Monitoring of Natural Gas Pipeline Leaks

Pipe is a very important tool for long-distance transportation of nature gas. In the long-term running, there will be inevitably an appearance of a rupture, leak or damage usually caused by manmade event or by nature disaster. Leaks may generate dangerous clouds of gas escaping from the high-pressure pipe and produce serious incidences involving fire and explosion endangering the life and property safety of people in and around the area. Monitoring of natural gas pipeline leaks will timely find out and locate these dangerous occurrences and reduce loss. Within the leak monitoring, the core contents are the accurate location of leaks as well as the rapid identification of different signal sources reducing false alarm ratio. Once a leak occurs, the supersonic jet of escaping gas can generate a non-linear & chaotic negative pressure wave signal based on static pressure measurement and an acoustic signal based on dynamic pressure measurement [1]. By properly interpreting these two kinds of signals together, it is possible to detect and locate the leak along the pipe. However, useful signals usually mix in the powerful backdrop signals and noises. In order to resolve the problem, the wavelet packet decomposition technique [2] is used to reduce the noises and get the feature signals of negative pressure wave and acoustic wave. Furthermore, a lot of different condition regulating signals for instance compressor start-stop, valve adjusting and gas turbulence can interfere with the accurate identification of leaks and result in false alarm. It is quite required to classify these similar signals. Thus, BP neural network [3] is used to quickly recognize the different pressure fluctuation signals. Finally, an integrated system developed by LabView is introduced to timely monitor the operation condition and locate the leak. Field tests indicate this system using negative pressure wave method, acoustic wave method, wavelet packet decomposition technique as well as BP network has a good effect.