Multi-Electrical Signal Analysis Based Bearing Fault Diagnosis for Permanent Magnet Synchronous Machines Under Uncertain Controller Bandwidths

The analysis of electrical signal signatures has become the state-of-the-art solution in bearing fault diagnosis of permanent magnet synchronous machines (PMSMs). However, signatures in electrical signals of PMSMs are easy to vary with the bandwidths of PI controllers, which will affect the accuracy of fault diagnosis. Aiming to solve this issue, a bearing fault diagnosis method considering the interaction among controller bandwidths and electrical signals has been proposed in this paper, which has superior performance in automatically diagnosing various fault patterns. Firstly, it uses the variational mode decomposition to extract the fault harmonic components from electrical signals under various controller bandwidths, by which the fault features can be obtained afterwards. The above fault features under various controller bandwidths are then employed as inputs for the training of a variable prediction model (VPM). The VPM after training will be finally used for real-time diagnosing the patterns of bearing faults. The proposed method is tested on PMSMs equipped with three types of faulty bearings and shows good accuracy in recognizing various fault patterns under uncertain controller bandwidths.