NEURAL NETWORK PROCESSING OF ELECTROMAGNETIC ACOUSTIC SIGNALS TO IDENTIFY THE STRESS-STRAIN STATE AND DAMAGE OF POWER EQUIPMENT

Abstract

The purpose of the study is to develop and train an artificial neural network to identify the stress-strain state and damage to the metal of power equipment based on the values of the parameters of the harmonic components of the electromagnetic-acoustic transducer signal.
 
 Materials and methods. Experimental study of the relationship between the parameters of the harmonic components of the signal of an electromagnetic-acoustic transducer with the stress-strain state and damage to the structure of standard metal samples, development of an artificial neural network and methods for its training to identify the stress-strain state and damage to the structure of the metal according to the loading diagram.
 
 Results. Analysis of changes in the microstructure and frequency models of standard steel samples used in power engineering confirmed the possibility of identifying the stress-strain state and damage to the structure of metals based on the values of the parameters of the harmonic components of the electromagnetic-acoustic transducer signal. To solve this problem, an artificial neural network has been developed and trained. After training, the effectiveness of the network in identifying the stress-strain state and damage to the structure of metals reached 92.16%, which is acceptable for the tasks of recognizing the technical condition of metal structural elements of electrical installation equipment.
 
 Conclusions. The use of an artificial neural network to identify the stress-strain state and damage to metal structures based on the harmonic parameters of the electromagnetic-acoustic transducer signal enables to identify areas of concentration of mechanical stress and damage to the metal structure at the early stage of development, thereby increasing reliability and safety operation of electrical equipment.