Rail wear in curves at the tramway of Amsterdam is a considerable problem since it consumes a significant part of the maintenance budget. The wear mechanism is not fully understood because some curves could suffer much more than others, while the cause for the difference is unknown. GVB, the public transport operator of Amsterdam, wants to obtain better insight into this problem to optimise maintenance processes and reduce maintenance costs.

First, a literature review was conducted to assess knowledge on this issue and analyse methods to model train-track systems. Next, based on the literature review, parameters were identified which could influence the rail wear process, e.g. tram velocity, primary yaw stiffness or rail hardness. An evaluation of available data on those parameters at GVB was made. Data that was available and deemed essential was aligned and later used as input for a rail wear prediction model for the tramway of Amsterdam. The essential data analyses are the velocity analysis and wheel qR decay analysis. A tool was developed to obtain a detailed profile of driven velocities at the tram network, based on massive daily data from all the trams. Wheel measurement data from all the trams in the network were analysed to assess the wear of the tram wheels. Statistics were obtained over six years of wheel measurement data measured twice per year per tram. Besides, the usage data processed for inclusion in the model are tonnage, vehicle type distribution, amount of coupled vehicles and average vehicle loading.

Furthermore, more than six hundred simulations were performed at DEKRA rail with GVB tram models in VAMPIRE, to obtain more insight into tram curving behaviour. Based on the simulation outcomes, relationships between energy dissipation (\(T_{\gamma}\)) and radius, velocity, flange angle, vehicle loading and -type were derived. Those relationships were obtained per wheel and aggregated per tram passage. Finally, a rail wear model was made, which combines the relationships derived from the simulations, track characteristics and usage data about the curves. The energy dissipation (\(T_{\gamma}\)) was used as a wear indicator by the prediction model.

From the velocity analysis, trams generally keep to the speed limits, but trams drive too fast at specific curves. The wheel analysis showed qR at the tram's front wheels decays considerably faster than other wheels. Combino tram type's wheel decay was poorer than the older BN tram type. From the simulations, wheel wear has the most adverse effect on rail wear in curves. Increased vehicle velocity or amount of passengers also has a considerable negative influence on rail wear. The expected wear rises exponentially if the curve radius decreases. Also, the dependency on velocity increases exponentially when the curve radius decreases…