Research On Leak Detection Technology Of Pipeline Based On RBF Neural Network And Wavelet Transform

Because of the special advantage pipeline plays a important role in the long distance transportation of crude oil or refined oil. But with aging, erosion and man-made damage, the leakage occurs frequently. Especially, in recent years, the people organized to drilling Oil Stolen seriously interfere with the normal oil transportation, which not only causes great loss of economy, but also leads to serious pollution to the environment. So, to detect pipeline leakage timely and locate accurately have great significance of economic and social for protecting both environment and national property.The technology of pipeline leakage detection which is researched in this thesis is a major topic in the field of fault diagnosis. Several typical methods of pipeline leakage detection are discussed, and their advantages and disadvantages are compared. On the basis of this, the new method based on RBF neural network combined with wavelet transform is proposed.Firstly, aimed at the problem during the pipeline leakage detecting, the nearest neighor clustering algorithm of the negative gradient descent algorithm are proposed, which can not only update the center, width and weight in-line, but also determine the structure of RBF neural network automoatically. Thus the learning accuracy of RBF is improved effectively. A large number of simulations show that the new algorithm is effective in pipeline leakage detection.Secondly, in the view of the problem of the RBF algorithm, the wavelet transporm denoising method is proposed to eliminate the noise. This method puts forward a new threshold function which is continuously updated according to real-time data. At the same time, decomposition levels are adjusted by coarse and fine adjustment, which can not only accurately determine the decomposition level, but also reduce the decomposition time. Experiments using MATLAB prove that the algorithm has better effects than traditional methods.Finally, according to the research above, the method is proposed to detect leakage based on RBF neural network combined with wavlet. RBF neural network is also imported to find the accurate location of leakage with the help of Negative pressure wave technique and to analyse working condition. Lots of simulations prove that the pipeline leakage can be detected accurately by the method proposed in the thesis.According to the data obtained from the Sinopec Group, lots of simulations are carried out, the results show that the method which is proposed in the thesis can detect the leakage in the pipeline timely and accurately and have good practicability and feasibility. This method will play a more and more important role in the pipeline leakage detection with further development of the theory and practice.