PROBLEMS AND PROSPECTS OF THE USE OF NEURAL NETWORKS IN MONITORING TASKS OF HELICOPTERS AIRCRAFT ENGINES TECHNICAL STATE IN FLIGHT MODES

Abstract

To improve the quality indicators of automatic control systems, it is urgent to develop new algorithms for the identification and diagnostics of technical objects. One of the ways to solve the problem is the use of artificial neural networks. In order to study the problems of using artificial neural networks for the identification and diagnostics of technical objects, in particular, helicopters aircraft engines, an analysis of scientific works on this topic in recent years has been carried out. The existing approaches to the construction of fault diagnostics systems, as well as automatic control systems based on artificial neural networks, are considered. The results of the above analysis can be used in the development of new methods and algorithms for identification and diagnostics of technical objects based on neural network analyzers. In this paper, as an example, a generalized neural network model of helicopter aircraft engines is presented, which is used to monitoring their technical state in the helicopter flight mode. Based on this model, the paper formulates the problem of identifying the technical condition of helicopter aircraft engines. This paper provides an example of the use of neural network technologies in the control problems of helicopter aircraft engines in flight modes. The structure of the model for controlling helicopters aircraft engines in flight modes using a neural network of the perceptron architecture has been developed. The results obtained indicate the advantages of using neural networks in the problems of controlling helicopters aircraft engines in flight modes over other methods, for example, a control method using PID controllers. On the part of modern software and hardware, there are no restrictions on the complexity of the algorithms used, however, to realize the significant potential capabilities that control systems based on artificial neural networks have, it is necessary to develop conceptually new approaches to the construction of such systems.