An intelligent sliding mode controller for vibration suppression in flexible structures

A novel sliding mode (SM) control system with an embedded neuro-fuzzy approximator is developed in this paper to provide more effective vibration suppression, especially in flexible structures. It aims to force system state to move to, and maintain on, the defined sliding surface without chattering. A new hybrid training technique based on an extended gradient method is proposed to optimize the neuro-fuzzy system to approximate unknown nonlinear functions and to enhance control performance. When the principle of the terminal attractor is incorporated into the classical gradient method and/or SM control systems, some implementation problems arise especially when the error is close to its origin. The proposed extended gradient method can enhance the SM control to not only speed up convergence but also overcome the existing implementation problems of the terminal attractor. The Lyapunov stability analysis demonstrates that the approximation with the proposed hybrid training technique is stable and can converge to the optimal approximation. The effectiveness of the developed control system and the hybrid training technique is verified experimentally corresponding to nonlinear and time-varying system control.