Leakage Detection of Water-Induced Pipelines Using Hybrid Features and Support Vector Machines

Abstract

Pipelines are significant parts of water distribution systems for life and manufacture. Any leakage occurring on those can result in a waste of resource and finance; consequently, detecting early faults for them become necessary. Nowadays, there are many approaches to deal with this problem; however, their results still have limitations. This paper proposes a pattern recognition method that first extracts time-domain and frequency-domain features from vibration signals to represent each fault distinctly, and these features are then utilized with a classifier, i.e. support vector machine (SVM), to classify fault types. To verify the proposed model, the experiments are carried out on different samples of fault in various operating conditions such as pressure, flow rate and temperature. Experimental results show that the proposed technique achieves a high classification accuracy for different leakage sizes, which can be applied in real-world pipeline applications.