Liquid pipeline leakage detection based on moving windows LS-SVM algorithm

Abstract

Aiming at the problem of liquid pipeline leak detection, this paper proposed a leakage detecting method based on moving windows least square support vector machine algorithm (MWLS-SVM). Least square support vector machine algorithm (LS-SVM) classification algorithm is one of the improved algorithms of support vector machine (SVM). The main idea of the LS-SVM classification algorithm is to change the nonlinear constraint in SVM to linear constraint and to apply the sum of square errors as the empirical loss function of the training set to improve the training speed. The moving windows method is used to update the dynamic pipeline leak detection model. In this paper it also uses the negative pressure wave method for pipeline leak detection. The simulation experiment results of the liquid pipeline leakage data show that this proposed method has higher accuracy than support vector machines and neural networks methods.