BIC-Based Data-Driven Rail Track Deterioration Adaptive Piecewise Modeling Framework

• Main Authors: Yaqin Yang, Peng Xu, Guotao Yang, Long Chen, Junbo Li
• Format: Article
• Language: English
• Published: Frontiers Media S.A. 2021-02-01
• Series: Frontiers in Materials
• Subjects: maintenance activities identification; Bayesian information criterion; track irregularity; adaptive thresholding; dynamic programming
• Online Access: https://www.frontiersin.org/articles/10.3389/fmats.2021.620484/full

The records of maintenance activities are required for modeling the track irregularity deterioration process. However, it is hard to guarantee the completeness and accuracy of the maintenance records. To tackle this problem, an adaptive piecewise modeling framework for the rail track deterioration process driven by historical measurement data from the comprehensive inspection train (referred to as CIT) is proposed. The identification of when maintenance activities occurred is reformulated as a model selection optimization problem based on Bayesian Information Criterion. An efficient solution algorithm utilizing adaptive thresholding and dynamic programming is proposed for solving this optimization problem. This framework’s validity and practicability are illustrated by the measurement data from the CIT inspection of the mileage section of K21 + 184 to K220 + 308 on the Nanchang-Fuzhou railway track from 2014 to 2019. The results indicate that this framework can overcome the disturbance of contaminated measurement data and accurately estimate when maintenance activities were undertaken without any historical maintenance records. What is more, the adaptive piecewise fitting model provided by this framework can describe the irregular deterioration process of corresponding rail track sections.