Adaptive super-twisting sliding mode control with neural network for electromechanical actuators based on friction compensation

Mengmeng Cao,Jian Hu,Jianyong Yao
DOI: https://doi.org/10.1177/09544062241271676
2024-08-19
Proceedings of the Institution of Mechanical Engineers Part C Journal of Mechanical Engineering Science
Abstract:Proceedings of the Institution of Mechanical Engineers, Part C: Journal of Mechanical Engineering Science, Ahead of Print. Parameter uncertainties in the electromechanical actuator system and the obvious friction nonlinearity in the low speed stage will greatly deteriorate the control performance and even lead to system instability. In this paper, an adaptive super-twisting sliding mode controller with neural network (ASTSMNNC) is proposed for the electromechanical actuator system. The LuGre model is used to describe the nonlinear friction, a nonlinear dual-observer is designed to observe the LuGre model internal friction state, a parameter adaptive law is designed to estimate the unknown parameters existing in the system, the time-varying disturbance in the system is estimated by using the universal approximation property of neural network. The feedforward compensation technology is used to compensate the estimated errors of parameters and the observed error of disturbance, the second-order nonlinear sliding mode is designed to compensate the residual estimated errors of parameters and neural network, and the chattering phenomenon caused by the sliding mode control can be reduced at the same time. What's more, the controller theoretically guarantees a prescribed tracking performance in the presence of various uncertainties, which is very important for high-accuracy control of motion systems. Lyapunov stability theorem is used to prove that the proposed controller can achieve the bounded stability of the system. Extensive comparative experimental results are obtained to verify the high-performance nature of the proposed control strategy.
engineering, mechanical
What problem does this paper attempt to address?