Multiscale morphology analysis and its application to fault diagnosis

A novel approach to fault diagnosis is proposed using multiscale morphology analysis to extract impulsive features from the signals with strong background noise. Multiscale morphology is applied to one-dimensional signal by defining both the length and height scales of structuring elements (SEs). A local-peak-value based adaptive algorithm is also introduced. The new approach makes the selection of SEs more transparent and is independent of empirical rules. Both simulated impulsive and vibration signals of two defective roller bearings are employed to validate the proposed algorithm. The roller bearing faults presented in the validation include both inner and outer race faults. The test results show that the multiscale morphology analysis is effective and robust to extract morphological features.