A composite iterative neural network sliding mode control for hybrid reluctance actuator

In the hybrid reluctance actuator (HRA)-based stage (HRA-BS), the uncertainties can be divided into a repetitive part and a non-repetitive part. To this end, this paper proposes an improved adaptive sliding mode control (ASMC) method with an adaptive-iterative composite neural network (AICNN) compensator for the closed-loop control of HRA-BS. Firstly, an AICNN-ASMC structure is constructed, where a sliding mode control (SMC) is used for guaranteeing the closed-loop control performance, and an adaptive-iterative neural network (NN) compensator and an adaptive switching controller are used to tackle the repetitive and non-repetitive uncertainties, respectively. Secondly, composite kernel functions are designed for the NN to enhance the compensation ability for the repetitive uncertainty, where an inverse hysteresis operator (IHO) is incorporated into the kernel functions to characterize the hysteresis effect. Thirdly, a predictive error (PE) term is embedded into the iterative learning law to accelerate its convergence. The convergence of the closed-loop control system is analyzed by the Lyapunov theorem in the sense of the \(L^2 \) $L^2$ -norm with a composite energy function. Comparative experimental results show that the AICNN-ASMC method can guarantee high control performances for the HRA-BS.