Uncertainty Propagation Assessment in Railway-Track Degradation Model Using Bayes Linear Theory

Abstract

Periodic inspection could be regarded as a form of insurance to check for variation in the performance of the empirical-based track degradation model. Excessive track inspection, however, hurts economic maintenance and probably occurs because of inadequate knowledge to manage uncertainty in the track degradation model. Through regularly assessing uncertainty propagation in the model parameters, the relative importance of inspection decisions can be evaluated, and this may remove the need for excessive inspection(s). Because the evaluation process has a strict time limit, (i.e., before the next inspection arrives), a simple assessment method is required. Thus, this paper introduces a semiprobabilistic method driven by the Bayes linear theory to assess uncertainty propagation in the parameters of the linear model of track-geometry degradation. Through the assessment, two quality measures, namely, partial size and bearing adjustment of expectation of prior belief, iteratively display how parametric uncertainty propagates from one inspection to the next. As a result, in general, the proposed method shows the existence of a transition point in the inspection plan, i.e., a splitting of inspection decisions into two categories: exciting, and as expected. Although track inspection is recursive, a revision is suggested for inspection decisions under the latter category. A practical use of the proposed method is presented using real data and discussed in this paper.