Wireless Transmission of Acoustic Emission Signals for ... - Springer Link

6 downloads 128849 Views 2MB Size Report
values and use wireless transmission for communication, as well as a control ... many advantages, such as concealment, ease of deployment, timeliness of data ...
KSCE Journal of Civil Engineering (2011) 15(5):805-812 DOI 10.1007/s12205-011-0899-0

Geotechnical Engineering

www.springer.com/12205

Wireless Transmission of Acoustic Emission Signals for Real-Time Monitoring of Leakage in Underground Pipes Bui Van Hieu*, Seunghwan Choi**, Young Uk Kim***, Youngsuk Park****, and Taikyeong Jeong***** Received June 9, 2009/Revised March 2, 2010/Accepted June 17, 2010

···································································································································································································································

Abstract In this paper, we propose a system combining various methods for monitoring leakage of underground pipelines. The system uses a wireless connection for communications and uses acoustic emission effects to detect and locate leakages. The system is easily deployed, with a flexible configuration that requires less maintenance because a wireless connection is used. Moreover, the system is accurate, simple, and inexpensive, because it is based on acoustic emissions. Experiments determining the wireless connectivity of the proposed system are also presented. Keywords: wireless sensor, monitoring system, underground, acoustic emission sensor, leakage monitoring ···································································································································································································································

1. Introduction Currently, Real-time Monitoring Systems (RMS) are important in many areas, such as transportation, airports, and civil construction industries. A Wireless real-time Monitoring System (WMS) is a RMS that uses wireless communication. A WMS includes a network of sensor nodes, which measure monitored values and use wireless transmission for communication, as well as a control center that gathers all of the values, processes them, and alerts users if there is a problem in the monitored system. Recently, WMS has been extended to monitor underground applications, e.g., an environment with soil, soil-mixture and non-soil components. Compared with monitoring systems using a wire for communication, WMS for underground systems offers many advantages, such as concealment, ease of deployment, timeliness of data, reliability, coverage density, and quality of service, etc (Akyildiz and Stuntebeck, 2006). Pipeline systems are used as the main transportation system for many applications, such as water distribution, oil and gas transportation, metropolitan heat systems, etc. Leakage in these pipeline systems can cause environmental and chemical damage. Leakage not only wastes resources but also creates environmental problems (Wu and Wang, 2006). Hence, monitoring systems for pipeline leakages are being increasingly demanded. A leakage monitoring system involves two critical tasks: detecting leakage and locating a leak position. Some studies use

pressure, flow, and temperature variables to detect pipeline leaks, but these methods are only able to detect large leaks. Also, they are difficult to deploy and have difficulty locating the leak position (Silva et al., 2005; Wu and Wang, 2006). The current proposed method uses acoustic emission sensors and is designed to detect an accurate point of leakage while the system is running. With only two sensors at two pipe ends, leakage can be determined, and the leakage position can be resolved. Furthermore, the acoustic emission sensor is small, lightweight, and mountable on outside pipe surfaces. For those reasons, an acoustic emission detection method is suitable as a pipeline leakage monitoring system, especially when the pipeline is buried underground. In this paper, firstly leak detection and locating methods based on acoustic emissions are discussed first. Next, special issues related to sensor nodes for underground applications and their solutions are presented. Then, we propose a WMS that uses an acoustic emission leak detection method to monitor an underground pipeline leak.Finally, experiments show the reliability of the proposed system. The rest of the paper is organized as follows. In Section II, leakage sound features are discussed. Then, leakage detection and leakage locating methods are presented in Section III and section IV, respectively. Next, underground acoustic emission sensor node issues are presented in Section V. In Section VI, we present our proposed monitoring system. Section VII shows the experimental results, and section VIII is the conclusion.

