Rotary Machines Fault Diagnosis based on Principal Component Analysis

Abstract

Rotating machines are commonly used in industrial applications. Mechanical faults such as rotor unbalance, shaft misalignment, pulley misalignment, structural looseness, and bearing faults leading to unplanned shutdown based on the severity of these faults. The condition monitoring technique based on vibration analysis has the potential to detect and diagnose a great number of early stage faults. However, some mechanical faults have correlated vibration features leading to ambiguous diagnosis to identify and distinguish these faults. In this paper, a proposed method based on the Principal Component Analysis (PCA) is presented to produce uncorrelated Principal Components (PCs) to identify the healthy and different faulty cases. A test rig was prepared to simulate a group of mechanical faults such as rotor unbalance, pulley misalignment, belt damage, combined unbalance with pulley misalignment, and combined unbalance with belt damage. The conventional vibration measurements were collected for each case and their features were extracted and used to produce the equivalent PCs. It was found that the produced uncorrelated PCs have the superior to distinguish the majority of simulated faults which have correlated vibration features as presented in the rest of paper.