Diagnosis of partial blockage in water pipeline using support vector machine with fault-characteristic peaks in frequency domain

Abstract

Partial blockages in water pipe network can cause waste of energy and poor hygiene. Therefore, periodic diagnosis of water pipe state is necessary to maintain or replace blocked pipe when the blockage size is larger than a threshold. This work proposes a nondestructive diagnosis scheme that estimates the partial blockage in water pipe by classifying pressure signals in the frequency domain. Pressure data were collected with normal and two different fault states. A peak search algorithm is proposed to identify the ‘fault-characteristic’ peaks (FC-peaks) relevant for each blockage size. Support vector machine (SVM) classifier for each blockage was constructed with the FC-peaks as input. The SVM scores of different blockage sizes are used for diagnosis of partial blockage. The partial blockage can be diagnosed by comparing the SVM scores of different blockage sizes. The SVM classifier was able to successfully classify and diagnose three model pipes with normal state, moderate, and severe blockages.