*Master Student, Dept. of Electronic Engineering, Myongji University, Yongin 449-728, Korea (E-mail: [email protected]) **Principal Researcher, Korea Electric Power Research Institute, Daejeon 305-380, Korea (E-mail: [email protected]) ***Member, Professor, Dept. of Civil and Environmental Engineering, Myongji University, Yongin 449-728, Korea (E-mail: [email protected]) ****Member, Professor, Dept. of Civil and Environmental Engineering, Myongji University, Yongin 449-728, Korea (E-mail: [email protected]) *****Member, Assistant Professor, Dept. of Electronic Engineering, Myongji University, Yongin 449-728, Korea (Corresponding Author, E-mail: [email protected]) − 805 −

Bui Van Hieu, Seunghwan Choi, Young Uk Kim, Youngsuk Park, and Taikyeong Jeong

2. Acoustic Leak Signal When a leak happens in a pipeline, it generates acoustic sound. This acoustic signal propagates along the pipeline so we can detect the leak by detecting the acoustic signal. The following paragraphs discuss features of an acoustic leak signal. A leak’s frequency features, as well as its amplitude, depend on many factors, for instance, the size of leak, the type of transported fluid (i.e., water, oil), and pipeline pressure. If the pipe is large in diameter or less solid, then the leak sound contains lowerfrequency components. On the contrary, if pressure is higher, then higher-frequency components dominate. The amplitude of the leak sound is higher if the pressure or flow speed is higher or if the leak is large, but not very large (Hunaidi and Chu, 1999; Hunaidi et al., 2004). If operational conditions of the pipeline, such as temperature, pressure, and flow do not change, then the leak sound is assumed to be a stationary signal, a signal whose frequency components do not change over time. Propagation of a leak signal also depends on many factors, such as type of pipeline (i.e., PVC, iron) and the pipe’s surrounding environment. A leak signal propagates well with metal pipe but attenuates greatly with plastic or concrete pipe (Hunaidi and Chu, 1999; Muggleton and Brennan, 2004). When pipe is buried underground, a leak signal is attenuated more than when the pipe is above ground. The signal is less attenuated in sandy soil, asphalt, and concrete but attenuated greater in clayey or grass areas (Hunaidi and Chu, 1999). Moreover, attenuation of the leak sound, while it propagates along the pipe, is not linear, i.e. attenuation factors of different frequencies are different (Muggleton and Brennan, 2004). While the leak sound signal propagates, noise interferes with the leak signal. Noise generates from the flow of fluid inside pipes, wind, and construction sound. Noise can be modeled as two kinds: background or white noise and burst noise. Background/ white noise is a stationary signal component, and burst noise is a non-stationary one. Stationary noise often has low amplitudes, whereas non-stationary noise often has high amplitudes and narrow bands. We performed experiments to check leaks above ground and underground. Devices used in the experiment are shown in Fig. 1. A metal pipe was used to transport water and simulate leak signals. An acoustic emission sensor sensed sound signals and was connected with an analyzer for gathering data. These data were transmitted to a laptop over an Ethernet connection for processing. The metal pipe length was 6 m, and its diameter was 3 cm. A leak was made on the pipe by drilling a 1.5-mm diameter hole. The acoustic emission sensor was a sensor integrated preamplifier. Its specifications are shown in Table 1. Leak signals were measured by mounting a sensor on the surface of the metal pipe, as shown in Fig. 2(a). First, the pipe was placed on the ground to measure leak signals above ground. Then the pipe was buried underground at a depth of 30 cm to simulate a leak underground. Fig. 2(b) shows the pipe setup before being covered by soil.

Fig. 1. Experiment Devices Table 1. Acoustic Emission Sensor Specifications Name

Value

Peak sensitivity, ref V/(m/s)

120 dB

Peak sensitivity, ref V/ubar

-28 dB

Frequency range

10-70 kHz

Resonant frequency, ref V/ (m/s)

25 khz

Resonant frequency, ref V/ubar

31 kHz

Directionality

+/- 1.5 dB

Temperature range

-35o - 70oC

Fig. 2. Outdoor Experimental Setup on Pipe Line: (a) Above Ground Pipe Setup, (b) Underground Pipe Setup

In both cases, most working conditions, such as the pipe, fluid pressure, and leak size, were identical. The only difference was in the placement of the pipe above aground or underground. However, the measured leak signals shown in Fig. 3 are different. The above-ground leak signal had higher amplitudes and changed more suddenly than the underground leak signal because the attenuation above ground was lower and noise was higher than underground.

