the current electrodes, the arrangement shown in Fig. 1(b) can be used successfully. If buried length, b, of electrodes into soil is small compared to their separation d and c, then the measured resistivity ρ can be calculated as follows: ρ = πc(c+d)R/d

ELECTRICAL RESISTIVITY IMAGER Electrical resistivity measurements are required to determine the resistivity distribution in the sub-surface of a test location both in lateral (horizontal) and transverse (vertical) directions. Single measurements at a fixed spacing of electrodes (e.g. Wenner method) can therefore be misleading, and instead, several sets of arrays of different spacings have to be obtained to produce a one-dimensional plot of ρ. This leads to a cumbersome procedure. However, advanced field technology now allows a much greater number of measurements in a shorter time. Sophisticated inversion strategies based on finite element analysis and making use of increased computing performance have been invented to obtain high-resolution spatial images of subsurface resistivity distributions. This technology, known as Electrical Resistivity Tomography (ERT), is a very fast and cost-effective technique for measuring resistivity profiles of earth for delineating areas of changing resistivity. 261 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


The ERT instrument works on the principal of 4-electrode technique mentioned earlier but using induced polarization method. To carryout measurements, a multielectrode cable sets (MECS) are laid on the ground and connected to electrodes which are placed in a straight line at pre-determined spacings. The instrument selects four equally spaced electrodes at a time and sends current signal pulses to the two outer electrodes and receives back the voltage pulses from the inner two electrodes. Thus, when the instrument is switched ON it will scan the first four electrodes (e.g. 1,2,3, and 4), then the computer will move to the next set of electrodes (i.e. 2,3,4, and 5) and so on. If the spacing of these electrodes is 1m, these iterations will generate lateral resistivity profile encompassing a depth of 1m. After going up to the last connected electrode, the computer will then shift to scan lateral profile of soil resistivity at a depth of 2m. In this case, it will select the electrodes (1,3,5, and 7), (2,4,6, and 8), (3,5,7, and 9) and so forth till the last available electrode in the cable set is reached. Similarly to scan the lateral resistivity profile at a depth of 3m it will scan electrodes (1,4,7, and 10), (2,5,8, and 11), etc., while for a depth of 4m, electrodes (1,5,9, and 13) and (2,6,10, and 14) will be scanned. With MECS containing 16 electrodes, 5 layers of earth will be investigated for soil resistivity. Then using these measured resistivity values and software inversion tools, a 2D-plot of earth resistivity with lateral distance and depth can be obtained. Fig. (2) exhibits a 5-layer resistivity profile of a site in Riyadh area, while Fig. (3) provides its 2D-inverted plot. With such an imager, 3-D plots of ρ can also be obtained if the measurements at different lines are carried out and proper software tools are available.

Fig. (2): Five layer resistivity profile of a site in Riyadh area. 262 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007

N.H. Malik. , et al

Fig. (3): Inverse model 2D-plot of data shown in Fig. (2).

RESISTIVITY MEASUREMENTS IN SAUDI ARABIA Using an Electrical Resistivity Imager, earth resistivity profiles were determined at 40 locations covering main cities as well as various geographical regions of Saudi Arabia. For each test site, layered resistivity profile as well as 2D inverse models were derived. Fig. (4) shows an example of 2-D resistivity variation for a test site in Dammam near the Arabian Gulf.

Fig. (4): Inverse model 2D-plot for Damman 1. 263 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


Besides the 2-D and 3-D resistivity tomography, the resistivity data can be used to determine variation of average earth resistivity with depth. Different types of soils display different behaviors in this respect. Measurements show that in areas near sea or water channels (e.g. Dammam, Jeddah, Jizan, Hafoof, Khafji) the average resistivity strongly decreases with depth. In other areas, resistivity can increase with depth, it can decrease with depth or it can approximately stay the same. This mainly depends upon the geology of the test site. Fig. (5) shows same examples of resistivity variations with depth at few selected sites in the Kingdom of Saudi Arabia. Since earthing installations usually make use of vertical rods that are driven into earth for few meters only, the resistivity of earth for a 6m depth was determined at all test locations. It was found that some sites have average resistivity of as low as 2Ωm, whereas other sites had average resistivity of > 106 Ω.m. Reference [6] recommends typical guidelines for soil resistivity range classification. Using this guideline, the soil resistivity is classified as ρ ≤ 100 Ω.m. = Low, 100 Ω.m. < ρ ≤ 300 Ω.m. = Medium, 300 Ω.m. < ρ ≤ 1000 Ω.m. = High and ρ > 1000 Ω.m. = Very High. Table (1) shows some examples of measured ρ and resistivity classifications. In areas that have low resistivity, rod earthing electrodes may be sufficient for achieving satisfactory earthing installation. For areas having medium soil resistivity, it may become essential to use grounding grids. However, in areas that have high or very high earth resistivity, special measures such as use of low resistivity material surrounding the earthing electrodes will be required in many cases. Such materials are being developed and investigated at the H.V. laboratory of King Saud University so that these may be used by Saudi Electricity Company in earthing applications.


N.H. Malik. , et al

Soil Resitivity (Makkah 1) 150

Soil resitivity (Ohm - Meter)

120

90

60

30

0 3

6

9 Depth (m)

12

15

Soil Resistivity (Sabt Al-alaya)

Soil Resitivity (Dawadmi 1) 600

200 180

525 Soil resitivity (Ohm - Meter)


160 140 120 100 80 60

450 375 300 225 150

40 75

20 0

0 3

6

9 Depth (m)

12

3

15

Soil Resistivity (Mujardah)

6

9 Depth (m)

12

15

12

15

Soil Resistivity (Najran)

150

750 675

125 Soil resitivity (Ohm - Meter)


600

100

75

50

25

525 450 375 300 225 150 75

0 3

6

9 Depth (m)

12

15

0

3

6

9 Depth (m)

Fig. (5): Variation of average resistivity at selected sites in Saudi Arabia.



Table (1):

SEC Operating Areas

Central

Western

Eastern

Southern

Measured values of minimum, maximum and average soil resistivity at 6m depth.

Measurement Site Name (Location) Riyadh−1 Riyadh−3 Riyadh−4 Dwadmi−1 Dwadmi−2 Kharj−1 Kharj−2 Gaseem−1 Gaseem−2 Jeddah−4 Makkah−1 Dammam−2 Hafoof−1 Hafer-AlBatin−1 Hafer-AlBatin−2 Khafji−2 Atawala Mandaq Mikhwa Sabt-Al-Alaya Shudiq Mujardah Abu Areish Byesh Najran Sharourah

Min.

Soil Resistivity (Ωm) Max. Ave.

Soil Type

20 7105 70 22 − − 18 −

69 8026 275 98 − − 20 −

30 7658 131 46 754 439,800 19 > 629000

3 3 111 7 9 85

5 4 150 9 12 152

4 4 126 8 10 128

Hard rocky soil. Rocky soil. Sandy soil. Hard sandy soil. Soft sandy soil. Stony soil. Hard clay. Very loose sandy soil. Moist sandy soil. Sandy soil. Soft soil. Agricultural soil. Agricultural soil. Stony sand.

69

115

90

Stony sand.

17 121 357 174 257 106 71 55 41 200 −

26 161 473 297 634 266 103 98 63 1787 −

21 135 412 229 416 172 85 74 55 534 9800

Compact sand. Agricultural soil. Rocky soil. Stony soil. Sandy soil. Stony soil. Rocky soil. Mountain soil. Clay + sand. Stony soil. Very loose sand.


Average Resistivity Classification Low Very High Medium Low High Very High Low Very High Low Low Medium Low Low Medium Low Low Medium High Medium High Medium Low Low Low High Very High

N.H. Malik. , et al

CONCLUSIONS Earth resistivity measurements methods including modern Electrical Resistivity Imagers have been discussed. Using such tools, earth resistivity have been measured at 40 locations covering various geographical as well as geological regions of Saudi Arabia. The results and analysis show that Saudi soils have a very large range of earth resistivity. Stony soils and soils of very soft loose sand can have from medium to very high soil resistivity, whereas areas near sea have normally low resistivity. In areas that have higher soil resistivities, grounding network designers need care and may involve use of low resistivity materials.

ACKNOWLEDGEMENT The authors would like to thankfully acknowledge Saudi Electricity Company for sponsoring this work through SEC Research Project No. S0803-R. Thanks are also extended to Dr. Y. Khan and Eng. Nisar R. Wani for their support during this work. REFERENCES [1]

ANSI/IEEE Std. 80-1986, IEEE Guide for Safety in Substation Grounding, 1986.

[2]

ANSI/IEEE Std. 142-1982, IEEE Recommended Practice for Grounding of Industrial and Commercial Power Systems, 1982.

[3]

ANSI/IEEE Std. 141-1986, IEEE Recommended Practice for Electric Power Distribution for Industrial Plants, 1986.

[4]

NFPA-70, "National Electrical Code", 2002 Edition.

[5]

A.T. Johns and D.F. Warne (Editors), "Advances in High Voltage Engineering", Chapter 8 by H. Griffiths and N. Pilling, "Earthing", IEE Power and Energy Series 40, UK, 2005.

[6]

ANSI/IEEE Std. 81-1983, IEEE Guide for Measuring Earth Resistivity, Ground Impedance, and Earth Surface Potentials of a Ground System, 1983.

[7]

F. Wenner, "A Method of Measuring Resistivity", National Bureau of Standards, Scientific Paper, No. S-258, 1916, pp. 469.


E-READINESS ASSESSMENT CASE-STUDIES USING THE STOPE APPROACH

Abdulmohsen Alheraish2, Saad Haj Bakry3 Khalid Al-Osaimi1, College of Engineering, King Saud University 1: General Directorate of Civil Defense, [email protected] 2: College of Engineering, King Saud University, [email protected] 3: College of Engineering, King Saud University, [email protected]

ABSTRACT The STOPE-based e-readiness assessment approach is used here for practical casestudies. The approach integrates and evaluates the various e-readiness assessment issues, over its well-structured domains: strategy, technology, organization, people, and environment. The case-studies considered are concerned three Saudi organizations, which include: a "government" organization, a "bank", and a private sector "company". The results obtained provide e-readiness indicators that illustrate the strengths and weaknesses of each organization with regards to e-readiness "strategic" development issues, information and communication "technology" (ICT) issues, management and "organization" issues, "people" issues, and "environment" issues within which the organization operates. The work shows how the approach can be practically used, and it also provides practical results that direct the organizations concerned toward enhancing their e-readiness in response to the current trend of globalizing e-business.

KEY WORDS e-Readiness assessments, STOPE, organizations, Framework. 1.

Introduction

With the current trend toward globalizing e-business, e-readiness assessments are becoming of increasing importance for diagnosing related problems, and directing future development efforts [1]. A recent paper has developed a comprehensive approach for e-readiness assessment that can be applied both to countries and to organizations [2]. The approach is based on Bakry's STOPE (strategy, technology, organization, people, and environment) view. The approach identifies the e-



readiness factors, using previous studies; and it groups them into issues associated with the five STOPE domains. The factors are identified using ten important previous e-readiness assessment studies, and these studies are given in the following. • Three studies were associated with academic centers: one in the USA [3], another in Russia [4], and the third in the UK [5]; • Three studies were supported by international organizations: one was supported by the UNDP [6], another by the ITU [7], and the third by “infodev” of the Worldbank [8]; • Two studies were given by specialized firms: one by McConnell International [9], and the other by Bridges [10]; • Two studies were concerned with e-readiness assessments in firms, with each supported by the government concerned [11, 12]. This paper provides practical e-readiness assessment applications of the approach. These applications illustrate the use of the approach and clarify its benefits; and they also produce useful results to the organizations of the applications concerned. The applications are associated with three Saudi organizations: a "government" organization, a "bank", and a private sector "company". The paper presents an overview of the approach, explaining the STOPE architecture of its e-readiness issues. The applications are then described, and their results evaluated and compared, with the differences emphasized and discussed. Finally, views and comments on the future development and application of the approach are presented. 2. The STOPE Approach An illustrative representation of the e-readiness assessment STOPE approach is given in Figure 1 [2]. The Figure shows how each domain is divided into a number of issues. In the following, the main domains together with their issues and assessment factors are described. In addition, the questions of how the factors, issues and domains of the STOPE structure are assessed, and how the overall STOPE e-readiness indicator is derived, are addressed. This provides us with the necessary background to present the target case-studies in the subsequent section. The Strategy Domain The strategy domain is identified as “the directions, commitments and plans concerned with ICT development and utilization”. Table 1 presents the issues of this domain and gives the number of factors associated with each issue. A brief explanation of the factors involved is also given. It is shown that assessing this domain requires the assessment of "26 factors" distributed over "2 issues".


Khalid Al-Osaimi , et al

Figure 1: The STOPE framework for e-readiness assessment 271 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


The Technology Domain The technology domain is identified as “the current state of issues concerned with ICT facilities”. Table 2 presents the issue of this domain and gives the number of factors associated with each issue. A brief explanation of the factors involved is also given. It is shown that assessing this domain requires the assessment of "43 factor" distributed over "4 issues". Table 1: e-Readiness “strategy” issues STRATEGY “Directions, commitments and plans concerned with ICT development and utilization” FACTORS’ ISSUES FACTORS NUMBER Vision: directions priorities, initiative; Commitments: involvement of top administration, position of CIO ICT 14 (chief information officer), e-business teams; ICT Leadership managers: qualifications and responsibilities of ICT managers; Government support. Technology (ICT) plan; (ICT) Organization plan; Future (ICT) HR (Human Resources) plan; Related (non12 Development ICT) plans: associated with the cultural and business Plans environment

Table 2: e-Readiness “technology” issues TECHNOLOGY “Current state of issues concerned with ICT facilities” ISSUES

FACTORS

FACTORS’ NUMBER

Communications & Basic Information Infrastructure

Availability, services and performance: computers, fixed telephones, cellular phones, highspeed lines, Internet & Intranet (for organizations).

18

ICT e-Services Infrastructure ICT Provisioning

ICT Support

Availability, services and performance: webs, portals & services (internal e-transactions: within the organization; external: B2Bservices, B2C,..) Market: availability of ICT products; Provisioning: contracts, delivery and upgrading. Standards: availability and use of local, national and international standards; Operation & Maintenance: availability of sources of support; Performance: use of measures.


11 8

6


The Organization Domain The organization domain is identified as “the current state of issues concerned with ICT regulations and management”. Table 3 presents the issues of this domain and gives the number of factors associated with each issues. A brief explanation of the factors involved is also given. It is shown that assessing this domain requires the assessment of "35 factors" distributed over "3 issues". The People Domain The people domain is identified as “the current state of issues concerned with ICT users and skills”. Table 4 presents the issues of this domain and gives the number of factors associated with each issue. A brief explanation of the factors involved is also given. It is shown that assessing this domain requires the assessment of "18 factors" distributed over "4 issues".

Table 3: e-Readiness “organization” issues ORGANIZATION “Current state of issues concerned with ICT regulations and management” FACTORS’ ISSUES FACTORS NUMBER Basic ICT regulations: legal framework, standardization, computer crimes and software privacy, investment, competition, tariffs; Internet services ICT 12 regulations: registration of domain names, Regulations authorization of ISPs (Internet Services Providers); Ebusiness services regulations: digital signature, PKI (Public Service Infrastructure), taxation. Knowledge sharing and innovation: cooperation (industry, education & research); Partnerships & ICT 9 Cooperation services: customers, suppliers, outsourcing (local, national, international) Measures: evaluation measures; Change: response to requirements; Quality: competition, use of modern ICT 14 Management techniques; Cost & affordability: cost of facilities, cost of access, use and maintenance (relative to income).



Table 4: e-Readiness “people” issues PEOPLE “Current state of issues concerned with ICT users and skills” FACTORS’ ISSUES FACTORS NUMBER ICT literacy and acceptance; General education 5 ICT Awareness system support; Media support ICT Education & Training ICT Skills & Jobs Management of ICT Skills

ICT available qualifications; Availability of eeducation & e-learning. ICT skills: availability and need of ICT skills; ICT jobs: ICT skills in ICT & non-ICT jobs Performance: productivity of skills; Satisfaction: retaining ICT skills

5 6 2

The Environment Domain The environment domain is identified as “the current state of non-ICT issues surrounding and affecting the current state of ICT”. Table 5 presents the issues of this domain and gives the number of factors associated with each issue. A brief explanation of the factors involved is also given. It is shown that assessing this domain requires the assessment of "24 factor" distributed over "4 issues". Assessment Method The above shows that the STOPE view of the assessment has the following four main levels: • the top level is the STOPE level, which provides a single e-readiness indicator that integrates the indicators of its five domains. • the domain level, which has five domains "strategy, technology, organization, people, or environment", and which provides an indicator for each domain integrating the indicators of its issues; • the issue level, which has seventeen issues distributed over the five domains, and which provides an indicator for each issue integrating the indicators of its assessment factors; and • the bottom level is the factor level, which has "146 factors" distributed over the seventeen issues, and which provides an indicator for each factor.



Table 5: e-Readiness “environment” issues ENVIRONMENT “Current non-ICT issues surrounding and affecting the current state of ICT” FACTORS’ ISSUES FACTORS NUMBER Culture: profile, technology literacy, knowledge of 7 Knowledge English; Education: quality, science and technology, research, science parks and incubators Infrastructur Basic services: electricity, transportation, postal system, 5 e health care. Natural resources; Revenues & profitability; Trade: Resources & 4 import & export; Income & standard of living Economy Government regulations: rule of law, business opportunities; Cooperation: local, national, 8 Management international; Management: professional culture, response to technology & change, use of modern management techniques, retaining skills.

The assessment starts at the bottom level, where each “factor” is "measured", or graded, and assigned a "weight" relative to its estimated effect on the case study considered. The assessment then steps up to higher levels assessing “issues”, “domains”, and the “top level”. Table 6 shows how these levels can be assessed using the assessment of its lower levels. The Table recommends the use of a spreadsheet software for organizing the results, and the use of radar graphs for illustrating these results. 3. Practical Assessments An e-readiness assessment questionnaire, based on the above STOPE approach, has been designed considering all "146 factors", and their distribution over the issues and domains. The questionnaire was used for assessing e-readiness in three different well-established organizations in the Kingdom of Saudi Arabia (KSA). In the following the organizations concerned are introduced, and the results of the assessments are presented, with comments on the e-readiness strengths and weaknesses of each organization, considering the STOPE domains and their associated issues and factors.



Table 6: Levels and method of assessment STOPE (TOP) LEVEL MAIN Strateg DOMAIN y SYMBOL S i=1 INDEX M[1] MEASURE w[1] WEIGHT STOPE ASSESSMENT

Technology

Organization

People

Environment

T i=2 M[2] w[2]

O i=3 M[3] w[3]

P i=4 M[4] w[4]

E i=5 M[5] w[5]

=

STOPE

i=5

∑

w [ i ]. M [ i ] =

i =1

i=5

∑

D [i]

i =1

DOMAIN LEVEL: S, T, O, P, or E Any of the five main STOPE domains, identified as: D [i] DOMAIN i: gives the domain index: 1 = < i < = 5 (STOPE) INDEX ji represents the sub-domain (issue) index: 1 ≤ ji ≤Ji M [i, ji] WEIGHT w [i, ji] MEASURE DOMAIN ASSESSMENT

D [i] =

ji = J

∑

i

ji = 1

w [ i , j i ]. M [ i , j i ] =

ji = J

∑

ji = 1

i

D [ i , ji ]

SUB-DOMAIN (ISSUE) LEVEL Any of the issues of a domains, identified as: D [i, ji] ISSUE kji: represents the factor index of domain i and issue ji: INDEX 1 ≤ kji ≤ Kji (Kji: number of factors in issue j of domain i) M [i, ji, kji] w [i, ji, kji] MEASURE WEIGHT k

ISSUE ASSESSMENT

D [i, ji ] =

j i

= K

∑

k

ji

= 1

j i

w [i, ji , k

j i

]. M

[i, ji , k

j i

ASSESSMENT METHOD: BOTTOM-UP

SCALE

The evaluation (assessment) starts at the "factor" level, followed by the "issue" level, the "domain" level, up to the STOPE level which provides the overall e-readiness indicator. MEASURE Four possibilities: 0 to 3 WEIGHT INDICATOR

RESULTS

TABLES: SPREADSHEETS RADAR GRAPHS

Five levels: 0 to 4 Relative to 100 Numerical results for factors, issues, domains, with measures, weights and indicators, leading to the overall STOPE indicator. Domains represented by their issues. STOPE represented by the domains



The Organizations Considered The organizations considered include: a Saudi "government organization"; a Saudi "commercial bank"; and a Saudi private sector "company". The three organizations have the following common features: • more than "500 employees"; • more than "100 branches", distributed over the regions of the country; and • more than 20 years in business. All three organizations have an IT department. The IT managers of the government organization, and of the private sector company, have answered the e-readiness assessment questionnaire; while the IT manager of the commercial bank's central region was the one who answered the questionnaire.

