The apparent decrease in fatigue strength of welded components with increasing size was introduced in Chapter 1. The significance of the effect is that fatigue design curves based on tests on relatively small-scale laboratory specimens could prove unconservative for larger components. In the UK design guidance the thickness effect is catered for by an empirical -0.25 power law (Equation (1.6)), as suggested by Gurney [2.6]. This relationship was originally based on several reviews of available data [2.7-2.9] for plate thicknesses up to 50mm. A comprehensive investigation into the effect of thickness up to 100mm was initiated in 1982 as part of the UKOSRPII [2.2] programme. This investigation indicated that the exponent
of -0.25 in Equation (1.6) was perhaps not sufficient to describe the effect over the greater range of thicknesses examined. The current review of the UK design guidance also proposes a modified thickness exponent [2.1 0].
From the UKOSRPI&n projects and other researches, a reasonably clear explanation of the effect of section thickness has emerged. The thickness effect in welded plate joints is primarily dependent upon the following;
- the local stress concentration factor - the through-thickness stress gradient
- the ratio of attachment to base plate thickness - the extent of crack initiation
The current thinking on each of these factors is reviewed below.
2.2.1 The effect of the local SCF at the weld toe
Using fracture mechanics analyses of various notched components, Haagensen et al [2.11] identified the local weld stress concentration factor (SCF) as being mainly responsible for the thickness effect. In welded joints, the local SCF is predominantly controlled by the degree of blending of the weld toe pass, which in general does not change with absolute joint size. This produces higher local SCFs in larger components due to incomplete scaling [2.12] as shown in Figure 2.2; this results in earlier crack initiation and faster initial crack growth rates, and hence shorter life. This has led to the development of modified weld profiles such as the AWS [2.13] profile for increasing thicknesses; this has been claimed to virtually eliminate the thickness effect. Experimental evidence of the influence of local SCF on the thickness effect was obtained by Vosikovsky and Bell [2.14], who used the local SCF as a basis for comparison of fatigue endurances obtained from 45“ and AWS weld-profiled T-plates of thicknesses from 26-102mm. Joints with a 45“ weld and equal base and attachment thicknesses gave the strongest thickness effect, together with the greatest corresponding increase in local SCF with thickness. Joints having the AWS profile exhibited virtually no thickness effect, and only a very slight increase in local SCF.
2.2.2 The influence of through-thickness stress gradient
Other investigations have demonstrated the thickness effect to be primarily dependent upon the stress gradient which exists through the base plate. For example, a thick joint loaded in bending and having the same surface stress as a smaller but geometrically similar joint, will have a reduced through-thickness stress gradient compared to the smaller joint, as illustrated in Figure 2.3. Berge [2.15] proposes that cracks in thicker joints experience a higher stress intensity in the early stages of growth due to the reduced through-thickness stress gradient. Gurney’s [2.16] tests on a series of joints with different plate thicknesses, but the same weld size, produced a reduced thickness effect compared to that for geometrically-scaled joints. A thickness effect has also been demonstrated [2.10] on 16-160mm thick
plates with no weldment tested in 3-point bending.
A thickness effect is observed in both the initiation life and the crack propagation life of welded joints [2.17]. It would seem then that both the local SCF and the through-thickness stress gradient contribute to the observed effect. This hypothesis is supported by the work of Webster and Walker [2.18] who carried out a comparison of the results of tests on welded and similar cast joints in the range 50- 100mm thick. The welded joints showed an effect of thickness on fatigue threshold which the cast joints did not. This effect on threshold must be due to the local SCF, since the cast joints were designed to give minimal local stress concentration. For similar initial defect sizes in both thick and thin joints, the local stress in the thicker joints would have to be greater. In addition, the endurance results for the cast joints showed a thickness effect which was similar to that for the welded joints, which points to the through-thickness stress gradient as the source of the effect.
