Life Prediction Method Based on LDA-ELM for Mechanical Components of Door Systems

Abstract

The life of door system is closely related to the capacity of rail vehicle safe operation and maintenance. Rolling pin is a built-in mechanical component of the rail vehicle door system. Its wear degree is difficult to measure so that its lifetime is hard to predict in real time. In order to predict the life of rolling pin online and provide decision support for active maintenance, this paper proposes a data-driven life prediction method based on Linear Discriminant Analysis (LDA) and Extreme Learning Machine (ELM). Firstly, features related to the wear state of the rolling pin are extracted from raw data collected from motor of the door. Then, with the LDA, the features are fused to filter out the redundant features. Finally, the ELM model for predicting the diameter of small end is built, and the life of rolling pin is calculated according to the relationship between the run times and the diameter of small end. The simulation results show that the method enables to accurately predict the life of the product, which has reliability and important engineering application value.