Pipeline Leak Fault Detection And Location Based On Wavelet Analysis

As the safe and economic transport tools for conveying resources such as oil and gas,pipeline transportation has a natural advantage. Crude oil pipeline transportation in Chinadeveloped rapidly to form a large pipeline network. However, pipeline leak caused by wearof long-time running, natural aging of equipment, changes of geography and atmosphereand man-made damage occur frequently. Therefore, it is of great economic importance andsocial significance to establish pipeline leak detection system.The research on pipeline leakage detection and localization is carried out based ondetailed analysis of pipeline fluid flow characteristics. The main efforts accomplished canbe summed up as follow:1. The pressure signals are pre-processed based on wavelet analysis: The liftingwavelet packet transform is applied to decompose the pressure signals. The waveletcoefficients are de-noised by adaptive threshold method. The pressure signals are de-noisedby adaptive threshold lifting wavelet package.2. The pipeline leakage is detected using wavelet transform modulus maxima method:The wavelet transform is applied to decompose the pressure signals upstream anddownstream in different scales. Those high frequency parts of the wavelet coefficients areaveraged. The averages are compared with the given threshold to detect whether thepipeline leakage happened.3. The pipeline leakage is located using improved wavelet neural network: Thepressure signals formed when the pipeline leaks in different positions are taken as thesample. The wavelet transform is applied to extract fault feature. The feature vectors are sent into wavelet neural network, and the wavelet network is trained. The test pressuresignal is identified using the trained network to locate the position of the pipeline leakage.Wavelet neural network combined neural network and wavelet analysis. Twoimproved points of traditional algorithm of wavelet neural network are proposed. Firstly,inertial coefficient has been added into correcting process of network weights in order toreduce the vibratory trends in learning process and increase the training speed. Secondly,local adaptive learning rate algorithm based on gradient symbol change is used in thelearning process, which can adjust learning rate dynamically by itself according to each ofadjustable parameters’ state information. The wavelet network is trained to locate theposition of the pipeline leakage. Final test results show that the improved algorithm ofwavelet neural network is of great training speed, high identified accuracy and strongstability and play an important role in the pipeline leak localization.