The main goal for this thesis has been to perform fatigue life predictions on large components.Dierent methods were to be chosen and reviwed critically. Experimental datafor girth welded pipes was assessed by means of a version of BS7910 which were modeledfrom scratch. This version was veried with results from CrackWise, and used to predictthe fatigue life for loade cases matching the selected experimental data. The fatigue assessmenttool P-FAT was used to predict the fatigue life for the same experimental data,and the results were compared. Except for some deviations recorded for embedded cracks,a generally good agreement was found between the two tools.Both the modeled BS7910 and P-FAT predicted conservative but accurate fatigue livesfor the experimental data that had signicant defects. They predicted non conservativeresults for specimens that did not have signicant defects, when the initial crack depthwas set to 0.1 mm. When the initial crack depth was set to 0.5 mm the modeled versionpredicted fatigue lives in agreement with the experimental results.Fatigue life predictions were also performed by assistance of S-N curves provided by DNVRP-C203.The results from these predictions were in agreement with the experimentaldata, except for four specimens. These four specimens had signicant aws, which reducedthe fatigue life signicantly.A review of a two-phase fatigue assessment model was made. This model uses a strainedbased approach to assess the fatigue crack initiation phase, and the fracture mechanicapproach suggested in the BS7910 to assess fatigue crack growth. This method haveshown promising results for fatigue assessment of llet welds in the literature, but themodel needs further investigation and calibration to be used to predict fatigue life ofgirth welded structures. A specic area to investigate is the proposed transition depth of0.1 mm. This depth is dubious based on the results