NEURAL NETWORK APPROACH TO CLUSTERIZATION OF CONTROLLED PARAMETERS AS ONE OF THE STAGES OF AUTOMATION OF THE IDENTIFICATION OF DIFFICULT TECHNICAL OBJECTS 

Andrey I. Loskutov, Doctor of technical sciences, professor, head of sub-department of telemetry systems, complex processing and protection of information, Мilitary Space Academy named after A. F. Mozhaisky (13 Zhdanovskaya street, St. Petersburg, Russia), E-mail: vka@mil.ru
Vladimir A. Klykov, Candidate of technical sciences, teacher, sub-department of telemetry systems, complex processing and protection of information, Мilitary Space Academy named after A. F. Mozhaisky (13 Zhdanovskaya street, St. Petersburg, Russia), E-mail: vka@mil.ru
Ekaterina A. Ryakhova, Adjunct, sub-department of telemetry systems, complex processing and protection of information, Мilitary Space Academy named after A. F. Mozhaisky (13 Zhdanovskaya street, St. Petersburg, Russia), E-mail: vka@mil.ru
Andrey V. Stolyarov, Adjunct, sub-department of telemetry systems, complex processing and protection of information, Мilitary Space Academy named after A. F. Mozhaisky (13 Zhdanovskaya street, St. Petersburg, Russia), E-mail: vka@mil.ru
Olga L. Shestopalova, Candidate of technical sciences, associate professor, dean of faculty "Test of aircraft", Voskhod branch Moscow Aviation Institute (5 Gagarina street, Baikonur, Kazakhstan), E-mail: vka@mil.ru 

629.7.017 

10.21685/2307-5538-2021-1-1 

Background. The article discusses the issues of automation of the process of identification of complex technical objects (STO). The problem of automating the process of identification of complex technical objects is formulated on the basis of solving the problem of clustering of controlled parameters. The aim of the work is to use the neural network clustering method in automating the stage of the process of building a model of the SТО functioning.
Materials and methods. An approach to the clustering of controlled parameters using a self-organizing neural network (NN) of Kohonen is presented. The novelty of the approach lies in the application of the well-known neural network method of data clustering in a new area, namely, the automation of the process of one of the stages of mathematical modeling of the STO functioning.
Results. An algorithm for clustering the controlled parameters according to the states of systems has been developed. The generalized block diagram of the automatic identification system using the NN is presented. The results of modeling were evaluated based on the comparison of the obtained clusters of controlled parameters with the data of the verbal model of STO.
Conclusion. The analysis of the results obtained in the course of the study showed that the use of a neural network in the clustering of controlled parameters is possible as an implementation of one of the stages of automation of the STO identification process. 

complex technical object, neural network, identification, clustering