Gas Pipeline Leakage Detection And Location Based On HHT

As the high efficient,convenient and environment-friendly transport tool,pipeline transportion has unique advantages especially in conveying gas and liquid.It has become one of the five important transportations,in parallel with railway,highway,aviation and water means.However,due to inevitable aging and erosion,as well as natural disasters and contrived damage and so on,pipeline leakage accident happened frequently,which had lead to a great loss of property and serious environmental pollution as well.Therefore,pipeline leak detection and location has become an urgent problem to be solved.With the development and application of computer technology,signal processing,pattern recognition,artificial intelligence and other disciplines,software-based pipeline leak detection has become one of the most popular methods because of detection sensitivity,low cost,simple and other advantages.In this background,this paper makes deep research and analysis to the technology of pipeline leakage detection and location to meet the needs of the actual detection of pipeline.By adsorbing and integrating the previous research results,the thesis mainly discusses the method of Hilbert-Huang Transform,support vector machine and neural network.The main work of this paper has the following several aspects:To establish and solve gas pipeline model: On the basis of detailed analysis of the pipeline fluid dynamics,the steady-state and dynamic model of gas pipeline is established.Then using the fourth-order Runge-kutta method and characteristic line method respectively to solve the model,the pressure and flow value is get.The pipeline leakage is detected by the method of Hilbert-Huang Transform and SVM:The pipeline operation conditions are classified into three types: normal state,leakage state and valve adjusting state.Firstly,the feature vectors of these three type signals are extractedby the Hilbert-Huang Transform.Secondly,the support vector machine is chosen as the classifier whose input is the extracted feature vectors and then is trained by history signals.Lastly,the detecting signal is distinguished as a leakage signal,a valve adjusting signal or a normal signal by the well-trained support vector machine.The result shows it has better effect of classifying than BP network.The pipeline leakage position is located by the method of Hilbert-Huang Transform and the BP neural network with particle swarm optimization: Firstly,the negative pressure wave positioning principle and formula are introduced.According to the complexity of revised negative pressure wave formula,the leak position formula is fitted by BP neural network whose parameter is optimized by particle swarm optimization algorithm(PSO).Then,analysis of the accuracy of time delay estimation by HHT transform and wavelet transform,the HHT transform is used to obtain the abrupt change time of leakage signal.