ЗАСТОСУВАННЯ НЕЙРОННИХ МЕРЕЖ У ЗАДАЧІ ПРОГНОЗУВАННЯ ТЕХНІЧНОГО СТАНУ АВІАЦІЙНОГО ДВИГУНА ТВ3-117 У ПОЛЬОТНИХ РЕЖИМАХ

Abstract

The subject matter of the article are the methods and models for the identification of the technical state of the aircraft engine TV3-117. The goal is to develop an on-board system for identification of the technical state of the aircraft engine TV3-117, one of the solved tasks is the prediction of its technical status in real time. The tasks to be solved are: to development of methods and algorithms for forecasting the technical state of the aircraft engine TV3-117 in flight modes based on neural network technology. The methods used are: methods of probability theory and mathematical statistics, methods of neuroinformatics, methods of information systems theory and data processing. The following results were obtained: the application of the proposed neural network prediction method based on the approximation and extrapolation of the processes of changing the gas dynamic parameters of the aircraft engine TV3-117 on fixed segments of the time window (within the «sliding time window») allows effectively solving the problems of forecasting its technical state. The analysis of the effectiveness of the application of the neural network method for forecasting the technical state of the aircraft engine TV3-117 under the conditions of random interference has shown its advantages in comparison with the classical prediction methods, which consist in providing higher prediction accuracy for different forecasting intervals (short-, medium-, long-term forecasting). Application of the developed neural network method makes it possible to detect the moments of the time series disorder, that is, the appearance of the trend of the parameters of the aircraft engine TV3-117, which is a consequence of the qualitative change in the characteristics of the engine, which allows timely making operative decisions on changing its operation mode. Conclusions. The scientific novelty of the results obtained is as follows: the method of solving the problem of forecasting the technical state of the aircraft engine TV3-117 with the help of neural network technologies has been further developed, the accuracy of which in the short-term medium and long-term forecast is significantly higher compared with the use of polynomial regression models, the method of exponential smoothing, moving average, which indicates that the use of neural network technologies makes it possible to detect the appearance of the trend of the parameters of the aircraft engine TV3-117, which allows is to make timely operational decisions to change its mode of operation