2.2.3 The effect of attachment size
The UK guidance on the thickness effect assumes that the base plate and attachment are of the same thickness (1:1). However, geometric scaling of joint has been shown to be detrimental to fatigue life, thus producing an exaggerated thickness effect [2.19]. There is evidence that if the attachment is maintained as a relatively thin plate and the base plate thickness varied, the thickness effect is substantially reduced [2.14], largely as a result of the decreased local SCF [2.20]. AWS D l.l [2.13]
assumes that the attachment thickness is the controlling parameter, since increases in the base thickness are offset by the improved weld profile. Most of the improvements in fatigue life obtained from use of the AWS profile and with reduced attachment thickness are observed in the very early stages of crack growth (a/T<0.1) [2.14].
2.2.4 The influence of multiple crack initiation
Another factor influencing the variation in fatigue strength of welded joints with thickness relates to crack shape development. The final failure of welded plate joints is often defined as the number of cycles required to propagate a crack to a depth equal to half the base plate thickness. At this stage the overall joint deflections may become unmanageable or fracture may occur. In reaching this depth a crack will typically have initially propagated as a surface crack and become a full width edge crack at some stage during the fatigue life. The transition to an edge crack is an important stage in the fatigue life, since the stress intensity factor for an edge crack is greater than that for a surface crack of the same maximum depth. Thus, early transition to an edge crack will generally lead to a reduced crack propagation life. This is often the case for thick joints of moderate plate width (see Figure 2.4), leading to an exaggerated thickness effect; in some studies the plate width is increased in proportion to the thickness thus avoiding this problem [2.21]. Early transition to an edge crack is promoted by crack initiation occurring at a large number of sites along the weld. The transition to an edge crack has been demonstrated to occur progressively earlier with increasing stress and increasing plate thickness [2.22, 2.14], as can be seen from Figure 2.5. The notch stress field and residual stresses can also influence the transition to an edge crack. The strength of these stress fields decreases with crack depth, which may lead to faster growth at the surface than at the point of maximum depth, resulting in particularly long shallow cracks [2.23]. Various fracture mechanics models have been developed to cope with the phenomenon of multiple crack initiation [2.23, 2.24]. The model developed by Bell and Vosikovsky [2.24] was successfully used to predict the thickness effect observed in the Canadian fatigue programme. The model estimated a thickness correction exponent of -0.3 for simple welded joints rather than -0.25
as proposed by Gurney. The recent review of the UK guidance with respect to thickness effects in welded plates also proposes an exponent of -0.3 as a better fit to the available data.
2.2.5 Thickness e^ect in tubular joints
As discussed in Chapter 1, a thickness effect has also been observed in tubular joint test data. Many of the explanations described above for the origin of the thickness effect in simple welded joints also apply to tubular joints, except those relating to the interaction of cracking with finite plate width. Very few results exist for tubular joints of chord thickness greater than 50mm, thus making the effect difficult to define over the whole range of thickness. The effect can be expected to depend on parameters such as brace-chord diameter ratio and loading mode, in addition to brace-chord thickness ratio and absolute size. It has been suggested that tubular joints are more susceptible to local variations in weld toe geometry, producing a lot of scatter in fatigue test results; this masks the thickness effect [2.2], making
interpretation difficult. This may be due in part to the fact that initiation in tubular joints is confined to a relatively small length of weld situated in the hot-spot region. In this case, the probability of encountering extremes of weld toe profile in such a small region is reduced in comparison to that in planar joints, which have an approximately uniform distribution of stress along the entire length of the weld.
Another difference between planar joints and tubular joints is the greater proportion of the fatigue life spent in propagating large cracks in the tubular joints. For a given hot-spot stress range, the average growth rate during the propagation phase of tubular joint tests has been shown to exhibit a significant thickness effect [2.5]. This observation is interesting since it implies that the thickness effect in tubular joints is related more strongly to global geometric properties than to weld toe geometry. In the absence of more complete information, the forthcoming design guidance [2.10] recommends a thickness exponent of -0.3 for all welded joints including tubulars.