Polishing Step Purification of High-Strength Wastewaters by ... - MDPI

membranes Article

Polishing Step Purification of High-Strength Wastewaters by Nanofiltration and Reverse Osmosis Jinxiang Zhou, Brian O. Baker, Charles T. Grimsley and Scott M. Husson * Department of Chemical and Biomolecular Engineering and Animal Co-Products Research and Education Center, Clemson University, Clemson, SC 29634, USA; [email protected] (J.Z.); [email protected] (B.O.B.); [email protected] (C.T.G.) * Correspondence: [email protected]; Tel.: +1-86-4656-4502 Academic Editor: Clàudia Fontàs Received: 12 January 2016; Accepted: 29 February 2016; Published: 10 March 2016

Abstract: This article reports findings on the use of nanofiltration (NF) and reverse osmosis (RO) for secondary treatment of high-strength rendering facility wastewaters following an ultrafiltration step. These wastewaters present significant challenges to classical treatment technologies. Constant-pressure, direct-flow membrane filtration experiments were done to screen for flux and effluent water permeate quality of ten commercial NF and RO membranes. All membranes tested were effective in reducing total dissolved salts (TDS) and chemical oxygen demand (COD); however, only two membranes (Koch MPF-34 and Toray 70UB) gave sufficiently stable flux values to warrant longer term cross-flow filtration studies. Cross-flow flux measurements, scanning electron microscopy (SEM), X-ray dispersive spectroscopy (EDS), and attenuated total reflectance-Fourier-transform infrared spectroscopy (ATR-FTIR) indicated that both membranes were eventually fouled by organic and inorganic foulants; however, the Toray 70UB RO membrane yielded a capacity of 1600 L/m2 prior to cleaning. A preliminary economic analysis compared the estimated costs of energy and consumables for a dual-stage UF/RO membrane process and dissolved air floatation (DAF) and found membrane process costs could be less than about 40% of the current DAF process. Keywords: impaired water; membrane fouling; sustainability; water treatment

1. Introduction The rendering industry recycles perishable materials from livestock, meat/poultry processing, food processing, and supermarkets and restaurants into valuable products. However, it also generates large volumes of high-strength wastewaters. Previously, we highlighted the advantages of membrane processes over other wastewater treatment technologies such as coagulation, flocculation, air flotation and gravity separation that are used to treat high-strength wastewaters such as those in the rendering industry [1,2]. Whereas dissolved air flotation (DAF) typically requires the use of chemicals or polymer additives to adjust the pH and improve flocculation of the solids for improved removal efficiency [3], membranes provide a positive barrier for rejection of solids that does not require such additives. Results of our work with ultrafiltration (UF) membranes [1,2] showed that it is effective in achieving significant reduction in turbidity, chemical oxygen demand (COD), and total dissolved solids (TDS) at high flux. Ultrafiltration membranes, however, are not designed for removal of low molecular weight compounds and salts that contribute to COD and TDS. Rather, they provide an initial purification step that can be followed by a polishing step such as nanofiltration (NF) or reverse osmosis (RO) to recover clean water for direct discharge or beneficial use. The use of such membrane cascades is common, and seawater desalination is one example where UF followed by RO is used in practice. In this contribution, we evaluate the use of commercial NF and RO membranes following

Membranes 2016, 6, 19; doi:10.3390/membranes6010019

www.mdpi.com/journal/membranes

Membranes 2016, 6, 19


2 of 12

2 of 12

UF in an additive-free, membrane-based wastewater wastewater treatment process for high-strength wastewaters following UF in an additive-free, membrane-based treatment process for high-strength (see Figure 1). wastewaters (see Figure 1).

Figure 1. 1. Two-stage Two-stagemembrane membranetreatment treatment process rendering facility wastewater. first stage Figure process forfor rendering facility wastewater. The The first stage uses uses membrane UF. This paper provides performance and economic data for the second “polishing” membrane UF. This paper provides performance and economic data for the second “polishing” stage stage using P represents a pressure using RO orRO NF.or P NF. represents a pressure gauge.gauge.