3. Leak Detection Procedure with Signal Transform The early leak detection method based on acoustic emission is very straightforward: a person uses a device to hear the acoustic sound and determine whether there is a leak. The results depend

− 806 −

KSCE Journal of Civil Engineering

Wireless Transmission of Acoustic Emission Signals for Real-Time Monitoring of Leakage in Underground Pipes

Fig. 4. Leak Detection Procedure

Fig. 3. Leak Signal in Different Locations: (a) Leak Signal Above Ground, (b) Leak Signal Underground

mostly on the experience of the user. Later, many studies used a computer to analyze the spectral features of an acoustic sound to discover leaks (Zhan-Hui et al., 2005; Jin et al., 2008; Liu and Zhao, 2008). The detection method in these studies generally includes two stages: extracting signal features and recognizing the signal base on extracted features. Extracting signal features means to extract special signal characteristics (Kang et al., 2009). Extracted features are used to represent the signal and replace the original signal in the following process. Leak signal varies under different conditions, as discussed in the previous section. For accurate results, the system must extract features which reflect the signal exactly and distinguish between leak signal and noise. Fourier transform is a general method to extract frequency features of a signal and is used in many applications, including sometimes as a way to extract leak signal features. Using Linear Prediction Cepstrum Coefficients (LPCC) is another way to extract signal features (Changsheng et al., 2006). Recently, another scheme, which is promising and interesting to many researchers, is to use wavelet transform coefficients to represent signal features. Compared with Fourier transform, wavelet transform computation is less and it reflects both the frequency and timing of the signal, whereas Fourier transform only reflects the frequency of the signal. After obtaining signal features, a recognizing model uses these features to determine whether a signal is a leak signal. The model Vol. 15, No. 5 / May 2011

can be a predetermine as maximum modulus (Zhan-Hui et al., 2005), Hidden Markov Machine (HMM) (Changsheng et al., 2006), or base on training values in a support vector machine (Liu and Zhao, 2008; Huali et al., 2004), or Neural network (Jiao et al., 2006). These methods have trade-offs between complexity and accuracy. Among them, the Support vector machine balances computation complexity and result accuracy. With only two steps, leak detection results are very good under laboratory conditions, but results will be less efficient in practice. These methods do not deal with noise, and noise can make the result incorrect. Therefore, to increase accuracy, we propose a new detection procedure which has three stages, as in Fig. 4. One more stage is added before the Feature extracting stage. This stage will reject noise from the leak signal. As mentioned in section 2, there are two kinds of noise: stationary and non-stationary noise. Stationary noise may be removed implicitly by the recognizing model. Non-stationary noise is an important agent affecting the recognizing results. Wavelet transform can be used effectively to remove the nonstationary noise and is presented in (Daneti, 2008).

4. Leak Locating Method To locate a leak in the pipeline, we need input from two sensors with the condition that the leak must happen between two sensors, as in Fig. 5. From above figure, we have: L = Li + Lj = 2Li – ( Li – Lj )

(1)

As a leak happens, it propagates and reaches sensor i and sensor j at times ti and tj, respectively. Assuming that the leak signal propagation p, which can be predetermined easily, is constant along the pipe, Eq. (1) becomes: L = 2Li – p ( ti – tj ) = 2Li – p ∆tij

(2)

If the difference in arrival time ∆tij = ti – tj can be determined, then leak location can be derived by the simple Eq. (3) With Eq. (3), leak locating requires determining ∆tij. To determine ∆tij, the general autocorrelation (Knapp and Carter,

− 807 −

Fig. 5. Description of Leak Assumption Point

Bui Van Hieu, Seunghwan Choi, Young Uk Kim, Youngsuk Park, and Taikyeong Jeong

1976) is often used. This method models the receive signals at two sensors as below: ⎧ si ( t ) = l ( t ) + ni ( t ) ⎨ ⎩ sj ( t ) = α l ( t + ∆tij ) + nj ( t )

