Inductive learning method for control of intelligent structures

摘要

Abstract: An inductive learning method is used for online control of a structure. The controller has the benefit of being designed without the need for a system model and is able to adapt to varying system parameters. The learning method is empirical in nature where the trials and errors of the controller generate a stimulus-response function which is used to improve the performance of the system. Numerical experiments were performed with the quantized inductive learning (QIL) algorithm on simple linear systems and a simulation of a simply supported aluminum beam. In both cases, the algorithm controlled the dynamic response of the system from an arbitrary initial condition. The QIL algorithm learned the control function without access to the computer model. Other issues associated with the development of this algorithm were examined concurrently. The effects of various performance indices, varying the sampling periods, and changing the levels of quantizations were determined and evaluated. In addition, QIL was used to reject sinusoidal disturbances on these systems. Finally a comparison of the QIL algorithm with state feedback was made to compare the effectiveness of this method with a standard model-based approach. !8