TWI Industrial Member Report Summary 244/1984
By I J Smith and S J Hurworth
Background
The present fatigue design rules are based upon a large series of experimental investigations. These investigations can be expensive and time-consuming to perform. Fracture mechanics based predictions of fatigue strength require an accurate numerical (usually finite element) solution to the relationship between stress intensity factors, external loading and crack length. It is now possible to determine this relationship quickly and relatively cheaply at least for two-dimensional elastic models. Thus it has become possible to investigate a wide range of possible combinations of geometry changes on the fatigue behaviours of common welded joints. The results of such an investigation are given in this paper, and the use of these results in helping to select the most cost-effective experimental test programme is discussed. Although the accuracy of any absolute fracture mechanics based fatigue strength prediction depends upon certain experimentally determined parameters such as initial defect size, relative predictions can be made and compared with experimental results. Likewise, the stress intensity factor v. crack length results can be used for defect assessment and fitness-for-purpose studies. For the joints studied in the work, the normalised stress intensity factor v. normalised crack length relationships were calculated using a finite element program developed especially for this type of work. The calculations were repeated for changes in the model analysed, such as attachment thickness to load-carrying plate thickness ratio or relative weld leg length. From these basic data, using various assumptions, fatigue strengths were then calculated. These numerically predicted fatigue strengths were then compared with the design curves and the adequacy of the various assumptions considered.