(4)

with si(t) and sj(t) as the received signals at sensor i and sensor j, respectively. l(t) is the leak signal, α represents signal attenuation, and ni(t) and nj(t) represent noise. With this method, noise must be a stationary signal, otherwise the leak locating result is incorrect. Other methods rejecting noise and determining ∆tij concurrency are presented in Yumei et al. (2004) and Daneti (2008). When noise ni(t) and nj(t) including stationary and nonstationary signals, are removed, time lag ∆tij can be determined easily by a cross-correlation method as discussed in section 3. Firstly, the cross-correlation of two signals, si(t) and sj(t), are calculated, as in Eq. (5). Then, the t value that makes Rsi,sj maximum is determined. This value is the time lag, ∆tij. RS , S ( t ) = i

j

Fig. 6. ATMega128L Microcontroller Embedded on MICA2 Board Table 2. MICA2 Transmission Features Name



∫ ( si*( δ ).sj( t + δ ) ) dδ

(5)

–∞ * i

where s denotes the complex conjugate of the si signal. In summary, to localize the leak position, we need data signals from two sensors. First, noise including background and burst noise is removed. Next, cross-correlation of the two sensor data is calculated. Then, combining the position which make the cross correlation maximize and the sample rate, we determine the time lag in arrival time of the two sensors. Finally, the leak point is determined by Eq. (3).

5. Underground Sensor Node In a wireless sensor network, a sensor node collects sensor information and communicates wirelessly with other nodes in the network. The most popular sensor node platform used in overground applications is a MICA2 board, shown in Fig. 6. A MICA2 board has an expansion connector which connects with sensors. The central part of the MICA2 is a ATMega128L microcontroller that controls the sensor gathering and communicating processes. Power is provided by two AA batteries, and an additional power supply connects to an external power connector, which only needs to connect with an antenna to communicate wirelessly. Its main transmission features are shown in Table 2. With a data rate of 38.4 kBaud and an outdoor range of 150 m, the MICA2 is suitable for most aboveground wireless sensor network applications. Although the MICA2 is not developed for underground applications, most current underground wireless sensor networks use it. Unfortunately, underground environments with different combinations of sand, soil, clay, and moisture are unfavorable for Electromagnetic (EM) wave transmissions which are used for communication in MICA2 boards. Some studies model the effect of the underground environment on EM, and all of them show that EM transmission attenuation is very high in underground

Value

Output gain

5 dBm

Receive sensitivity

-98 dBm

Signal rate

38.4 Kbaud

Outdoor range

150 m

Center frequency

300-915 MHz

environments (Dam et al., 2005; Li et al., 2007). Experimental results show that the wireless communication distance cannot exceed 7 m horizontally and 0.6 m vertically for the MICA2 board (Stuntebeck et al., 2006). With these distance limitations, MICA2 is difficult to use for applications in which the sensor nodes are buried in the ground. To increase underground communication distance, other transmission techniques are considered in Vasquez et al. (2004) and Sojdehei et al. (2001), but more investigation is needed before they are practical. Hence, currently, reasonable underground communication is still EM; therefore, MICA2 is still the appropriate selection for underground wireless communication. This is the limitation of underground wireless sensor networks at this time and it needs to be considered when designing underground monitoring systems. Another issue with buried sensor nodes is the power supply. To communicate underground, sensor nodes must transmit signals with higher energy than above ground; hence, they need more power. Sensor nodes are usually battery-powered, but changing batteries is sometimes impossible for buried sensor nodes. The battery has to provide enough power for the sensor node during the entire operating time. To provide enough energy for longterm operations, the battery size may become too big to be deployed with a sensor node. One feasible solution is to integrate a device into the sensor node that can generate electrical energy from other sources with the sensor node (Stordeur and Stark, 1997; Ping and Yumei, 2005). This source would be an external power supply that provides additional power for the sensor node. Moreover, choosing a suitable sensor for a sensor node is an

− 808 −

KSCE Journal of Civil Engineering

Wireless Transmission of Acoustic Emission Signals for Real-Time Monitoring of Leakage in Underground Pipes

important issue. There are many types of acoustic sensors such as piezoelectric or fiber bragging with different shapes, sizes and sensitivities. Fiber bragging sensors have many advantages, such as accuracy, electromagnetic immunity, and harsh environment immunity, but they are difficult to integrate with wireless sensor nodes, which require low power consumption, simplicity, and cost efficiency. With suitable features, such as small size, light weight, and simple connection, piezoelectric sensors are the most appropriate for acoustic sensor nodes buried underground. Piezoelectric acoustic sensors use piezoelectric effects to transduce acoustic sound into electricity. The frequency response of this sensor type can vary with frequency. Different sensors have differences in sense frequency range. With various expected detection leaks, frequency ranges are different. Consequently, the chosen sensor must have the suitable frequency response. The wide band sensor can sense a broad range of frequencies; therefore, it has the ability to detect many types of leak, but it makes processing more complex because it also includes abundant noise.

