Control system design of mechanical systems with time varying mechanical and control parameters

Abstract

A new control framework of motion and/or vibration control system design for mechanical systems with a time varying mechanical and control parameters is studied. As examples of such time varying mechanical and control parameters, we can assume a variable damping or stiffness parameter of semi-active control devices and a time varying weighting function in a generalized plant respectively. The mechanical system in the present study is assumed to have also an actuator for active motion and/or vibration control. The active control law to drive the actuator is obtained by a gain-scheduling controller based on linear matrix inequalities (LMIs) so that the closed-loop system is stable for all assumed values of the time varying mechanical and control parameters in the generalized plant. We use the adjustability of the time varying mechanical and control parameters in the closed-loop system to realize given control specifications. As the control law of the time varying parameters in the closed-loop system, a multi-layered feedforward artificial neural network (ANN) is designed as a dynamic map from available signals in the control system to time varying mechanical and control parameters. Design parameters of the ANN are optimized with a genetic algorithm (GA). With a design example of an active positioning and vibration control of a mechanical system with variable damping coefficient, the proposed design approach is shown to be capable of achieving highly sophisticated control specifications that are hard to be satisfied by conventional control methods.