In a manner similar to that of Gurney and Maddox who blended fatigue data from widely divergent tests into guidelines which form the basis of British Standard, BS153, the authors have analyzed a body of existing data on fatigue tests of transverse butt welded joints as a prelude to formulation of design guidelines for such joints in offshore structures.
The fatigue life data from 759 individual tests of butt welds taken from sixteen different references are reanalyzed as a group to sort out the influence of various parameters on fatigue strength. Summaries of these data wi11 appear in the paper. The parameters which seemed to have little or no influence on fatigue strength at two million cycles are yield strength of the material, specimen configuration, welding process, joint configuration, electrode type, cyclic loading rate, residual stress, and R ratio (i.e., ratio of minimum to maximum cyclic stress). In contrast, the effects of small weld defects or poor reinforcement profile are significant. It was found that the data could be grouped into five families of S versus N relationships which resulted in much less scatter than that present when all of the data were blended together.
Design guidelines for fatigue loading of welded structures have been formulated by a number of standards and specifications groups. Most of these guidelines are based upon data obtained from numerous laboratory-type fatigue tests involving a wide variety of materials, specimen configurations, and test procedures. In particular, the guidelines of British Standard BSl53 are of interest because of their applicability to certain designs of offshore structures. The basis for BSl53 is a compilation of data by Gurney and Maddox(l). These authors blended data from numerous investigations into design curves to form the basis of the BSIS3 design guidelines.
Fatigue data characteristically involve considerable scatter from random uncontrolled variables. In spite of the scatter Gurney and Maddox made a commendable attempt to blend the available data into meaningful S versus N (Stress versus number of cycles to failure) relationships based upon weld joint classifications. One weakness in their system however, is the lack of data to support the Class D design curves (Le., transverse butt welds made by the shielded metal arc process). Because these data are extremely useful in some cases, the study described herein was undertaken to provide meaningful S versus N families for this class of joints. Data from sixteen references on transverse butt welds were gathered, analysed, and blended into S versus N families.
In view of the lack of shielded metal arc welded specimen data (Class D weldments) in the Gurney and Maddox paper, data from a number of references were compiled and analyzed in this study. In order to decide how best to blend these data for statistical analysis, each set of data was considered on the basis of facts available in the accompanying reference.