6. Proposed Testing Platform As mentioned in section III, sensor nodes for underground systems have more issues that need to be considered. Hence, we propose a new structure for sensor nodes of leakage monitoring systems for underground pipelines (Fig. 7). The acoustic emission sensor is a piezoelectric acoustic sensor which converts acoustic emissions to electrical signals. A MICA2 board connects with the sensor, gathers sensor data, and then wirelessly transfers data to the other node. Power is provided by a rechargeable battery. There is an added power generator which transforms other energies to electricity to provide for the operation of the sensor node. Because this power is not stable, it cannot supply power directly to the sensor node. Instead, this power is stored in a rechargeable battery. A battery charging control circuit controls the charging process and ensures that the battery is not overcharged. The battery and power generator ensures that the sensor node has enough power for its operational lifetime. Based on analysis of the leak acoustic emission processing and wireless underground sensor node, we propose a WSN monitoring system for monitoring underground pipeline leak, as shown in Fig. 8. The system includes sensor nodes that are buried underground. The sensor is mounted on a pipe surface to sense acoustic emis-

Fig. 8. The Proposed Testing Platform Structure

sions by converting them to electrical signals. The MICA2 board converts the analog electrical signal to a digital number, adds a time stamp, stores it in flash memory and then packets and transfers them to another node. Because underground wireless communication distance is relatively short, as mentioned in section III, the topology of the WSN has to be designed carefully. One MICA2 board, a wireless repeater, is placed on the ground above every buried sensor node. The MICA2 board receives packets from underground sensor nodes. Then, they forward these packets to a predetermined MICA2 board, called a Data Acquiring Station (DAS). DAS receives all sensor data and then transmits it to a computer by an RS232 or LAN interface. At the computer, leak detection, location, and monitoring processes are performed. The monitoring process at the computer involves many steps. First, noise in the sensor data is rejected. Next, the Wavelet transform is applied to extract signal features. Signal features are the input for a Support vector machine to determine if a leak occurred. If there is a leak, then a leak location procedure is performed. The process is performed many times, as configured by the user, to ensure that the result is correct. If the leak results are consistent, an alarm is activated. The pseudo-code of the process in the computer is shown in Fig. 9. With the proposed network topology, the system is easy to deploy, configure, and maintain. One DAS can collect data from sensor nodes in a radius of 150 m. If greater distance is needed, it can be easily reached by placing one more MICA2 boards between DAS and the MICA2 board placed above the buried sensor node. The limitation of our proposed system is that the depth of the monitored pipeline must be less than or equal to 0.6 m because of the limits of EM transmission in underground environments.

Fig. 9. The Pseudo-code of the Monitoring Process

Fig. 7. The Proposed Sensor Node Structure Vol. 15, No. 5 / May 2011

− 809 −

Bui Van Hieu, Seunghwan Choi, Young Uk Kim, Youngsuk Park, and Taikyeong Jeong

7. Wireless Connectivity Experiments The wireless connectivity of our proposed testing platform was verified by experiments. Experiments were performed outside at the Myongji University campus. The first experiment was performed to determine the communication distance between DAS and MICA2 boards, which were placed above ground in our proposed system. The experimental plan is illustrated in Fig. 10. Two MICA2 boards were placed above ground. One board always sent data packets, while the other board detected the packet. The distance between the two boards was increased until they couldn’t communicate or the packet wasn’t received. Input strength was measured, while the distance was increased (Fig. 11). Input strength decreased almost linearly with distance. At a distance of 10 m, input strength was under the limit (i.e., -95 dBm) of the MICA2 board. At distances greater than 10 m, the two boards could not communicate. The second experiment was performed to determine the limit of the communication distance between the MICA2 board on the ground and the MICA2 board buried underground. The experi-

