A fault prediction method for CMOR bearings based on parameter-optimized variational mode decomposition and autocorrelation function

The CFETR Multi-Purpose Overload Robot (CMOR), a 7- DOF manipulator with 7 joints, serves as a crucial component in the China Fusion Engineering Test Reactor (CFETR). Each joint integrates electrical and mechanical components seamlessly including sensors, motors, and planetary reducers. The robot is characterized by its heavy-load capacity, high precision, and compact structure, placing significant demands on both the bearings and the gears of the reducers. Therefore, fault prediction for bearings of CMOR is essential and beneficial for normal operation of the whole machinery. This paper presents a novel fault prediction method for CMOR bearing based on parameter-optimized variational mode decomposition (VMD) and autocorrelation function (ACF). Firstly, a new fitness function is defined based on ACF. Then, optimize parameters of VMD by slime mold algorithm (SMA). Next, VMD is performed on the signal and suitable modes are selected for signal reconstruction. Then Hilbert transform (HT) is employed to the reconstructed signal for fault prediction. Finally, in order to evaluate the effectiveness of each envelope spectrum, a quantitative indicator (Q) is defined to measure the proximity between the theoretical fault characteristic frequency (or its 2nd to 4th harmonics) and the center frequencies computed in the spectrum. Comparative experiments through the quantitative indicator Q demonstrate that the superiority of the proposed method has validated over two other methods. It is also demonstrated that the proposed method significantly benefits CMOR bearings' future fault prediction.