Experience has shown that in many instances designers have reacted to service fatigue failures in welded structures in the wrong way, with the result that design changes made have not had the desired effect. It is the intention of the present paper to demonstrate that fracture mechanics techniques, which are now well established, can be used to provide an appreciation of those aspects of a design which are important with respect to the fatigue performance of the structure. This understanding provides an excellent basis for evaluating the benefit of proposed design changes when service problems are encountered.

Analysis is based on stress-intensity solutions derived by two-dimensional linear elastic finite element modelling. Fatigue lives are predicted by integration of the appropriate fatigue crack growth law on the basis that the life of a welded joint is dominated by the propagation of small pre-existing defects.

An important consequence of this feature of their behavior is that the fatigue lives, and hence fatigue strengths, of joints are strongly influenced by factors which affect the rate of crack propagation. Results are given which illustrate that factors such as weld profile, the tensile strength of the parent material, and stress relief by post weld heat treatment have little effect on fatigue strength, although historically designers have often looked to modify one of these in the face of service problems. Other factors such as the detailed design of joints and the reduction of stress concentration factors are shown to be more significant.

It is also demonstrated that even the most fundamental step in the design process, that is, the selection of scantlings, may not have the expected effect on fatigue life since the fatigue strength of a joint may be reduced by increasing the plate thickness.