Government

Bank

IT Leadership

Company

IT Leadership

73

80

77 Future Development Plans

IT Leadership

84

55

61 Future Development Plans

Future Development Plans

Figure 2: "Strategy" domain assessment results

The Strategy Domain Figure 2 illustrates the results obtained for the e-readiness assessment issues of the "strategy" domain. The results show the following: • the "bank" enjoys the highest scores in both: "ICT leadership" and the future “development plans”; these scores are 84 % and 80 % respectively; • the "government" organization has higher scores than the private sector "company" in both issues; • a common problem found in both the "government" organization and the private sector "company" is limited IT funding; • unlike the other two organizations, the president of the "company" does not directly receive e-readiness reports from his e-business team; • an important strength associated with the "bank" is giving "priority" to ereadiness. The overall "strategy" domain, non-weighted, scores for the organizations concerned are: 75 % for the "government" organization, 82 % for the bank", and 58 % for the private "company".



Government

Bank

ICT Basic Infrastructure

Company ICT Basic Infrastructure

ICT Basic Infrastructure

73

89 56

ICT e-services Infrastructure

61

ICTICT e-services Support Infrastructure

ICT e-services ICT Infrastructure Support 100

69

72 60 ICT Provisioning

33 0

ICT Support 94

63 74 ICT Provisioning

ICT Provisioning

Figure 3: "Technology" domain assessment results

The Technology Domain Figure 3 gives the results obtained for the e-readiness assessment issues of the "technology" domain. The results illustrate the following: • the "bank" has the highest scores in this domain, including in: "ICT basic infrastructure", "ICT e-services infrastructure", "ICT provisioning", and "ICT support", where the scores are 89 %, 69 %, 74 % and 100 % respectively; • the "government" organization has the least score in "ICT basic infrastructure", "ICT provisioning ", and in "ICT support"; • the "company" has the least score in "ICT e-services infrastructure"; • the gap is not so big among the three organizations, with regards to "ICT provisioning". The overall "technology" domain, non-weighted, scores for the organizations concerned are: 62 % for the "government" organization, 84 % for the bank", and 66 % for the private "company". The Organization Domain Figure 4 provides the results concerned with the "organization" domain. The results show the following. • the "government" organization and the "bank" have close scores with regards to "ICT regulations", "ICT cooperation" and "ICT management"; • with regards to "ICT management", the "company" also has a close score to the other two organizations; • the "company" enjoys highest score in "ICT cooperation", but it has the least score in "ICT regulations".



The overall "organization" domain, non-weighted, scores for the organizations concerned are: 70 % for the "government" organization, 72 % for the bank", and 67% for the private "company".

Government

Bank

Company ICT Regulations

ICT Regulations

ICT Regulations

100

72

71

51

0

69

70

70

ICT Cooperation

68

91

74

ICT Cooperation ICT Management

ICT Management ICT Cooperation

ICT Management

Figure 4: "Organization" domain assessment results

Government

Bank

ICT

Company

ICT Education & Training

ICT

ICT 67

56

65

80 Skills ICT Education 79 67 Management & Training

Skills Management

80 SkillsICT Education & Training Management

17

100 96

67

61

ICT Qualifications & Jobs ICT Qualifications & Jobs

ICT Qualifications & Jobs

Figure 5: "People" domain assessment results

The People Domain Figure 5 shows the results of the "people" domain. The results illustrate the following: • in "ICT awareness" the bank has the highest score of 67 %; • all three organizations receive close scores in "ICT education and training", that is around 80 %; • in "ICT qualifications and jobs", the "government" organizations has the highest score of 96 %; • in "ICT skills management", the bank scored 100 %, while the company received only 17 %. The overall "people" domain, non-weighted, scores for the organizations concerned are: 77 % for the "government" organization, 78 % for the bank", and 58 % for the private "company". 279 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


The Environment Domain Figure 5 gives the results of the "environment" domain. The results show the following: •

the "bank" enjoys the highest scores in "knowledge" and in general "infrastructure", where these scores are 82 % and 63 % respectively;

•

the "company" enjoys the highest score in "resources and economy" at 69 %;

•

both the "company" and the "bank" have equal scores for "management", at around 64 %;

•

the company has the least score in the general "infrastructure, that is around 49%, while the "government" organization is behind the high scores of all issues, but not by far. The overall "environment" domain, non-weighted, scores for the organizations concerned are: 64 % for the "government" organization, 70 % for the bank", and 66 % for the private "company".

Government

Bank

Knowledge

Company

Knowledge

Resources & Economy

Knowledge 80

82

76 58 59

Infrastructure Resources & Economy

67 0

63

Resources 69 Infrastructure & Economy

61 64 Management

Management

Figure 6: "Environment" domain assessment results


Infrastructure 49 64 Management


Table 7: The STOPE weighted indicator

Domai n S T O P E Indicat or

Government M W IND. (0(of 1) (%) 3) 2.24 0.2 75 1.85 0.2 62 2.09 0.2 70 2.29 0.2 77 1.91 0.2 64

Bank

Company

M (0-3)

W (of 1)

IND. (%)

M (0-3)

W (of 1)

IND. (%)

2.45 2.51 2.16 2.34 2.09

0.24 0.19 0.19 0.19 0.19

82 84 72 78 70

1.75 1.99 2.02 1.73 1.97

0.2 0.2 0.2 0.2 0.2

58 66 67 58 66

69

77

63

The STOPE Indicator The results presented above are associated with the “issues” of the STOPE “domains”. These results are concluded from those taken from the answers of the questionnaire for the “factors” of the issues, that is the level below these issues, using the method of Table 6. In addition, the results can also be represented at the STOPE level, by collecting the results of each of the STOPE domains, from the above, and using the method of Table 6. Table 7 gives the results concerned with the STOPE level, and Figure 7 illustrates these results. The overall STOPE ereadiness indicator, for each organization, is also given. For the "bank" the indicator is 77 %; for the "government" organization it is 69 %, and for the private sector company, it is 63 %. These results are illustrated in Figure 8.

Government

Bank

STRATEGY

Company STRATEGY

STRATEGY 82

75

58

TECHNOLOGY ENVIRONMENT

TECHNOLOGY

64

62

70 ORGANIZATION

77

ENVIRONMENT TECHNOLOGY

84

PEOPLE ORGANIZATION

70

72

78

ENVIRONMENT

66

67 PEOPLE

66

0

58

ORGANIZATION

PEOPLE

Figure 7: STOPE assessment results



Government 69

77 Bank

63

Company

Figure 8: The overall STOPE e-readiness indicators for the three organizations.

4. Discussions and Conclusions In this paper, the STOPE e-readiness, multi-level, assessment approach has been used for three Saudi organizations: a "government" organization; a "bank"; and a "company". Four assessment levels were involved: the e-readiness "factor" level, which is the lowest level; the group of related factors or "issue" level; the group of issues or "domain" level, which involves the five STOPE domains; and the "overall" STOPE level that considers all the domains as an integrated whole. The presented results did not go as deep as showing the "factor" level, as this would require a large space, not usually available to a single paper, since there are over "146" factors involved. However, the factor level assessment has been performed for providing the given results at the "issue" level, and then at the "domain" level and at the "overall" STOPE level. The presented results have given the obtained score for each of the "17" issues that include: "2" strategy issues; "4" technology issues, "3" organization issues; "4" people issues; and "4" environment issues. Out of assessing the "17" issues, for the three organizations, only the issue, of the "people" domain, concerned with "ICT skills management" received the score of 100 %; and this was only for the "bank". The rest of the scores ranged from 17 % to 94 % illustrating the state of the issue in each of the organizations concerned, thus indicating where future e-readiness development should focus. It is hoped that the work presented in this paper would be used in the future for other practical e-readiness assessment case studies, not only concerned with 282 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


organizations, but also with countries. Following the STOPE integrated approach in future assessments will ease evaluations and comparisons, and would establish an assessment standard that can be periodically updated and upgraded depending on development in ICT and in the applications of ICT.

5. References 1.

Bakry SH. Transformation to the Knowledge Society; King Abdulaziz Public Library: Riyadh, Saudi Arabia, 2006 (in Arabic). 2. Al-Osaimi K, Alheraish K, Bakry SH. "An integrated STOPE framework for ereadiness assessment", Proceeding of the 18th National Saudi Computer Conference, March 2006, pp. 19-29. 3. Geoffery S, Carlos A, and Jeffrey D. The networked readiness index: measuring of nations for the networked world. Center for International Development at Harvard University 2002; Massachusetts, USA. 4. Sergey S. Russia e-readiness assessment. Institute of Information Society, Moscow 2004; Russia: www.russia-gateway.ru/index.php. 5. Ali Al-Solbi, Evaluating and improving e-readiness assessment methods and tools: Questionnaire will be distributed in the Kingdom of Saudi Arabia, University of East Anglia 2004; Norwich, UK. 6. ICT for Development. The United Nations Development Program 2004: www.undp.org. 7. World telecommunication development report. International Telecommunication Union (ITU) 2003; Geneva, Switzerland. 8. Mohsen A, Michael F, Bruno D, Vivek C. INDIA: E-Readiness Assessment Report, Thompson Press, InfoDev, Worldbank 2003: www.infodiv.org. 9. Ready net go: partnerships leading the global economy. McConnell international 2002: www.mcconnellinternational.com. 10. e-Readiness assessment: who is doing what and where: an open content report. Bridges 2002: www.bridges.org. 11. The wisdom exchange e-business readiness assessment. Leading Growth Firms Series 2001; Queen's Printer for Ontario, Ontario, Canada: www.ontariocanada.com. 12. Dawn J, Peter B, and Jasbir D. Government support for e-readiness of small and medium sized enterprises(SMEs). Proceedings of the 35th Hawaii International Conference on System Sciences 2002; IEEE Computer Society.


PREPARATION OF FUTURE PRACTICAL ENGINEERS - CASE STUDY: LABORATORY EXPERIMENTS ON POWER CABLES 1

2

3

M.S. Al-Saud , M.A. El-Kady , R.D. Findlay 1: McMaster University, Hamilton, Canada, [email protected] 2: King Saud University, Riyadh, Saudi Arabia, [email protected] 3: McMaster University, Hamilton, Canada, [email protected]

ABSTRACT Future practical engineers in the Kingdom of Saudi Arabia have to acquire the know-how as well as sufficient field and experimental skills to enable them to succeed and excel in the job market. In addition, collaboration and cooperation with other international universities and training institutions are also very valuable in acquiring advanced technologies and learn from the existing world-wide experiences in this area. This paper presents the results of a case study in which international cooperative efforts were made to build a practical laboratory experimental setup at King Saud University and conduct advanced measurements for the purpose of assessing the performance of underground Power cables used to transmit electrical energy to municipal and rural consumers. It is well known that the cable performance in practice is critically influenced by thermal properties of the medium in which it is placed, as well as the thermal properties of the cable it self. The heat generated by an electrical power transmitting conductor has to dissipate through the cable insulation and surrounding backfill. The current carrying capacity of buried cables depends to large extent on thermal conductivity of the surrounding medium. In fact results reported in previous publications indicate that the sensitivity of cable temperature to variation of thermal conductivities of the surrounding medium is much greater than the effect caused by variation of other parameters. The results of the paper are believed to be extremely important in demonstrating how international cooperation can be useful in establishing practical testing facilities in the Kingdom for better training and preparation of skilful and qualified engineers.

KEY WORDS Underground cables, ampacity, laboratory measurements, finite-elements, thermal field.

1. INTRODUCTION Future practical engineers in the Kingdom of Saudi Arabia have to acquire the know-how as well as sufficient field and experimental skills to enable them to


PREPARATION OF FUTURE PRACTICAL ENGINEERS - CASE STUDY: LABORATORY

succeed and excel in the job market. In addition, collaboration and cooperation with other international universities and training institutions are also very valuable in acquiring advanced technologies and learn from the existing world-wide experiences in this area. An example of such international cooperative efforts is to utilize the existing international experience and expertise in building advanced practical laboratory experimental setups. One of the most important engineering fields in Saudi Arabia's power industry is the underground power cables, which are used extensively over vast areas of the Kingdom. In this regard, the determination of thermal behavior of underground power cable plays an important role in the design and manufacture processes of the cables as the heat generated inside the cable may lead to the cable break down [1]. The use of various material in the composition of the cables and the backfill in contact with them under certain conditions of use, can give rise to temperature increase above levels that the cable insulation can withstand without deterioration. As a consequence, under normal use, these installations are used below their real load possibilities. However, given the high cost of such installations, it would be useful to make optimum use of them so that the maximum possible current can circulate without exceeding the temperature limit for insulation deterioration. Consequently, it is necessary to know the temperature distribution around buried cables and situations that can arise during use as accurately as possible [2]. In the field, soil temperature and moisture changes with depth as well as with seasons. Thus, a one time on-site measurement gives a set of parameters true only for that location and time. However, with a through laboratory analysis of the undisturbed sample one can simulate other conditions and their influence. This paper presents the results of a case study in which international cooperative efforts were made to build a practical laboratory experimental setup at King Saud University and conduct advanced measurements for the purpose of assessing the performance of underground Power cables used to transmit electrical energy to municipal and rural consumers. It is well known that the cable performance in practice is critically influenced by thermal properties of the medium in which it is placed, as well as the thermal properties of the cable it self. The heat generated by an electrical power transmitting conductor has to dissipate through the cable insulation and surrounding backfill. The current carrying capacity of buried cables depends to large extent on thermal conductivity of the surrounding medium. In fact results reported in previous publications indicate that the sensitivity of cable temperature to variation of thermal conductivities of the surrounding medium is much greater than the effect caused by variation of other parameters. The laboratory experimental setup was developed in order to investigate the soil thermal specifications of any soil type. The thermal properties of a soil sample used as backfill in the underground power cables in Saudi Arabia were investigated using the developed laboratory testing facility as reported in this paper. 286 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007

M.S. Al-Saud , et al

The laboratory test facility includes a 154×111×124 cm main container unit of an isolated wooden box, a heat source along one of its sides and a heat sink along the opposite side face to face with each other, and a set of sensitive thermocouple probes and electrical measurement devices. There are 48 probes (45 distributed in the surrounding soil in addition to three more out of the thermal model, two of which are attached to the container wall to observe the ambient temperature for the isolated boundary and one at the top of the roof to observe the ambient for the convectional boundary. In addition to the 48 probes, there is one in the middle of the cable insulation and another on the cable surface. The total number of probes used in this experiment is therefore 50. The results of the paper are believed to be extremely important in demonstrating how international cooperation can be useful in establishing practical testing facilities in the Kingdom for better training and preparation of skilful and qualified engineers. 2. LABORATORY EXPERIMENTAL SETUP A field test facilities were required to simulate the operation of buried electrical transmission cables includes: 1.

The main unit is the container, which was filled with the soil containing the tested cable. It is a specific wooden box having approximate dimensions of 154×111×124 cm (W×H×D) placed in the laboratory and designed to be portable, with a removable roof to allow the accessibility to the box internal contents, as shown in Figures 1 ,2 and 3.

2.

Thermal specification of the used soil was tested by stimulating high gradient along the container width using heat-source heat-sink provided in the container design.

3.

Actual cable samples XLPE 15 KV was laid out inside the container. Cable was placed in the soil filled container. Then the cable was subjected to a certain load (300A and 600A) in order to examine the thermal behavior inside the cable and at the surrounding of the cable.

4.

High current transformer connected through circuit breakers was used to load the tested cable. The current transformer was used to step up the supply current to the required high current level (300A and 600A) needed to test the cable.

5.

Sensitive thermocouple probes (50 probes) were used to record the temperature of specific points of the cable (external surface, insulation) as well as the surrounding soil and ambient. The thermocouple probes were distributed inside the container in three different similar layers each with a different height level. Probes are suspended on the intersections of the shown insensitive grid in Figure1. 287



Heat Sink

Tested Cable

Compressor

Heat Source

Suppl

Supply Transformer

(a) Soil box configuration

Cable Heat Sink

Heat Source

~220V

Soil Box

Probes

1/12.5 ~220V

b) Schematic component diagram Fig. 1 Experiment set up and system configuration 288 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


154 53 111 27.5

32.4

22.3 77 Figure 2 System container front views ١٥ ٣٣٫٥ ٢٨٫٥

١٢٤

٢٢٫٣ ٣٢٫ Fig. 3 System container top views

3. EXPERIMENTAL MODEL AND SIMULATIONS After completing the model of experimental set up, embedding the tested cable in the container, filling with tested soil and shorting the tested cable (loop configuration) - to be connected as secondary side of the transformer, which is connected through the primary to the supply- the cable was loaded at 300A and the cable insulation, cable surface, ambient and other grid probe temperatures were recorded continuously several times per a day over 570 hours before the steady state was achieved. Then, the cable load was increased to 600A and observed daily over 230 hours before the steady state was achieved at the new operating conditions to complete 800 hours of continuous experiment run. The temperature observation of the cable insulation is shown in Figure 4. The proposed finite element algorithm is applied to calculate detailed and accurate schemes of cable temperatures and the 289 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


surrounding soil environment in ordered to be compared with the recorded experimental readings captured by the insensitive grid of the used thermo-couple props. The experimental setup shown in Figure 1 was simulated using finite element grid consisting of 1,116 nodes which includes the internal cable material portions, used filled soil, and the container boundaries. Figure 5 shows the used finite element grid and illustrates positions of the mid layer probes used for the comparison. As it can be noted from Figure 1, container body is constructed to maintain as much as possible the heat generated inside by the mean of insulating its body edges. Consequently the sides and lower edges will be simulated as heat flux boundaries. The perfect insulating or a zero heat flux is difficult to implement in the laboratory and a tolerance of a heat losses will be allowed thought these insulating edges. The upper roof of the container is open and is adjacent to the ambient environment and will be modeled as a convective boundary condition.

Fig. 4 Selected probes temperatures and cable current against time



Fig. 5 Simulation finite elements grid Unfortunately, the thermal behavior of the cable is strongly dependent on the loading conditions, thermal parameters of the cable materials as well as thermal specifications of the soil surrounding, ambient environment and boundary conditions. In practices, the cable loading is monitored and observed during the experiment run while the cable specification is provided by manufacturers. On the other hand remaining parameters, which associated with the surrounding, are either difficult to obtain or can be subjected to variations which may affect the cable thermal field predictions. In this regard, the thermal conductivity which may be considered as one of the most dominant effaceable part in the thermal circuit [3] was tested and investigated precisely in the laboratory before the tested cable was loaded using the proposed heat source-heat sink temperature gradient stimulator constructed in the container design as can be seen from Figure 1. The theoretical background for the experimental set up, methodology and analysis of results are provided in [4-7]. The heat coefficient losses at the convective boundaries and the heat losses at the isolated boundaries were investigated using the proposed combined finite element- gradient optimization method, which is based on matching the computed thermal field to that obtained from the experimental measurements. 291 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


Due to the mentioned thermal circuit parameters investigation methodologies, the container soil surrounding the cable are divided into three sub regions as can be seen from Figure5, each with a specific thermal resistivity. In practice, when the cable is loaded, the soil around the cable affected strongly by the intensive thermal field more than the far soils, consequently, more deviation in their thermal resistivity will be expected. The finite element-gradient optimization method presented in this paper allows more flexible and precise modeling of the circuit thermal parameters, including the thermal resistivities of the sub-divided soils, heat loses at the insulated boundaries and the heat convection losses factor, as shown in Table 1. Figure 6 shows the thermal field temperature solution using the proposed finite element mesh. On the other hand, Table 2 shows a comparison between the recorded experimental temperatures and the calculated values from the above simulation. The cable surface and insulation temperatures as well as the other probes reading distributed in the surrounding around the cable is shown in Figure 7 for both the measured and computed cases. It is observed that temperatures distributions, at different points on the cable surface and cable insulation as well as other distributed probes around the cable, matches the associated simulation results with high accuracy, especially at the cable media and most of the distributed point at the surrounding soils. Some points have more deviation but with less than 1.5 %, these probes are subjected to move from their specific coordinate (used in simulation) during filling the container with the soil and may cause such errors. TABLE 1: EXPERIMENTAL MODEL PARAMETERS SPECIFICATIONS Conductor cross section area (mm2) XLPE insulation thickness (mm) Rated voltage (Kv) Conductor losses (W/m) Dielectric losses (W/m) Shield losses (W/m) Conductor thermal conductivity W/( oC. m) Dielectric thermal conductivity W/( oC. m) Soil cable adjacent thermal conductivity W/( oC. m) Soil mid layer thermal conductivity W/( oC. m) Soil outer layer thermal conductivity W/( oC. m) Ambient temperature oC Heat convection losses coefficient W/( oC. m2) Left boundary heat losses (W/m) Right boundary heat losses (W/m) Bottom boundary heat losses (W/m)


185 4.5 15 13.5741 0.0963 0.6787 385 0.2857 0.4700 0.4703 0.4952 24.0786 11.9634 -0.8797 -1.6668 -2.1985


Fig. 6 Thermal field solutions



TABLE 2: COMPARISON OF THE PROBE EXPERIMENTAL READING WITH THE SIMULATION RESULTS

Probe group

Probe No.

