Research On The Ball And Beam System Control Algorithm Based On Model Predictive Control

With the successful application in the industrial process control, Generalized Predictive Control (GPC) algorithm has attracted increasing attention in fields of control theory and engineering for its property of robustness and disturbace ability. However, practical processes usually are always complex with multiple inputs multiple outputs, strongly nonlinear, strongly coupling features. The GPC algorithm can hardly been applied directly. Over the past several years, neural networks have been widely applied to indentification and control of nonlinear system.The model predictive control which based on neural networks gradually become the important method to solve the complicated nonlinear control problems and has attracted more and more attation. And its good control performance has shown in the industrial process control.The ball and beam system is the research object in this thesis. The characteristics of the ball and beam system, composition and working principle are firstly introduced, and then the ball and beam system model is established, which lays a foundation for the design of the system controller. Both the generalized predictive control method and neural network model predictive control method are in-depth study.Secondly, the basic principles of model predictive control is expounded, on this basis, the description of the generalized predictive control algorithm and analysis are introduced in detail, and implicit generalized predictive control is analyzed. In order to validate the algorithm, the simulation results of the two control methods show that the result has a good effect.In order to make predictive control more widely used in nonlinear systems, the neural network is introduced into the predictive control. The basic principles of neural network and BP neural network are introduced in this thesis, and then the neural network prediction model is applied to generalized predictive control, because the system requires rapid response Therefore the LM algorithm is used to optimize neural network model. The description of the neural network predictive control algorithm and analysis are introduced in detail. And have a simulation for neural network model predictive control algorithm.The generalized predictive controller and implicit generalized predictive controller and neural network model predictive controller are designed in the experimental platform of the ball and beam system, each controller of the Matlab simulation results show that these types of Algorithm can get better control effect. And then have a debugging on the ball and beam system coming with PID controller and neural network model predict controller, the simulation results show that neural network model predictive control can effectively improve the control results and achieve the desired goal.Finally, the thesis summarizes the work has done, and put forward the prospect of further research directions and problems.