Plastic Bearing Fault Diagnosis Based on a Two-Step Data Mining Approach

Abstract

Plastic bearings are widely used in medical applications, food processing industries, and semiconductor industries. However, no research on plastic bearing fault diagnostics using vibration sensors has been reported. In this paper, a two-step data mining-based approach for plastic bearing fault diagnostics using vibration sensors is presented. The two-step approach utilizes envelope analysis and empirical mode decomposition (EMD) to preprocess vibration signals and extract frequency domain and time domain fault features as condition indicators (CIs) for plastic bearing fault diagnosis. In the first step, the frequency domain CIs are used by a statistical classification model to identify bearing outer race faults. In the second step, the time domain CIs extracted using EMD are developed to build a k-nearest neighbor algorithm-based fault classifier to identify other types of bearing faults. Seeded fault tests on plastic bearing outer race, inner race, balls, and cage are conducted on a bearing diagnostic test rig and real vibration signals are collected. The effectiveness of the presented fault diagnostic approach is validated using the plastic bearing seeded fault testing data.