NF membranes are effective in removing divalent salts and organic molecules with high enough NF membranes are effective in removing divalent salts and organic molecules with high enough molecular weights. Compared to RO membranes, NF membranes have lower rejection of monovalent molecular weights. Compared to RO membranes, NF membranes have lower rejection of monovalent salts. Typical pore sizes of NF membranes are around 1 nm, and molecular weight cut-offs (MWCOs) salts. Typical pore sizes of NF membranes are around 1 nm, and molecular weight cut-offs (MWCOs) are about 300–500 Da [4]. RO membranes are characterized by MWCO of about 100 Da [5]. Previous are about 300–500 Da [4]. RO membranes are characterized by MWCO of about 100 Da [5]. Previous studies reported in the literature [6–8] have successfully applied RO and NF membranes to treat studies reported in the literature [6–8] have successfully applied RO and NF membranes to treat different types of high strength wastewaters, including those from the chemical industry, food different types of high strength wastewaters, including those from the chemical industry, food industry, industry, and agricultural wastewaters, among others. Treatment of rendering wastewater using RO and agricultural wastewaters, among others. Treatment of rendering wastewater using RO or NF or NF membranes has not been reported. membranes has not been reported. The objectives of this study were to determine the performance of commercial RO and NF The objectives of this study were to determine the performance of commercial RO and NF membranes for secondary treatment of rendering facility wastewater following an ultrafiltration step, membranes for secondary treatment of rendering facility wastewater following an ultrafiltration step, to characterize the membrane surfaces pre- and post-filtration to assess fouling, and to perform a to characterize the membrane surfaces pre- and post-filtration to assess fouling, and to perform a preliminary cost analysis for operating a dual-stage UF/NF or UF/RO membrane separation process. preliminary cost analysis for operating a dual-stage UF/NF or UF/RO membrane separation process. Direct- and cross-flow membrane filtration experiments using wastewater provided by a rendering Direct- and cross-flow membrane filtration experiments using wastewater provided by a rendering facility were carried out, and membrane performance was evaluated by measuring productivity (i.e., facility were carried out, and membrane performance was evaluated by measuring productivity the volumetric filtrate flux), capacity (i.e., the total volume processed per unit membrane area before (i.e., the volumetric filtrate flux), capacity (i.e., the total volume processed per unit membrane area the membrane must be cleaned), and effluent water quality (COD, turbidity, TDS). Membrane fouling before the membrane must be cleaned), and effluent water quality (COD, turbidity, TDS). Membrane was detected using scanning electron microscopy (SEM), energy-dispersive X-ray spectroscopy fouling was detected using scanning electron microscopy (SEM), energy-dispersive X-ray spectroscopy (EDS) and attenuated total reflection Fourier-transform infrared spectroscopy (ATR-FTIR). The cost (EDS) and attenuated total reflection Fourier-transform infrared spectroscopy (ATR-FTIR). The cost analysis compared the energy and consumables costs of the membrane processes to DAF. analysis compared the energy and consumables costs of the membrane processes to DAF. 2. Results and Discussion 2. Results and Discussion In the thecurrent current work, NF five andRO five RO membranes were evaluated forstep polishing step In work, fivefive NF and membranes were evaluated for polishing purification purification of wastewaters generated in rendering facility following treatment UF.1Table of wastewaters generated in rendering facility following primary primary treatment by UF. by Table lists 1 lists the RO and NF membranes tested, including their specifications: membrane type, material of the RO and NF membranes tested, including their specifications: membrane type, material of construction, manufacturer, and pure water permeability. Membranes were selected to provide a construction, manufacturer, and pure water permeability. Membranes were selected to provide range of manufacturer-reported characteristics, as there often is a trade-off between effluent quality a range of manufacturer-reported characteristics, as there often is a trade-off between effluent quality and productivity. Measured pure water permeability reported in Table 1 was found to match closely and productivity. Measured pure water permeability reported in Table 1 was found to match closely with manufacturer-reported values. with manufacturer-reported values.


3 of 12


3 of 12

Table 1. Nanofiltration and reverse osmosis membranes used in the study.

