Research And Implementation Of Water Supply Pipeline Leak Detection And Location Based On SVM And Improved EWT

Water resource is one of the basic resources in nature,and water supply pipeline is an important facility to guarantee the transportation of water resource.However,the pipeline will inevitably suffer from natural wear,erosion and other reasons in the long time of fluid transportation,leading to the leakage of water supply pipeline often occurs,which not only brings huge resource loss to the society,but also causes damage to the foundation around the pipeline and endangers public safety.Therefore,the research on pipeline leak detection and location has great application value.In this paper,an experimental platform is set up to monitor the state of water supply pipeline by using acoustic emission technology,and the leakage detection and location of water supply pipeline are realized by using data analysis of signals.The main research contents of this thesis are as follows:1.In practice,the problem of low identification rate of water pipeline leakage due to the influence of working conditions,the causes of pipeline leakage and the principle of acoustic signal generation are analyzed.The wavelet packet entropy characteristics,kurtosis characteristics and sample entropy characteristics of signals under normal working conditions,leakage working conditions,knocking working conditions and passing working conditions are extracted for identification by using time-domain and time-frequency domain feature analysis methods.Finally,combining multiple features into a feature vector and using it as input of Support Vector Machines(SVM)significantly improves the identification rate compared to a single feature vector in the case of testing based on small samples,and the final identification rate of signal state is95%.2.An Improved Empirical Wavelet Transform(EWT)method is proposed to solve the problem that t the decomposition scale and spectrum segmentation interval need to be preset manually in the traditional EWT method.Firstly,the wavelet packet decomposition scale is determined adaptively by the proposed method,then the wavelet packet energy spectrum is obtained by wavelet packet analysis of the signal,and then the local minimum of the energy spectrum is used to adaptively divide the segmentation interval of the frequency band.Finally,the empirical wavelet filter is designed according to the segmentation interval,which can realize the adaptive decomposition of the signal.3.For the problem that the signal is disturbed by environmental noise,which causes a large error for time delay localization.After obtaining a number of components by empirical wavelet packet decomposition,the correlation coefficient is used to filter out the effective components for reconstruction to obtain the reconstructed signal,and then the band with a horizontal trend in the mutual spectrum phase difference spectrum is band-pass filtered to the reconstructed signal to remove the interference noise and extract the signal components containing the leakage component,which further reduces the The error of the time delay is further reduced,and the accuracy of the leakage localization is significantly improved,and the leakage localization error using this method is 1.6289%.4.A water supply pipeline leak detection and location system is developed based on the mixed programming design of Qt platform and Matlab under the environment of Ubuntu platform,.The system demonstrates and analyzes the water supply pipeline leak detection and location algorithm based on SVM and improved EWT,verifies the feasibility of the system and the effectiveness of the algorithm,and realizes the edge-side application of water supply pipeline leak signal detection and location.