Anomaly Detection in Bearing Inner Race via Spectral Kurtosis (SK) and Autoregressive (AR) Modeling

Today, fault diagnosis plays an essential role in industrial systems by improving equipment reliability and ensuring the safety of both machines and people. Advanced monitoring methods have become indispensable, and among them, vibration analysis stands out in condition-based maintenance for its high reliability. This technique is used to detect and diagnose faults in rotating systems. This article focuses on a vibration analysis technique and signal preprocessing methods that allow for diagnosing ball bearing faults, particularly inner ring defects. Three techniques are discussed: envelope analysis, the most widely used method in vibration analysis, as well as two signal preprocessing techniques the autoregressive (AR) model and spectral kurtosis (SK), chosen to illustrate the ball bearing diagnostic process. These techniques are applied to real datasets provided by Case Western Reserve University. The comparison of results highlights a complementarity among the different techniques, which offer reliable results and are positioned among the recommended tools for monitoring ball bearing faults.