Дистанционный контроль частоты вращения коленчатого вала двигателя трактора с использованием алгоритма машинного обучения

Abstract

Diagnostics of the technical condition of agricultural machinery based on machine learning with artificial intelligence employs the accumulated experience to localize the malfunction and makes it possible to evaluate its technical condition in the shortest possible time. New computing devices (built-in diagnostic tools) store and process large amounts of information and reduce the time needed to assess the technical condition of the equipment. To predict malfunctions, the authors analyzed the crankshaft speed of a tractor engine using the Random Forest machine learning algorithm. They developed a counter-indicator and software for the remote control of the engine crankshaft speed. The developed prototype counter-indicator software was tested on the D‑243 engine. As a result, remote diagnostics of agricultural machinery revealed the main causes of malfunctions affecting the engine crankshaft speed. The Random Forest algorithm made it possible to “predict” malfunctions with acceptable accuracy: it calculated all values correctly from a sample of 13 values and made 4 errors from a sample of 51 values. Diagnostics with the help of a machine learning algorithm made it possible to assess the technical condition of the equipment in real time without making fundamental changes to the design, and to give forecasts and suggestions for its maintenance and repair.