An Intelligent Vehicle Oriented EMC Fault Shooting Method Based on Semi-supervised Learning

Along with the development of intelligent vehicle, the complexity of electrical architecture results in an increase of fault shooting difficulty. As the theory of machine learning becomes well-rounded gradually, learning based fault shooting approach has been appeared. However, due to sample shortage, the available algorithm is limited, while fault shooting performance is barely satisfactory. Hence, in this paper, a prior-knowledge based method is proposed to realize sample data augmentation, while a semi-supervised leaning algorithm, which combine density clustering approach with TSVM - namely DB-TSVM, is proposed. Experiment results show that proposed method performs a higher accuracy rate of fault classification, and verify its effectiveness.