Adaptive Control of a Flexible Structure

Abstract

The design of a high-performance adaptive controller is discussed for an experimental lightly-damped flexible structure. The plant, the Stanford Four-Disk System, sustains large step changes in parameters as payload mass is added. New methods that make use of limited a priori knowledge of the plant are shown not only to greatly improve the quality of the identification in the presence of measurement noise and other disturbances, but greatly reduce the number of parameters that need be identified. The methods make use of known plant dynamics, frequency bands where unknown, modeled dynamics may exist and the natural damping of such dynamics. A method of dynamic pole-placement that ensures good step response, a high degree of active damping of modeled modes, modest amounts of control effort and low computational intensity is presented. A controller structure that ensures bumpless transfer between leapfrogging adaptive controllers and a standby robust controller is discussed. The performance of the adaptive non-colocated compensator, a robust colocated compensator and a robust non-colocated compensator are compared. The experimental adaptive controller dramatically outperforms the robust compensators within several seconds after a major change in plant parameters.