Bearing Fault Diagnostics Based on Reconstructed Features

Rolling-element bearings are widely used in various mechanical and electrical systems. A reliable bearing fault diagnostic technique is critically needed in industries to recognize a bearing fault at its early stage so as to prevent system's performance degradation and malfunction. In this work, a genetic programming based feature reconstruction approach is proposed for bearing fault diagnostics. A new fitness measure is proposed to improve the GP operations in feature formulation. The original features are from the modified kurtosis ratio and the one-scale wavelet analysis. Investigation results show that the proposed method is an effective feature formulation tool; the reconstructed features are more robust against the variations in bearing geometry and operating conditions. The corresponding fault diagnostic reliability can be enhanced significantly. As a result, this work provides a promising technique and tool for bearing condition monitoring for real-world applications.