Adaptive neural-bias-sliding mode control of rugged electrohydraulic system motion by recurrent Hermite neural network

Abstract

This paper deals with design and implementation of a real-time position control scheme based on the synergistic combination of a recurrent neural network, an integral sliding mode controller and a bias controller for a rugged electrohydraulic actuation system. The controller design is based on recurrent Hermite neural network comprising a single hidden layer with orthonormal Hermite polynomial basis functions as activation functions for each hidden neuron and an integral sliding surface as the input. The bias controller is designed as a hyperbolic tangent of the error. Additionally, an adaptive scheme has been formulated based on the Lyapunov criterion and its convergence has been established. The performance of the proposed scheme has been evaluated on a laboratory scale single-rod electrohydraulic actuation system with a large dead band (∼10%) proportional valve in real-time. The experimental results suggest a significant improvement in the position tracking performance of the system for conventional tracking trajectories compared to other established methodologies.