Effects of plate stiffness on the fatigue resistance and failure location of pipe-to-plate welded joints under bending

A series of tests have been carried out using specimens made of a tube, having a thickness of t=10 mm, joined to a plate by fillet welding. Two different kinds of specimen were employed, differing in the plate geometry (stiffness). Both kinds of specimen were tested under bending (prevalent load) and shear loading in as welded conditions. Different initiation regions for the fatigue cracks were found and significantly different fatigue resistances were obtained for the two geometries in terms of the nominal stress approach (or in terms of applied load vs cycles to failure). Two local methods for the fatigue life assessment were then applied to independently analyse the experimental results: the fictitious notch rounding approach proposed by Radaj, which is also recommended by some international standards and the more recently proposed peak stress method, which is based on the NSIF concept. It is shown that the nominal stress method, which is by far the simplest method among those recommended in standards for analysing the joint under study, fails to explain the observed different endurances. On the other side, the methods based on local stresses account for the different joint stiffness and provide a reduced scatter in the results. However, even if local approaches, accounts for differences in the structural behaviour of the joint, the knowledge of the actual geometry of the weld need to be accounted for, in order to be able to identify the fatigue crack initiation region. For a design purpose, a safe prediction of the fatigue endurance of the joint can be obtained by all the analysed methods, if the corresponding recommended design curve is used.