Fault-tolerant control for levitation systems of high-speed maglev train based on diversified basis neural networks

To address the effects of system parameter perturbations, actuator faults, and left and right electromagnet coupling in high-speed maglev vehicles during long-term service, the mutual coupling relationship between left and right electromagnets and the actuator faults in the joint-structure of the maglev vehicle suspension system during the process of connecting the vehicle bodies were analyzed. A neural network method for adaptive fault-tolerant suspension control based on diversified basis functions was proposed. Diversified basis functions were introduced into neural networks, and an upper norm boundary processing method for neural networks was incorporated to address complex and discontinuous issues in the control process. Through Lyapunov functions, the fault tolerance of the proposed method against failures and its robustness to uncertain system dynamics were verified. On this basis, the ultimately uniformly boundness of the control method was proven. Experimental results indicate that when partial failure occurs in electromagnets, adaptive variables are adjusted according to fault conditions and affect control currents, thereby achieving fault tolerance performance. For steady-state signal tracking, left and right electromagnets show maximum errors of 0.2 and 0.1 mm and average errors of 0.14 and 0.09 mm, respectively. For sinusoidal signal tracking, left and right electromagnets demonstrate maximum errors of 0.2 and 0.1 mm and average errors of 0.18 and 0.10 mm, respectively. For square wave signal tracking, left and right electromagnets exhibit maximum errors of both 1.1 mm and average errors of 0.18 and 0.14 mm, respectively. The proposed method displays adaptability to faults in both left and right electromagnets, enabling rapid tracking of desired signals and ensuring operational reliability and safety of maglev vehicles.