Aplikace neuronových sítí pro řízení laboratorních modelů

Abstract

In the first part of thesis deals with theory about neural networks. This is primarily a choice of architecture neural network, learning algorithms, initialization weight parameters and identification using neural networks. Another section is devoted to the application of neural networks for control a laboratory model of DC motor. Carried out was identification of DC motor using neural networks and subsequent application of control algorithms using neural networks to the real model. For this purpose, were developed simulation models for the different methods in Matlab/Simulink. Primarily used method of Inverse Control Method, Internal Model Control and Adaptive Control using Orthogonal Network. Each method is evaluated in terms of quality control criterion IAE.