ment plan is shown in Fig. 12. The distance between the two sensors was increased until they could not communicate. The received signal strength was measured and shown in Fig. 13. Attenuation in the underground environment was much higher than in the aboveground environment. Received signal strengths reduced rapidly. The depth limit of a buried sensor is 30 cm based on our real experiment. In practice, the position of the MICA2 board above ground may be slanted with the board buried underground. An experiment was performed to determine this effect. The experiment plan is shown in Fig. 14. This experiment was similar to the second experiment. The MICA2 board buried underground was set up at some determined depth, and the MICA2 board on the ground was moved away from the buried sensor position. The MICA2 board buried underground was placed at three depths, 10 cm, 15 cm, and 20 cm. The received signal strength was measured and shown in Fig. 15. As shown in that figure, the distance limit at a depth of 10 cm

Fig. 10. Horizontal Alignment Experiment Plan

Fig. 13. Result of Vertical Alignment Experiment

Fig. 14. Skewed Alignment Experiment Plan Fig. 11. Result of Horizontal Alignment Experiment

Fig. 15. Result of Skewed Alignment Experiment

Fig. 12. Vertical Alignment Experiment Plan − 810 −

KSCE Journal of Civil Engineering

Wireless Transmission of Acoustic Emission Signals for Real-Time Monitoring of Leakage in Underground Pipes

was 40 cm. At a depth of 15 cm, the limit was 30 cm, and, at a depth of 20 cm, it was 22 cm. Applying the Pythagorean Theorem in the experiment illustrated in Fig. 14, we have: 2

w = r +h

2

(6)

where w is the direct underground connection between the two sensor boards. We can see it as the underground transmission distance. However, this distance decreases if the sensor is buried deeper underground. Maximum values of this distance are 65 cm, 76 cm, 38 cm and 28 cm if the buried sensor board is placed at depths of 10 cm, 15 cm, and 20 cm, respectively.

8. Conclusions In this paper, we discussed techniques using acoustic emission effects to recognize a leak and locate its position. We proposed removing noise before processing leak signals. We also analyzed issues of underground wireless sensor nodes that need to be considered and proposed a structure for underground sensor nodes. Based on these analyses, we proposed a monitoring system that uses acoustic emission effects to monitor underground pipeline leakages. Our proposed system uses wireless communication so that it can be easily deployed, configured, and maintained. Experiments showed that the communication limits of our proposed system were 10 m horizontally and 30 cm vertically.

Acknowledgements This work is supported by National Science Foundation (NRF) grant funded by the Korea gov. (MEST) (No.20090069991)and also supported by the 2010 research fund of Myongji University in Korea. This work is supported by the Korea gov. Ministry of Knowledge and Economics (MKE) under the grant No. I-20101-012 of the Electric Power Industry Tech. Evaluation and Planning Center (ETEP). The circuit was designed at IC Design Center.

References Akyildiz, I. F. and Stuntebeck, E. P. (2006). “Wireless underfround sensor networks: Research challenge.” Journal of Ad Hoc Networks, Vol. 4, pp. 669-686. Changsheng, A., Honghua, Z., Rujian, M., and Xueren, D. (2006). “Pipeline damage and leak detection based on sound spectrum LPCC and HMM.” Sixth International Conference on Intelligent Systems Design and Applications, Vol. 1, ISDA 06, pp. 829-833. Dam, R. L. v., Borchers, B., and Hendrickx, J. M. H. (2005). “Methods for prediction of soil dielectric properties: A review.” Proceedings of the SPIE, Vol. 5794, pp. 188-197. Daneti, M. (2008). “A practical preprocessing treatment for pipeline leak locating improving.” IEEE International Conference on Emerging Technologies and Factory Automation, pp. 9-12.

