Development of a neural network approach for risk management in helicopter technical condition diagnostics

The object of the study is the quality of helicopter maintenance based on digital diagnostic tools. To ensure the required quality, quantitative risk assessment models for the in-depth and express diagnostics system of helicopter gas turbine engines in a neural network environment are proposed. The assessment of diagnostic efficiency is based on the analysis of probable control risks by standard deviations, which distinguishes the proposed approach from the traditional one. Two diagnostic modes are considered: rapid diagnostics exemplified by vibration diagnostics, and in-depth diagnostics, including both vibration diagnostics and pyrometric control. These diagnostic methods make it possible to implement a remote monitoring system at aircraft repair facilities, which significantly reduces maintenance labor intensity. As a result, it was found that control risks depend not only on the metrological level of measuring instruments but also on a combination of the statistical nature of control agents in their system composition according to the following characteristics: statistical parameters of the controlled indicator, distribution laws, and values of uncertainty of measuring instruments, as well as the uncertainty of control standards (tolerances). In the modeling process, risks were assessed as a function of the ratio of uncertainties of measuring instruments to the uncertainty of the controlled parameter, with varying values of the standard (tolerance). This approach will allow, in practice, the creation of a more effective system for monitoring and collecting statistical information on the operational reliability of the Mi-8 helicopter engine, where the quality of control is predicted to a greater extent based on the metrological indicators of the measuring instruments and methods