Multi Fault Diagnosis of the Centrifugal Pump Using the Wavelet Transform and Principal Component Analysis

Abstract

In this paper, the features of vibration signals from normal and faulty conditions of a centrifugal pump were extracted from time-domain data using the discrete wavelet transform (DWT). The DWT with Multi Resolution Analysis (MRA) was used to pre-process raw vibration signals prior to extraction of statistical features. The features obtained were used as input to Principal Component Analysis (PCA). A method based on PCA was then developed to build a framework for multi-fault diagnosis of centrifugal pumps by using historical normal conditions. The fault detection was determined using T ²-statistics and Q-statistics while fault identification was carried out through the combination of loadings and scores of principal components (PCs). The normal and faulty conditions of the centrifugal pump were collected from the Spectra Quest Machinery Fault Simulator. Various fault conditions were investigated in the experiment including cavitation, impeller fault, and combination of impeller fault and cavitation. The results showed that combined wavelet-PCA can be used to detect multi-faults in the centrifugal pump. Furthermore, the combination of loadings and scores of PCs was demonstrated which showed effective fault identification.