Hybrid Collaborative Diagnosis Method for Rolling Bearing Composite Faults

Traditional composite faults diagnosis approaches applied in isolating rolling bearing mixed faults are not effectively.To overcome this defect, this paper proposes a hybrid collaborative diagnosis method for rolling bearing composite faults.Firstly, by using second order blind identification method, we calculate the whiten matrix of fault signals.Then, by minimizing the degree of diagonalization function, we can realize joint diagonalization on whiten fault matrix.Finally, we use singular value decomposition for the orthogonal matrix, further we can estimate the signal matrix of fault sources from the compound fault signals.But through the above procedure, we can only get the out-of-order signals.If there are some similar signals, it is hard to recognize them by our intuitions.So we next apply the short-term Fourier transform to solve this problem, by comparing the time-frequency spectrograms trends between isolated and original signals, we can clearly identify the fault types.At last, to validate its effectiveness, we put the fault data from the Key Laboratory of Fault Diagnosis of Petrochemical Equipment in Guangdong Province into this proposed method, the results show its practical value in mechanical engineering.