Rolling Bearing Fault Diagnosis Under Different Severity Based on Statistics Detection Index and Canonical Discriminant Analysis

Bearing failures are the most frequent causes of breakdowns in rotating machinery. Different levels of severity in these failures exhibit distinct fault characteristics in the vibration signal. This paper presents a bearing fault diagnosis method that considers different severity levels, involving the selection of statistics detection index symptom parameter and the application of canonical discriminant analysis (CDA). Initially, kurtosis is employed to detect abnormalities in the bearing. Subsequently, statistical analysis theory is utilized to extract efficient symptom parameters from the time domain and frequency domain vibration signals. As a statistical analysis method, CDA can discriminate between different signals by maximizing the between-group difference and minimizing the inter-group difference. By analyzing the distribution of CDA canonical scores, bearing faults can be intuitively diagnosed. The proposed method is validated using vibration signals obtained from an experimental bench with three different bearing conditions (normal, inner race fault, outer race fault) exhibiting varying severity levels. The results demonstrate the effectiveness and feasibility of diagnosing faults under different severity levels.