THE USE OF NEURAL NETWORKS FOR ISOLATION OF COMPONENTS IN REAL TIME 

Priymak Anton Aleksandrovich, Postgraduate student, Penza State University (40 Krasnaya street, Penza, Russia), 
chen-wei.yang@ltu.se

681.3.01

10.21685/2072-3059-2018-2-2

Background. The object of the study are fast-changing processes occurring in various systems of diagnosis and monitoring. The subject of the study is an algorithm based on neural networks designed to simplify the procedure of adaptation of extreme filtering to the real-time mode. The purpose of the work is to create and study an algorithm that implements extreme filtering using neural networks, and its adaptation to work in real time.
Materials and methods. The algorithm was studied in Matlab and Simulink. The simulation of the algorithm operation was carried out on the samples of the signal received in real time.
Results. The possibility of using neural networks for the implementation of the algorithm of extreme filtering is substantiated. In Matlab and Simulink environment the program implementing the algorithm of extreme filtering on neural networks is created.
The proposed algorithm was tested on samples of a real signal and the was proposed algorithm is compared with analogues. The method of adaptation of the proposed algorithm to the real-time mode using the frame by frame method is shown.
Conclusions. The use of the algorithm of decomposition built on neural networks, allows you to successfully allocate the components of the signal, for which parameters can be calculated, allowing to judge the changes taking place in the system.
Adaptation to the real - time mode of the algorithm of decomposition signal, built on neural networks, using the method of frame by frame, makes it possible to allocate all the components of the signal in real time. Thus, the algorithm of decomposition signal, built on neural networks, in combination with the method of frame by frame, can successfully solve the problem of diagnosis, monitoring of the state of systems operating in real time.

real-time, objects, neural networks, extreme filtering, frame-byframe processing method

1. Myasnikova N. V., Beresten' M. P., Tsypin B. V. Ekspress-analiz signalov v inzhenernykh zadachakh [Express analysis of signals in engineering tasks]. Moscow: Fizmatlit, 2016, 184 p.
2. Priymak A. A. Inzhenernyy vestnik Dona [Engineering bulletin of Don]. 2017, no. 2, pp. 2–10. Available at: ivdon.ru/ru/magazine/archive/n2y2017/4175.
3. Krichevskiy M. L. Metody issledovaniy v menedzhmente: ucheb. posobie [Research methods in management: teaching aids]. Moscow : KnoRus, 2016, 296 p.
4. Zenov A. Yu., Myasnikova N. V. Izvestiya vysshikh uchebnykh zavedeniy. Povolzhskiy region. Tekhnicheskie nauki [University proceedings. Volga region. Engineering sciences]. 2012, no. 3 (23), pp. 3–7.
5. Priymak, A. A. Sovremennye tekhnologii v zadachakh upravleniya avtomatiki i obrabotki informatsii: sb. tr. XXV Mezhdunar. nauch.-tekhnich. konf. [Modern technologies in the management of automation and information processing: proceedings of
XXV International scientific and engineering conference]. Moscow, 2016.