Using multiple adaptive regression to address the impact of track geometry on development of rail defects

Abstract This paper presents an application of multivariate regression splines (MARS) to a model looking at the railroad track defect behavior. MARS is a non-parametric function estimation technique that shows great promise for fitting non-linear multivariate functions. The MARS approach was used here, together with traditional regression analysis techniques, to develop equations to predict the life (in Millions of Gross Tons or MGT) of a rail defect in the presence of track geometry defects. The MARS approach, using extensive input data, identified and accounted for the key variables contributing to a reduction in rail life. The MARS technique allows for easy interpretation of the relative importance of the different input parameters, to include defect type, and resulted in the development of rail defect life predictive equations as a function of these key parameters. As part of this study, it was determined that there is a reduction in rail life of approximately 30%, when track geometry defects are present.