Analysis of the Possibility of Using Neural Networks to Monitor the Technical Efficiency of Diesel Engines During Operation

Abstract

The aim of the research was to analyse the possibility of using neural networks to determine the parameters of the chemical composition of exhaust gases as a function of engine performance parameters obtained from the on-board diagnostics system such as crankshaft speed and engine load index. The subject of the study was a Fiat Panda car equipped with a 1.3 Multijet diesel engine and powered by pure diesel. The tests used the MAHA MET 6.3 exhaust gas analyser and the on-board diagnostics system OBD II. The obtained values of NO x ,O 2 ,CO 2 and PM measured behind the DPF were analysed. For the purpose of building a neural network model, preliminary studies were carried out in non-urban traffic (high-speed route). On the basis of the data obtained, processes of learning neural network structures with approximate properties with backward propagation of errors were carried out. Subsequently, tests were performed on the operational parameters of the vehicle and the chemical composition of exhaust gases in urban traffic. Analysis of the obtained values of the average parameters obtained during the measurement and on the basis of the prepared neural models allows determining the relative differences at the level of not more than 10 percent.