Failure mode and effect analysis by exploiting text mining and multi-view group consensus for the defect detection of electric vehicles in social media data

Nowadays, the electric vehicle industry is developing rapidly. However, the frequent reports of electric vehicle accidents in the news also make consumers doubt its safety performance. Due to the complexity of the use environment, a large number of potential failure modes are difficult to detect before leaving the factory. There are real-time feedback from consumers on social media, which can help enterprises collect product defect information in time. Therefore, we propose an improved failure mode and effect analysis (FMEA) method that combines text mining and multi-view group consensus. On the one hand, it utilizes text mining and deep learning methods to build the product component dictionary and the failure word dictionary to classify defect-related texts and identify failure modes in social media data. On the other hand, considering the cardinal consensus and the classification consensus, the minimum adjustment consensus model is constructed in the proposed FMEA method to obtain the decision results of failure modes with the consistency of evaluation information and classification results. Finally, a case study of Tesla electric vehicles and comparison analyses are presented to show the effectiveness of the proposed FMEA method. With respect to general machine learning methods, the proposed FMEA method can find defect-related texts in social media data more accurately and comprehensively. Compared with the previous FMEA methods, the proposed FMEA method can not only identify failure modes in social media data, but also obtain consistent decision-making results with multi-view group consensus.