Variational Mode Decomposition and Artificial Lemming Algorithm-Optimized Back Propagation Neural Network-Based Inter-Turn Short-Circuit Fault Diagnosis for Permanent Magnet Synchronous Motors

To address the issues of local optimum trapping and slow convergence speed of back propagation (BP) neural networks in the inter-turn short circuit (ITSC) fault diagnosis of permanent magnet synchronous motors (PMSMs), this paper proposes a novel fault diagnosis method integrating variational mode decomposition (VMD) and the artificial lemming algorithm (ALA)-optimized BP neural network. Firstly, stator current signals of PMSMs under diverse fault states are acquired, and the third harmonic components are extracted via VMD as fault features to achieve effective ITSC fault identification. On this basis, a fault feature dataset is established based on the extracted characteristic parameters, which is subsequently imported into the ALA-BP model for quantitative classification of ITSC fault severity. Simulation results verify that the proposed ALA-BP model achieves a diagnostic accuracy of 98.5%. Compared to the PSO-BP, NGO-BP and GWO-BP models, the proposed method improves diagnostic accuracy by 3.2%, 4.5% and 5.3%, respectively, and exhibits a substantially superior convergence performance.