Extreme asymmetric condition-based fault diagnosis and stability assessment in EV-driven IM using DEGSANN and ANRLSFIS

The performance of electric vehicles mainly depends on the stability, efficiency, and safety of the motor drive system. Hence, Artificial Intelligence (AI)-based fault detection and diagnostics have become an important research area in EV-driven Induction Motors (IMs). Yet, existing research works failed to focus on the Extreme Asymmetric Condition (EAC), reducing diagnostic reliability under complex operating conditions. Therefore, this paper proposes an EAC-based FD and Stability Assessment (SA) in EV-driven IMs using Deep Elman Gradient and Spatial Attention Neural Network (DEGSANN) and Adaptive Neuro Regularised Least Squared Fuzzy Inference System (ANRLSFIS) approaches. Initially, motor data are generated using a Variable Frequency Drive (VFD) simulation model. The generated data are pre-processed through Missing Value Imputation (MVI) and non-normality correction. Subsequently, the K-Quartic and Shrinkage Mahalanobis Function (KQSMF) method is employed for data grouping and graph construction. Features and EAC information are extracted using Negative Sequence Second Ordered Component Analysis (NSSOCA). Based on the extracted information, fault diagnosis and stability assessment are performed. Experimental results demonstrate that the proposed DEGSANN-based framework achieves an accuracy of 99.7453%, outperforming existing methods in terms of fault diagnosis reliability and stability assessment, thereby enhancing the operational robustness and safety of EV-driven induction motor systems.