| Summary: | [Objective] In contrast to the planned and faulty maintenance modes commonly used in modern metro vehicles, condition-based repair can capture the health status and operation condition of the vehicle system while optimizing maintenance strategies and costs. To realize condition-based repair, the health status of vehicle system needs to be accurately evaluated. [Method] Focusing on metro vehicle bogie system as object, the evaluation method of its health status is studied in-depth. Based on the hierarchical characteristics of bogie system composition, with reference to the degree of deterioration, combining the analytical hierarchy process method and variable weight theory are combined, to establish the variable weight fuzzy comprehensive evaluation model of metro vehicle bogie based on deterioration degree according to using the idea of fuzzy comprehensive judgment, and the modelling process is proposed. By evaluating the health status of the selected metro vehicle bogie system with the health status evaluation indexes, sample data is obtained and used to train three different artificial neural networks, namely, BP (back propagation) neural network, support vector machine and random forest. Then, the evaluation performance of the three different types of artificial neural networks is judged with actual test data. [Result & Conclusion] The random forest model has the strongest ability to identify the health status of the metro vehicle bogie system and realizes the intelligent health assessment of the metro bogie system.
|
|---|