Civilization has relied on welded structures to facilitate fabrication and improve our quality of living for the past century. Welds are used in our production of energy, to create infrastructure that we rely upon such as bridges and building, and to fabricate the equipment that makes all of this happen. In short, the joining of two metals through welding has contributed immensely to our society.

One problem that has plagued welds is their susceptibility to fatigue failure due to cyclic loading. Fatigue in welded joints is a complicated phenomenon and the subject of fatigue of welded structures been the subject of great study. The goal of the research presented in this dissertation is to improve fatigue life prediction capability by incorporating the effect of the welding process prior to making the structure.

The first area examined in this study is the residual stress that is induced during the welding process. If the goal of virtual design and verification of welded structures is to become a reality the residual stress state needs to be known prior to making a product. Computational welding simulation can be used to predict the residual stress state of the welded structure prior to the manufacturing of any part. In order to use computational welding simulation in fatigue life predictions the validity of the results need to be confirmed. This was done in the following dissertation work in two steps, initially by using 3D image correlation to measure the full field displacement of a structure as compared to simulation, and secondly by using neutron diffraction to measure the residual stress after welding as compared to the computational welding simulation results. The results showed that the residual stress state could be predicted with enough accuracy to be used in fatigue life predictions.

It is known that the residual stresses redistribute during cyclic loading which can have an impact on their effect on the fatigue life of the structure. The third area this dissertation looks at is the redistribution of the residual stresses during cyclic loading, where residual stress is measured as a function of cycles, again using neutron diffraction. This analysis provides an understanding of how much of an effect the residual stress redistribution has on the residual stress state during the majority of the cycles experienced by a part undergoing cyclic loading.

The last section combines the results of these earlier studies to suggest a methodology to predict the distribution of the fatigue life for welded structures that accounts for the welding manufacturing process. This is achieved by accounting for distribution of the local geometry, the residual stress present, and the material properties. By using a Monte Carlo simulation a predicted distribution for fatigue life is obtained, which is then then compared to experimental fatigue test data to test the validity of the proposed methodology.