Automatic leakage detection for water supply systems using Principal Component Analysis

The purpose of this study is to propose an easy and stable automatic leakage detection method using acoustics, and obtain higher accuracy than our previous study. In this study, a Support Vector Machine (SVM) was used for pattern recognition. Approximately, 8 leakage sounds, and 8 pseudo-sounds were used to train the SVM which was then tested using the remaining sounds (at present about 90 data samples have been collected). Eigenvalues which were derived from Principal Component Analysis were used in the SVM as feature vector components. The PCA was done on data representing specific range of frequency of the Power Spectral Density functions as obtained from the divided data. Using the proposed method, the classification accuracy of 90.4% was obtained. This high accuracy result showed that possibility of automatic leakage detection.