Modeling Rail Fatigue Behavior with Multiple Hazards

Abstract

Rail fatigue failures are of great concern to railroad companies because they may cause derailments. Therefore, the railroad companies' objective is to predict the locations and times of potential failures and perform preventive maintenance. To predict when and where rail fatigue failures will occur, it is necessary to establish the relationship between fatigue failures and factors affecting fatigue. This paper specifies a statistical rail fatigue model in four stages: first, the model describes the relationship between probability of failure and usage; second, the model is extended to cases of multiple types of failures; third, the model is derived for discrete usage categories or time periods with a continuous spatial representation of rail units; and finally, explanatory variables capturing geometry, maintenance, and material properties are included to explain differences in fatigue behavior among heterogeneous rail units. The model is estimated using available rail fatigue data. The modeling framework can be applied to other multitype failure behavior and to reliability problems such as vehicle breakdown.