INFORMATION TECHNOLOGIES WITH THE USE OF NEURAL NETWORKS IN THE EXPERIMENTAL STUDY OF VIBRATION SIGNALS OF ROTARY OBJECTS

Abstract

The paper shows the relevance of the use of neural networks in the control of vibration parameters of rotary objects. In this case, neural network methods allow an empirical approach to the problem of classification. Neural networks require fewer implementations than statistical analysis of input signals. In the general case, the inverse error propagation network allows to achieve fewer classification errors than any of the ART variants. The aim of the work is to substantiate the choice of the method of formation of primary informative features of vibration signals of control objects with the use of tunable matrix spectral operators and the use of neural networks in order to control vibroparameters. The methodology for solving this problem is to use the methods of spectral analysis and neural network methods to solve the classification of vibration signals of rotary objects, which provides control of the parameters of objects during their operation. Scientific novelty. For the first time, the use of rearranged spectral operators is proposed to form a control sample that provides a classification that can be compared in efficiency with the classification using a two-layer neural network, while providing much less learning time. Conclusions. The use of information technology with the use of rebuilt spectral operators and neural networks allows you to effectively control the parameters of rotor objects to assess their functional state by the vibration signal, providing much less training time.