Wheelset bearing fault detection using morphological signal and image analysis

The detection of wheelset bearing faults is of extreme importance for railway vehicle operation safety. Wheelset bearing faults induce impulses in vibration signals, which are hard to detect because of signal modulation and environmental noise. To suppress noise and identify these impulses effectively, we propose a method that leverages morphological signal and image processing techniques. The proposed method mainly includes two aspects: a novel double cross-correlation operation for noise reduction and an improved image processing algorithm for highlighting the fault features. The performance of the proposed method was tested using real vibration signals collected from a wheelset bearing test rig and compared with other advanced methods reported in the literature. By analysing two wheelset bearing faults, namely, an outer race fault and a pin roller fault, the proposed method is demonstrated to be more effective in the detection of wheelset bearing faults.