Measured value o C

Calculated value o C

Difference o C

Percentage difference (%)

1

27.30

27.528

-0.2284

-0.836

2

29.30

29.392

-0.0928

-0.316

3

30.20

30.184

0.0158

0.052

4

28.30

28.376

-0.0767

-0.271

5

27.20

26.896

0.3039

1.117

6

28.60

28.864

-0.2645

-0.924

7

32.60

32.436

0.1631

0.500

8

36.50

36.215

0.2842

0.778

9

28.20

27.952

0.2473

0.876

10

28.30

28.368

-0.0685

-0.242

11

29.90

29.916

-0.0162

-0.054

12

30.20

30.402

-0.2021

-0.669

13

29.10

29.155

-0.0554

-0.190

14

27.70

27.734

-0.0346

-0.125

Insulation

15

44.10

44.243

-0.1431

-0.324

Cable Surface

16

46.0

45.885

0.1146

0.249

Classification

Upper layer

Grid Probes

Middle layer

Lower layer

Cable Probes



Fig. 7 Probes temperatures measured against calculated

4. CONCLUSIONS An example of the fruitful collaboration and cooperation with other international universities and training institutions in the preparation of future practical engineers in the Kingdom of Saudi Arabia was presented in this paper, where such international cooperative efforts were made to utilize the existing international experience and expertise in building advanced practical laboratory experimental setup. The experimental setup pertains to one of the 295 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


most important engineering fields in Saudi Arabia's power industry, namely underground power cables, which are used extensively over vast areas of the Kingdom. In this paper, a full size experimental setup was constructed and instrumented in order to analyze the behavior of the underground cables directly buried in sand and soil when loaded. In this experiment, in addition to the cable probes, an insensitive grid of thermocouple was embedded all around the cable in the surrounding soil to study the thermal circuit behavior including surrounding cable medium. The developed finite element thermal model used to simulate the experimental model, and the results match to an acceptable accuracy the experimental results.

REFERENCES [1]

I. Kocar and A. Ertas, 2004, "Thermal analysis for determination of current carrying capacity of PE and XLPE insulated power cables using finite element method," Proceeding of 12th IEEE Mediterranean Electrotechnical Conference, MELECON-2004.

[2]

C. Garrido, A. F. Otero, and J. Cidras, , 2003, "Theoretical model to calculate steady-state and transient ampacity and temperature in buried cables", IEEE Transactions on Power Delivery, Vol. 18, pp. 667-678.

[3]

M. A. El-Kady, 1984, "calculation of the sensitivity of power cable ampacity to variations of design and environment parameters", IEEE Transaction on power apparatus and Systems, Vol. PAS-103, pp. 20432050.

[4]

M. S. Al-Saud, M. A. El-Kady, and R.D. Findaly, 2006, “Application of finite-element sensitivities to power cable thermal field analysis”, Proceeding of IASTED International Conference on Power and Energy Systems (EuroPES, 2006), Rhodes, Greece, June 26 -28.

[5]

M. S. Al-Saud, M. A. El-Kady, and R.D. Findaly, “Accurate assessment of thermal field and ampacity of underground power cables”, CCECE06 Conference, Canada May,2006.



[6]

M. S. Al-Saud, M. A. El-Kady, and R.D. Findaly, 2006, “Power cable thermal field sensitivities using finite-Elements”, Proceeding of IASTED International Conference on Power and Energy Systems (PEA 2006), Gaborone, Botswana, September 11-13.

[7]

J. A. Williams, J. H. Cooper, T. J. Rodenbaugh, G. L. Smith, and F. Rorabaugh, 1999, "Increasing cable rating by distributed fiber optic temperature monitoring and ampacity analysis", Proceeding of Transmission and Distribution Conference (IEEE-1999).


AUTOMATED OIL PALM FRUIT BUNCH GRADING SYSTEM USING DENSITY OF COLOR (RGB)

1

Meftah Salem M Alfatni, 1Abdul Rashid Mohamed Shariff, 1Helmi Zulhaidi Mohd Shafri, 1Osama M. Ben Saaed, 2Omar M. Eshanta and 2Mohammad Abuzaed 1 Faculty of Engineering, Universiti Putra Malaysia, 43400, Serdang, Selangor, Malaysia. 2 Faculty of Engineering, Universiti Kebangsaan Malaysia, 43600, Bangi, Selangor, Malaysia. e-mail:[email protected]

ABSTRACT: This research deals with the ripeness grading of oil palm fruit bunches. The current practice in the oil palm mills is to grade the oil palm bunches manually using human graders. This method is subjective and subject to disputes. In this research, we developed an automated grading system for oil palm bunches using the RGB color model. This grading system was developed to distinguish between the three different categories of oil palm fruit bunches. The maturity or color ripening index was based on different color intensity. Our grading system employs a computer and camera to analyze and interpret images equivalent to the human eye and brain. The colors namely red, green and blue (RGB) of the palm oil fruit bunch were investigated using this grading system. The computer program developed and used the mean color intensity to differentiate between the different color and ripeness of the fruits such as oil palm FFB. The program results showed that the ripeness of fruit bunch could be differentiated between different categories of fruit bunches based on RGB intensity.

Keywords: Automated, Grading System, Ripeness, Density of color and Oil Palm Bunch.


Meftah Salem M Alfatni. , et al INTRODUCTION: The oil palm grading system uses the fruit color to determine the ripeness of fruit. Digital image processing will be main method to retrieve the sample without any physical experiment on the sample. The data can be obtained by looking at the image or picture of the product with the help of software that converts the image into digital number (DN). The color is good indicator of the ripeness of palm oil fruits. That is with comparing the digital number (DN) of the samples taken by the digital camera with the available standards. Color is one of the most significant criteria related to fruit identification and fruit quality. The color of an object is determined by wavelength of light reflected from its surface. In biological materials the light varies widely as a function of wavelength. These spectral variations provide a unique key to machine vision and image analysis. They play a role in substituting human labour with precision with the use of machine vision, NIR analysis, and automation technologies. LITERATURE REVIEW: The grading system which gives us many kinds of information such as size, color and shape. Many grading systems have been developed and practically used for fruits and vegetables. Many researches explored to establish relationship between object shape and its boundary values in Fourier spectrum. Sudhakara et al, (1999) developed an on-line apple grading system, he also present some new approaches using correlation techniques, graphical analysis of radius and area signature, directional change of contour and boundary Fourier coefficient to extract the shape. Devrim and Bernard, (2004) provided comparisons of several feature selection algorithms and classifiers. Computer vision based automatic quality sorting of apple fruits is a hard but necessary task for increasing the speed of sorting as well as eliminating the human error in the process. In this field Unay and Gosselin, (2004) classified 'jonagold' apples of varying size and types to pre-determined categories as correctly and as quick as possible. (Yang et al, 2000) developed an (ANN) artificial neural network to classify images taken from the field and detect the presence of weeds. The recent development and application of image analysis and computer machine vision in sorting of agricultural materials and products in the food industries was represented by (Raji and Alamutu, 2005). Basic concepts and technologies associated with computer vision, a tool used in image analysis and automated sorting is highlighted. Fluorescence techniques have shown great potential for detecting animal feces on foods. [Moon et al, 2005] developed field portable multispectral fluorescence imaging system to use in acquire steady-state fluorescence images of feces 300 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


contaminated apples. Polder and et al (2002) recorded and analyzed spectral image of five ripeness stages of tomatoes. Andrew et al, (2003) reported that the evolution of trichromatic colour vision by the majority of anthropoid primates has been linked to the efficient detection and selection of food, particularly ripe fruits among leaves in dappled light. Unay and Gomez, (2004) reported that Color in tomato is the most important external characteristic to assess ripeness and post harvest life, and is a major factor in the consumer’s purchase decision Saad et al, (2003) determined the color band from image analysis that best correlate against the oil content. Rashid et al, (2002) investigated the correlation between oil content in the oil palm fruit against the color of the oil palm fruit. This study helps in increasing the efficiency of quality harvesting and grading of oil palm fresh fruit bunches (FFB). Sudhakara et al, (1999) developed an on-line apple grading system based on some of the most important external parameters including the fruit’s surface color. The specially developed C++ system software package collects the scene for further processing and analysis using RGB mode. Wan Ishak et al, (2000) investigated the use of a machine vision system to identify the oil palm fruit bunch by color and trigger a signal to enable the robot to pick the bunch to differentiate between the colors of the oil palm bunch with the other objects available in the oil palm plantation.


Meftah Salem M Alfatni. , et al

METHODS AND MATERIALS: This study is confined to the whole bunch of an oil palm fruit. The samples of this study are based on three categories of ripeness of fruits which are ripe, under ripe and over ripe. The images covered the whole sides of the fruit by using 2 cameras to get 2 different images simultaneously. This study did not involve the analysis of texture and the shape of the oil palm bunch. Figure 1 shows the grading system setup (Raji and Alamutu, 2005) and figure 2 shows the integration of software and hardware.

Output

Camera Light Computer

Sample

Figure 1 Fruit grading system setup

The instrumentation for the grading system can be divided into hardware and software. The hardware (Figure1) consists of 2 cameras. These cameras are connected with computer by cables. The cameras were used to capture images of the whole bunch. The software consists of MATLAB version 7 running on windows XP OS. The MATLAB software was used to perform analysis on the colors of the images.



Palm oil Fruit Grading System

Computer + MATLAB 7

Pictures Taking

NO

Training ok

YES

Testing

Samples

Digital Number (DN)

Mean of Red, Green and Blue

Standard range of

2 Pictures

Digital Number (DN)

Mean of Red, Green and Blue Ripeness testing

Ripe, Over ripe and under ripe

Decision making

Figure 2 Integration of software and hardware 303 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


Data Collection: Image data of oil palm bunches are collected from a plantation. Two pictures for each maturity category (ripe, under ripe and over ripe) were taken simultaneously from opposite sides for each oil palm bunch. These images are stored in the hard disk of a high speed PC. Sample the original data for each bunch is shown in Figure 3.

Figure 3 The original image for each bunch

Noise Removal: Color density is analyzed to determine the ripeness of each bunch. However, the background color of the images may mislead the RGB color density of each bunch. For this reason, we need to work on the bunch without its background color. This is done by removing the background color. Figure 4 shows the images after the background has been changed to white color.

Figure 4 The data without background for each bunch



The images are now ready to be analyzed. The DN value of each bunch can be obtained and ripeness of the bunch can be known. Training Stage: In this project the computer program was written using MATLAB 7 to analyze the RGB data of the images. First of all, the program reads the data from the specific file and changes it to a DN matrix. The computer program used the mean color intensity to differentiate between the different colors or ripeness of the fruits by using these 3 formulas. Mean of Red = R / number of pixels Mean of Green = G / number of pixels Mean of Blue = B / number of pixels

(1) (2) (3)

Where R = Red pixel, G = Green pixel and B = Blue pixel This step was individually performed on each image of the oil palm bunch. The mean for the whole bunch is obtained by calculating the mean for both images. This procedure is repeated with all the samples which we used them as training samples for our program. The color of oil palm FFB was classified as ripe, over ripe and under ripe. After that we obtained the range value (minimum and maximum mean) of RGB density for each category (ripe, over ripe and under ripe). This range value is used as the reference and standard value of ripeness for our test program. Testing Stage: In this step the computer program calculates the mean color intensity for RGB which is related to picture 1 and picture 2 individually. After that calculating of the mean color intensity of RGB for the whole bunch is performed and the mean of picture 1 and 2 computed. Next, the computer program test for the ripeness of the bunch based on the standard mean of the ripeness according to the training program. This is done by using IF statement. Finally, the ripeness is obtained and the result displayed. RESULTS AND DISCUSSIONS: In this section, we discuss about the results obtained from training program and the testing program which is dependent on the training program.


Meftah Salem M Alfatni. , et al Training Program: The computer program calculated the mean color intensity to differentiate between the different colors or ripeness of the fruits. This method is done on 30 pictures for 15 samples. This represents 5 samples for each category (ripe, under ripe and overripe). The mean RGB for each category of the oil palm fruit bunch sample maturity is shown in Table 1, 2 and 3. Table 1 Mean RGB intensity of the sample for ripe category Samples 1 2 3 4 5

Red 88 93 87 97 75

Mean of RGB Green Blue 38 50 43 56 25 41 50 61 51 56

Table 2 Mean RGB intensity of the sample for under ripe category samples 1 2 3 4 5

Mean of RGB Red Green Blue 60 33 39 60 33 39 55 30 36 60 35 40 60 28 33

Table 3 Mean RGB intensity of the sample for over ripe category samples 1 2 3 4 5

Red 47 58 67 53 50

Mean of RGB Green Blue 35 33 35 44 43 52 40 46 42 45



After obtaining the RGB intensity of the samples for each category we compared them with each other in order to calculate the range of RGB intensity of the oil palm fruit bunches for each category as shown in the Table 4. These ranges are placed in the testing program as reference of the ripeness for testing of bunch.

Table 4 The range of RGB intensity of the sample of categories.

category Ripe Under ripe Over ripe

RGB intensity range Red Green Blue min max min max min max 75 97 25 51 41 61 55 60 28 35 33 40 47 67 35 42 33 52

Testing Program: This part explains two stages which are the program’s result and the program’s interface.

The Program’s Results: The computer program used the mean color intensity to differentiate between each ripeness category. Test program is executed in order to show steps of the life cycle for the data inside the program. First of all, the program reads the data and changes it into DN and work with it as matrix of DN. In order to eliminate the effects of the white color background of the pictures, the program will trace the matrix of DN and use its function to convert the white color (255) to black color (0). Because the RGB of white color equal to 255 with high effect, but the RGB of black color equal to 0 with no effect. The results of program will show the original image with white background and converted the white background into black background as shown in the Figure 5 respectively.



Figure 5 The original image and Converted image with black background

The program checks image1 and image 2, pixel by pixel by using FOR LOOP function in order to calculate the mean (mean) of color components (RGB) for each image. The mean of RGB for each picture has already been obtained through the procedure that has been described. The mean of RGB for the whole bunch is obtained by calculating the mean value for the two pictures which are shown in the Table 5. After we obtained the mean of RGB for each color, the program will show us comparison between the means RGB of the colors for whole bunch as shown in the Figure 6.

Figure 6 Comparisons between the means RGB of the colors for whole bunch. Finally, the computer program will test the ripeness of our bunch depending on the standard mean of the ripeness. 308 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


Table 5 The mean of RGB intensity of the tested bunch. Tested image Picture 1 Picture 2 Whole bunch

Red 88 87 88

Mean of RGB Green Blue 36 50 39 51 38 50

The mean values are calculated; the ripeness of the bunch is tested by comparing between the results of the Table 5 the results of standard samples in Table 4. The final results are in Figure 7. From our experience in this research, the accuracy of the results obtained depends on the accuracy of the data source and the number of training samples.

Figure 7 Result of the ripeness for the bunch

The User Interface’s Results: The program has already been developed and the results of the fruit’s ripeness have already been obtained. In order to manage them and make them easily understood by 309 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007

Meftah Salem M Alfatni. , et al the users, we have created the user interface. The interface is designed using forms depending on our requirement in the program. One of the forms will contain the main menu which contains information and some buttons such as help button which consists of guidance instruction to our system and start button of our program. The main form of the interface is shown in figure 8. After we executed the grading system's user interface. Pressing the start button will open the second form which was our way to run the test program. The second form of the system is shown in the Figure 9. It contains the actual steps and results along with the methodology

Figure 8.The main form of the interface



Figure 9 The steps and results of the grading system with their methodology

The second form will appear with different boxes which are image1 box, image2 box, inputs box, outputs box and the box of buttons (default, run, reset and end). To run the program, press default button. Program automatically enters the number of bunch's pictures as 2 which is the default number in our system. The number of picture will appear in the input box immediately. Image1 and image2 boxes will show the pictures of tested bunch individually with its mean for RGB. The ripeness of bunch is tested and its RGB mean will appear in the outputs box. Finally, we can use the reset button to clear all the results in order to prepare for new testing or use the end button.

CONCLUSIONS: This project was conducted to determine and differentiate between the colors properties of oil palm fruit bunch. The grading system was able to detect and differentiate the ripeness of oil palm FFB between the different categories of oil palm FFB. The computer program using MATLAB 7 software was written to analyze the RGB data of the images. The computer program used the mean color intensity to differentiate between the different color and ripeness of the fruits such as oil palm FFB. The color of oil palm FFB was classified as ripe, under ripe and over ripe. The range of RGB intensity of the oil palm fruit bunches came out as result of 311 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007

Meftah Salem M Alfatni. , et al training program to be as a reference of the ripeness in the test program. The test program in this prototype system is successfully able to differentiate between the three ripeness categories of the oil palm fruit bunches.

REFERENCES: López Camelo, A.F. Gómez, P.A..(2004). Comparison of color indexes for tomato ripening. Horticultura Brasileira, Brasília, v.22, n.3, p.534-537, jul-set. Yang C. C, Prasher S.O, Landry J. A, Ramaswamy H.S. and Ditommaso A.(2000). Application of artificial neural networks in image recognition and classification of crop and weeds. Canadian Agricultural Engineering Vol. 42, No. 3. Devrim U. Bernard G. (2004). Stem-End/Calyx Detection In Apple Fruits Comparison of Feature Selection Methods and Classi_ers. TCTS Lab., Faculté Polytechnique de Mons, Multitel Building, Avenue Copernic 1, Parc Initialis, B7000, Mons, Belgium Sudhakara P. Rao, A. Gopal, Revathy R. Meenakshi K. (1999) Colour Analysis Of Fruits Using Machine Vision System For Automatic Sorting And Grading. J. Instrum. Soc. India 34 (4) 284-291 Sudhakara P. Rao, A. Gopal, S. Iqbal Md. Revathy R. Meenakshi. K. (1999) Classification Of Fruits Based On Shape Using Image Processing Techniques. J. Instrum. Soc. India 34 (4) 277-239 Rashid S, Nor A A, Radzali M, Shattri M, Rohaya H, and Roop G. (2002). Correlation between oil content and DN values.GISdevelopment.net A. Raji and A. Alamutu(2005). “Prospects of Computer Vision Automated Sorting Systems in Agricultural Process Operations in Nigeria”. Agricultural Engineering International: the CIGRJournal of Scientific Research and Development”. Vol. VII. pages 1 - 12 Invited Overview. Saad A.A, Rashid S, Halim S, Thomas C and Fakhrul-Razi A. (2003) Optimizing the Correlation between percentages of the Oil Content in palm oil fruit lets and Digital Number of Images System. Faculty of Engineering, University Putra Malaysia Wan Ishak W, Mohd Z B and Abdul Malik A H. (2000). Optical Properties For Mechanical Harvesting Of Oil Palm Ffb. Journal of Oil Palm Research Vol. 12 No. 2, p. 38-45 Moon S. K. Alan M. L. Yud-Ren C. Yang T.(2005). Automated detection of fecal contamination of apples based on multispectral fluorescence image fusion. Sciencedirect. Journal of Food Engineering 71 85–91



Polder G., Van der Heijden G. W. A. M.and Young I. T.(2002). Spectral Image Analysis For Measuring Ripeness Of Tomatoes. American Society of Agricultural Engineers ISSN 0001–2351. Vol. 45(4): 1155–1161 Andrew C S. Hannah M B. Alison K. S. Daniel O. and Nicholas I M. (2003). The effect of colour vision status on the detection and selection of fruits by tamarins (Saguinus spp.). The Journal of Experimental Biology 206, 3159-3165. the Company of Biologists Ltd Unay D and Gosselin B. (2004). A Quality Sorting Method For ‘Jonagold’ Apples. 2TCTS Lab.-Multitel, Faculté Polytechnique de Mons, Av. Copernic 1, Parc Initialis, B- 7000, Mons, Belgium. ACKNOWLEDGMENTS: We would like to thank the Malaysian Palm Oil Board (MPOB) and the Malaysian Center for Remote Sensing (MACRES) for the invaluable support rendered in carrying out this research


LOW JITTER WIDE BAND CLOCK RECOVERY CIRCUIT USING DUAL LOOP PLL A. Telba, A. Almaroo, M. A. El Ela, and B. Almashary 1 School of Engineering, Design and Technology, College of Engineering, Electrical Engineering Department, King Saud University, Riyadh 11421, Saudi Arabia Abstract: Clock recovery is needed in all data communication systems for the purposes of synchronization. In general a PLL (phase Locked Loop) circuit is used to extract the clock signal from the input data stream. The recovered clock signal is always jittered and have to be adjusted by using a dejitter circuit. Tracking these errors over an extended period of time determines the system stability. One source of jitter is the PLL circuit itself due to some ac components at the VCO input which modulate its output frequency. A narrow band PLL may be used after the recovery circuit to minimize the jitter associated with the recovered clock. This second PLL has to be locked at the same frequency as the recovered clock. Other solution is by Using a single PLL with Controlled Crystal Oscillator (VCXO) whose centre frequency is equal to the data bit rate and a wide band loop filter. The conversion gain of the VCXO (Hz/V) is very small so a narrow band PLL is resulted without scarifying the dynamic behaviour. This work presents a proposed system for low jitter clock recovery circuit using two cascaded PLLs which enables to generate several clock frequencies using single VCXO .The system described is tested by simulation and experimentally. The experimental results confirm the simulation. I Introduction The major circuit in any synchronization scheme is the clock recovery circuit. The performance of this circuit is measured by its ability to regenerate a clock signal whose frequency exactly equals the clock frequency of the transmitted data. The Phase-Locked Loop (PLL) is designed to simplify different tasks such as clock recovery, data retiming, frequency translation and clock smoothing applications. The output signal from a given PLL suffers from an associated jitter [1 − 4] especially at high bit rate resulting in bit errors at the receiver side and may cause malfunctioning for the all network if this error exceeds a certain threshold level. As the recovered clock signal is always jittered it is needed to be adjusted by using a de-jitter circuit. Jitter is defined as the misalignment of edges in a sequence of data bits from their ideal positions, it can cause data errors. Different types of jitter had been reported such as: Cycle-to-Cycle Jitter [5], period jitter [6] and long term jitter [7].


