Design Of Ball And Beam Control System Based On Neural Network

Ball and beam system is usually used to test the effect of control strategy, because it is a typical nonlinear and unstable system. In control area, there are many difficulties in controller design and nonlinear and unstable system modeling. How to make a combination of intelligent control to play their respective advantages is an important subject in science.This thesis establishes nonlinear model of the ball and beam system with the dynamics, analysises the qualitative of the system based on the linear theory, points out that the ball and beam facilities is unstable, the equilibrium point has a strong self-excited oscillation characteristics, and to make it as a study to design PID controller in Matlab (Simulink), the data acquisition cards and external systems connected to achieve a computer-controlled digital control platform.Then the thesis introduces the basic principles of neural network and adjusts PID parameters on line by adjusting the neural network weights and threshold values, designs the BP neural network PID controller by using Matlab M-function, simulates the ball and beam system mathematical model, achieves better effect than PID controller’s.The application for the BP algorithm is easily trapped in local minimum, and sensitive to the initial values as limitations, the thesis presents genetic algorithm neural network PID controller using the genetic algorithm improves BP network genetic, and simulates the ball and beam system model to compare and study, the result shows that after improved by the conventional genetic neural network PID controller, the system is improved its the robustness and dynamic performance, the control of the results improved effectively, the system is achieving the expected purpose.Finally, the ball and beam system physical model is simulated by using Matlab under the different controllers, the system’s static and dynamic characteristic are compared and analyzed, the simulation result shows that ball and beam syetem under the control of the genetic neural network PID controller effectively enhances its the adaptability of system and improves the system’s dynamic and static quality.