Adaptive Bearing Failure Diagnosis of CNC Machine Tools Based on BSS and Nonlinear-Convolutionary Modeling in ICA for Separation of Vibrations

The paper presents the application of Independent Component Analysis (ICA) - a Blind Source Separation (BSS) technique - for the vibration-based fault diagnosis of rolling element bearings. A geodesic ICA algorithm based on Lie group optimization was implemented, ensuring stable convergence and accurate signal separation. The proposed approach allows for the extraction and enhancement of fault-related signal components, improving the detectability of characteristic defect frequencies such as BPFI and BPFO. A new diagnostic indicator, the Locally Normalized Fault Frequency Amplitude (LNFFA), was introduced to assess the presence of specific bearing faults in localized frequency bands. Experimental results obtained from a rotor-bearing test rig with simulated defects demonstrated a clear enhancement and separation of diagnostic indicators after ICA processing compared to original signals. The findings confirm that ICA-based signal decomposition can increase the sensitivity and reliability of vibration-based bearing diagnostics, particularly in industrial applications where multiple fault sources coexist.