LOW JITTER WIDE BAND CLOCK RECOVERY CIRCUIT USING DUAL LOOP PLL

Fig. 1 Typical long term jitter ` 1.1 Jitter Minimization Techniques Different techniques have been reported for the design and implementation of jitter minimization circuits. The following techniques are: a. b.

Modifying the filter design to narrow band the PLL bandwidth and make the phase noise at the VCO input as low as possible [10]. Reducing power supply noise [8-9]. Heydari et al proposed a mathematical model for calculating the power supply noise inducing timing jitter in PLLs. Eliminating ground bounce and using additional filter to minimize the effect of the sudden changes of the supply voltage and to have a good grounding to discharging the unexpected charges to ground [11-12].

c.

Using dual phase frequency detector [13]. This design uses two phase detectors (PD), the first one is multiplier (PD) and the second is phase and frequency detector. This design is proposed for high frequency and gets a low phase noise.

d.

Using modified charge pump as phase frequency detector [14]. In this design a low power consumption charge pump improves the jitter characteristics of a Phase-Locked Loop (PLL) by blocking the control voltage leakages.

e.

Using PLL with Voltage Controlled Crystal Oscillator (VCXO) [15].


A. Telba, A. Almaroo , et al

An example of a de-jitter PLL circuit is shown in Fig. 1.2. The design objective of this circuit is to generate a stable, low-jitter clock based on either the recovered receive clock or transmit clock input.

Fig.2 Block Diagram of dejitter circuit

One disadvantage of such technique is that we have to use a VCXO with the same frequency of the input jittered clock. This means that for each carrier system we need to build a VCXO working at the same clock frequency. For example for (E1) telephone carrier system VCXO has to provide exactly 2.048 MHz signal. II Dual PLL System The two main solutions usually used for reducing jitter are those given in a , e . Both techniques work properly and succeeded in reducing the jitter appreciably especially the second one. On the other they have one common problem: each of them may be used for only one bit rate (clock frequency). The user has to change the circuit design for each data rate. For example in the second technique we need a VCXO for each bit rate. This work presents a proposed system for low jitter clock recovery circuit with limited jitter. Two cascaded PLL,s are used as given by the block diagram given in Fig.3. The recovered clock from the first loop is jitter bounded due to the effect of the VCXO. The second loop is a wide frequency band (wide Lock- in range) to cover several standard data rates. In steady state operation and while the two loops are in locking conditions we may write: fVCXO/N1 = fin /M1 and fVCXO/N1M2 = fout/N2 Combining the above two equations results in fout = fin (N1 N2 /M1M2 ) The above equation shows that the output frequency fout may be programmed through selecting (N1, N2 , M1 and M2) to be equal to fin independent on fVCXO when N1 N2 = M1M2.This means that we can generate several clock frequencies using single VCXO .The system described by Fig.3 is tested by simulation and experimentally. The experimental results confirm the simulation. 317 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


Fig.3 dual loop PLL clock recovery circuit III- Simulation and Experimental results 3-1 System structure The first PLL is built around TRU 050 IC chip, from Vectron Semiconductor devices, Inc. It is used as a low jitter clock recovery circuit. The complete details of (PLL-1) are shown in Fig.4. If the chip is hard wired, it will be suitable only for one DATAIN rate.

Fig.4 Basic building block of TRU-50 (VCXO) 318 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


The second PLL uses wide range Voltage Control Oscillator and is built around the 74HC 4046 integrated circuit that contains all parts necessary to make a PLL except for the loop filter components. The programmable frequency dividers needed to complete the dual loop structure are built using the FPGA prototype board type (SPATAN_XC10TMTM).

3.2 Simulation results For different values of N we chose N=130,150,170,193 as shown in figure 5.10, the RMS jitter was measured at the output from the first and second loops respectively. The relationships between the RMS jitter and noise power are plotted for different values of N for both outputs and is shown in Fig.5-a and Fig.5-b respectively. We can notice that the final output jitter is almost constant irrespective of the input phase noise.

3-3 Experimental results

Fig.5-a Jitter at PLL1 output

Fig.5-b Jitter at PLL2 output

The dual loop system described above was built and tested using Le-croy Oscilloscope "Wave Runner" model 6100 "1GHz Sampling Oscilloscope". The JTA2 software package for LeCroy Oscilloscopes provides advanced jitter and timing analysis capabilities. It uses LeCroy’s long memory and Zoom architecture to capture and precisely measure thousands of cycles of timing information and then present the results with three different views. By applying a high jittered signal to 319 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


the system and measuring the output jitter at frequency 1.544 MHz (T1 Carrier) we can deduce the jitter behaviour of the system .The output from the oscilloscope is as shown in Fig.6 , where the upper part shows the carrier frequency. The middle portion of the figure shows the Jitter track. The lower portion is the Jitter histogram. The Jitter statistics were taken mainly (cycle to cycle jitter, RMS jitter and accumulated jitter). The TIE has also been measured. It is clear from the figure that the value of the jitter was reduced dramatically from 5.6 nano seconds in input to 37.5 pico seconds in output. The test is repeated at different carrier frequencies and the output RMS is plotted versus the carrier frequency in Fig.7.

Fig.6 Jitter measurements of T1carrier frequency



Fig.7 Output Jitter at different carrier frequencies.

IV Conclusion Jitter is almost non useful noise that exists in both digital and analogue signals. Most telephone carriers nowadays use digital signals, and such carriers are meeting different standards (European, American and Japanese). This work presents a proposed system for low jitter clock recovery circuit with limited jitter. Two cascaded PLL,s are used where, the recovered clock from the first loop is jitter bounded due to the effect of the VCXO. The second loop is a wide frequency band (wide Lock- in range) to cover several standard data rates. Simulation and experimental results shows that the jitter behaviour is almost directly proportional to the carrier frequencies from 1.54 MHz up to almost 10MHz, and then it remains almost constant for values of frequency beyond 10 MHz up to 24.77 MHz. The experimental results confirm the simulation and prove the validity of the proposed system in increasing the operating frequency range of a given low jitter circuit.



REFERENCES [1] P. Y. Kim, “Bit error probability of QPSK with noise phase reference,” IEE Proceedings Volume 142 Oct 1995, pp.292-296. [2] W. E. Thain and J. A. Connelly, “Simulating Phase Noise in Phase-Locked Loops with a Circuit Simulator,” Proc. of IEEE International-Symposium-on Circuits and ystems, vol. 3, 1995, pp. 1760-1763. [3] Jitter in PLL-Based Systems: Causes, Effects and Solutions (Cypress Semiconductor Corporation 1997) [4] P. Heydari and M. Pedram, “Analysis of Jitter due to Power-Supply Noise in Phase-Locked Loops Custom,” Proc. Of IEEE Integrated Circuits Conference, May 2000, pp. 443 - 446 [5] F. Herzel and B. Razavi, “A study of Oscillator Jitter Due to Supply and Substrate Noise,” IEEE Transactions on Circuits and Systems II: Analog and Digital Signal Processing, vol. 46, No. 1, Jan. 1999, pp. 56 - 62. [6] D. Wolaver, Phase-Locked Loop Circuit Design, Prentice Hall, 1991. [7] Roland E. Best, Phase-Locked Loops: Design, Simulation, and Applications, McGraw-Hill, NY 1999. [8] M. Kiha, S. Ono and P. Eskelinen Digital clocks for synchronization and ommunications, Artech House, 2003. [9] T. Pialis and K. Phang, “Analysis of Timing Jitter in Ring Oscillators Due to Power Supply Noise,” Proc. of IEEE International Symposium on Circuits and Systems, Bangkok, Thailand, vol. 1, May 2003, pp. 685-688. [10] A. Hajimiri and T. H. Lee, “A General Theory of Phase Noise in Electrical Oscillators,” IEEE Journal of Solid-State Circuits, vol. 33, No. 2, Feb. 1998, pp. 179 – 194. [11] P. Heydari and M. Pedram, “Jitter-Induced Power/Ground Noise in CMOS PLLs: A Design Perspective Computer Design,” Proc. of International Conference ICCD, Sept.2001, pp 209 - 213. [12] “DART Dejitter PLL Design and Analysis” AN-531 Application note, DART Device TXC-02030-AN2 TranSwitch Corporation, 2004. [13] K.M. Ware, H.-S. Lee, and C. G. Sodini, “CMOS phase-locked loop with dual phase detectors,” Proc. of 36th IEEE International Solid-State Circuits Conference, Digest of Technical papers, Feb. 1989, pp. 192 – 193. [14] W. Yan and H. Luong, “A 2-V 900-MHz monolithic CMOS dual-loop frequency Synthesizer for GSM receivers,” IEEE Journal of. Solid-State Circuits, vol. 36, pp. 204-216, Feb.2001. [15] “DART Dejitter PLL Design and Analysis” AN-531 Application note, DART Device TXC-02030-AN2 TranSwitch Corporation, 2004.


ATTENUATION OF ELECTROMAGNETIC RADIATION BY POLYMERIC THERMAL INSULATION MATERIALS

Majeed A. Alkanhal1, Saleh Alshebeili1, Ashraf Mohra1, and Wahid El-Masry2 1 Dept. of Electrical Eng., King Saud University, [email protected] 2 Dept. of Chemical Eng., King Saud University, [email protected] P. O. Box: 800, Riyadh 11421, Saudi Arabia

ABSTRACT This paper reports a laboratory study on electromagnetic (EM) energy attenuation through selected polymeric materials that are currently used for thermal insulation in buildings and typically available in local markets in Saudi Arabia. The experiments were performed in the anechoic microwave laboratory in the department of Electrical Engineering at King Saud University. The tested specimens of the foam thermal insulation material show no significant attenuation of EM energy at the frequency range (700-2700) MHz. The density and thickness variations of these specimens, also, indicate no noticeable variation in the attenuation of the EM fields through them. To enhance the electromagnetic insulation of the typical foam thermal-insulating materials, metal inclusion must be considered in the fabrication process of these materials. Covering these materials with Aluminum foil from only one single side, indeed, makes a good electromagnetic insulation. The results of this study are beneficial to public, construction engineers, and operators of mobile communications systems. Key Words Electromagnetic measurements.

radiation,

polymer

materials,

attenuations,

microwave

1. INTRODUCTION There has been a considerable rise in official and public awareness and concern in recent years regarding biological effects of EM fields on humans. In this regard, several standards and research studies have been published and enforced officially [1-5]. Based on that, the investigation of electromagnetic isolation that some materials used in buildings construction may have become more important. This is especially vital to people living close to mobile base stations and/or other communication transmitters [5-7].



The local Saudi market is witnessing an increase in the number of polymeric materials that are currently used in buildings for thermal and water insulations. The thermal and water insulating capabilities of these materials are well understood in the Saudi market due to its long experience in that field. However, the effect of these insulating materials on radio wave propagation is yet to be explored. This paper performs a laboratory study on electromagnetic (EM) propagation through thermal insulation materials currently available commercially in the market. Experiments were performed on thermal insulation materials. Each specimen of the material under study was placed in a special test range consisting of transmitter (with horn antenna) and receiver (with LPDA antenna), both working in the operating range of mobile phone frequencies. The foam materials were placed on a wooden holder between the transmitter and the receiver. All the measurements were performed inside a microwave chamber to avoid any electromagnetic interference from other sources. The vector network analyzer was used to measure the S-parameters for the transmitted and received signals when the foam material is present or in the case when the foam is absent (reference signal). Measurements of power loss (attenuation) as a function of the material type are conducted at 700-2700 MHz. This electromagnetic spectrum covers 900-1800 MHz range, which is currently allocated for the operation of Global System for Mobile Communications (GSM).

2. Materials against Electromagnetic Radiations Materials for EM shielding are increasingly needed to reduce the interference of radio frequency (RF) and microwave (MW) devices with computers, transformers, cables and medical devices, and minimize the effect of EM radiation on human health. Some materials commonly used for biological and industrial safety are as follows: Polymer-matrix composites containing conductive fillers are attractive for shielding due to their processability (e.g., moldability), which helps to reduce or eliminate the seams in the housing that is the shield [8]-[10]. Cement is slightly conducting, so the use of a cement matrix allows the conductive filler units in the composite to be electrically connected, even when the filler units do not touch one another[10]-[15]. Metals are more attractive for shielding than carbon based materials due to their higher conductivity, though carbons are attractive in their oxidation resistance and thermal stability. Fibers are more attractive than particles due to their high aspect ratio. Thus, metal fibers of a small diameter are desirable. Nickel filaments of diameter 0.4 µm (as made by electroplating carbon filaments of diameter 0.1 µm) are particularly effective [16]-[17]. Continuous fiber polymer-matrix structural composites that are capable of Electromagnetic insulation (EMI) shielding are needed for aircrafts and electronic 324 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007

Majeed A. Alkanhal. , et al

enclosures [18]-[19]. The fibers in these composites are typically carbon fibers, which may be coated with a metal (e.g., nickel) or be intercalated (i.e., doped) to increase the conductivity. Flexible graphite is a particularly attractive EMI gasket material, which is a flexible sheet made by compressing a collection of exfoliated graphite flakes (called worms) without a binder. Colloidal graphite is fine graphite powder suspended in a liquid carrier (such as water and alcohol), together with a small amount of a polymeric binder. Colloidal graphite is used for shielding in television scopes. Foam materials are inorganic, in contrast to polymer foams. The inorganic nature makes the materials water resistant and fire resistant. Foam building materials include foam glass, foam gypsum and cellular concrete [8,13]

3. Foam Material Properties The foam materials are mainly used for thermal and acoustic insulations in buildings and they are not intended to be electromagnetic insulators. There are four major rigid plastic foam insulations commonly used for residential, commercial and industrial thermal insulation • Extruded polystyrene (XEPS) • Expanded polystyrene (EPS) • Polyurethane (PUR) • Polyisocyanurate (PIR) Each type has its individual characteristics and specific advantages and disadvantages for particular building applications. Nonetheless, the stable properties of polystyrene, when combined with a unique foam extrusion process, produce an exceptionally useful product with benefits for nearly all construction and engineering applications. Extruded polystyrene has a well established reputation for long-term reliability and superior resistance to the elemental forces of nature: time, water, cold, heat, and pressure. 3.1 Foam Materials in Local Markets Foam materials that are widely available in the Saudi market can be, basically, classified as shown in Table 1. In fact, these foam materials are mainly used for the isolation of heat, and sometimes for sound isolation. Their performance in the isolation of electromagnetic (EM) radiation will be investigated in the following section.



Table 1: Selected foam materials widely available in the local Saudi Market. Material

Thickness (Cm)

Given Label

20 20 35 26 26 32

2 4 2 4 5 4

A B C D E F

60 32

5 5

G H

Density 3

( Kg / m ) Extruded Polystyrene Extruded Polystyrene Extruded Polystyrene Extruded Polystyrene Extruded Polystyrene Extruded Polystyrene Rock Wall ( covered with adhesive Aluminum foil) Extruded Polystyrene

4. Measurement Setup The measurements of electromagnetic energy attenuation (insertion loss) through selected foam thermal insulation material are carried out in the microwave anechoic chamber at the microwave laboratory at the department of electrical engineering, College of Engineering, King Saud University. The system as shown in Fig.1, consists of a Horn antenna (transmitter), a LPDA (Receiver), Vector Network analyzer, Microwave amplifier, Cables and adapters as well as Holders (for the transmitter, receiver and foam material ). The distance between the transmitter and the receiver was set to 3.5 meter in order to position the foam material in the farfield region of the transmitted antenna [20,21].



Foam Holder

Receiver

Transmitter

From

To Vector Network

Fig.1: System set-up used for attenuation (insertion loss) measurements.

5. Measurement Results The measurements are carried out in the microwave range (0.7-2.7 GHz). First measurements are done without foam and use these measurements as a reference for the measurements of the foam materials in free space. The results of the attenuation (insertion loss) measurements in dB versus frequency for selected material specimens of those listed in Table 1 are illustrated in Figs. 2 -4. The tested specimens (except material G) of the foam thermal-insulation material show no significant attenuation of EM energy at the frequency range (700-2700) MHz. The density and thickness variations of these specimens, also, indicate no noticeable variation in the attenuation of the EM fields through them. However, one specimen of the thermal insulation materials (material G) which is made of Rockwall and covered by adhesive thin metal layers from both sides has a high EM insulation. This, obviously, is attributed basically to the wave reflection 327 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


mechanism from the thin metal layers covering (containing) the thermal-insulation material. A one sided or two sided covering of the material board with a very thin aluminum foil as that used typically in Rockwall material is an alternate to metal inclusion and may be the best suggestion for good performance electromagnetic and thermal insulation too. Fig. 5 illustrates the effect of covering one side with an aluminum foil for material (c).

Attenuation for Material (A) 0.2 0

Att.(dB)

-0.2 0.7

0.9

1.1

1.3

1.5

1.7

1.9

2.1

2.3

2.5

-0.4 -0.6 -0.8 -1 -1.2 -1.4 -1.6 -1.8 Frequency (GHz)

Fig.2: EM attenuation of the foam material (A).


2.7


Attenuation for Material (C) 0.2 0 -0.2 0.7

0.9

1.1

1.3

1.5

1.7

1.9

2.1

2.3

2.5

2.7

Att.(dB)

-0.4 -0.6 -0.8 -1 -1.2 -1.4 -1.6 Frequency (GHz)

Fig.3: EM attenuation of the foam material (C).

Attenuation For Material (G) 0 -5

0.7

0.9

1.1

1.3

1.5

1.7

1.9

2.1

2.3

2.5

2.7

Att.(dB)

-10 -15 -20 -25 -30 -35 -40 Frequency (GHz)

Fig. 4: EM attenuation of the foam material (H)



Attenuatoion for for Material (E) covered with Aluminum Attenuation material (C) covered with Foil from single side

aluminum foil from one side

0 -5

0.7

0.9

1.1

1.3

1.5

1.7

1.9

2.1

2.3

2.5

2.7

Att.(dB)

-10 -15 -20 -25 -30 -35 -40

Frequency (GHz) Frequency (dB) Fig.5: EM attenuation for the foam material (E) when covered with aluminum foil from single side 6. Conclusions and suggestions The tested specimen of the foam thermal-insulation material show no significant attenuation of EM energy at the frequency range (700-2700) MHz. The density and thickness variations of these specimens, also, indicate no noticeable variation in the attenuation of the EM fields through them. An average of only 1 dB attenuation is contributed to the foam boards inserted in free space environment between transmitter and receiver in the measurement frequency band (700-2700MHz). To enhance the electromagnetic insulation of the typical foam thermal-insulating materials, metal inclusion must be considered in the fabrication process of these materials. This must be studied not only from the electromagnetic insulation point of view, but also from the point of view of thermal insulation performance too. Good thermal insulation must be maintained and cost effectiveness must be observed in all design structures. Metal wires of half-wave length or more should be distributed randomly but homogeneously in the hosting dielectric material to have good electromagnetic insulation. Thin and long wires could be inserted along and/or across the foam board with 10 cm or less distance between wires would, indeed, improve the electromagnetic insulation of the foam material. A one sided or two sided covering of the material board with a very thin aluminum foil as that used typically in Rockwall material is an alternate to metal inclusion and may be the best suggestion for good performance electromagnetic and thermal insulation too. Using material with better (higher) dielectric constant or increasing the density of these 330 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


foam materials to be in the range of (250-450 Kg/m3) may, also, contribute to a slightly better electromagnetic insulation.

References 1. Safety Levels with Respect to Human Exposure to Radio Frequency Electromagnetic Fields, 3 kHz to 300 GHz, ANSI/IEEE-C95.1, 1992. 2. Human Exposure to Electromagnetic Fields, High Frequency (10 kHz to 300 kHz), European Community Prestandard ENV 50 166-2, 1995. 3. “Guidelines for limiting exposure to time-varying electric, magnetic and electromagnetic fields (up to 300 GHz),” Documents of the Health Phys., vol. 74, pp. 494-522, 1998. 4. Evaluating compliance with FCC-specified guidelines for human exposure to radiofrequency radiation, U.S. FCC, Office of Engineering and Technology, Washington, DC, OET Bulletin 65, 1997. 5. James C. Lin, “Microwave Exposure and safety associated with personal wireless telecommunication base station,” IEEE Microwave Magazine, vol. 3, pp. 28-31, Sept. 2002. 6. R. C. Petersen and P. A. Testagrossa, “Radio-frequency electromagnetic fields associated with cellular-radio cell-site antennas,” Bioelectromagnetics, vol. 13, pp. 527- 542, 1992. 7.

