Case studies determining the effects of track stiffness on vehicle dynamics

Vehicle/Track dynamic simulation programs are now used routinely in the rail industry in areas such as vehicle and track design, derailment investigation, and prediction of vehicle/track behavior under a variety of conditions. In these investigations the accuracy and validity of the simulation highly depends on information input into the program pertaining to both the vehicle and the track. Track geometry cars provide a variety of track measurements which are used to characterize the track (e.g. gage, cross level, profile, curvature, etc.) but these measurements do not include track stiffness information. For lack of better data these simulations generally assume constant track stiffness and this can lead to significant inaccuracy. The University of Nebraska (UNL), under sponsorship from the Federal Railroad Administration, has developed a method to estimate vertical track deflection and stiffness from a rail car traveling at revenue speed. The system has been tested on thousands of miles of track. It is believed that these vertical track stiffness estimates, along with the standard track geometry data can now be used to improve vehicle/track simulation results. In this paper, sites from recent testing of the University of Nebraska system were chosen to be studied with the vehicle dynamics software VAMPIRE®. These case studies demonstrate scenarios where the variation of track stiffness may significantly affect vehicle dynamics performance in different aspects. Results suggest that wheel unloading may be increased by up to 20% by the inclusion of vertical stiffness data in the simulations. Car body vibration and dynamic wheel loads also tend to be affected by the stiffness variation. The proposed approach has limitations because stiffness estimates are made without direct measurements of wheel loads which leads to possible errors in the estimate. Also, the inadequacy of the Winkler model used in the process of stiffness estimation may also cause some inaccuracies. These limitations are further explored in this paper.