Modeling Approach for Transit Rolling-Stock Deterioration Prediction

Abstract

An effective public transportation management system (PTMS) requires accurate and efficient models for the prediction of rolling-stock conditions. If the state of any given rolling-stock unit is known, its future condition can be predicted from the corresponding deterioration curves. The purpose of this study was twofold: first, to evaluate and model the relative importance of factors causing deterioration of rolling stock and, second, to provide projections of future condition to be used in transit capital programming. A methodology was developed for the estimation of rolling-stock deterioration models from condition rating data. Using a rolling-stock inspection data set from Indiana, the capabilities of the proposed methodology are empirically demonstrated. This ordered probit-based methodology provides models that are intuitively appealing, fundamentally sound, and a useful and easy-to-use tool in projecting future rolling-stock condition. The models presented in this paper are a part of the public transit management system being developed in Indiana for determining optimal rolling-stock maintenance, repair, and replacement strategies.