Research On Pipeline Leakage Detection Based On Neural Network

With the development of China's social economy,people's dependence on oil and natural gas energy in daily life has become more and more important,and it has become another lifeline for human survival.Pipeline transportation is the fifth largest mode of transportation.Its safety and economy play an indispensable role in transporting oil and gas.However,at present,there is no way to monitor and detect the pipeline once and for all,mainly because some small leaks and leaks are difficult to detect.Most of the pipeline leaks that occurred were due to leaks and small leaks that were not detected and eventually caused irreparable damage.Therefore,many researchers have been working on the research of pipeline leak detection to find a way to effectively monitor and detect pipelines and reduce the occurrence of leakage accidents.In this paper,the problem of signal denoising and leakage signal detection for pipeline leakage condition signal and normal working condition signal is mainly carried out:To solve the problem of denoising the collected pipeline condition data including interference signals,A wavelet packet optimization local mean decomposition denoising method based on wavelet packet analysis and local mean decomposition method is proposed in this paper.Due to the denoising method using wavelet analysis,the resolution of the low frequency time domain and the high frequency frequency domain of the decomposition is low,that is,the low frequency time domain signal characteristic component will contain signal information of multiple frequencies,and the frequency of the signal is no longer decomposed in the high frequency domain..The method of empirical mode decomposition and local mean decomposition will produce more obvious end effect and modal aliasing.The empirical mode decomposition is more serious than the local mean decomposition,and its end effect and modal aliasing are more serious.Therefore,this paper uses the wavelet packet to optimize the local mean decomposition method to reduce the noise of the pipeline working condition signal,and the noise reduction effect is significantly improved.To solve the problem of low accuracy of pipeline leak detection,A BP(Back Propagation)neural network pipeline leak detection method optimized by Genetic Algorithm(GA)is proposed in this paper.The method uses GA to solve the optimal weight and threshold of BP neural network,and constructs the pipeline leakage detection model of BP neural network.By extracting the feature of leakage pressure signal,the leakage feature vector is established,and this vector is used as the input of BP neural network.The leakage condition category serves as the output of the BP network.By comparing with the accuracy of leak detection based on Particle Swarm Optimization(PSO)algorithm BP neural network,this method can accurately detect and identify leaked signals.