НЕЙРОСЕТЕВАЯ СИСТЕМА ЛАЗЕРНОЙ ДИАГНОСТИКИ ЭЛЕМЕНТОВ ОСТЕКЛЕНИЯ КАБИН САМОЛЕТОВ

Abstract

Assessing the technical condition of glazing elements in the cockpits of operational-tactical aircraft still remains the most important task in ensuring flight safety. To increase the efficiency of operations for non-destructive testing of glazing elements using the speckle structure method of optical radiation, the authors propose to use neural network technologies to automatically identify controlled areas in the cockpit. Artificial intelligence technologies have been used to realise this task. They are based on algorithms of semantic segmentation, classification and detection of monitored areas according to the established markers on the cabin due to the application of convolutional neural network on YOLOv8. The application of machine vision technology have made it possible real-time measurement of the glazing exit from the termination when overpressure has been created inside the cabin. This reduces the time for technical condition assessment by at least 10 times. The use of machine vision technologies have made it possible to measure the value of the glazing outlet from the sealing in real time when creating excessive pressure inside the cabin and thereby reduce the time to assess the technical condition by at least 10 times. The authors have established the reason for the discrepancy between the results of using the speckle-structure method of optical radiation in determining the value of glazing yield from the termination and the "tape" method and developed recommendations to reduce measurement errors.