Research On Genetic Optimization_Based Neural Network Control Strategy

With the development of the automatic science, and increasing demands on the control system's performance, the traditional control strategies reveal much more limitations to the complex systems which are non-linear, uncertainty, or hardly in mathematical modeling, meanwhile the intelligent control theory's progress may provide some new research ideas for solving these problem. This paper studies the genetic optimization_based neural network compound inverse control strategy, using the fusion of the genetic algorithm and the artificial neural network to solve the current difficult problem in control field with. Owing to its well characters in nonlinear approximation, parallel processing, self-learning and strong robustness, etc., the artificial neural network has been widely used in the fields of the complex non-linear system's modeling, parameter optimization, control and so on.This paper launches research about the learning issue of the BP neural network. For the BP algorithm ranged to the gradient descent has the problem of local convergence, the genetic algorithm is introduced to improve learning method of the BP neural network's weight and threshold values, and the genetic optimization learning algorithm is proposed to enhance the precision of BP neural network's learning. And guided by the inverse system control theory, proposes the genetic optimization_based neural network compound inverse control strategy.In this paper, a material testing machine which is a typical electro-hydraulic servo system is treated as the controlled object, applying the genetic optimization_based neural network compound inverse control strategy to the material testing machine's position servo system, wherein the neural network is used to identifying the system's dynamic inverse model, through simulation in MATLAB environment and experiment with the computer control system developed by Labview, and compared with the traditional PID control strategy, the result indicates that the genetic optimization_based neural network compound inverse control strategy satisfies the request the material testing machine's dynamic characteristics well.