A Novel Lazy Learning-based Data Driven Fault Diagnosis Method for Public Transport Vehicles

In this work, a new data driven online local modeling method, which combines relevance vector machine and Lazy Learning (RVM-LL) is proposed for fault diagnosis of public transport vehicles. In proposed RVM-LL, LL builds multiple local models at each query point based on the data query, and determines the best data model in the current query through a selection mechanism. First, the method establishes the similarity function and selects the local nearest neighbor points. Secondly, a local model is built based on RVM algorithm. Third, the optimal local model is selected by leave-one-out cross validation method. The advantages of this method are: LL does not need to establish an accurate bus fault diagnosis model, and it has the characteristics of small sample local online modeling, which does not need to spend a lot of time on training, saves time and is adaptive. In addition, the background of this paper is based on practical applications, and the data used are also from real public transport vehicles, and the accuracy of the method is verified by simulation experiments.