Research On Control System With Neural Networks And RTX-Based Software Design For Three-axis Table

Three-axis table is one of the key measuring equipment of the inertial component. With the development of aerial industry, the requirement of performance of the three-axis table is higher and higher. This thesis is written on the background of the research project of three-axis table. To make use of analysis and design on the theory of neural network, it mainly focuses on the analysis of frictional disturbance and designing neural-PID controller and the real time system software on the base of Windows 2000-RTX operation system.Firstly, the mathematical model of three-axis table is presented by the theory analysis, and is simplified according to practical parameters. The state equation of three-axis table is established. According to the adverse system theory the uncoupled ability of the network is proved and linear uncoupling network is also obtained. The model is transformed into three single input and single output linear systems, and then it can be analyzed and designed by the linear system theory.Secondly, based on affection of the machine and molecular the friction which is the mainly disturbance in the control accuracy is introduced. The useful stribeck friction model is given.After discussing the characteristic and algorithm of BP neural network, the applications of neural network, using in three-axis table system, are discussed. The method of compensation is discussed, and then made the theoretic analysis and simulation. The PID controller based on BP neural networks is designed to realize control parameter self-learning and self-adjusting. Both convergence and stability of this system are analyzed and simulation results are presented in the article.Finally, for the three-axis table real time control requirement the control software was specifically designed and accomplished on the base of Windows2000 plus RTX real-time operating system.