An IGSA-VMD method for fault frequency identification of cylindrical roller bearing

Abstract

The number of intrinsic mode function (IMF) and quadratic penalty factor (QPF) are two important parameters in variational mode decomposition (VMD) for fault frequency identification of cylindrical roller bearing, but it is difficult to obtain the optimal values by experiential knowledge. Accordingly, an improved gravitational search algorithm (IGSA) with nonlinear decreasing inertia weight factor is integrated into the VMD, namely as IGSA-VMD, for adaptive selection of the two parameters. Firstly, a target function of minimum envelope entropy is defined in the IGSA to optimize the IMF number and QPF. Secondly, the optimized two parameters are employed to decompose the measured vibration signals into several IMF components by utilizing the VMD. Finally, the feasibility of proposed IGSA-VMD method is validated through the fault frequency identification of cylindrical roller bearing and the benchmark test dataset from Case Western University. Both the simulative and experimental results show that the extracted fault frequencies are highlighted more clearly by utilizing the IGSA-VMD comparing to the PSO-VMD and the traditional VMD. The proposed IGSA-VMD is a more efficient and effective vibration diagnostic approach for outer race and roller fault frequencies extraction from original vibration signals of cylindrical roller bearings in rotary machines.