Neural Network Predictive Control for Piezoelectric Smart Structures

Abstract

The piezoelectric smart structure is a force-electric coupling structure, and piezoelectric patches can not be patched ideally, so it is difficult to build the accurate mathematical model of piezoelectric smart structure. The traditional vibration control methods depend on the structural mathematical model, and the control result is unsatisfactory. Considering this problem, this paper introduces the nonlinear generalized predictive control algorithm based on neural network predictive model into piezoelectric smart structure. Because of the difficulties of building the mathematical model and extracting dynamic data from experiment, the finite element software (ANSYS) is employed to analyze and obtain the dynamic response data of piezoelectric smart structure through modal analysis and transient analysis. Neural network predictive model of structure is built through off-line training on the basis of the data. The nonlinear generalized predictive control based on neural network has a better ability to solve complex nonlinear problem. Then the author introduces the Neural Network Based System Identification Toolbox (NNSYSID) and Neural Network Based Control System Design Toolkit (NNCTRL), which are two special toolboxes for designing neural network control system and can save lots of time for designers who can commit themselves to sixty-four-dollar question. At last, the author shows the method through a case. A cantilever beam which surface is boned piezoelectric patches used for sensor and actuator respectively is analyzed by ANSYS and controled by the neural network predictive control algorithm on the platform of NNSYSID and NNCTRL. This is a simple and effective method for designers to solve the vibration control problem of piezoelectric smart structure.