Thickness Effect

A size effect was well known in notched machined components whereby fatigue strength decreases with increase in size. This effect was also found in welded joints in 1970s. Fatigue design curves for offshore structures were developed from the test results of laboratory specimens which generally are less than 25mm thick (typical thickness= 12.7mm and diameter=500mm) due to the limitation of loading capability of testing machines. However, the size of offshore structures are usually larger than that of these laboratory specimens. Therefore, the size effect issue has been the subject of considerable research over the past decade. Significant advances have been made in the understanding of the mechanisms which govern this apparent reduction in fatigue strength with increasing size.

Plate material in the as-rolled condition also shows a thickness effect which is somewhat less than that for welded joints, but still significant. It can be explained statistically in that the number and severity of flaws is likely to increase with size. However, for welded joints, this statistical explanation is probably of lesser significance. Of particular importance is the influence of weld joint dimensions on stress concentration and through wall stress distribution. This geometry effect has been generally recognised as the major cause to thickness effect and can be explained by the following factors(Berge 1985):

1) the magnitude of the stress concentration at the weld toe which is mainly determined by the local weld geometry. The notches at the weld toe in large joints are relatively sharper. The toe of the weld has nearly always the same radius due to the welding process.

2) the gradient of the stress in the plane of crack growth which is mainly determined by the plate thickness. The stress gradients are less steep in large size joints, which means that in a large joint a crack of a certain dimension is in a higher strained area than in a smaller joint. Moreover the plastic zone size will be larger.

3) the number of cycles in crack growth through the region of a steep stress gradient, relative to the total number of cycles to failure which is mainly determined by the size of the initial crack and the crack ellipticity.

Based on theoretical fracture mechanics calculation and considering the effect of joint dim ension on SIF and crack growth rates by using weld toe SIF correction factor( ) for fillet welded joints, Gum ey(1979) proposed the following model:

s = s.

W = ( 1 - 5 5 )

where Sg is the stress range and Ng is the fatigue life at the reference thickness, Tg, while S is the stress range which results in the same fatigue endurance and N is the fatigue life which has the same stress range at a thickness T.

Using a simple fracture mechanics model, Berge(1985) got the same answer as Gurney by making following assumptions:

1) W elded joints of the same type in various plate thickness are geometrically similar. 2) Initial conditions of fatigue crack growth are independent of plate thickness( a. constant ). 3) Furthermore, the notch root at the weld toe may be of constant radius, p = constant

instead of ^ = constant as they are largely determined by the condition of the last pass at the weld toe.

The experimental data from some plate as-welded joints with thickness from 16mm-100mm confirm ed the above model. Supported by the experimental data, the model suggested by Gurney and Berge has been implemented in UK Department of Energy fatigue guidance in 1984. (B=22mm for welded plates and 32mm for tubular joints). Insufficient tubular joint test data (UKOSRP-I thickness 6.3mm -32m m diameter 168mm-914mm as-welded condition and some ECSC data) exist to investigate this effect. However, it is assumed that tubular joints will behave in the same way as other welded plates with respect to the influence of thickness.

W ith more data produced from extensive research programs in Europe including UKO SRP II data (chord diam eter 1830 and thickness 75mm) available, it seems that it is possible to have a more reasonable model. By fitting S-N curves to these data at each thickness, a conservative thickness correction exponent of 0.3 is proposed in current fatigue guidance:

T ( 1 - 56 )

A dditionally, the base line thickness for thickness correction was extended to 16mm for tubular joints.

There is a debate about whether the thickness effect can be largely decreased or eliminated by improving weld profiles between the researchers in Europe and US. In the US view, the thickness effect can be compensated for by improving weld profiles in thick sections. Weld improvement leads to an increase of fatigue strength and a decrease of the thickness effect. This means that increasing rate of fatigue strength due to weld improvement becomes larger as the plate thickness increases. Use of an improved AWS weld on joints with equal attachment as base-plate thickness reduces the effect of thickness from a factor of over four to a factor of about two.

European research, especially from UK and Norway, found that the thickness effect is essentially the same for a wide range of welded joints and weld geometry. Welds with improved weld toe profile showed higher fatigue strength, however the strength decreased when the thickness was scaled up. For joints improved by weld toe grinding or TIG re­ melting the fatigue strength in general is improved. However, test data indicate that the thickness effect is of the same magnitude as for unimproved joints.

Apart from this debate, it is noted that there are some secondary factors from research programmes and these are listed as follows:

1) attachment thickness

There is a reduction of the thickness influence if the attachment thickness is kept constant, but it is only secondary effect on fatigue strength compared to the effect of main plate thickness because the dominant parameter affecting weld toe SCF is the ratio of weld toe radius to main plate thickness, (-^ ), which is not influenced by attachment thickness. The

experimental data on tubular joints indicated that the fatigue endurance is proportional to and it is small compared to that for chord thickness T~^'^.

2) Loading mode

The thickness effect is stronger under bending than under axial loading. 3) High/low SCFs

Experimental results and fatigue predictions indicate that the thickness exponent decreases with decreasing value of the SCF. Welded joints with low SCFs(SCF<1.5) give a smaller thickness effect as compared with as-welded joints.

4) Fatigue Life

The thickness effect depends on fatigue life, and increases as the fatigue life becomes longer. In particular, this trend is remarkable in the as-welded proportional joints. Both crack initiation and propagation lives are reduced with increasing thickness for welded

joints. The thickness effect is larger for crack initiation life than for failure life, the thickness effect on crack propagation life is very small. The thickness effect of a welded joints is supposed to be determined by crack initiation behaviour and the growing process of a shallow crack to the depth of about 1mm.

However, the thickness effect appears to be the same, regardless of local weld geometry and loading. Also it represents the average performance of welded joints and therefore is not too conservative when the thickness effect is evaluated for individual data sets. So the design rules for welded joints should be amended to include some secondary factors such as a less severe thickness penalty for low SCF joint and a stronger thickness effect for high SCF joints. It seems that the best solution is to establish the accurate fracture mechanics model that can take account the thickness effect. Thus an accurate thickness exponent can be applied depend on local geometry and loading.