Membrane Table 1. Nanofiltration and reverse osmosis membranes usedPure Water Permeability in the study. Type Material Manufacturer Product (Lm−2· h−1· bar−1) MPF-34 NF Proprietary Koch Membrane 1.9Permeability Membrane Pure Water Type Material Manufacturer ´2 ¨ h´1 ¨ bar´1 ) Product (Lm SB90 NF Cellulose Acetate Blend TriSep 0.9 MPF-34 Proprietary Koch Membrane TS80 NF NF Polyamide TriSep 5.81.9 SB90 NF Cellulose Acetate Blend TriSep 0.9 NF90 NF NF Polyamide Dow 10.55.81 TS80 Polyamide TriSep NF90 Polyamide Dow HL NF NF Proprietary GE Osmonics 3.7 10.5 1 HL NF Proprietary GE Osmonics 70UB RO Polyamide Toray 2.83.7 70UB RO Polyamide Toray 2.8 X201 RO RO Polyamide-urea TriSep 4.14.1 X201 Polyamide-urea TriSep ACM4 Polyamide TriSep ACM4 RO RO Polyamide TriSep 4.94.9 SG RO Proprietary GE Osmonics 2.4 SG RO Proprietary GE Osmonics 2.43.3 BW30FR RO Polyamide Dow BW30FR 1 Permeability RO 3.3 for thisPolyamide membrane was calculated basedDow on data provided by the supplier. 1

Permeability for this membrane was calculated based on data provided by the supplier.

2.1. Membrane Performance during Direct-Flow Filtration Direct-flow filtration filtrationwas was used for membrane screening to compare performance for ofa used for membrane screening to compare performance for a variety variety of membranes. performance interest for direct-flow are membranes. Membrane Membrane performance metrics of metrics interestof for direct-flow studies studies are flux— flux— volume/area/time—versus volume processed and selectivity—high rejection of dissolved volume/area/time—versus volume processed and selectivity—high rejection of dissolved solids. In solids. In this UF pretreated wastewater used asstream. the feed stream. this study, UFstudy, pretreated wastewater was usedwas as the feed Figure 2a,b show membrane flux versus permeate volume for direct-flow filtration rendering wastewater by the NF and RO membranes. Table 2 provides the specific transmembrane facility wastewater membranes. Table pressure used for each membrane. These These pressures pressures were were selected selected to provide the same starting starting flux these membranes, membranes, there was significant significant flux reduction based on pure water permeability data. For all these from initial fluxes, which is attributed to the additional resistance induced by membrane fouling and concentration polarization. From Figure 2a, the MPF-34 membrane had the highest stable flux of any of the NF membranes. From Figure 2b, the 70UB membrane had the highest stable flux of any of the RO membranes. 25

25

a

2

2

15

20 Flux (L/m ·h)

20 Flux (L/m ·h)

b

MPF-34 SB90 TS80 NF90 HL

10

15 10 70UB X201 ACM4 SG BW30FR

5

5

0

0 0

5

10

15

20

permeate volume (mL)

25

30

0

10

20

30

40

permeate volume (mL)

Figure 2. (a) Figure 2. (a) Flux Flux measurements measurements by by direct-flow direct-flow filtration filtration using using NF NF membranes membranes and and (b) (b) flux flux measurements by direct-flow filtration using RO membranes. Transmembrane pressure (TMP) was measurements by direct-flow filtration using RO membranes. Transmembrane pressure (TMP) was adjusted to achieve adjusted for for each each membrane membrane to achieve a a constant constant initial initial flux flux for for all all of of the the membranes. membranes. Using Using constant constant initial flux enables direct comparisons among membranes. initial flux enables direct comparisons among membranes.


