Friction stir welding is a relatively new advanced joining technique that requires minimal power input, ultimately leading to less inherent residual stresses and distortion. The process involves a spinning tool which first plunges into the surface of the, to be welded assembly and then traverses along the joint. Frictional heat is generated, softening the material at temperatures significantly below the melting temperature of the parent material. As the tool traverses along the joint at a predetermined speed, the assembly is joined by means of a plastic straining process. This advanced welding technology has been validated for various aluminium alloys but it is only recently, due to advances in tool technology, that the possibility of joining mild steel using friction stir welding has become a viable option. This study looks into friction stir welding of mild steel and develops simplified numerical methods for the prediction of thermal gradients, residual stresses and deformation. In principle the process modelling requires a multi-disciplinary approach involving coupled thermofluid, microstructural-structural modelling process. Much of the latest thermo-mechanical studies of friction stir welding rely on a number of over simplifications particularly related to the heat flux distribution across the tool shoulder, and also on the backing plate boundary conditions. The objective of this paper is to scrutinise the effects of modelling in more detail and establish the most important factors leading to accurate yet computationally efficient prediction of thermal gradients and inherent residual stresses. The results show that both the heat input and