Open AccessProblems of using neural networks
Author(s) -
M. Yu. Mikheev,
Yu. S. Gusynina,
T. A. Shornikova
Publication year - 2020
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1661/1/012104
Subject(s) - artificial neural network , computer science , task (project management) , identification (biology) , artificial intelligence , machine learning , software , systems engineering , engineering , botany , biology , programming language
Currently, the urgent task is to develop new algorithms for identifying and diagnosing technical objects. Often artificial neural networks are used in solving such problems. In these tasks, most often, there is no setting the task in an explicit formal form to describe objects and phenomena, and neural networks do well in this case to solve them. The article considers various existing approaches of identification of objects based on neural networks, as well as diagnostics of technical objects based on artificial neural networks. A vision of the prospects for further development of such tasks is offered. Various methods for training the network in the case of problems of identification and diagnosis of faults are described. The task of identifying technical objects is reduced to the task of approximating data, and the task of diagnosing the state of the system to the task of recognizing images. The algorithms for selecting the optimal network structure are investigated. The advantage of using software and hardware technologies in the construction of neural network algorithms has been evaluated. The results of the article can be used in constructing a scheme of actions of identification and diagnostics of technical objects using artificial neural networks.