Vol. 15, No. 5 / May 2011

Huali, C., Hao, Y., Chen, L. V., and Hongyu, S. (2004). “Application of support vector machine learning to leak detection and location in pipelines.” The IEEE Conference on Instrumentation and Measurement Technology, Vol. 3, IMC 04, pp. 2273-2277. Hunaidi, O. and Chu, W. T. (1999). “Acoustical characteristics of leak signals in plastic water distributions pipes.” Journal of Applied Acoustic, Vol. 58, No. 3, pp. 235-254. Hunaidi, O., Wang, A., Bracken, M., Gambino, T. and Fricke, C. (2004). “Acoustic methods for locating leaks in municipal water pipe networks.” International Conference on Water Demand Management, pp. 1-14. Jiao, Y., Yang, Q., Li, G., and Zhang, J. (2006). “Acoustic emission source identification technique for buried gas pipeline leak.” International Conference on Control, Automation, Robotics and Vision, pp. 1-5. Jin, Y., Yumei, W., and Ping, L. (2008). “Leak acoustic detection in water distribution pipelines.” World Congress on Intelligent Control and Automation, pp. 3057-3061. Kang, J., Han, M. M., and Jeong, T. T. (2009). “A context-aware computing methodology and development for mobile handhelds.” Journal of Science, Measurement & Technology, IET, Vol. 3, No. 4, pp. 317-323. Knapp, C. and Carter, G. (1976). “The generalized correlation method for estimation of time delay.” IEEE Transactions on Acoustics, Speech and Signal Processing, Vol. 24, No. 4, pp. 320-327. Li, L., Vuran, M. C., and Akyildiz, I. F. (2007). “Characteristics of underground channel for wireless underground sensor networks.” Annual Mediterranean Ad Hoc Networking Workshop, Corfu, Greece. Liu, N. and Zhao, Y. (2008). “Application of wavelet packet and support vector machine to leak detection in pipeline.” International Colloquium on Computing, Communication, Control, and Management, Vol. 1, CCCM 08, pp. 66-69. Muggleton, J. M. and Brennan, M. J. (2004). “Leak noise propagation and attenuation in submerged plastic water pipes.” Journal of Sound and Vibration, Vol. 278, No. 3, pp. 527-537. Ping, L. and Yumei, W. (2005). “Self-Powered wireless sensor by collecting electromagnetic energy in inhomogeneous structure.” The 4th IEEE Conference on Sensors, pp. 28-31. Silva, H. V. D., Morooka, C. K., Guilherme, I. R., Fonseca, T. C. D., and Mendes, J. R. P. (2005). “Leak detection in petroleum pipelines using a fuzzy system.” Journal of Petroleum Science and Engineering, Vol. 49, Nos. 3-4, pp. 223-238. Sojdehei, J. J., Wrathall, P. N., and Dinn, D. F. (2001). “Magneto-Inductive (MI) communications.” MTS/IEEE Conference and Exhibition OCEANS, Vol. 1, pp. 513-519. Stordeur, M. and Stark, I. (1997). “Low power thermoelectric generatorself-sufficient energy supply for micro systems.” International Conference on Thermoelectrics ICT '97, pp. 575-577. Stuntebeck, E. P., Pompili, D., and Melodia, T. (2006). “Wireless underground sensor networks using commodity terrestrial motes.” 2nd IEEE Workshop on Wireless Mesh Networks, pp. 112-114. Vasquez, J., Rodriguez, V., and Reagor, D. (2004). “Underground wireless communications using high-temperature superconducting receivers.” IEEE Transactions on Applied Superconductivity, Vol. 14, No. 1, pp. 46-53. Wu, M. and Wang, W.-q. (2006). “Application of wavelet to detect pipeline leak point.” International Conference on Intelligent Systems Design and Applications, Vol. 2, ISDA 06, pp. 779-782.

− 811 −

Bui Van Hieu, Seunghwan Choi, Young Uk Kim, Youngsuk Park, and Taikyeong Jeong

Yumei, W., Ping, L., Jin, Y., and Zhangmin, Z. (2004). “Information processing in buried pipeline leak detection system.” International Conference on Information Acquisition, pp. 489-493. Zhan-Hui, L., Xiao-Dong, N., Guang-Yu, M., Xin, W., Hong, Z., and

− 812 −

Yun-Jie, P. (2005). “Method for acoustic leak detection of fast reactor steam generator using maximum modulus based on wavelet transform.” International Conference on Machine Learning and Cybernetics, Vol. 2, pp. 737-741.

KSCE Journal of Civil Engineering