8.

9.

10.

11.

12.

13.

P. Bernardi, M. Cavagnaro, S. Pisa, and E. Piuzzi, “Human exposure to radio base-station antennas in urban environment,” IEEE Trans. Microwave Theory Tech., vol. 48, pp. 1996-2002, Nov. 2000. C. Jackson B and G. Shawhan, “Current review of the performance characteristic of conductive coating for EMI control,” Proc. of the 1998 IEEE Int. Symp. On Electromagnetic Compatibility, vol. 1, pp. 567-572, 1998. L.G. Bhatgadde and S. Joseph, “Electroless technique for EMI shielding,” Proc. of the IEEE Int. Conf. on Electromagnetic Interference and Compatibility, pp. 443-445, 1997. A. Sidhu, J. Heike, U. Michelsen, R. Messinger, E. Habiger and J. Wolf, “Metallization of plastics for shielding,” Proc. of the IEEE Int. Symp. Electromagnetic Interference and Compatibility, pp.102-105, 1997. M.S. Bhatia, “A technique for depositing metal layers over large are for EMI shielding,” Proc. of the 1995 4th IEEE Int. Conf. Electromagnetic Interference and Compatibility, pp. 321-324, 1995. B.C. Jackson and P. Kuzyk, “A practical guide on the use of Electroless coating for EMI shielding,” Proc. of the 9th IEE Int. Conf. on Electromagnetic Compatibility, pp. 119-124, 1994. C. Nagasawa, Y. Kumagai, K. Urabe and S. Shinagawa, “Electromagnetic shielding particle board with Nickel-plated wood particles” J. Porous Mater, 6(3), pp. 247-254, 1999. 331



14. D.S. Dixon and J. Masi, “Thin Coating can provide significant shielding against low frequency EMF/magnetic fields,” Proc. of the 1998 IEEE Int. Symp. on Electromagnetic Compatibility, vol. 2, pp.1035-1040. 1998. 15. P. J. D. Mason, ‘Technologies and markets for thin film EMI/RFI shielding,” Proc. of the 37th Annual Tech. Conf. Soc. Vacuum Coaters, pp.192-197, 1994. 16. C.A. Grimes, “EMI shielding characteristics of permalloy multilayer thin films,” Proc. of the 1994 IEEE Aerospace Applications Conf., pp. 211-221, 1994. 17. W.J. Biter, P.J. Jamnicky and W. Coburn, “Shielding improvements by use of thin film multilayer films,” Proc. of the 1994 7th Int. SAMPE Electronics Conf., vol. 7, pp. 234-242, 1994. 18. D.A. Olivero and D.W. Radford “A multiple protection approach to EMI shielding composites incorporating conductive fillers,” J. Reinforced Plastics & Composites, 17(8), pp. 674-690, 1998. 19. M.S. Kim, H.K. Kim, S.W. Byun, S.H. Jeong, Y.K. Hong, J.S. Joo, K.T. Song, J.K. Kim, C.J. Lee and J.Y. Lee, “PET fabric/polypyrrole composite with high electrical conductivity for EMI shielding,” Synthetic Metals, vol.126 (ER2-3), 233-239, 2002. 20. Ray-Rong Lao; Jenn-Hwan Tarng; Chiuder Hsiao, “Transmission coefficients measurement of building materials for UWB systems in 3 -10 GHz,” The 57th IEEE Semiannual Vehicular Technology Conference, VTC 2003-Spring, vol. 1, pp. 11–14, April 2003. 21. P. Ali-Rantala, L. Ukkonen, L. Sydanheimo, M. Keskilammi, M. Kivikoski, “Different kinds of walls and their effect on the attenuation of radiowaves indoors,” IEEE Antennas and Propagation Society International Symposium, APS. 2003, vol. 3, pp. 1020 - 1023, June 2003.

Acknowledgement The authors would like to acknowledge the financial support provided by SABIC and the support provided by the Research Center in the College of Engineering at King Saud University for this project under grant number 26/425.


MATRIX MULTIPLICATION USING FIELD PROGRAMMABLE GATE ARRAYS: A STATE-OF-THE-ART REVIEW

Syed Manzoor Qasim, Bandar Almashary and Shuja Ahmad Abbasi Department of Electrical Engineering, VLSI Design Research Lab. King Saud University, College of Engineering King Saud University P.O.Box-800, Riyadh 11421 {smanzoor,bmashary,abbasi}@ksu.edu.sa

ABSTRACT Matrix multiplier is used as a basic building block in many signal and image processing applications. The computational complexity of matrix multiplication algorithm is O(n3) on a sequential processor and O(n3/p) on a parallel system with p processors and matrix dimension of n. Field Programmable Gate Array (FPGA) based realization of matrix multiplier provides a significant speed-up in computation time as compared to software based realization. This paper presents a comprehensive state-of-the-art review on the FPGA implementations of matrix multiplication.

KEY WORDS FPGA, Matrix multiplier, Review, State-of-the-art.

1.

INTRODUCTION

With the recent technological advancement, Field Programmable Gate Arrays (FPGAs) have improved considerably in logic density, functionality and speed, thus making them ideal for System–on–Chip (SoC) designs for wide range of applications. Today, FPGAs are large and fast enough for use in multimillion gate designs running at speeds exceeding 500 MHz. FPGAs now exceed the capacity and speed requirements of the vast majority of Application Specific Integrated Circuits (ASICs). Being dynamically reconfigurable, the same FPGA can be used for wide range of applications [1].



As the complexity of FPGAs is continuously increasing, and entire system can now be implemented on them with minimal off-chip resources, they provide an ideal platform for implementing computation intensive algorithms from real-life applications including communications, encryption, video and image processing, image reconstruction, medical imaging, network security and numerical computations that can be benefited from hardware realization [2]. Most of the computation-intensive algorithms such as those used in digital signal, image and video processing, numerical analysis, computer graphics and vision, etc. involve matrix multiplication. Matrix multiplier is a basic building block in such applications. The use of an optimized matrix multiplier will thus result in optimized realization of computation intensive algorithms. In practice, computationally intensive tasks are often performed by parallel processing systems which distribute computations among several processors in a way that leads to significant speedup gains. There are many implementations of matrix multiplication which is O(n3) operation, where n is the dimension of input matrices. These implementations differ mainly in terms of algorithms or the hardware platforms. In recent years, there has been a considerable interest in accelerating the matrix multiplication routine using FPGA technology. FPGA-based implementation of matrix multiplication is a fairly new approach. Research work related to the implementation of matrix multiplication on FPGAs are scattered throughout the literature. In this paper, an attempt is made to organize the literature and provide a comprehensive review on the state-of-the-art FPGA implementation of matrix multiplication. It is based on the careful analysis of the most recently available literature in this area. We summarize the related work done by various researchers covering various aspects of matrix multiplication implemented on FPGAs. The rest of the paper is organized as follows. Section 2 gives a brief description of FPGA technology and design flow for realizing FPGA based designs. An overview of the matrix multiplication process is then provided in section 3. A comprehensive review of fixed and floating point matrix multiplication implemented on FPGA is discussed in sections 4 and 5 respectively. Finally, the paper is concluded with some remarks.


Syed Manzoor Qasim. , et al

2.

FPGA TECHNOLOGY

An FPGA is a digital integrated circuit (IC) comprising of Configurable Logic Blocks (CLBs) connected with software configurable interconnections. FPGA combines the integration of an ASIC with the flexibility of user-programmable logic. Application specific hardware circuits can be created on demand to meet the computing and interconnect requirements of an application [1]-[3]. 2.1

Features of Xilinx FPGA

The fundamental architecture of Xilinx FPGA consists of an array of CLBs, which are the basic elements that can be programmed to perform various logic functions. Each CLB is coupled with a programmable interconnect switch matrix that connects the CLB to adjacent and nearby CLBs. Each CLB contains four logic slices, where each logic Slice usually consists of two four-input Look Up Tables (LUTs), two configurable flip-flops, some muxes, and other control logic. In addition to the CLBs and the switch matrices, the FPGA have a number of higher level logic blocks such as block RAMs (BRAMs), 18-bit multipliers, digital clock managers (DCMs), CPUs and a programmable interconnection network arranged in a rectangular grid pattern. 2.2

FPGA Design Flow

In order to realize an algorithm on an FPGA it must be programmed first. To achieve this, a design methodology is adopted as shown in Figure 1. Usually, design entry is done using schematic or hardware description language (HDL) such as Very High Speed Integrated Circuit HDL (VHDL) or Verilog.



Figure 1. FPGA Design Flow

The objective is to make the system description independent of the physical hardware such that it can be used on other FPGAs and even on ASICs. Once a design has been completed it is simulated to verify the correct operation. A vendor specific netlist file which is essentially text based descriptions of the schematic is generated from the design and is mapped onto the FPGA using synthesis, place and route and optimization tools. Mapping produces a bit-stream file that is used to program the FPGA [3].



3.

MATRIX MULTIPLICATION

The matrix multiplication C = AB of two matrices A and B is conformable, if the number of columns of A is equal to the number of rows of B. The ijth element of C is given by, n

cij =

∑a k =1

ik

b kj

(1)

For n = 3, the matrix product is given by, ⎡ a11 ⎢a ⎢ 21 ⎢⎣ a31

a12 a22 a32

a13 ⎤ ⎡ b11 b12 a23 ⎥⎥ ∗ ⎢⎢b21 b22 a33 ⎥⎦ ⎢⎣b31 b32

b13 ⎤ ⎡ c11 c12 b23 ⎥⎥ = ⎢⎢c21 c22 b33 ⎥⎦ ⎢⎣c31 c32

c13 ⎤ c23 ⎥⎥ c33 ⎥⎦

(2)

Where, ⎡ c11 c12 ⎢c ⎢ 21 c22 ⎣⎢c31 c32

c13 ⎤ ⎡ a11b11 + a12b21 + a13b31 a11b12 + a12b22 + a13b32 c23 ⎥⎥ = ⎢⎢a21b11 + a22b21 + a23b31 a21b12 + a22b22 + a23b32 c33 ⎦⎥ ⎣⎢ a31b11 + a32b21 + a33b31 a31b12 + a32b22 + a33b32

a11b13 + a12b23 + a13b33 ⎤ a21b13 + a22b23 + a23b33 ⎥⎥ a31b13 + a32b23 + a33b33 ⎦⎥

In general, cij = ai1b1 j + ai 2b2 j + K + ain bn j

(3)

For matrices of size n x n, this algorithm requires n3 multiplications. Hence this is O(n3) operation.

4.

FIXED POINT IMPLEMENTATION

FPGA based designs are usually evaluated using three performance metrics: speed (latency), area, and power (energy). Fixed point implementations in FPGA are fast and have minimal power consumption. Additionally, a fixed point matrix multiplier unit often takes fewer gates in an FPGA or ASIC than its floating-point counterpart. The limitation of fixed point number is that very large and very small numbers cannot be represented and the range is limited to bit-width of the number. There has been extensive previous work in the area of designing an FPGA based system for the computation of fixed point matrix multiplication. Amira et al. presented a novel architecture based on systolic architecture for a matrix multiplication [4]. A serial-parallel matrix multiplier based on the Baugh337 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


Wooley algorithm has been used. The design based on the systolic architecture has been implemented using a Xilinx XCV1000E of Virtex-E FPGA family. Amira et al. designed a parameterizable system for 8-bit fixed point matrix multiplication using FPGA [5]. Their design used both systolic architecture and distributed arithmetic design methodology for the implementation of matrix multiplication. The architecture proposed in this paper was targeted to Xilinx XCV2000E of Virtex-E FPGA family. The results presented in this paper showed better performance than the architecture presented in [4] in terms of area and speed. For n = 4, distributed arithmetic based design used 57 Slices as compared to 72 slices used in [4] and operated at a maximum frequency of 166.47 MHz as compared to 58.302 MHz used in [4]. Distributed Arithmetic based design provides better performance in terms of speed and area as compared to Systolic Array based design. The I/O bandwidth required by the design is directly proportional to the problem size. The designs presented in [4]-[5] were restricted to smaller matrices. For multiplying large matrices (n = 128, 256 and 512), Bensaali et al. designed an FPGA based coprocessor [6]. The designed coprocessor first partitions the input matrices into smaller sub-matrices and then calculates the product. In [7], Mencer et al. implemented the matrix multiplication on Xilinx XC4000E FPGA. Their design employs bit serial multipliers using Booth encoding. They focused on tradeoffs between area and maximum running frequency with parameterized circuit generators. Their design was improved by Amira et al. in [8] using modified booth encoder multiplication along with Wallace tree addition. For n = 4, 296 CLBs were used to achieve a maximum operating frequency of 60 MHz using Xilinx XCV1000E FPGA. Jang et al. improved the design in [7] and [8] in terms of area, speed [9] and energy [10] by taking advantage of data reuse. They reduced the latency for computing matrix product by employing internal storage registers in the processing element (PE). The algorithms need n multipliers, n adders, and total storage of size n2 words. For 4 × 4 matrix multiplication, the latency of the design in [7] is 0.57µs, while the design in [9]-[10] uses 0.15µs utilizing 18 % less area as compared to [7]. Belkacemi et al. [11] presented the design and implementation of a high performance, fully parallel matrix multiplication core. The core is parameterized and scalable in terms of the matrix dimensions (i.e., number of rows and columns) and the input data word length. Fully floor planned FPGA configurations are generated automatically, from high-level descriptions of the matrix multiplication operation, in the form of EDIF netlist in less than 1 sec. These are specifically optimized for Xilinx Virtex FPGA chips. By exploiting the abundance of logic resources in Xilinx Virtex FPGAs (LUTs, fast carry logic, shift registers, flip flops etc.), a fully parallel implementation of the matrix multiplier core is achieved; with a full matrix result 338 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


being generated every clock cycle. A 3 × 3 matrix multiplier instance consumes 2,448 Virtex slices and can run at 175 MHz on an XCV1000E-6 Virtex-E chip. Traditionally, the performance metrics for FPGA based designs have been latency and area. However, with the proliferation of portable, mobile devices, it has become increasingly important that the systems are also energy efficient and consume low power. In FPGA devices, major chunk of power is consumed by the programmable interconnects, while the remaining power is consumed by the clocking, logic, and I/O blocks. Another source of power dissipation in FPGAs is resource utilization and switching activity [12]. Research efforts towards the design of energy efficient matrix multiplier have been reported in [10], [13-15]. Most of the previous work in fixed point matrix multiplication focused only on reducing the latency and the area. Choi et al. developed new designs and architectures for FPGAs which minimize the power consumption along with latency and area [14]-[15]. They used linear systolic architecture to develop energy efficient designs. For linear array architecture, the amount of storage per processing element affects the system wide energy. Thus, they used maximum amount of storage per processing element and minimum number of multipliers to obtain energy-efficient matrix multiplier. Partially reconfigurability feature was exploited for the first time for the computation of matrix multiplication by Jianwen et al. in [16]. Partially reconfigurable devices offer the possibility of changing the design implementation without stopping the whole execution process. The matrix multiplier was implemented in Xilinx Virtex-II device, which supports partial reconfiguration. The design was evaluated in terms of latency and area and it was found that area is reduced by 72% - 81% for matrix sizes between 3 × 3 and 48 × 48 as compared to [9] and the performance further improves for larger matrices.

5.

FLOATING POINT IPLEMENTATION

Many high performance computing applications such as weather forecasting, computational fluid dynamics etc. require high performance floating point matrix multipliers. Previously, implementing floating point matrix multiplier in FPGAs was not feasible because it consumed a major portion of the resources available on FPGAs [17]. However, with the recent advancement made in FPGA technology [18], it is feasible to implement floating point matrix multipliers. The performance metric used for floating point matrix multiplier is GFLOPS. Research efforts to achieve this have started recently and some promising results have been reported in 339 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


[19]-[23]. In this section, we will provide a review on the state-of-the-art FPGA implementation of floating point matrix multiplier. The authors in [19] studied the effect of the floating point multiplier and adder on the performance of matrix multiplication. It was concluded that floating point matrix multiply requires more space than fixed point matrix multiply because the floating point adder is significantly bigger than the fixed point adder. They used four 4062XL-3 FPGAs achieving a performance of 192 MFLOPS for multiplying 1024 × 1024 matrices as compared to a theoretical peak of 264 MFLOPS. In [18] Underwood et al. discussed matrix multiplication as part of the BLAS (Basic Linear Algebra Subprograms). They studied the performance of FPGA on computing double precision (64-bit) floating point matrix multiplications and compared it with that of general purpose processors. Dou et al. proposed a parallel algorithm for matrix multiplication in [20], which consists of master processor and multiple slave processors. The master controls the slave processors. It distributes the data from the source matrices to the slaves and each slave computes different blocks of the resultant matrix in parallel. This algorithm is suitable for matrices of large dimensions. The authors of [22] extended the architecture in [9] for floating point matrix multiplication. They proposed two algorithms in which the number of floating point units and the storage size are independent of the problem size. In [23], Zhuo et al. provided parameterized and optimized designs for floating point matrix multiplication. The design was implemented on Xilinx Virtex-II Pro device and Cray XD1 high performance reconfigurable computing system. The designs achieved a sustained performance of 2.06 GFLOPS on a single node of XD1.

6.

CONCLUSIONS

This paper has attempted to review the research on the implementation matrix multiplication routine on FPGAs described in the open literature. This paper serves as a resource to the design engineers and researchers working in this area. The contributions of this paper are that a state-of-the-art review has been presented with an extensive literature survey covering various aspects of matrix multiplication.

REFERENCES 1.

Maxfield, C. M., 2004, The Design Warrior’s Guide to FPGAs, Elsevier Publishers, New York, USA. 340



2.

3.

4.

5.

6.

7.

8.

9.

10.

11.

12.

13.

14.

Todman, T.J., Constantinides, G.A., Wilton, S.J.E., Mencer, O., Luk, W. and Cheung, P. Y. K., 2005, “Reconfigurable computing: architectures and design methods”’, IEE Proceedings: Computer and Digital Techniques, Vol. 152, No. 2, pp 193–207. Qasim, S. M. and Abbasi, S. A., 2006 “A New approach for Arbitrary Waveform Generation using FPGA and Orthogonal Functions”, Proc. of 6th IEEE Int. Workshop on Real-time System on Chip Applications, Dec. 2006, pp. 28-32. Amira, A., Bouridane, A., Milligan, P. and Sage, P., 2000, “A High Throughput FPGA Implementation of a Bit-Level Matrix Product”, Proc. of IEEE Workshop on Signal Processing Systems, Oct. 2000, pp. 356–364. Amira, A. and Bensaali, F., 2002, “An FPGA based parameterizable system for matrix product implementation,” Proc. of IEEE Workshop on Signal Processing Systems, Oct 2002, pp. 75-79. Bensaali, F., Amira, A. and Bouridane, A., 2003, “An FPGA based coprocessor for large matrix product implementation,” Proc. of IEEE Int. Conf. on Field Programmable Technology, Dec. 2003, pp. 292–295. Mencer, O., Morf, M. and Flynn, M. J., 1998, “PAM-Blox: High performance FPGA design for adaptive computing,” Proc. of IEEE Symp. on FPGAs for Custom Computing Machines, April 1998, pp. 167–174. Amira, A., Bouridane, A. and Milligan, P., 2001, “Accelerating Matrix Product on Reconfigurable Hardware for Signal Processing,” Proc. of 11th Int. Conf. on Field Programmable Logic and Applications, Aug. 2001, pp. 101-111. Jang, J., Choi, S. and Prasanna, V. K., 2002, “Area and Time Efficient Implementations of Matrix Multiplication on FPGAs,” Proc. of IEEE Int. Conf. on Field Programmable Technology, Dec. 2002, pp. 93–100. Jang, J., Choi, S. and Prasanna, V. K., 2005, “Energy and Time Efficient Matrix Multiplication on FPGAs,” IEEE Trans. on VLSI Systems, Vol. 13, No. 11, pp. 1305–1319. Belkacemi, S., Benkrid, K., Crookes, D. and Benkrid, A., 2003, “Design and implementation of a high performance matrix multiplier core for Xilinx Virtex FPGA”, Proc. of IEEE Int. Workshop on Computer Architectures for Machine Perception, March 2003, pp. 156–159. Shang, L., Kaviani, A. and Bathala, K., 2002, “Dynamic power consumption in Virtex-II FPGA family”, Proc. of ACM/SIGDA 10th Int. Symp. on Field Programmable Gate Arrays, Feb. 2002, pp. 157–164. Jang, J., Choi, S. and Prasanna, V. K., 2002, “Energy efficient matrix multiplication on FPGAs,” Proc. of 12th Int. Conf. on Field Programmable Logic and Applications, Sept. 2002, pp. 534–544. Choi, S., Prasanna, V. K. and Jang, J., 2002, “Minimizing energy dissipation of matrix multiplication kernel on Virtex-II,” Proc. of SPIE, Vol. 4867, July 2002, pp. 98–106.



