Fault mode detection of a hybrid electric vehicle by using support vector machine

Abstract

Hybrid electric vehicle (HEV) is one of the ideal transportation tools to face the challenge of ‘Carbon peak carbon neutralization’. The high complexity of the latest HEVs result to the difficulties in vehicle fault diagnostics, which is considered to be the main factors of vehicle durability. In this study, a multi-fault mode detection method of a P2 diesel HEV was investigated by using support vector machine(SVM). The HEV physical model was built and validated by using the experimental data under China typical urban driving cycle (CTUDC). The SVM algorithm was coded in Matlab/Simulink environment. The training data vectors for normal mode and fault mode were acquired from the HEV model. Through the analysis of failure mode, the OVO-SVM method was proved as the best accuracy, the success rate of diagnosis reaches 98%. Case studies of single fault mode and multi-fault mode fault detection were conducted. Offline and online level test were performed to show the effectiveness of the detection algorithm. The study may support the development of intelligent diagnostic methods for the different types of HEVs.