Early Crack Fault Detection of Rotor Systems Based on an Optimized VMD

Rotor crack is a common and potentially dangerous fault in rotating machinery, which may cause catastrophic accidents if not detected in a timely manner. Early detection of rotor cracks is of great significance to ensure the safe and reliable operation of rotating equipment. As the response induced by early crack is very weak, it is difficult to extract fault feature information directly from the nonlinear and non-stationary vibration signal of rotor systems. Variational mode decomposition (VMD) can reduce the non-stationarity of signals with nonlinearity and enhance the weak features in the signal. However, the decomposition performance of the VMD is affected significantly by the mode number K and the quadratic penalty factor α, which often need to be manually selected. In this article, the model number K is determined based on the center frequency observation method and an optimization method for α parameter is proposed with sample entropy as the objective function. Then, the vibration response signals of a rotor system are decomposed into K IMF components by the parameter optimized VMD. Finally, the 2× and 3× super harmonic components of the IMF components are obtained as the early crack fault characteristic indexes. The correctness of the proposed method is verified by the fault diagnosis of a rotor system with a crack depth of 0.01D ~ 0.05D at 1/3 subcritical speed.