A comparative study of neural and conventional adaptive predictive controllers for vibration suppression

摘要

Two adaptive predictive control systems are developed for vibration suppression of nonlinear and time varying smart structures. The neural adaptive predictive controller employs a nonlinear neural network autoregressive external input plant model. The adaptive generalized predictive controller is based on a linear autoregressive external input plant model. Efficient algorithms are used for online plant identification and performance index minimization to achieve real time control of plants with relatively fast response time. Controller performances are validated using an experimental setup comprising a cantilevered plate with surface bonded piezoelectric actuators. The plant response is nonlinear due to the relatively large magnitude of the input sine wave disturbance (combining the first and second natural frequencies). Adaptiveness of the controllers is tested through changes in the plant dynamics and external disturbance during experiments. Experimental results demonstrate very good performance and robustness of the developed controllers. The neural adaptive predictive controller shows superior performance over the adaptive generalized predictive controller.