A segmental evaluation model for determining residual rail service life based on a discrete-state conditional probabilistic method

Abstract

Because steel rail is one of the most fundamental components of railway operations, the accurate estimation of residual rail service life is of great significance in ensuring the safe operation of railways. In addition, maintenance expenses must be minimized in a manner that allows limited railroad resources to be optimally allotted. In this study, the typical types of continuous rail segments on a rail line are classified into non-sharply curved rail segments and sharply curved rail segments. Using these classifications, a model for estimating the residual service lives of rail segments using a discrete-state conditional probability method is proposed based on an analysis of rail deterioration characteristics. The model considers several heterogeneous factors to determine their influence on the deterioration process and is shown to be capable of estimating the residual service lives of rail segments. Finally, the model is validated through a case study of the Beijing Metro, using inspection records of rail defects in conjunction with heterogeneous factor data to predict the service life of the rail, which is then compared with its actual service life. The model is found to show good agreement with the rail inspection and maintenance records of the Beijing Metro, indicating its appropriateness for use by railroad management in allocating future rail maintenance resources.

Improvement recommendations for railway infrastructure maintenance

“…Around the world, researchers are investigating the possibility of early detection of imperfections in "Infrastructure" subsystem and improving maintenance efficiency [12][13][14][15][16][17][18][19][20]. The authors of this paper are engaged in the development of methodology for nondestructive inspection of superstructure and substructure condition (Figure 5).…”

A rail service life prediction model based on BP neural network

“…Corotis 4 used Weibull probability distribution to model the risk rate function of rail fatigue defects and exponential probability distribution to model the failure time of rail fatigue defects, and determined the economic service life of the rail based on the goal of minimizing the total life cycle cost. Wen Bai et al 5 proposed a discrete-state conditional probabilistic model to evaluate the residual rail service life based on the classification of continuous rail segments.…”

Heterogeneity‐oriented ensemble learning for rail monitoring based on vehicle‐body vibration

“…To help understand the reality and seriousness of this issue, Figure 2 illustrates it with a brief example. Even for the same task, when specific line sections with influential factors (Bai et al., 2019) change, completion requires adaptation. In line section (b), defects are mainly generated on the outer rail where the curve is located, particularly for a small‐radius curve, which is partly identical to the defect formation hypothesis.…”