Research And Application Of Leak Detection And Location Based On VMD And Cross-spectrum Analysis In Water Supply Pipeline

The water supply pipeline network is an essential infrastructure of social existence and development,and the pipeline leakage has been a very serious problem that our country is facing.Therefore,the leak detection and location have great important practical significance.There are many leak detection methods for water supply pipelines,and the method of acoustic detection is widely used.This Thesis takes the water supply pipeline leakage as the background,and takes the leakage sound signal as the object.The Signal-Noise Ratio(SNR)of the leakage signal under low SNR is improved by using the location method based on variational mode decomposition(VMD)and cross-spectrum analysis,and then the location accuracy is also improved.The feature vectors obtained by the VMD decomposition of the signals in different working conditions are combined to improve the accuracy of leak recognition.The main research contents include:1.Aiming at the problem that the large leak location errors under the low SNR of leak acoustic signal in water supply pipeline,the leak location method based on the combination of cross-spectral and VMD analysis is proposed.Firstly,the characteristic frequency band of leakage signal can be found using cross-spectrum analysis,and then the leakage signal is decomposed into several IMF components using VMD.The energy ratio of the IMF component in the characteristic frequency band is used as the selection criterion to determine the effective IMF components for signal reconstruction.Finally,the time delay of the reconstructed signal is estimated to determine the location of the leak.Experimental results show that the average relative location error of the proposed method is 2.530%,which is smaller than the cross-correlation of 8.621% and the combination of VMD and correlation coefficient of 16.860%.2.Considering the non-linear and non-stationary characteristics of pipeline leakage signals,the pipeline signals such as the leakage,knock,normal and crossing are decomposed by using VMD,and then feature vectors such as kurtosis,sample entropy and center frequency can be extracted.The experimental result shows that the recognition rate of leakage detection which is based on single feature or feature vector is not high.The final leak detection recognition rate is 98.75% by combining the VMD kurtosis,VMD sample entropy and VMD center frequency as the input of the support vector machines(SVM).Compared with the highest recognition rate of a single feature vector which is 91.25%,it is significantly improved.3.Using mixed programming of Qt and MATLAB,the proposed method of leak detection and location based on combination of VMD and cross-spectrum analysis in water supply pipeline will be displayed on the interface.The software system includes login management,information management and data processing function modules,which realizes the convenient operation and friendly interface of leak detection and location.Relevant research contents were verified on specific cases,and the experimental results show that the software can achieve the function of leak detection and location in the water supply pipeline proposed in this thesis.