SELF-DIAGNOSIS OF THE COMPONENTS OF THE CONTROL UNIT OF A HEAVY-DUTY VEHICLE

Verkner A. S., Gorlova K. O., Akimov D. A., Guryanova E. O., Mayak A. A.

The paper considers the issues of automatic classification of vibration states of cars based on the use of convolutional neural network processing of vibration measurement data presented in spectral form and processed by a modified t-distribution method, and the knowledge of experts with experience in interpreting spectrograms characterizing the vibration states of composite aggregates of heavy-duty vehicles. The developed spectrogram analysis model allows monitoring the condition of heavy-duty vehicles of various models in automatic mode and timely notifying the driver of signs of pre-emergency situations, as well as the type of possible malfunctions. The data samples used in training the neural network classifier during experimental studies were formed on the basis of available archive files containing complete aperture data from car vibration sensors and information about the malfunctions detected in them.

Keywords:big data, vibration diagnostics, heavy-duty vehicles, transmission, internal combustion engine, intelligent system, software, deep machine learning, neural network, recurrent network, fault classification, t-distribution.