Rotor crack and bearing fault diagnosis based onVMD-ICA

Abstract The rotor-bearing system is vulnerable to rotor cracks, bearing failures, or even multiple failures at the same time during long-term service, and it is more difficult to diagnose the faults under rotor crack and bearing fault coupling conditions. This paper proposed a signal reconstruction method to classify the rotor crack, the bearing outer ring fault and their coupling faults based on Variational Mode Decomposition (VMD) and Independent Component Analysis (ICA). Firstly, the original vibration signal collected on a rotor-bearing test bench is decomposed with VMD, and several intrinsic mode functions (IMF) are obtained. Then, the correlation kurtosis CK M which introduces the characteristics of periodic correlation function is to evaluate the decomposed IMF component, and screen the rotor crack information component. Finally, the rotor crack information component is input into a virtual channel signal and separated with the original vibration signal with ICA, and then the separated two sets of signals are superimposed to obtain a vibration signal that simultaneously highlights the information of the rotor crack and the bearing outer ring fault. In addition, the time domain signal is directly input into a LightGBM to diagnose different faults. The research results show that this method can simultaneously diagnose the rotor crack, the bearing outer ring fault and the coupling fault.