15. Choi, S., Scrofano, R., Prasanna, V. K. and Jang, J., 2003, “Energy efficient signal processing using FPGAs,” Proc. of ACM/SIGDA 11th Int. Symp. on Field Programmable Gate Arrays, Feb. 2003, pp. 225–234. 16. Jianwen, L. and Chuen, J. C., 2004, “Partially Reconfigurable Matrix Multiplication for Area and Time Efficiency on FPGAs,” Proc. of Euromicro Symposium on Digital System Design, 2004, pp. 244–248. 17. Ho, C. H., Leong, M. P., Leong, P. H. W., Becker, J. and Glesner, M., 2002, “Rapid Prototyping of FPGA Based Floating Point DSP Systems,” Proc. of 13th IEEE Int. Workshop on Rapid System Prototyping, July 2002, pp.19–24. 18. Underwood, K. D. and Hemmert, K. S., 2004, “Closing the Gap: CPU and FPGA Trends in Sustainable Floating-Point BLAS Performance,” Proc. of 12th Annual IEEE Symposium on Field-Programmable Custom Computing Machines, April 2004, pp. 219–228. 19. Ligon III, W. B., McMillan, S., Monn, G., Schoonover, K., Stivers, F. and Underwood, K. D., 1998, “A Re-evaluation of the Practicality of Floating-Point Operations on FPGAs,” Proc. of IEEE Symposium on FPGAs for Custom Computing Machines, April 1998, pp. 206-215. 20. Dou, Y., Vassiliadis, S., Kuzmanov, G. K. and Gaydadjiev, G. N., 2005, “64-bit Floating point FPGA matrix multiplication,” Proc. of ACM/SIGDA 13th Int. Symp. on Field Programmable Gate Arrays, Feb. 2005, pp. 86–95. 21. Govindu, G., Zhuo, L., Choi, S. and Prasanna, V., 2004, “Analysis of HighPerformance Floating-Point Arithmetic on FPGAs,” Proc. of 18th Int. Parallel and Distributed Processing Symposium, April 2004, pp. 149–156. 22. Zhuo, L. and Prasanna, V. K., 2004, “Scalable and Modular Algorithms for Floating-Point Matrix Multiplication on FPGAs,” Proc. of 18th Int. Parallel and Distributed Processing Symposium, April 2004, pp. 92–101. 23. Zhuo, L. and Prasanna, V. K., 2007, “Scalable and Modular Algorithms for Floating-Point Matrix Multiplication on Reconfigurable Computing Systems,” IEEE Trans. on Parallel and Distributed Systems, Vol. 18, No. 4 pp. 433–448.


SURFACE-MOUNTED PMBM SENSORLESS CONTROL BASED ON EMF EXTRACTION AND SIMPLE DIGITA FILTERING CASE I: PERMANENT MAGNET SYNCHRONOUS MOTOR

Osama M. Arafa1, Osama A. Mahgoub2, Mona N. Eskander3 1,3 Researcher, Prof. Researcher Power Electronics Dept., Electronics Research Institute, National Research Center, Dokki, Cairo 12622, Egypt , [email protected] , [email protected] 2: Prof. Faculty of Engineering, Elect. Power and Machines, Cairo Uni. Giza, Egypt [email protected]

ABSTRACT In this paper, a powerful method for sensorless control of permanent magnet brushless motors (PMBM) is explained. The method doesn’t include any modification to the classical construction of such motors and can be easily integrated within the drive electronics. It is applicable to both motor types, i.e. those with sinusoidal or trapezoidal back EMF. The method is independent of excitation current profile. It is also virtually independent of system parameters. The back EMF waveform is extracted on-line using simple and integration-free algebraic expression through a newly proposed measurement setup. Appropriate simple digital filters to make it directly usable for calculating rotor position and speed are then used to filter the extracted waveforms. This paper discusses application of the method on the PMBM whose back EMF waveforms are sinusoidal; its application on the trapezoidal back EMF machine is left for a separate paper. System performance using the proposed method is tested using simulation and experimentation. Results obtained in both cases ensure the effectiveness of the proposed technique.

KEY WORDS Sensorless control, permanent magnet synchronous motor.

I-

INTRODUCTION

Permanent magnet brushless motor (PMBM) is widely used as a basic element in high performance servo drive systems due to its high efficiency and torque density. Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007

SURFACE-MOUNTED PMBM SENSORLESS CONTROL BASED ON EMF EXTRACTION

Recently, a lot of research has been focused on the sensorless control of PMB motors, which normally operates using position and speed sensors. Two different types of surface-mounted (SM) PMBM are available, the sinusoidal back EMF type and the trapezoidal back EMF type. The first type is frequently referred to as permanent magnet synchronous motor (PMSM) and the second type is frequently referred to as brushless DC motor (BDCM). Sensorless control is an attractive solution to mechanical sensor problems, which include extra cost, wiring complexity, reduced reliability and demand for additional mounting space. The authors of this study believe that space limitations [1], incompatibility with some harsh working environment (e.g. inside refrigerant compressors) [2] and weight reduction considerations (e.g. aviation industry) represent the most apparent motives behind sensorless control research efforts. This is particularly true if one of the potential fields of application of such motors is considered, i.e. disk drives where miniaturization and reliability come as top design priority [3]. Sensorless control schemes for surface-mounted (SM) PMBM are mostly depending on direct or indirect back EMF detection since the other schemes employing inductance changes with rotor position are applicable only to interior magnet (IM) PMBM [3]. Previous works developed sensorless control schemes for BDCM based on direct EMF detection include [3],[4],[5] and [6] where EMF zero crossings provide position and speed information six times the number of pole-pairs every rotor cycle. The major drawback of such technique is the availability of speed information once every one sixth of the electrical cycle. This further degrades the drive’s dynamic performance, which is readily suffering from large torque dips during commutation intervals. The same concept equally applies to the alternative method proposed in [1], which extracts the rotor position information from the conduction state of the inverter’s free wheeling diodes. Trying to provide continuous position and speed information, Hamid A. Toliyat et al suggested a novel scheme for indirect back EMF detection [7], [8]. The scheme is well functioning even at very low rotational speed. However, it involves some modification to the classical structure of such motors regarding the coils’ arrangement and the number of leads to be brought out of the motor. This is generally inconvenient for automatic winders; also phase-splitting requires a nonstandard production of PMSM. The scheme is therefore restricted to customdesigned PMSM and can’t be extended to the case of BDCM as well. Serving the same target, a comparative study [10] have proven that substantial improvement in BDCM performance can be realized by adopting an extended Kalman filter (EKF) algorithm utilizing the open phase voltage measurement in its correction mechanism [9]. According to the proposed scheme, continuous position and speed estimation is possible. Unfortunately the implementation is dedicated to BDCM only and can’t be extended to PMSM. Silverio Bolognani, et al [11] introduced a high-performance sensorless PMSM drive based on EKF covering most crucial aspects like the impact of motor parameter changes on performance, and the randomness of sense of 344 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007

Osama M. Arafa , et al

rotation upon startup etc. The proposed remedies are excellent and well applicable in this current case; therefore they will not be repeated. However EKF methods are characterized by intensive math manipulations and are some how difficult to tune. Parallel to the benefits gained from inductance dependence on rotor position in sensorless control of IM PMBM. It is quite normal to make benefits from its independence of rotor position in sensorless control of SM PMBM. In this study a trial shall be made to gain benefits from the special construction of PMBM as is, thus avoiding any structural modifications. A unified and general approach for position and speed estimation capable of handling both PMSM and BDCM is introduced. The proposed technique is also independent of the excitation current profile since PMSM can be excited using sinusoidal or rectangular current waveforms. The proposed technique is based on a newly proposed method for indirect back EMF detection. Part I through VI of this paper has the theory and part VII has the simulation and results while part XX introduces the experimental results.

II- SM PMBM MODEL The effective air gap of SM PMBM is virtually large and uniform due to the considerable magnet depth in the radial direction and due to the fact that magnet materials has permeability close to that of free air [12]. The large air gap helps maintain the magnetic circuit virtually free from saturation and keeping the winding’s inductance nearly constant and independent of rotor position and excitation current. Current control loop can further support the assumption of the linearity of magnetic circuit by enforcing current limits that maintain magnetic circuit’s linearity. A quite general voltage equation of both surface-mounted PMBM motor types can be written as:

0 ⎤⎡pia⎤ ⎡ea⎤ ⎡va⎤ ⎡rs 0 0⎤⎡ia⎤ ⎡L−M 0 ⎢v ⎥=⎢0 r 0⎥⎢i ⎥ +⎢ 0 L−M 0 ⎥⎢pi ⎥+⎢e ⎥ ⎢ b⎥ ⎢ s ⎥⎢ b⎥ ⎢ ⎥⎢ b⎥ ⎢ b⎥ ⎢⎣vc ⎥⎦ ⎢⎣0 0 rs ⎥⎦⎢⎣ic ⎥⎦ ⎢⎣ 0 0 L−M⎥⎦⎢⎣pic ⎥⎦ ⎢⎣ec ⎥⎦

(1)

where : va,vb,vc : stator terminal voltage of phases a, b and c respectively in Volt. ia, ib, ic : currents of phases a, b and c in Amp p : differential operator. L : self-inductance per phase in Henry. M : mutual inductance in Henry. rs : stator resistance per phase in Ohm. ea, eb , ec : back EMF of phases a, b and c respectively in Volt, they are given by



⎡ cos θ r ⎢ 2 ⎢ cos θ r − 3 π ⎢ cos θ r + 2 π 3 ⎣

⎤ ⎥ ⎥ ⎥ ⎦

⎡e a ⎤ ⎢ e ⎥ = λ' dθ r m ⎢ b⎥ dt ⎢⎣ e cs ⎥⎦ Where: λ 'm

: the EMF constant in volt.s/elect. rad.

θr

: Electrical angular displacement in rad.

dθ r dt

: Derivative of displacement rad/s.

( (

) )

(2)

Electromagnetic torque Te is given in N.m by :

Te = λ' m

( )( i P 2

3 2 a

cosθ r +

where P

3 2

(ib − ic ) sinθ r )

(3)

: Number of poles.

The mechanical system dynamics are given by: Te = TL + Bω rm + J pω rm (4) : electromagnetic developed torque and load torque in N.m Te , T L

ωrm

respectively. :damping coefficient in N.m.s/rad. : rotor’s mechanical speed , ω = P . d θ r . rm

J

moment of inertia in N.m.s2/rad

B

dt

Equation (1), which is valid under the condition of isolated neutral point, shows that the three motor phases seem as if they were three independent coils of equivalent inductance (L-M). The three windings subject to the flux of a rotating magnet which induces a back EMF of value eph . III- BACK EMF DETECTION III-1 Theory i

Rx

Lx

Ext. Inductor A

P

Rph

Lph

EMF

Phase Winding

N Neutral

Fig. 1: External small inductor connected to each phase of PMBM In the circuit shown in Fig.1, let the circuit part from P to N represent one of these three independent coils of the PMBM, a small external coil is placed between points



I (inverter’s feeding point) and P (motor terminal). Writing the voltage drop equations for the two parts:

VPN = R ph .i + L ph .di / dt + e Vip = Rx.i + Lx.di/ dt Where: VPN Vip e Lph Rph Lx,Rx i

(5) (6)

: voltage drop from P to N in Volt : voltage drop from P to I in Volt. : back EMF generated due to permanent magnet excitation in Volt. : equivalent inductance (L-M) of the phase winding in Henry. : resistance (rs) of the phase winding in Ohm. : inductance and resistance of the external inductor respectively. : phase current in Amp.

Solving the two previous equation for e yields: e=−

L ph Lx

Vip + (

R x L ph Lx

− R ph )i + V PN

(7)

The previous equation can be applied for each motor phase independently. For PMSM a total of four voltage measurements is needed since the EMFs of the three phases ideally sum to zero. III-2 Conditions of validity Equation (7) can be used on line to calculate the instantaneous value of the back EMF waveform to a reasonable degree of accuracy if the following presumptions could be satisfied. 1. The motor magnetic circuit is operated in the linear region. Recall that SM PMBM easily satisfies this condition. 2. Inductive voltage drop across windings’ inductance are too large compared with the resistive voltage drop. This is also real when current shaping is done using a high frequency switching process. This condition is to ensure that the method is virtually insensitive to resistance variation due to thermal drift. 3. Differential voltage measurement at the selected points in Fig.1 is simultaneous and fast enough relative to the current control switching frequency. This done by simultaneous sampling 4. The external inductors are identical , made with low-loss core, with constant inductance, minimum resistance and should have no-magnetic coupling



IV-

NOISE ANALYSIS & PROPOSED FILTER

In the described system, noise is expected as: ♦ white noise due to surrounding environement. ♦ impulse noise due to high frequency current switching. White noise (similar to white light) occurs with similar amplitudes over a wide frequency spectrum. Impulse noise is a momentary perturbance, limited in the frequency band, and often at saturation or maximum signal height permitted. While simple finite impulse response (FIR) digital filters can be used to effectively filter white noise, impulsive noise can’t be filtered by FIR without introducing large phase shift and amplitude attenuation. Median filters are more suitable to deal with impulsive noise since they introduce very little phase shift and ideally no magnitude attenuation. Also it’s particularly useful in preserving the flat top of the back EMF waveform in BDCM types. A combination of both types of filters is used in the next simulation runs to produce minimum position estimation error. IV.1 Digital FIR Filter The mathematical formula describing this filter (Fig. 2) is given by the following equation: N

Y = ∑ ai X (n − i)

(8)

i =0

Where Y is the current filter output, the X(n-i) are current or previous filter inputs. The ai’s are the filter’s feed forward coefficients corresponding to the filter’s zeros. If this filter is excited with an impulse, the output is present for only a finite (N) number of computational cycles. Due to its all zero structure (no poles at all), the FIR filter has a linear phase response in most standard filter applications.

Fig. (2): FIR Filter implementation. IV.2 Median Filter For an input signal V, the vector U is an n-element vector whose elements are taken from the elements of V by a window sliding over the input signal V where n is the window size of the median filter. Normally n is chosen to be an odd number, then if Y = median (U ) and S is the sorting of the elements of U by value. The output Y is the central element of the sorted vector S as follows 348 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


(9) Y = S ((n + 1) / 2) For on-line filtering this implies that the last n samples of a noisy signal should be always kept in memory and refreshed in a first-in first-out dropping style.

V- DIRECT EMF-BASED POSITION & SPEED CALCULATION The transformation of the extracted three phase EMF from stationary frame to the two phase stationary frame results normally in a vector whose magnitude represents the rotor electrical speed and whose angle represents the electrical rotor position. V.1 Related Problems Since both vector’s magnitude and phase would be affected by filtering, two problems arise. • A speed calculation error due to amplitude attenuation caused by filter. • A position calculation error due to phase lag introduced by filter. Another problem is invited if the drive operation requirements involve speed reversing. As the speed gets closer to zero, position estimation error gets higher due to poor signal to noise ratio, and eventually any trustable information about the real rotor position may be lost. Therefore a modified estimation technique should be provided for position and speed in the vicinity of zero speed under speed reversal conditions. A similar situation at drive’s startup can be fixed by open loop motor acceleration. V.2 Problems Fixing • Speed Error Compensation The first problem can be resolved by considering a hypothetical value of the permanent magnet flux coefficient λ' m . This hypothetical coefficient’s value is initially assumed equal to the machine EMF constant upon drive startup. Then it can be tuned on line by properly differentiating the position signal over integral electrical cycles to get the actual speed under steady state conditions. The hypothetical magnet flux coefficient is then refreshed using a fresh value of the motor actual speed and a fresh value of the EMF’s vector magnitude in the α-β frame after filtering. • Position Error Compensation Through off-line simulation, the relation between phase lag introduced by any given filter combination with known parameters and speed can be predicted starting from standstill and up to rated motor speed. It has been shown by simulation that it has a linear change profile, which makes on-line compensation easier. If for given filter parameters and at given speed the phase lag is δ, then the following transformation 349 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


given in equation (10) can be used to revert the filtered back EMF vector obtained after filtering Eα’β’ to its companion Eαβ before filtering to correct the position error. ⎡ E α ⎤ ⎡ cos δ ⎢E ⎥ = ⎢ ⎣ β ⎦ ⎣ − sin δ

sin δ ⎤ ⎡ E α ' ⎤ ⎢ ⎥ cos δ ⎥⎦ ⎣ E β ' ⎦

(10)

• Estimation at speed’s zero-crossing The motor speed changes linearly with time in the vicinity of zero speed during speed reversal under almost all loading conditions. The rate of change can be captured on-line; extrapolation can be used to predict the motor speed once the estimated speed drops under a predefined threshold. The rotor position can be estimated by simple integration process (e.g. trapezoidal rule) applied on the predicted speed.

VI- EXPERIMENTAL RESULTS VI.1 On-line EMF detection Due to space limitations, simulation results are not shown, only some experimental results will be shown. The motor used is 2-poles, 400 watt, 5800 R.P.M PMSM. In Fig. (3) the upper yellow colored trace shows the differential voltage measurement across the external inductor connected to phase “A”, the high frequency voltage component is due to the inductor reaction to the fast switching process of the inverter while the lower frequency voltage changes is due to the inductor reaction with the fundamental component of the winding current. The next blue colored trace shows the terminal voltage measured at the motor terminal of phase A. The trace resembles portions broken out of a sinusoid wave and impaired by some offset and high frequency noise. The third pink colored trace shows the extracted EMF resulting from applying the algebraic filtering formula given in equation (7), the improvement established in the waveform is evident since it gets closer to a sinusoid without introducing any phase shift or magnitude attenuation. The fourth green colored waveform shows the EMF resulting after applying simple Finite Impulse Response (FIR) digital filtering, it is clear that some phase shift is now introduced due to this additional filtering. The filter’s math is executed on line upon acquiring each new fresh sample. Fig. (4) gives a closer view on the time axis of Fig. (5). Fig. (7) further illustrates phase lag due to additional filtering. Consequently, the shift is reproduced in the estimated position as shown in Fig. (8), where the green colored trace represents the measured position, the blue colored trace represents the position 350 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


estimated from the filtered EMF and the yellow colored upper trace represents the algebraic difference between them. The apparently high error pulses near the 2π transitions actually have a low value (by dropping 2π representing the complete cycle). VI.2 Open Loop Acceleration Although the proposed technique is capable of starting the drive in sensorless mode from zero speed, the resulting sense of rotation may conform to the commanded one or not based on the initial rotor position (unknown). Therefore, the first step in any running session is to provide an open loop acceleration period to serve two purposes: • Guiding the sense of rotation to ensure conformance with the commanded one. • Reaching a minimum speed to make the EMF detection fairly accurate Fig. (7) illustrates the time profile of open loop acceleration, the upper trace shows the actual rotor position, while the lower trace shows the rotor position considered within the open loop acceleration routine to provide the excitation voltage.. When the acceleration rate is well tuned to the motor and load inertia, the error between the assumed position and the actual position decreases and the expected transient due to switching from open loop to closed loop sensorless acceleration gets more tolerable.



Fig.( 3): Voltage waveforms / phase “A”

Fig. (4): A closer view of Fig.(5

Fig. (5): Phase shift added by subsequent digital Filtering

Fig. (6): Phase lag due to filtering



Fig. (7): Open loop acceleration, first cycles

Fig. (9): Speed step-response

Fig. (8): Rotor position during normal sensorless operation

Fig. (10): Three-phase stator current profile in transient



VI.3 Normal Sensorless Operation Once the motor acquires a reasonable speed, the estimated speed and position can be directly used for closing the current and the speed control loops. Fig. (8) gives a comparative vision of the actual (measured) rotor position and the estimated with motor loaded and running in sensorless closed loop control mode with phase shift compensator enabled. The high frequency ripples superimposed on the yellow trace are merely due to picked up noise. The third trace represents the error between the actual and the estimated position. The picked up noise is reproduced in the error signal accordingly, the coincidence between the two signals are otherwise excellent. Fig. (9) illustrates the speed step response, the yellow is the speed command, and the blue is the measured speed response. The open loop acceleration is clear at the early beginning where the motor builds up some speed, during which the estimated position gets more accurate and the sense of rotation is defined (clockwise rotation is selected here). Control is then transferred to closed loop mode, first the speed reference was step changed from zero to 59.6% of rated speed. Then a periodic change of speed reference from 59.27% to 24.6 % of rated speed and from the lower to the higher again is enforced. The motor takes longer to decelerate to the lower speed reference because it is left for free retardation during this period (neither dynamic nor regenerative braking is provided). Fig. (10) illustrates the dynamic reaction of the three-stator currents due to the step increase in speed reference.

