Reducing Air Pressure System Repair Costs in Scania Trucks through Deep Learning

Air pressure systems play a fundamental role in Scania trucks because the proper functioning of the brake and gear shifting systems of these vehicles relies on the health of the air pressure system. The presence of sensors in the air pressure system gathers various information about its status, which can be stored and analyzed in the form of datasets. Using machine learning algorithms to detect faults in the air pressure system prevents manual inspection at different time intervals, thus preventing material and time costs. Many efforts have been made to detect faults in the air pressure system through the collected datasets from sensors of the various components of Scania trucks using traditional machine learning algorithms such as decision trees, KNN, Random Forest and SVM, but they still lack sufficient accuracy and speed in data processing. However, since the number of records collected at different intervals by the Electronic Control Unit (ECU) is very high, novel machine learning algorithms like deep learning can be used to increase accuracy and speed in detecting faults in Scania truck air pressure systems. In this article, feature selection and deep learning algorithms have been used to detect and predict faults in the air pressure system of heavy trucks. The results and observations showed that the output of the evaluation parameters of the deep learning algorithm has an accuracy rate of ۹۸.۶۶%, a recall rate of ۶۳.۴۷%, and a f-measure rate of ۶۸.۹۹%.

Keywords:

Deep Learning , Air Pressure System , preprocessing , Feature selection , Dataset