The use of welding is widely adopted to assemble structural components in the construction industry for many years. To ensure safety of these welded components, many fatigue tests have been conducted on many different shapes and configurations of welded connections to precisely assess the fatigue life. However, testing real-size structures and specimens is very limited due to it’s high cost and in-applicability. Thus, these full scale specimens are cut down into small scale specimens to allow applicability for testing. Different characteristics are exhibited between the full and the small scale specimens, as there are major difference in residual stresses induced by welding and cutting, which may give non conservative predictions for fatigue life. In this thesis, the objective is to forecast the evolution of the residual stress field originated by welding of a full scale and small scale specimens of a cruciform joint at the weld toe after breaking it down into smaller specimens using a cutting process by performing a numerical analysis using Abaqus finite element analysis software. To achieve this goal, first, a thermo-mechanical welding simulation was performed to obtain a welding residual stress field on a 910 mm long cruciform joint, which is done in two main parts, starting with the thermal model in which a temperature field is analysed. The temperature field from the thermal model is used as an input for the mechanical model in which the residual stress field is produced due to the temperature change and restriction of movement of material due to the shrinkage and expansion. Secondly, the 910 mm long full-scale cruciform joint was cut into five shorter specimens of 500, 210, 120, 75, and 20 mm. The welding residual stress (WRS) levels at the weld toe for each specimen was recorded and showed large stress losses and relaxations as the specimen gets shorter in length. a major longitudinal stress loss of 97% and 77% loss of maximum principle stresses when cutting down the 910 to a 20 mm long specimen, making it almost free of WRS, but only a 5-6% loss of longitudinal and Max. principal stresses when going from 910 to 500 mm. Thirdly, after generating multiple welded specimens with different lengths, a tension load of 186.2 is set in the x-direction of the attachment plate of the cruciform joint, and the stress level at the weld toe was analyzed due to the applied load and the WRS. A 40% increase of the stress occur due to the applied load, but a very slight decrease in longitudinal stresses for the 910, 500 and 210 mm due to depicting a plate-like behaviour in contraction due to poisson’s effect. Finally, the same specimens were analysed under the 186.2 MPa load but without including the WRS. Different shapes of stress distributions were found, and differences in stresses when comparing the models with and without WRS in the models. The difference in longitudinal z-direction reached up to 282 MPa, while only 77 MPa in the transverse direction. The maximum principle stresses insured the importance of including WRS when performing fatigue assessment as it showed the fatigue failure to occur in the weld root with a crack to happen at the middle part. The specimens that exclude WRS would start cracking at the edges of the weld root, but in the central part of the weld seam when including WRS. The model that included WRS showed similar fracture location at the weld root as the fractured specimen performed in tests at TNO’s laboratories. The next steps in this research is the modification of a modelling methods. The results can be improved and smoothed by modelling using the effective notch method were a radius is introduced at the weld toe and the root to eliminate the stress singularities. The welding simulation can be improved to get better results by modelling the full cruciform joint without symmetry conditions, and include a weld order for all four welds with proper cooling time in between each weld.