Rolling bearing multi-fault diagnosis based on AE signal via ICA

An acoustic emission signal separation approach based on fast independent component analysis (ICA) is proposed for fault diagnosis of rolling bearing. When various faults exist, the AE sensor would collect a mixed fault acoustic emission signals. This paper firstly separates the AE signal sources by Fast ICA based on the largest negative entropy principle. Then the spectral features are extracted. Through feature comparison between the mixed multi-fault AE samples and the single fault samples, four running states of rolling bearing can be diagnosed, including the normal state and three fault states, i.e., the rolling element defect, the inner race defect and the outer race defect. The validity of the proposed method is proved by the simulation using actual experimental data of a rolling bearing.