4 of 12


4 of 12

Table 2. Performance data for nanofiltration and reverse osmosis membranes used in the study. MPF-34data SB90 TS80 NF90 HL 70UB ACM4usedSG Table 2. Performance for nanofiltration and reverse osmosisX201 membranes in theBW30FR study. Parameter

1 1 1 2 2 1 1 1 1 1 TDS Parameter MPF-3474% 1 SB9061% 1 TS8022% 1 NF9098% 2 HL 2 70UB X201 ACM4 SG 1 BW30FR 1 N/A 90%1 91%1 80% 1 90% 70% Reduction 74% 61% 22% 98% 90% 91% 80% 90% 70% TDS Reduction N/A COD 81% 81% 66%66% 24%24% 98% 95% 90% 80% COD Reduction 98% 81% 81% 95% 90% 80% 94%94% 92% 92% Reduction Steady State Steady State 0.76 0.10 0.57 0.29 0.39 0.90 1.03 0.91 0.33 0.75 Permeability (LMH/bar) Permeability 0.76 0.10 0.57 0.29 0.39 0.90 1.03 0.91 0.33 0.75 15.8 31.5 5.3 31.1 15.5 22.1 15.5 11.0 35.9 20.0 TMP (LMH/bar) (bar) 1 Feed TMPTDS (bar)level was 15.8 31.5and COD 5.3 was31.1 15.5 2 Feed 22.1TDS level 15.5 was 6800 11.0 mg/L 35.9 20.0 4060 mg/L 8300 mg/L. and COD was 86001 Feed mg/L. TDS level was 4060 mg/L and COD was 8300 mg/L. 2 Feed TDS level was 6800 mg/L and COD was 8600 mg/L.

Figure 3a illustrates thethe role that plays reduction of flux. Percentage Figure 3a illustrates role thatmembrane membrane permeability permeability plays onon reduction of flux. Percentage ´ 2 ´ −2·h−1),¨ and reduction of flux (Rflux was calculated thestable stable flux (Lm h 1 ), stbisisthe reduction of flux (R)flux ) was calculatedusing usingEquation Equation (1). (1). FFstb flux (Lm Fintand is Fint ´ 2 ´ 1 is thethe initial flux initial flux(Lm (Lm−2·¨hh−1). ). ˙ ˆ F ´F (1) Rflux “  Fint  F stb ¨ 100% R  intFint stb 100% flux

 

Fint

 

(1)

100%

100%

80%

80%

60%

RCOD

Rflux

60% MPF-34 SB90 TS80 NF90 HL 70UB X201 ACM4 SG BW30FR

40%

20%

MPF-34 SB90 TS80 NF90 HL 70UB X201 ACM4 SG BW30FR

40%

20%

0%

0% 0

2

4

6

8

10

12

0

2

4

6

8

10

12

permeability (Lm-2h -1bar-1)

permeability (Lm-2h-1bar-1)

(a)

(b)

Figure 3. (a) Percentage flux reduction for RO and NF membranes and (b) percentage COD reduction.

Figure 3. (a) Percentage flux reduction for RO and NF membranes and (b) percentage COD reduction.

Figure 3a shows that Rflux does not correlate strongly with the membrane permeability. Rflux is

Figure 3a shows that Rfluxfouling. does not strongly with theNF90 membrane R,flux is associated with membrane Thecorrelate most permeable membrane did havepermeability. the highest Rflux indicating this membrane wasThe moremost severely fouled than other membranes. general Rflux , associated withthat membrane fouling. permeable membrane NF90 didHowever, have theinhighest membrane typemembrane (i.e., NF or was RO) does seem to play athan significant role here. Figure 3b compares indicating that this morenot severely fouled other membranes. However, in general percentage COD reduction to membrane permeability. COD reduction is defined in Equation RO membrane type (i.e., NF or RO) does not seem to play a significant role here. Figure 3b 2. compares membranes (red symbols) had higher RCOD than NF membranes (black symbols) with similar percentage COD reduction to membrane permeability. COD reduction is defined in Equation 2. permeability. As mentioned earlier, RO membranes have lower MWCO [4,5], which may be able to RO membranes (red symbols) had higher RCOD than NF membranes (black symbols) with similar sieve larger organic molecules than the NF membranes, leading to higher RCOD. The most severely permeability. As mentioned earlier, RO membranes lower MWCO [4,5], which may be be able to fouled membrane NF90 had a high RCOD compared have to other NF membranes. This finding may sieveassociated larger organic molecules thanonthe membranes, leading RCOD . The severely with the foulant layer theNF membrane surface. While to thishigher layer increased themost transport fouled membrane NF90 had athe high RCOD compared other NF membranes. resistance, it also improved sieving capability and to increased COD reduction. This finding may be associated with the foulant layer on the membrane surface. While this layer increased the transport CODfeed − COD filtrate R = ( ) × 100% (2) COD resistance, it also improved the sieving capability increased COD reduction. CODand feed ˆ RCOD “

