Vibration-Based Diagnostics of Rolling Element Bearings Using the Independent Component Analysis (ICA) Method.

This manuscript presents a study on the application of blind source separation (BSS) techniques, specifically the Independent Component Analysis (ICA) method, for the detection and identification of localized faults in rolling element bearings. Bearing defects typically manifest as distinct harmonics of characteristic fault frequencies, accompanied by modulation sidebands in the vibration signal spectrum. The accurate extraction and isolation of these components are crucial for reliable fault diagnosis, particularly in systems where multiple vibration sources overlap. In this work, a linear ICA algorithm was applied to vibration signals acquired from a simplified rotating machinery setup designed to emulate common bearing fault conditions. The study investigates the effect of ICA-based signal decomposition on the statistical distribution of selected diagnostic indicators and evaluates its ability to enhance the detectability of fault-related components. The experimental results demonstrate that the application of ICA significantly improves the separation of vibration sources, leading to a more distinct representation of fault signatures. The findings confirm the effectiveness of blind source separation methods in vibration-based diagnostics and highlight the potential of ICA as a complementary tool for improving the accuracy and robustness of bearing fault detection systems in rotating machinery.