184 / 2025-06-13 10:23:51

Fault estimation of IMT systems in electric vehicles via KalmanNet

Integrated motor-transmission (IMT),Kalman filter,fault estimation,electric vehicles

Special Sessions > SSL 08 New perspective of intelligent detection, diagnosis, prognosis, and maintenance: Generative AI and Industrial Large Model

With increasing demands for efficiency in electric vehicle (EV) powertrains, integrated motor-transmission (IMT) systems have gained significant attention due to their compact configuration. To address the critical need for real-time state information in closed-loop control and onboard fault diagnosis of IMT systems, this paper studies KalmanNet - an innovative observer based on the Kalman filtering framework. While conventional Kalman filters (KF) provide accurate predictions for fully known linear Gaussian state-space(SS) models, they are limited when dealing with practical nonlinear systems with incomplete modeling. KalmanNet integrates model-based and data-driven approaches by employing deep neural networks to dynamically compute the Kalman gain, effectively reducing dependence on precise system models and eliminating the need for Jacobian matrix computations in extended Kalman filters (EKF). This method not only applies to partially known nonlinear systems but also has low computational complexity, high data efficiency, and interpretability. Experimental validation on an IMT system demonstrates that the proposed observer achieves accurate estimation of system inputs and state parameters, while successfully identifying actuator faults through state deviations.