An Optimal DAG-LSTSVM Algorithm Based on Imbalance Factor for Vehicle Fault Diagnosis

In recent years, health monitoring for vehicles has become increasingly important. The Internet of Vehicles (IoV) is inevitably susceptible to various faults due to its dynamic structure and variable environment, which has a serious impact on system performance and can even lead to traffic accidents. Therefore, it is critical to investigate the health management and fault diagnosis for vehicles. Conventional diagnosis schemes, such as signal-based methods, suffer from significant drawbacks because of poor coupling detection performances. Hence, data-driven methods have been proposed and widely applied. Among these, Support Vector Machine (SVM) and its variants, such as Directed Acyclic Graph (DAG) and Least Squares Twin (LST) SVM, show excellent performance in vehicle fault diagnosis. Nevertheless, those algorithms have weaknesses of low diagnostic efficiency and are incapable of dealing with the imbalanced classification. To address these shortcomings, this paper proposes a novel multi-classification algorithm for vehicle health management, called Imbalance Factor (IF) based LSTSVM. Firstly, to address the problem of multi-classification, the discriminable metric (DM) between different classes is proposed and utilized as the measurement to construct the optimal DAG structure. Further, considering the effect of imbalance classification, IF is employed in SVM classifiers to alleviate this issue, which leads to a higher accuracy rate. Moreover, taking advantage of LSTSVM, the proposed algorithm achieves high efficiency. Experiments on the typical datasets reveal its superiority in vehicle fault diagnosis. Experimental results show that compared with the existing methods, the proposed algorithm has obvious advantages in terms of classification accuracy and efficiency.