EXPERIMENTAL STUDY OF CONTROL METHODS VIBRATION PARAMETERS OF CONTROL OBJECTS USING NEURAL NETWORKS

Abstract

It is shown that when controlling vibration parameters of control objects, neural network methods allow an empirical approach to the classification problem. For learning neural networks, fewer implementations are required than for statistical analysis of input signals. In the general case, a back-propagation network allows for a smaller number of classification errors than any of the ART options. When using a representative control sample, the classification using tunable spectral operators is comparable in efficiency to the classification using a two-layer neural network, while at the same time providing significantly less training time.