CODfeed ´ CODfiltrate CODfeed

˙ ˆ 100%

(2)

Table 2 shows the characteristics of the UF pretreated wastewater samples used as a feed for the NF and RO membranes in the second polishing step, along with data from the permeate quality measurements. For all membranes, there were substantial reductions in COD and TDS. The turbidity


5 of 12

Table 2 shows the characteristics of the UF pretreated wastewater samples used as a feed for the NF and in the second polishing step, along with data from the permeate quality Membranes RO 2016,membranes 6, 19 5 of 12 measurements. For all membranes, there were substantial reductions in COD and TDS. The turbidity was reduced by nearly 100% for all the membranes tested. In general, RO membranes provided was reduced by nearly 100% for all theand membranes In general, RO membranes provided higher salt rejection (i.e., lower TDS) rejectiontested. of compounds contributing to COD. Tablehigher 2 also salt rejection (i.e., lower TDS) and rejection of compounds contributing to COD. Table 2 also summarizes summarizes the pseudo-steady-state permeability. RO membranes generally had higher pseudothe pseudo-steady-state RO membranes generally had higher steady-state permeabilitypermeability. than NF membranes, consistent with measured and pseudo-steady-state reported pure water permeability than NF membranes, consistent with measured and reported pure water permeability. Permeability depends on porosity and nominal pore size. It therefore is permeability. possible that Permeability depends on porosity and nominal pore size. It therefore is possible that differences in differences in these properties among the various membranes play a role in the measured these properties among the various membranes play a role in the measured permeability. permeability. 2.2. 2.2. Membrane Membrane Performance Performance during during Cross-Flow Cross-Flow Filtration Filtration

2.0

Permeability (L/m2×h×bar)

Permeability (L/m2×h×bar)

Based Based on on the the direct-flow direct-flow productivity productivity and and selectivity, selectivity, two two membranes membranes (MPF-34 (MPF-34 and and 70UB) 70UB) were were selected for further consideration in cross-flow studies. During cross-flow studies, we selected for further consideration in cross-flow studies. During cross-flow studies, we measured measured permeability permeability versus versus volume volume processed processed for for 96-h 96-h runs runs using using wastewater wastewater that that had had been been pretreated pretreated using using UF. For the MPF-34 and 70UB membranes, a TMP of 14.5 bar and 22.4 bar, respectively, was used UF. For the MPF-34 and 70UB membranes, a TMP of 14.5 bar and 22.4 bar, respectively, was used to to maintain maintain consistency consistency with with the the direct-flow direct-flow study. study. Figure Figure 44 shows shows long-term long-term (96 (96 h) h) cross-flow cross-flow filtration filtration results results for for MPF-34 MPF-34 (Figure (Figure 4a) 4a) and and 70UB 70UB (Figure 4b) membranes. Permeability data are given for runs with pristine membranes (1st run) (Figure 4b) membranes. Permeability data are given for runs with pristine membranes (1st run) and and after cleaning protocols were examined. For MPF-34, cleaning was done after cleaning. cleaning.Two Twodifferent different cleaning protocols were examined. For MPF-34, cleaning was using done ausing caustic solutionsolution (aqueous NaOH).NaOH). For 70UB, was donewas using a surfactant solution (aqueous a caustic (aqueous Forcleaning 70UB, cleaning done using a surfactant solution sodium dodecyl sulfate, SDS). In both cases, CIP (cleaning in place) was done by flowing (aqueous sodium dodecyl sulfate, SDS). In both cases, CIP (cleaning in place) was donethe bycleaning flowing solutions through the test cell for period 1 h. the cleaning solutions through thea test cellof for a period of 1 h.

