Rail wear can result in extensive costs for the track owner if it is not predicted and preventedin an efficient way. To limit these costs, one measure is to predict rail wear through wear simulations. The purpose with this work is to perform simulations of successive rail wear on the Swedish light rail line Tvärbanan in Stockholm, by means of the track-vehicle dynamics software GENSYS in combination with a wear calculation program developed in MATLAB.

The simulation procedure is based on a methodology with a simulation set design, where the simulations to be performed are selected through a parametric study. The simulations include track-vehicle simulations, where the wheel-rail contact is modelled according to the Hertzian contact theory together with Kalker’s simplified theory (including the numerical algorithm FASTSIM). The results from the track-vehicle simulations serve as input to the wear calculations. When modelling rail wear Archard’s wear model has been used, including wear coefficients based on laboratory measurements. The measurements have been performed under dry conditions, so the wear coefficients under lubricated conditions (both natural and deliberate lubrication) are reduced by factors estimated by field observations. After the wear depth calculations the wear distribution is smoothed and the rail profile is updated. The simulation procedure continues with a new wear step as long as the desired tonnage is not attained.

Four curves of Tvärbanan with different curve radii, ranging from 85 to 410 m, have beenstudied in this work. On three of the curves the high rail is deliberately lubricated, whereas no lubrication has been applied in the widest curve. The vehicle operating the light rail line is an articulated tram with two motor end bogies and one intermediate trailer bogie. The line was opened in August 1999 and extended in one direction one year later. Rail profile measurements have been carried out by SL since March 2002. The traffic tonnage at the selected sites from the opening of the line to the last measurement occasion (September2004) is at most 8.9 mega gross ton per track.

The results of the rail wear prediction tool are evaluated by comparing worn-off area of the simulated rail profiles with that of the measured rail profiles. Simulated and measured resultsdo not agree too well, since the simulated rail wear is more extensive than the measured one, especially on the outer rail. However, the shapes of the simulated worn rail profiles are comparable to those of the measured rail profiles.