Diesel engines are crucially important in various fields, particularly in the automotive sector, as they ensure a reliable supply of mechanical energy. However, injection system failures, which are among the most recurrent failures, can lead to performance deterioration and increased pollutant emissions and maintenance costs. Therefore, adopting an effective maintenance strategy to analyze and predict such failures would significantly improve the efficiency of these engines. Based on collected data from engines by reliable sensors, the application of predictive maintenance coupled with a machine learning model allows effective prediction of failures for optimal appropriate maintenance. This study presents an approach to diagnosing the injection system of automotive diesel engines using a test bench based on data from temperature sensors installed on engine cylinders. These temperature data exhibit unusual variations in the event of an injection system failure. The Random Forest (RF) algorithm was employed to analyze these data and establish a clear relationship between cylinder temperatures and failure. The proposed model can detect failures associated with the injection system. Performance evaluation, particularly after parameter tuning, underscores the model's efficacy, achieving an accuracy exceeding 97%.
Read full abstract