Vibration-Based Bearing Fault Diagnosis Using Reflection Coefficients of the Autoregressive Model

Bearing faults are one of the main reasons of rotary machines failure. Monitoring vibration signal is an effective method for diagnosing faulty bearings and preventing thus catastrophic failures. However, existing algorithms neither offer satisfactory accuracy nor are efficient for real-time implementation due to complexity in feature extraction part. In this paper, we propose an accurate method for bearing diagnosis customized for real-time implementation. The proposed system estimates power spectral density of vibration signal using an autoregressive model for feature extraction. This is a novel use of autoregressive model for fault diagnosis which reduces the dimensionality of vibration signal and captures its frequency contents simultaneously. The proposed system can diagnose different bearing faults under variable load conditions with above 99% accuracy.