The in-situ railway repair welding process is significantly different from other rail welding processes, such as thermite welding of rails or flash butt welding, as it consists of multiple heat cycles. In repair welding, the damaged part of the railhead is removed, and new material is added using fusion welding to, layer-by-layer, fill the gap. Therefore, the repair welding procedure can be described as an in-situ Wire Arc Additive Manufacturing (WAAM) process for the rail head. The similarities include the practical process, the complex microstructure evolution and the resulting residual stress state. In this study, a computationally efficient FE-simulation methodology for railhead repair welding of a pearlitic rail is developed. To accurately simulate the material behavior, the modelling includes phase transformation kinetics, cyclic hardening plasticity, transformation-induced plasticity (TRIP), and self-consistent multi-phase homogenization. Moreover, the simulation uses explicit modelling of a moving heat source and continuous addition of weld filler material. It is well established that detailed multi-pass and multi-layer welding (i.e. WAAM) simulations like these are associated with immense computational cost, and to address this, we explore a model reduction scheme. In this scheme, 3D and 2D models are combined to accurately model the residual stress state of the repaired rail. More specifically, using a 2D generalized plane strain model, extended with out-of-plane axial and bending stiffness, to simulate several rail cross-sections, a 3D residual stress state is obtained by interpolation between the cross-sections. Comparing the simulation results to that of a 3D reference simulation model and experimental measurements shows that high precision results are obtained at a fraction of the computational cost. Moreover, this allows for different mesh densities to be used in the process simulation and subsequent component operation simulations. To demonstrates the usability of the proposed simulation methodology, the risk for subsurface fatigue crack initiation is estimated by simulating the repaired section of the rail, including the residual stress field, being subject to loads corresponding to passing wheels in operation conditions. The risk of fatigue cracks due to the residual stresses interacting with the superimposed operational lodes is estimated using the Dang Van criterion.