Developing Finite Element Models to Examine Rail Defects Under Combined Loading

Abstract

One of the Federal Railroad Administration’s (FRA’s) current areas of research within its rail integrity research program includes investigating the defect growth behavior of modern rail steels. The modern rail steel research is a collaboration among several organizations: Thornton-Tomasetti, Arcelor-Mittal, Lehigh University, Harvard University, and the Volpe National Transportation Systems Center (Volpe). A companion paper to this one will describe the results of recently-completed mechanical testing, fracture toughness testing, fatigue crack growth rate calculations, and residual stress field characterizations performed in Phase I of this research. The behaviors measured in Phase I were examined under laboratory conditions. The effects of the service load environment, including thermal loads, track support conditions, wheel loading, internal defect position and geometry will also need to be investigated for their effects on defect growth. A candidate approach that can be used to investigate these effects is to employ the finite element (FE) method to simulate a variety of conditions. Several of the types of measurements made in Phase I, such as residual stress distribution, serve as inputs to an FE model. Additional inputs, such as the wheel load and support conditions on the rail would be defined based on typical values encountered in the railroad environment. Stress intensity factors can be calculated around each simulated crack front for a given combination of material inputs, load conditions, and defect geometry. These stress intensity factors can then be used to estimate the fatigue crack growth rate under the given conditions. The modeling approach described above can result in a model that contains several complicated behaviors, including wheel-rail contact, discrete rail supports, and modeling techniques allowing the calculation of stress intensity factors. Further, several of these behaviors require specialized meshing techniques or analysis procedures. Thus, it is essential that the credibility of the model be established through a process of model validation. This paper lays out a framework for examining individual modeling techniques employed in the model, using a “building block” approach. Rather than trying to assess the entire model of a wheel on a discretely-supported rail containing an internal defect against a test measurement of the same conditions, the model is broken down into several key behaviors that must be verified. These distinct model behaviors, such as the method of discrete support, are then individually compared to known results to develop confidence in the simulation’s ability to produce physically-realistic results. In this way, confidence can be developed in the overall, complete model by developing confidence in several of the distinct modeling techniques that are employed in the overall model. The modeling techniques described in this paper include modeling the discretely-supported rail under a wheel load, modeling the internal defect as a crack, and using a submodeling technique to combine areas of coarse and fine mesh in a computationally-efficient manner.