СЕЛЕКЦІЯ ВХІДНИХ ПАРАМЕТРІВ НЕЙРОННОЇ МЕРЕЖІ ПРОГНОЗУВАННЯ ВІДМОВ БОРТОВОГО ОБЛАДНАННЯ

Abstract

The article is devoted to the study of the urgent issue of forecasting onboard equipment failures. The effectiveness of combat missions by aircraft crews of the Armed Forces of Ukraine depends on the operational and high-quality preparation of on-board equipment for flights. The long life of the existing fleet of aircraft requires constant analysis of the technical condition of the equipment to prevent failures. Existing forecasting methods are associated with the analysis of statistical data and probability metrics. A possible alternative way to monitor the health of equipment is to use neural networks. The problems of constructing the topology of the failure forecasting neural network is associated with the need to select significant factors affecting the equipment during operation. The article discusses the most important factors of influence in terms of application conditions, quality and operating conditions. Dependencies between the parameters for one object of study and the parameters of objects of the same type under various conditions of application are revealed. The following equipment parameters are selected for use in the construction of a neural network: total equipment operating time, equipment operating time after the last restoration, total number of on-off cycles and a failure cycle coefficient. For each of the parameters, expressions are given that are formulated taking into account the characteristics of the technical and flight operation of aircraft. So, the value of the operating time of the facility is calculated as the total time of the equipment in flight and on the ground when performing all types of work. The number of on-off cycles is proposed to be calculated based on the minimum amount of equipment used during operation. A failure cyclicality factor is also introduced, which is proposed to be calculated on the basis of a frequency analysis of the serial number of the failure and the operating time of the object to the specific failure.