a

1.6 1.2 0.8 1st run after cleaning

0.4 0.0 0

20

40

60

Total Permeate Volume (L)

80

2.0

b

1.6 1.2 0.8 1st run after cleaning

0.4 0.0 0

20

40

60

80

100

Total Permeate Volume (L)

Figure 4. (a) Cross-flow filtration results for MPF-34 NF membrane. Each run was done for a period Figure 4. (a) Cross-flow filtration results for MPF-34 NF membrane. Each run was done for a period of 96 h. The vertical dashed line indicates the point where CIP was done with NaOH solution. (b) of 96 h. The vertical dashed line indicates the point where CIP was done with NaOH solution. Cross-flow filtration results for 70UB RO membrane. Each run was done for a period of 96 h. The (b) Cross-flow filtration results for 70UB RO membrane. Each run was done for a period of 96 h. The vertical dashed line indicates the point where CIP was done with SDS solution. vertical dashed line indicates the point where CIP was done with SDS solution.

Figure 4a shows that, despite its good performance in short-term direct-flow testing, the MPFFigure 4a shows that,significantly despite its good performance short-term testing. direct-flow testing,with the MPF-34 34 NF membrane fouls during long-termin cross-flow Cleaning caustic NF membrane fouls significantly during long-term cross-flow testing. Cleaning with caustic solution solution restores flux fully; however, the permeability again begins to decrease over time following restores flux fully; thefor permeability begins topost-cleaning decrease overistime the cleaning step. however, The reason the higher again permeability not following clear, but the maycleaning be due step. The reason for the higher permeability post-cleaning is not clear, but may be due to partial hydrolysis of the selective layer during exposure to the caustic solution. Figure to 4b partial shows hydrolysis of the selective layer during exposure to the caustic solution. Figure 4b shows that the that the 70UB RO membrane effectively maintains permeability during long-term testing. Its capacity 70UB RO membrane effectively maintains permeability during long-term testing. Its capacity prior 2 prior to cleaning is roughly 1600 L/m . Unlike cleaning done with caustic solution, cleaning withtoa 2 . Unlike cleaning done with caustic solution, cleaning with a surfactant cleaning is roughly surfactant solution1600 was L/m not effective for restoring permeability to the initial value. The cleaning solution was not effective for restoring to the scaling initial value. cleaning results indicate results indicate that fouling may be permeability due to inorganic ratherThe than fouling by organics. that fouling may be due to inorganic scaling rather than fouling by organics. Nevertheless, 70UB Nevertheless, 70UB approaches steady-state operation following the initial cleaning step. approaches steady-state operation following the initial cleaning step. 2.3. Membrane Fouling Characterization SEM and EDS were used to visualize the morphology and fouling of membrane surfaces and to perform elemental analysis of the membrane surfaces before and after filtration.


6 of 12

2.3. Membrane Fouling Characterization


6 of 12

SEM and EDS were used to visualize the morphology and fouling of membrane surfaces and to perform elemental analysis of the membrane surfaces before and after filtration. Figure Figure 55 shows shows SEM SEM images images of of the the MPF-34 MPF-34 membrane membrane surface surface before before and and after after filtration. filtration. Similar Similar results were obtained for the 70UB membrane. There is a visual change in the surface results were obtained for the 70UB membrane. There is a visual change in the surface morphology morphology of of the membrane surface after fouling. Table 3 summarizes the elemental composition of the surface from the membrane surface after fouling. Table 3 summarizes the elemental composition of the surface EDS BeforeBefore fouling, the surface comprises C, N, and O,and characteristic of the amide frommeasurements. EDS measurements. fouling, the surface comprises C, N, O, characteristic of the chemistry used as the selective layer. After fouling, N is no longer observed, indicating two things: amide chemistry used as the selective layer. After fouling, N is no longer observed, indicating two First, the foulant is not protein-based, and the base is coated with awith layer of foulant that things: First, the foulant is not protein-based, and themembrane base membrane is coated a layer of foulant is thicker than the penetration depth of the EDS measurement (