VII- CONCLUSION The advantages obtained by using simple digital filtering along with effective extraction of EMF from excited windings for PMSM sensorless control is discussed in this paper. Based on the concept presented here, power cable only is needed to connect the motor to the drive electronics; extra taps or winding modifications are not needed. Effective fixing procedures for canceling position and speed errors due to filtering are also proposed. These procedures try to avoid static dependence on motor parameters by on-line refreshing of those parameters affecting the estimated speed. The only noticeable drawback is the extra voltage amount that should be supplied by the inverter to compensate for voltage drop across the external inductors; selecting inductors of proper values can minimize this drawback. Very small inductor value would result in a low measurement accuracy due to poor signal to noise ratio. A compromise between reducing additional voltage and measurement accuracy improvement is best deciding the external inductor value. In the simulation a value of 10% of motor equivalent inductance per phase for external inductor was used. The on-line math power requirement can be satisfied by a suitable cheap microcontroller solution. 354 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


VIII-

REFERENCES

[1] Ogasawara S., Akagi H., 1991, "An Approach to Position Sensorless Drive for Brushless DC Motor", IEEE Transactions on Industry Applications, September/October, PP 928-933. [2] Kinichi IIzuka, Hideo Uzuhashi, Minoru Kano, Tsunehiro Endo, 1985 ‘Microcomputer Control for Sensorless Brushless Motor", IEEE Transactions on Industry Applications, Vol. IA-21, No. 3, May/June, PP 595-601. [3] L. Cardoletti, A. Cassat, M. Jufer, 1990, "Sensorless Position and Speed Control of a Brushless DC Motor from Start-up to Nominal Speed" ,EPE Journal, Vol.2, No.1, PP 25-34. [4] D. M. Erdman, H. B. Harms, J. L. Oldenkamp, "Electronically Commutated DC motors for the Appliance Industry", IEEE-IAS Conf. Rec., Oct. 1984, PP 13391345. [5] N. A. Demerdash, T. W. Nehl, E. Maslowski, "Dynamic modeling of brushless dc motors in electric propulsion and electromechanical actuation by digital techniques", IEEE IAC Conference Rec., 1980, PP 570-579. [6] P. Zimmermann, 1982,"Electronically commutated D.C. drives for machine tools", Motorcon Proceedings, PP 69-83. [7] Deepak S. Shet, Hamid A. Toliyat, Thomas A. Nondahl, "Position Sensorless Control of Surface Mount Permanent Magnet AC (PMAC) Motors at Low Speeds", Proceeding of the American Control Conf. San Diego, California, June 1999 PP 2141-2142. [8] Khawaja M. Rahman, Hamid A. Toliyat, "Sensorless Operation of Permanent Magnet AC (PMAC) motors with Modified Stator Windings”, Industry Application. Conf., 31st Annual IAS Meeting, 1996, vol.1 PP 326-333. [9] Paul Kettle, Aengus Murray, Finbarr Moynihan, 1998, "Sensorless Control of a Brushless DC Motor using an Extended Kalman Estimator", PCIM'98 Intelligent Motion. Nurenberg, May, Proceedings, PP 385-392 . [10] Petre Minciunescu, Tom Flint, Finbarr Moynihan, Paul Kettle, "Sensorless Control of Brushless DC Motors Using Extended Kalman Filter Estimator and Back EMF Integration Algorithm: A Comparison", Embedded Control Systems, Analog Devices Inc. Application Note



[11] Silverio Blognanai, Mauro Zigliotto, Marco Zordan, 2001, "Extended-Range PMSM Sensorless Speed Drive Based on Stochastic Filtering", IEEE Transactions on Power Electronics, Vol.16, No.1, Jan., PP 110-116. [12] T.W. Nehl, N.A. Demerdash, F.A. Fouad, 1985, "Impact of Winding Inductance and Other Parameters on the Design and Performance of Brushless DC motors", IEEE Transactions on Power Apparatus and Systems, Vol. PAS-104, No. 8, August, PP 2206-2211. [13] Peter S. Maybeck, 1979., "Stochastic Models, estimation, and Control” Volume 1, Academic Press Inc (AP) New York. [14] Katsuhiko Ogata, 2001,"Modern Control Engineering", Prentice-Hall Engineering-4th Edition.


A NEW NUMERICAL CALIBRATION MODEL FOR INDUSTRIAL EDDY CURRENT INSPECTION SYSTEMS

Ayed Algarni, Ibrahim Elshafiey, and Majeed Alkanhal Department of Electrical Engineering, King Saud University, Riyadh, Saudi Arabia [email protected]

ABSTRACT Eddy current techniques are widely used to inspect electrically conducting materials during the manufacturing process as well as the operation phase. Implementations of these techniques are found in nondestructive evaluation of aging aircraft, storage tanks, and natural gas pipelines. Recent industry trends are directed to increasing the sensitivity of detecting signals associated with subsurface cracks, while maintaining high spatial resolution for proper identification of multiple flaws. This paper presents a new tool to enhance the performance of eddy current inspection, compared to conventional implementation depending on single-frequency excitation and manual scanning of the specimen. The technique depends on the acquisition of eddy current signal in the form of c-scan images, which is becoming feasible with recent advances in instrumentation and in signal processing methods. Multifrequency excitation protocol is used, in combination with specially designed digital image analysis algorithms to enhance the acquired data and produce c-scan images of subsurface features. Simulation results are presented to demonstrate the feasibility of this tool in terms of increasing the reliability and sensitivity of industrial applications of eddy current nondestructive evaluation systems. Keywords: Nondestructive Evaluation, Eddy Current, and Image Analysis.

INTRODUCTION Eddy current (EC) testing technique works on the principle of electromagnetic induction. A coil is excited with sinusoidal alternating current to induce eddy currents in an electrically conducting material such as steel, aluminum, etc. The change in coil impedance that arises due to distortion of eddy currents at discontinuities and the associated magnetic flux linkages is measured and correlated with material features. The locus of impedance change formed during the movement of an eddy current probe coil over a test sample forms the eddy current signal. The amplitude of eddy current signal provides information about the severity of the



defect, while phase information with respect to a certain reference characterizes defect location or depth. Defects that cause maximum perturbation to eddy current flow produce large eddy current response and hence high sensitivity detection. On the contrary, defects that are parallel to eddy current flow are detected with poor sensitivity. Eddy current density is not uniform along the material depth. It is greatest on the surface and decreases monotonically with depth, due to the skin effect. In order to detect shallow defects in a material and also to measure thickness of thin sheets, high frequencies are to be used. On the other hand, detecting buried subsurface defects and testing thick materials, low frequencies should be adopted [1, 2]. Presenting inspection data as images provide a natural way for signal interpretation to inspection personnel. Various techniques have been developed to put eddy current inspection data in the form of c-scan images. Impedance components of eddy current probe can be associated with encoding data provided by an encoder to cast the data as a group of A-scan lines which can be arranged as two dimensional images [3]. New trends are directed to obtain a map of magnetic flux and present it as images. Magneto-optic/eddy current technology was suggested and developed for aircraft inspection applications [4]. Newly developed giant magneto-resistive sensors GMR have also been suggested for magnetic field mapping [5].

EDDY CURRENT FORMULATION Numerical modeling is invoked for applications for which analytical solutions cannot be performed. Mathematical description of eddy current physical process is expressed in the form of simulation programs [6]. The first step for developing a numerical model is to formulate differential equations that describe all the interactions, with the help of Maxwell's equations.

∂D ∂t ∂B ∇× E = − ∂t ∇⋅B = 0 B = µH J = σE

∇× H = J +

(1) (2) (3) (4)

Where H is the magnetic field intensity vector, J is the current density vector, D is the electric flux density vector, E is the electric field intensity vector, and B is the magnetic flux density vector. The parameter µ is the magnetic permeability, while


Ayed Algarni , et al

is the electric conductivity. A solution can be obtained by introducing potential functions such that

B = ∇× A ∂A E=− − ∇V ∂t

(5) (6)

Where A is the magnetic vector potential and V is the electric scalar potential. The differential equations for magnetic vector potential can be derived from the above equations as

∇ ×ν (∇ × A) − ∇ν (∇ ⋅ A) + σ

∂A + σ∇V = 0 (7) ∂t

⎛ ∂A ⎞ ∇ ⋅ ⎜σ + σ∇V ⎟ = 0 ⎠ ⎝ ∂t

(8)

Where the reluctivity ν is the inverse of permeability. For source regions, Equation 7 becomes

∇ ×ν (∇ × A) − ∇ν (∇ ⋅ A) = J s

(9)

Where Js is the current density in the source. The potential functions can be found by solving equations (7-9), with the appropriate boundary conditions. The finite element (FE) method is one of the best well known numerical methods that can be used to solve this problem. The FE does not offer a solution to the diffusion equation directly. Instead, the solution is obtained at discrete nodes in the solution region by formulating an energy functional equivalent to energy balance of Equations 7-9. On performing the minimization of energy functional, the problem is cast in matrix form for each element given as

[S ]T = [Q]

(8)

The matrices are represented by complex numbers. For each element in the mesh the [S] matrix is formed based on geometry as well as electric and magnetic properties of this element. [Q] is a vector formed from the complex current density at source elements. T vector represents the unknown magnetic and electric potentials, A and V. After forming the element matrix equations for all elements in the finite element region, a global matrix equation containing all of the elements matrices is solved after incorporating the necessary boundary conditions. The adopted method depends on solving circuit-coupled stranded coil as a source, for which the resulting global matrix is unsymmetric. 359 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


COMPUTATIONAL MODEL Analysis is performed under ANSYS environment, multi-physics finite element package [7], which allows modeling related to structural, thermal, mechanical phenomena in addition to electromagnetic formulation. A three dimensional model is built which as shown schematically in Figure 1. The model consists of several elements representing: coil, object under test, surrounding free space, and the interface boundary to infinity. The coil is designed to be of 10-mm outer radius, 5mm inner radius, and 5-mm height. The object is composed of aluminum with 100mm radius and 10mm thickness, and the lift-off between the probe and the object is of 1.1-mm. Two cracks are modeled. Type I is a surface crack of 1mm×1mm×1mm size, and Type II is a subsurface crack of 1mm×2mm×2mm size which is present 2 mm under the upper surface of the object. For accurate analysis the model building take into account that the center of excitation is the center of the model and the thickness of the open boundary almost equal to the thickness of the model.

1- Infinite Elements 2- Free Space 3- Probe 4- Object Under Test 5- Crack

Figure 1. Schematic finite element model of the numerical current inspection system. Among the element library in Ansys, with more than 100 element formulations, three elements are used: SOLID97, CIRCU124, and INFIN111. SOLID97 models 3-D magnetic fields based on the magnetic vector potential formulation for lowfrequency magnetic field analysis. The element is defined by eight nodes, and has up to five out of six defined degrees of freedom per node: the magnetic vector potential (AX, AY, AZ), the (VOLT) corresponding to time-integrated electric potential in classical formulation or electric potential in solenoidal formulation, the electric current (CURR), and the electromotive force (EMF). Two formulation options are available: classical and solenoidal. The adopted solenoidal formulation 360 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


automatically satisfies the solenoidal field condition. This element has a brick geometry that and the same node number can be applied to more than one node, to form a tetrahedral-shaped element, wedge-shaped element and a pyramid-shaped element. It is used to model the object, crack, and air around the model. It is also used in modeling the probe since it has electromagnetic-circuit field coupling capability. CIRCU124 is a general circuit element applicable to circuit simulation. The element may also interface with electromagnetic finite elements to simulate coupled electromagnetic-circuit field interaction. The element has up to 6 nodes to define the circuit component and up to three degrees of freedom per node to model the circuit response. INFIN111 is used to model an open boundary of a 3-D unbounded field problem. A single layer of elements is used to represent an exterior sub-domain of semi-infinite extent. This element is defined by either 8 or 20 nodes and has 3-D magnetic vector potential capabilities. After assigning the element attributes and material properties and setting mesh controls, the process for generating a mesh of nodes and elements is started. A number of 84060 elements is generated for acceptable accuracy, which is in Figure 2. The current and emf DOFs are coupled for all nodes of the probe. The boundary conditions are then enforced to the exterior INFIN111 elements.

The excitation is applied and the model is solved for various crack positions and also corresponding to various excitation frequencies. Execution is performed on a PC platform with 3.4GHz Pentium® CPU, and 2 GB RAM. Each run corresponding to 11 frequency load steps takes about 16 hours to be performed. The required storage for element results is about 5GB. Next section discusses the obtained results.



RESULTS Simulation results were conducted corresponding to the following 11 inspection frequency values: 100 Hz, 1 kHz, 5 kHz, 10 kHz, 20 kHz, 50 kHz, 100 kHz, 300 kHz, 500 kHz, 1M Hz, and 3 MHz. Sequence values of probe inductance, resistance, impedance, and phase are recorded corresponding to inspection frequencies. In order to form c-scan images, the model corresponding to various scanning positions of probe, and results are arranged in matrix array. Figure 3 shows the sensitivity of coil inductance to detect a crack. The figure presents comparison at the specified frequency values of inductance value of probe on an intact specimen and a specimen possessing subsurface crack described earlier. The figure reveals that the range from 5 kHz to 20 kHz is the most sensitive in detect the flaw.

Figure 2. Meshed model. (Coarse to be clear), showing the coil element and its surrounding elements.



Figures 4 and 5, present A-scan of the probe on a specimen with the two described crack types. The figures report resistance (Figure 4) and inductance (Figure 5) values at 16 positions with 2.08 mm distance separation between consecutive positions. Type I surface crack is present at position 2, and Type II subsurface crack is present at position 15. C-scan images of probe inductance values for Type I surface crack are presented in Figure 6, while probe resistance images corresponding to Type II subsurface crack are shown in Figure 7. The images are presented for 11 frequency values starting from the lowest frequency. The crack is adjusted at in the center of each image. Figure 8 shows images of the three components as well as the magnitude of the magnetic flux density around the subsurface crack. The images reveal the crack at around the center of each of the images.

Figure 3. Change in coil inductance in Henry due to presence of crack in the specimen.



Figure 4. Probe resistance for different frequency values in the presence of surface crack at position 2, and subsurface crack at position 15.

Figure 5. Probe inductance for different frequency values in the presence of surface crack at position 2, and subsurface crack at position 15.



(a)

(e)

(i)

(b)

(c)

(d)

(f)

(g)

(h)

(j)

(k)

Figure 6. Probe inductance images corresponding to 11 frequency values for material with the surface crack. Frequencies are: 100Hz (a), 1 kHz (b), 5 kHz (c), 10 kHz (d), 20 kHz (e), 50 kHz (f), 100 kHz (g), 300 kHz (h), 500 kHz (i), 1 MHz (j), 3 MHz (k).



(a)

(b)

(e)

(i)

(c)

(f)

(g)

(j)

(k)

(h) (h)

Figure 7. Probe resistance images corresponding to 11 frequency values for material with the subsurface crack. Frequencies are: 100 Hz (a), 1 kHz (b), 5 kHz (c), 10 kHz (d), 20 kHz (e), 50 kHz (f), 100 kHz (g), 300 kHz (h), 500 kHz (i), 1 MHz (j), 3 MHz (k).



(a)

(b)

(c)

(d)

Figure 8. Magnetic flux density images for material with the subsurface crack showing: x-component (a), y-component (b), z-component (c) and the magnitude (d).

CONCLUSIONS This paper presents the simulation results of eddy current inspection method for NDE. The results illustrate the advantages of eddy current imaging in providing natural way for signal interpretation. Probe inductance imaging is sensitive to material defects at low frequency values, while probe resistance imaging provides higher sensitivity at high frequency values. Integration of inspection images corresponding to different frequency values using digital image processing techniques can provide a robust way to reveal material defects. Imaging based on magnetic flux density maps provides a potential inspection technique that needs more investigation.

ACKNOWLEDGMENT This work was funded in part by the Research Center, College of Engineering King Saud University, Project Number 54/427.



REFERENCES [1]. H. J., Krause, G. Panaitov, and Y. Zhang, “Conductivity Tomography for Non-Destructive Evaluation Using Pulsed Eddy Current with HTS SQUID Magnetometer,” IEEE Trans. Applied Superconductivity, Vol. 13, 215-218, 2003. [2]. G. Y. Tian, A. Sophian, D. Taylor, and J. Rudlin, “Multiple Sensors on Pulsed Eddy Current Detection for 3-D Subsurface Crack Assessment,” IEEE Sensors Journal, Vol. 5, 90-96, 2005. [3]. L. Udpa, and I. Elshafiey, 2001, “WINSAS: A New tool for Enhancing the Performance of Eddy Current Inspection of Aging Aircraft Wheels,” Invited Paper, Proceeding of The 5th Joint NASA/FAA/DoD Conference on Aging Aircraft, Hyatt-Orlando, Orlando, Florida September 10-13, 2001. [4]. I. Elshafiey, 2003, “Transient Finite Element Modeling of Magneto Optic Imaging of Aging Aircraft,” 20th National Radio Science Conference, Cairo, Egypt, 2003. [5]. R. Rempt, “Scanning with Magneto resistive Sensors for Subsurface Corrosion”, Review of Progress in Quantitative Nondestructive Evaluation,” Vol. 21, 1771-1778, 2002. [6]. D. N. Dyck, G. Gilbert, B. Forghani, J. P. Webb, “An NDT Pulse Study With TEAM Problem 27,” IEEE Trans. Magnetics, Vol. 40, 1406-1409, 2004. [7]. ANSYS documentations. www.ansys.com.


SOLUTION TO INITIAL-VALUE 2nd ORDER ORDINARY DIFFERENTIAL EQUATIONS ON AN EXCEL© SPREADSHEET

Ahmed Mohameden Senior Lecturer, School of Science & Technology, University of Teesside, Middlesbrough TS1 3BA, United Kingdom, [email protected]

ABSTRACT Most engineering systems can be modeled by 2nd order differential equations (DEs); usually with some adjustments. Several programs, written in many computing languages, to solve DEs are available commercially as additional software. A selling point to some programs is their compatibility with MS Excel©, others come as an Add-In to Excel©. The assumption, sometimes stated, is that Excel© can not solve DEs. This paper presents a friendly method for solving initial-value DEs analytically on a MS Excel© workbook without the need for additional software. Upon entering the constant coefficients and boundary conditions, the type of solution is highlighted and the user can visualize the full solution accompanied by a graphical representation. Parameters can be varied and equations of interest, with solutions, saved on a separate worksheet for future use. The solver can be customized to adopt user’s terminology. This first version has been restricted to equations of the type ay” + by’ + cy = RSin(ωt + φ) where a, b, c, R, ω and φ are all constants and b or R can be zero. This tool is intended to be helpful for academics wishing to generate DEs for teaching and assessment purposes, but it may also be a handy aid to simulation and development engineers. Graphical artifacts and data misrepresentation have been explored and remedial procedures recommended.

KEY WORDS Initial-value problems, analytical solution, 2nd Order ordinary differential equations, aliasing.

INTRODUCTION Mathematical Education of Engineers (MEE) is of extreme importance to technological development; since engineering is the backbone of prosperity and quality of life. It is not surprising that MEE has attracted large funds for research and exploration. In the United Kingdom alone, consortia of universities have been set up for this purpose. The Institute of Mathematics and its Applications (IMA) runs Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007

SOLUTION TO INITIAL-VALUE 2nd ORDER ORDINARY DIFFERENTIAL EQUATIONS

regular conferences on MEE since 19941. Four years later, this need was recognized globally as suggested by the First International Conference on the Teaching of Mathematics in Greece2. The European Society for Engineering Education (SEFI) shares the same interests and concerns across continental Europe. SEFI and IMA are jointly hosting a conference on MEE, 6th – 9th April, 20083. Engineering Mathematics underpins almost all applications and spans all engineering disciplines. The depth of understanding of mathematical concepts often differentiates between peer engineers. Differential Equations (DEs) are common to all engineering and most science syllabi. Because DEs are not easily generated and solved, lecturers recourse to textbooks and laborious preparation work to develop suitable sets of DEs that illustrate particular points of interest and salient features in engineering applications. In response to this challenge, a range of software has been developed and sold commercially. Some are rather powerful (like Matlab™) and others are limited to specific tasks. However, all are additional software packages that need to be purchased, installed and users trained to manipulate. These requirements often deter potential users from benefiting of the features of such programs. Although the literature states that “EXCEL does not have the capabilities to solve systems of ordinary differential equations”4, this paper shows how it may, indeed, be used for this purpose. It presents an easy MS Excel© workbook that solves initial value 2nd order DEs with constant coefficients. The widespread use of MS Office on nearly every personal computer, at home or at work/study, justified the choice of Excel© as an application. This first version has been restricted to equations of the type

ay" + by ' + cy = RSin(ϖt + φ )

(1)

where a, b, c, R, ω and φ are all constants and b or R can be zero. The choice of equation (1) is based on the extensive use of this form in several fields of study, including electricity, vibration, dynamics, wave propagation and acoustics. In the following sections, the method of developing this program is described along with the validation process and resources and references used. Graphical representation and potential associated errors are explored before highlighting some of the features and scope for further development. Examples and illustrative screen shots are also presented.

METHOD OF INVESTIGATION Homogeneous and non-homogenous DEs have been solved symbolically through extensive transposition and algebraic manipulation of the relevant formulae. Logical functions and formatting features are used to improve the display and presentation of solutions.


