SYNTHESIS METHOD OF REGULATORS USING NEURAL NETWORKS FOR NONLINEAR OBJECTS
The main options for using neural networks to solve the problems of synthesis of automatic control systems are: using them to adjust the values of controller coefficients; the use of a neural controller, in which its training is carried out in various ways, in particular, according to pre-calculated data (errors and known control) or by optimization in order to minimize the error of the output signal of the object. A method for synthesizing a neural controller for non-linear objects based on the structure of a previously calculated controller for a linearized model of an object is proposed. The main steps of the proposed method are: synthesis of a controller for a linearized model of an object; obtaining the values of the controller signals during modeling using the «white noise» signal; training a neural regulator; successively performing the optimization procedure and increasing the setpoint range to the required values. An example of applying the methodology on an object of a model of the angle of deviation θ of an object of an inverted pendulum is given.
Authors: А. А. Voevoda, D. O. Romannikov
Direction: Informatics, Computer Technologies And Control
Keywords: Neural networks, regulation, closed systems, nonlinear objects, optimization
View full article