Early stage data-based probabilistic wear life prediction and maintenance interval optimization of driving wheels

This study presents an early stage data-based maintenance strategy of driving wheels that have different life distributions depending upon their location. Wear was predicted under the condition that the shape of the contact surface changes over time by an original method of back calculating degradation over time through the establishment of a basic wear model and a recursive function for wear progression. An accurate wear model was established and verified by an experiment. The variation in the profile of a wear-induced wheel was applied to the wear model. Furthermore, the model was combined with a recursive function and used to obtain the time-series degradation data. Subsequently, the factors which have a major influence on wheel production were analyzed, and a meta-model was configured using the response surface method. The degradation function and parameter distribution were estimated using uncertainty propagation, and the wear life distribution was derived using Bayesian inference and Markov chain Monte Carlo method. The reliability of driving wheels was obtained, and the maintenance interval was optimized under each maintenance conditions. Based on this novel method, the early stage data-based maintenance strategy was achieved, and the result of the wear life prediction was validated using the probability distribution analysis.