Ahmed Mohameden Seven possible cases were envisaged, grouped in Table 1, according to the nature of the roots of the characteristic quadratic equation (2) and the values of R and ω in equation (1). Initial values and manipulation of the solution to the relevant particular integral lead to determination of the full solution. Several examples from textbooks513 had been used to validate this program and cross-examine the solutions.

ar 2 + br + c = 0

(2) The roots of equation (2) are noted α and β if real, when complex α is taken to be the real part and β the imaginary part. Table 1: Summary of the seven possible cases of equation (2) R root Form of Solution s: R/C ? =0 R α = β, y = (At + B)eαt α ≠ β, y = Aeαt + Beβt =0 C y = eαt(ACosβt + BSinβt) >0 R α = β, y = (At + B)eαt + CCosωt + DSinωt α ≠ β, y = Aeαt + Beβt + CCosωt + DSinωt >0 C β = ω, α = 0, y = (A + Ct)Cosβt + BSinβt β ≠ ω, y = eαt(ACosβt + BSinβt) + CCosωt + DSinωt

Soluti on Code 1 2 3 4 5 6 7

R: real numbers, C: complex numbers, ω: the frequency in equation (2)

SOLUTION & GRAPHICAL REPRESENTATION The seven cases are coded 1 – 7 as shown in Table 1. Each solution is referred to by its code. These cases are, however, grouped under 5 variations of equation (1) in the main user interface page as shown in Fig. 1. Upon entering the input parameters (Fig. 1), a set of macros and other Excel© routines carry out algorithms and identify the correct solution code, which is then highlighted. On clicking the highlighted link, a new window is opened stating the lower limit is set to zero by default. The upper limit, however, should be selected with care. Depending on the type of solution, this limit can influence the apparent shape of the curve. Embedded buttons allows navigation between pages of the solver. 371 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


The Range Factor (RF) With exception of codes 1 & 2, all solutions exhibit some form of periodic function. In these cases the upper limit is selected as a multiple of the primary period. This multiple is obtained by multiplying the primary period by any positive real number (The Range Factor). Range Factor and period are double-circled in Fig.2. The primary period is taken as

T=

2π

(3)

ϖ

when β ≠ ω, otherwise, it is:

T=

2π

(4)

β

The increment in time (along the horizontal axis) is given by the formula:

step =

(Upper value − Lower value) 1000

(5)

detailed solution, giving values of all coefficients to two decimal places and displaying a 1000 data point graph of the solution over an adjustable range. The

Input

Fig. 1: a screen shot of the main interface page (switchboard) showing the 5 variations of equation (1) and the right solution highlighted.


Ahmed Mohameden In Fig. 2, the data points are plotted without a solid line to show the drop of density of data points per cycle as the upper limit increases, as expected from this particular function. This is important to reveal potential artifacts. Graphical Artifacts and Aliasing Since the range factor indicates how many primary periods are displayed using 1000 data points, it must not be an integer multiple of 1000. If it is, each cycle will be represented by a single point, thereby reducing the real graph to one edge of its envelope. This is a side effect of Sampling Theorem in the context of analytical solutions where finite elements represent continuous intervals. The RF is related to the Sampling Rate, number of data points per cycle, by the following formula:

Sampling Rate =

1000 RF

(6)

Fig. 2 displays the correct shape of the function:

y = (2 − 6.25t )Cos3t + 0.42Sin3t

(7)

Fig. 2: a screen shot displaying the solution of a code 6 (Table 1) 2nd Order DE with a graphical representation over 10 primary periods. Notice that the spacing between adjacent data points increases rapidly while the sampling rate remains constant. 373 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


This function represents a driven undamped vibration, an out of control system, a system in resonance. Its main feature is a steady growth of amplitude with time. However, uncalculated modulation of the RF can lead to variation of the range that in turn can entail erroneous graphs. Fig.2 shows 10 cycles as expected from the RF 10. It is noteworthy that spacing between consecutive data points is expected to increase as the amplitudes increases without altering the sampling rate (equation (6)). This is manifested on the graph by an apparent drop in “density” of points towards the upper limit of the range as shown on Fig. 2. When RF is set to 1000, the graph becomes that shown on Fig.3 (left): only one “edge” of the curve’s envelope is picked. That is the data points at the amplitudes- a sampling rate of 1. At RF = 500, when both amplitudes are hit, the graph turns into two symmetrical straight lines with the same absolute value gradient of 6.25, as shown in Fig. 3 (right). Now both edges of the envelope are revealed at a sampling rate of 2. The symmetry about RF = 500 suggests that qualitatively curves at [RF = n] and [RF = (1000 – n)] are similar in shape with difference in amplitudes and a phase of either 0 or 180°- this is known in Digital Signal Processing (DSP) as aliasing14. Without drifting into DSP or even Fourier Transforms, it is easy to visualize the correct shape of the graphs by selecting a low RF (say 5), then 995 to watch out for aliasing (180° phase change). Close to RF = 500, 250 or 750, serious aliasing can occur as illustrated in Fig.’s 4 and 5.

5000

6000

y = (2 − 6.25t )Cos3t + 0.42Sin3t

y = (2 − 6.25t )Cos3t + 0.42Sin3t

4000 y

Factor = 500 RF = 500

t

0 0

1000

y 0

-5000 -10000

2000

3000 2000

Factor = 1000 RF = 1000

-2000

t 0

200

400

600

800

1000

-4000 -15000

-6000

Fig. 3: Function on Fig. 2 displayed over 1000 (left) and 500 (right) cycles. The periodicity is masked by one data point per cycle (left). The curve is reduced to the bottom outline of its envelope (left). The envelope is revealed when data is sampled at 2 points per cycle (right).


Ahmed Mohameden Patterns in Fig.’s 4 and 5 can easily be distinguished from broadly similar genuine patterns generated by selected sampling rates like those shown on Fig. 6. The main feature is that the true graph retains its overall shape- the envelope. The knots in Fig.’s 4 and 5 are clearly alien to the real graph. Introducing a low but random factor in the step formula should prevent aliasing without compromising the shape of the curve. However, due to the randomness, there is always a chance of aliasing albeit a slim one.

step =

3000

(Upper value − Lower value) × Rand () 1000 8000

y = (2 − 6.25t )Cos3t + 0.42Sin3t

6000

2000

2000

y 0

-1000

200

400

600

800

1000

RF = =499 Factor 499 or 501 501 or

y 0

t 0

y = (2 − 6.25t )Cos3t + 0.42Sin3t

4000

RF = 499.78 Factor = 499.78 500.22 oror500.22

1000

(8)

t

1200 -2000 0

200

400

600

800

1000

1200

-4000 -2000

-6000

-3000

-8000

Fig. 4: Function on Fig. 2 displayed over 500±0.22 (left) and 500±1 (right) cycles. Departure from the correct shape is visible. Slight changes engender serious aliasing.

3000

4000

y = (2 − 6.25t )Cos3t + 0.42Sin3t

3000

2000

y = (2 − 6.25t )Cos3t + 0.42Sin3t

2000 1000

1000

y 0

-1000 -2000 -3000

t 0

200

RF = 250.4 Factor = 250.4 249.6 oror 249.6

400

y 0

600 -1000 0

-2000 -3000

t 200

400

600

RF = 248 Factor = 248 252 oror 252

-4000

Fig. 5: Function on Fig. 2 displayed over 250±0.4 (left) and 500±2 (right) cycles. Again, aliasing is apparent- note the phantom “harmonic cycles”. 375 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


This random factor should be independent for each step giving rise to up to 1000 different random factors in the steps. This measure is intended to keep aliasing at bay. If, in the other hand, the same random factor is kept for the entire 1000 data points, chances of aliasing are bigger since this is the same as slightly altering the step size. Replacing equation (5) by (8) is readily done by checking the “randomized steps” box in Fig. 2. Such a single click transforms Fig. 4 into Fig. 7. The envelope of the curve has been restored and the correct shape recovered. It is noteworthy that randomizing the steps can reduce the range as a result of progressively diminishing the step size. To counteract this effect, RF can be increased until the desired range is reached. Again, due to randomness of steps, this adjustment can only be best judged by users. 4000 3000

6000

y = (2 − 6.25t )Cos3t + 0.42Sin3t

2000

y = (2 − 6.25t )Cos3t + 0.42Sin3t

4000

Factor 278.2 RF = =278.2

2000

1000 y 0

t

-1000 0

200

400

600

y 0

-2000

t 0

400

800

-2000

RF = =471 Factor 471

-4000

-3000 -4000

-6000

Fig. 6: Function on Fig. 2 displayed over 278.2 (left) and 471 (right) cycles. These are genuine graphs preserving the envelope and shape of the curve. 6000

6000

y = (2 − 6.25t )Cos3t + 0.42Sin3t

4000 2000

Factor = 499 RF =499

2000

y 0

-2000

y = (2 − 6.25t )Cos3t + 0.42Sin3t

4000

Factor = 500.22 RF = 500.22

t 0

400

-4000

800

y 0

-2000

t 0

400

800

-4000 with randomized steps

-6000

with randomized steps

-6000

Fig. 7: The same data of Fig. 4 with a randomized step. The data point distribution and the envelope are restored. 376 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007

Ahmed Mohameden

Key Features of This Solver 1. 2. 3. 4.

Easy in-situ visualization of response to parameters variation. This is particularly useful when exploring responses close to critical values (e.g. natural frequency). Desirable DEs can be saved on a separate sheet with full solution and possibility to attach comments to appropriate parameters. The benefits of familiar Excel© environment. Flexible and open for user-defined contribution.

CONCLUSIONS & FURTHER DEVELOPMENT This work has demonstrated that MS Excel© can be useful in solving DEs and, potentially, widening participation into Mathematical Education of Engineers. Tackling DEs work in the familiar environment of Excel© can be expected to help lecturers and learners engage in a more interactive learning experience. The flexibility of an Excel© spreadsheet provides opportunities for innovation and improvement of this solver. It is hoped that advanced learners, simulation and modeling engineers may find in this solver a handy aid for quick tasks as well as an opportunity to expand and develop their own skills. It is planned to extend this solver to others DEs cases, where the right hand side of equation (1) is a polynomial, an exponential or other standard functions. The following challenge will be to write compact algorithms to solve simultaneous linear DEs. The author welcomes collaboration from interested colleagues in order to both improve and speed up the delivery of the next versions. This solver is not a “re-inventing the wheels” exercise despite the full range of commercial software to do the same task and more. This solver is rather seen as an inclusion tool to further the understanding of DEs, develop innovation, planning and optimizing skills of students; and involve them in creation of the solution. Students can excel and expand their working knowledge when encouraged to switch from passive operators to active designers... from mere consumers to confident producers.

REFERENCES 1.

Mustoe, L. R., and Hibberd, S., 1995, Mathematical Education of Engineers, Clarendon Press, Oxford



2.

Proceeding of The First International Conference on the Teaching of Mathematics; University of the Aegean, Samos, Greece, John Wiley, July 1998.

3.

http://mee2008.lboro.ac.uk/

4.

Cutlip, MB. (University of Connecticut, Texas) & Shacham, M. (Ben Gurion University of the Negev), in the summary of their software Polymath at www.che.utexas.edu/cache/newsletters/fall2005_revisedpolymath6.pdf

5.

Bird, JO, and May, A.J.C., 1994, Technician Mathematics, Longman, London

6.

Stroud K.A., 2001, Engineering Mathematics, Palgrave Macmillan, Basingstoke, Evans C.W., 1997, Engineering Mathematics: a programmed approach, Chapman & Hall, London

7.

Singh, K., 2003, Engineering mathematics through applications, Palgrave Macmillan, Basingstoke

8.

Kuhfittig, P.K.F., 1978, Introduction to the Laplace transform, Plenum Press, London & NY

9.

Grove, A. C., An 1991, Introduction to the Laplace Transform and the zTransform, Prentice Hall, London

10. Dyke, P.P.G., c2000, An introduction to Laplace Transforms and Fourier Series, Springer, London 11. Bolton, W., 1994, Laplace and z-Transforms, Longman, Harlow 12. Croft, A. and Davison, R., 2004, Mathematics for Engineers : a modern interactive approach, Pearson Prentice Hall, Harlow 13. Kreyszig, E., 1999, Advanced engineering mathematics, John Wiley, New York. 14. See for instance Steven, W. S., 2003, Digital Signal Processing: a practical guide for engineers and scientists, Newnes, London & Boston. ACKNOWLEDGMENTS The author wishes to thank Moctar S. O. Mohamed from Saint Etienne University (France) and Andy Campbell from Teesside University (England) for some stimulating conversations related to this article.


TRANSFER OF DEGREE OF REPARABILITY OF TWO TIME SCALE SINGULARLY PERTURBED SYSTEMS FOR THE ASSOCIATED REGULATOR H∞ LOOP SHAPING Abdessalem bsissa (1) , Souid souad (2) (1) : College Of Technology At Taif (2) : Ecole Nationale d’Ingénieurs de Gabés, 6029 Tunisia Email [email protected] – Tel. (00966) 564574970 Email [email protected] – Tel. (00216) 97 149 139

Abstract - In this communication we present the transfer of the property of reparability of the dynamics of singularly perturbed systems for the associated regulator at the time of a synthesis H∞ loop shaping. Keywords - two time scale singularly perturbed systems, Approach H∞ loop shaping, Transfer of reparability of the dynamics. 1. Introduction and presentation of the problem Generally, the mathematical models present the reality of process imperfectly, because of the uncertainties and the modeling errors (non linearity, variability of the structures and bad knowledge of some physical phenomena). Such an order will be said to be robust, since it will take into account the uncertainties and imprecision. In our present work we are interested in the singularly perturbed systems [1 - 11] presenting two separable dynamics. A robust control class essentially uses the space of the frequencies applied at the imperfect or uncertain models so-called approach H∞ in her version loop shaping [12 - 20] whose objective is to get at the same time the stability and the performance in a completely analytic way. In this work we propose the problem of control H∞ loop shaping of the two time scale singularly perturbed systems and we are interested specially in the transfer of the property of two time scale at the conceived regulator. 2. Two time scale singularly perturbed systems Consider the linear and stationary time singularly system described by the following state representation :


Transfer of degree of reparability of two time scale singularly perturbed systems for the associated regulator

With: Y

, U

, X

This system can be described by the following standard model singularly perturbed systems with two time scale [7,11]:

With: x1: slow part of the vector X. x2 : fast part of the vector X. µ : positive parameter translating the difference of the time scales of the two dynamics, it can be estimated by [8] :

The system (2) will have the property of two time scale if it verifies the condition carrying on the following norms [7 ,20]:

So in this moment it can be decomposed in a slow subsystem (Sl) and a fast other (Sr) as: (Sl) With: xl ∈ ℜ n1, ul ∈ ℜ m,yl ∈ ℜ p


Abdessalem bsissa , Souid souad

⎧ A = A − A A −1 A 11 12 22 21 ⎪ l ⎪ = − B B A A −1 B ⎨ l 1 12 22 2 ⎪ ⎪ C = C − C A −1 A 1 2 22 21 ⎩ l

⎧ x& r = A22 x r + B2 u r ⎩ y r = C 2 xr

(Sr) ⎨

(6)

3. H∞ control loop shaping The approach H∞ by loop shaping has been developed [14] from the notion of first factorization of a transfer matrix. The calculation of the order feedback is often considered as a process of formatting of this feedback, named the loop shaping, this one tends to respect the following objectives:

-

Closed-loop feedback stability. Robust in stability.

Attenuation of the perturbation. In the approach which is made by first factorization, the robust stability problem doesn't permit the introduction of parameters of performance; this inconvenience can be saved by a previous stage of modeling of loop shaping [17]. Let G(p) the transfer function of the system (1), the aim of the stage of modeling is to modify the model nominal G(p) by pre and/or post compensators W1 and W2 (figure. 1), in order to have a raised gain in low-frequency and a weak gain in high frequency, to have a good dismissal of the perturbation.

.

e1 s1 s2

e2 W1(p)

G(p

W2(p)

K(p)

Figure.1: Synthesis H∞ by loop shaping 381 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


It is good to signal that in the case monovariable only one compensation is sufficient, we will have an augmented system Ga(p) defined as:

G a (p) = W 1 (p)G(p) With : G

a

(p )=

C a ( pI − A a

(7)

)− 1 B a

After the phase of modeling we proceeds the stage of stabilization that consists in determining a corrector K(p) loop shaping that minimizes the transfer between the input (e1, e2) and the output (s1, s2) of the figure.1 this corresponds to the following H∞ problem:

Sa (p)

Sa (p)G a (p)

K(p)Sa (p) K(p)Sa (p)G a (p) ∞

≤γ

(8)

With:

Sa (p) : Sensitivity function of increased system as: S a ( p ) = [ I + G a ( p ) K ( p )] −1

(9)

K (p) : The regulator solution of the problem H∞ loop shaping described by the following state representation:

⎧ X& c (t ) = Ac xc (t ) + Bc y (t ) ⎨ ⎩u (t ) = Cc xc (t )

(10)

With:

⎧ A = A − BBT X + γ 2 Z YC T C , γ ⎪ c T 2 ⎪ Bc = γ Z γ YC ⎪⎪ T ⎨Cc = B X ⎪ 2 −1 ⎪Z γ = I + YX − γ I ⎪ ⎪⎩where γ = 1.1 1 + λsup (YX )

(

)



λsup

designate the sup of the eig and X,Y are respectively the solutions of the

algebraic Riccati equations as: XA + A T X − XBB YA

T

+ AY − YC

T

T

X + C TC = 0

CY + BB

T

(11)

= 0

4. Transfer of degree of reparability The method of synthesis which has just been presented will be put in application for a direct current motor [10] described by the following state model: ⎧ x& ⎪⎡ 1 ⎢ ⎪⎪ ⎣ x& 2 ⎨ ⎪ ⎪y = ⎪⎩

⎡− R ⎤ L ⎢ = ⎥ ⎢ Km ⎦ J ⎣ x [0 1 ] ⎡⎢ 1 ⎤⎥ ⎣x2 ⎦

− Kb − Kf

⎤ L ⎥ ⎡ x 1 ⎤ + ⎡ 1 L ⎤U ⎢ ⎥ ⎥⎢x ⎥ ⎢⎣ 0 ⎥⎦ J ⎦⎣ 2 ⎦

Where: y (t) = Ω(t) The physical parameters of the examined motor are the following: R = 4 Ω, L = 1 H : resistance and self of armature of motor. Km = 0.015Nm /A : coefficient of rotational torque. Kb = 0.1 fem constant. Kf = 0.2Nms: coefficient corresponding a rubbing. U=220V: supply voltage of the induced. J : moment of inertia of the load. The method of synthesis H∞ loop shaping will be applied to the system in 2 stages whose aim is to show the principle of transfer of degree of reparability of the system for with its regulator H∞ loop shaping associated.

4. 1. Stage 1: Synthesis H∞ loop shaping under composite shape: W1(p)=1 In this stage we are interested in the synthesis H∞ loop shaping; with only one compensator unit applied to the system for different values of J therefore of µ. Having for the continuation: G (p): the transfer function of the global system. - K∞(p) : the regulator H∞ loop shaping partner of the system. 383 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


-

µg : parameter of difference of the time- scales of the system. µk : parameter of difference of the scales of time of the regulator H∞ loop shaping of the system.

Table1. Survey of transfer of degree of separability of dynamics of the global system for the regulator H∞ loop shaping before compensation G(p) 3 p 2 + 44 p + 160.3

µg

K∞(p)

0.1004

p 2 + 44 p + 160.5

− 14

1.5 p 2 + 24 p + 80.15

0.2008

− 31.5 p − 8651.4 p 2 + 24.01 p + 80.25

0.75

− 27 p − 301

p 2 + 14 p + 40.07

p 2 + 14 p + 40.12

0.4015

0.375 p 2 + 9 p + 20..03

µk

0.803

− 21 p − 142.4 p 2 + 9 p + 20.06

0.1006 0.2011 0.4021 0.8037

The comparative survey illustrated by the table1, illustrates well that when we increase the µg values of the system the parameterizes difference of the scales of time of the regulator H∞ loop shaping will also increase and it will have the same value as µg parameter that is to say µk ≅ µg , which explains the transfer of reparability degrees of the global system’s dynamics to the regulator H∞ loop shaping associated before compensation 4. 2. Stage 2: Synthesis H∞ loop shaping after compensation In the continuation the synthesis H∞ loop shaping will be applied to the system for µg=0. 1004 (to see table. 1), this stage of synthesis will be put in curve for two different cases of compensation. Before compensation the state model of process is the following: ⎧ & − 0 .1 ⎤ ⎡− 4 ⎡1 ⎤ X + ⎢ ⎥u ⎪X = ⎢ ⎥ − 40 ⎦ ⎣ 3 ⎣0 ⎦ ⎪ ⎨ ⎡x ⎤ ⎡x ⎤ ⎪ y = [0 1 ]⎢ 1 ⎥ , X = ⎢ 1 ⎥ ⎪ x ⎣ 2⎦ ⎣x 2 ⎦ ⎩

The poles of the system are: 384 Proceedings of the 7th Saudi Engineering Conference, KSU, Riyadh, 2007


λ1 = -4.01 et λ2 = -39.99 According to the values of the poles of the system one notices that the process of survey presents two different dynamics. Let's note that the condition (4) existence of two scales of time is verified: A22 = −10 is inversible and 1 −1 −1 −1 A− 22 = 0.025