Preventing Internal Combustion Engine Failures by Including a Deep Learning Digital Analytical Module

Abstract

It has been established that monitoring and diagnostics of the technical condition and efficiency of an internal combustion engine (ICE) in real time are quite expensive not so much in data collection, but in the adequacy of the data processing model and their interpretation. An algorithm for creating the library of a deep learning program based on the existing database of ICE operation under various load conditions has been developed. As a result of ICE monitoring, a video endoscopy of the cylinder-piston group elements was performed to establish deviations from the standard state, and the data was processed using ELM327 and the Forscan program. The suggested data processing is carried out at the first stage using a combined method; identification of deviations is carried out on the basis of expert analysis, comparing them with the result of the decision-making by the digital module, which will allow an assessment of the validity of the decision-making by the artificial intelligence software module based on deep learning and will eliminate the occurrence of an erroneous decision.