SEGÚN EL OBJETIVO ESPECIFICO 2: ANÁLISIS DE LAS

The measured stress concentration factors have been quoted above to three significant figures. These values are quoted because they were used in setting up the scaling for the fatigue tests, and are not intended to quantify the accuracy of the measurements. It is estimated that the absolute accuracy with which stresses can be determined using foil strain gauges in conjunction with proprietary equipment is of the order of +/- 10%. This magnitude of measurement error is reflected in the scatter of the results, see Figure 4.4.

A peculiar feature of the stress analysis results is the difference in stress concentration factors quoted for the top and bottom saddle hot-spots. In many cases, the bottom hot-spot SCF is considerably greater than that at the top. It was discovered halfway through the " C series tests that this was due to the procedure used for recording strains. The effect is caused by zeroing the strain gauge signals in a no-load condition whilst the brace member is sagging under it’s own weight. When load is applied, the brace straightens and the measured strains incorporate this effect of brace bending, resulting in higher tensile strains being recorded for the bottom saddle than for the top. This effect was avoided in later tests by zeroing the strain gauges at a non-zero tensile load, and recording differences in strain from this preload. It can be seen that the effect has almost disappeared in the results for tests C4, D l, and D2. In an investigation into this effect, it was noted that the

average of the top and bottom saddle results was unaffected by the sagging of the brace, and for this reason the average SCF was consistently used in all subsequent data presentation and results analysis.

The parametric predictions show considerable variation, although none of these underestimates the hot-spot stresses. It is interesting to note that excepting the Hellier, Connolly, and Dover prediction, the variations in SCF prediction are actually smaller than the variation in measured SCF due to the differences in chord wall thickness. This is a very significant observation, bearing in mind the limited margin of conservatism in some of the parametric predictions for this geometry.

The effect of chord wall thickness on stress concentration factors is interesting. Although the likely influence of chord thickness is obvious from simple considerations, the magnitude of the effect is rather larger than might be expected. For a reduction in chord wall thickness from 17.4mm to 16mm (nearly 10%), the corresponding increase in measured hot-spot stress concentration factor is almost 25%. This is a consequence of the fact that hot-spot stress concentration in this geometry is primarily due to secondary bending stresses, such that the decrease in 2“** moment of area is more relevant than the decrease in area of the brace-chord

section. For a flat plate loaded in pure bending and subject to a constant moment, the increase in surface stress associated with a decrease in thickness of 1 0% would

be approximately 2 0%, which is closer to the measured observation.

The implications of the effect of wall thicknesses on stress concentration factors can be illustrated by consideration of the likely variations which would be encountered in actual structures. The thickness tolerances for rolled seamless tubes of the type used to fabricate Offshore tubular joints [4.11] are -12.5% / +15% of the nominal thickness. In practice, the tolerance is more likely to be an absolute error on thickness, rather than a percentage. This implies that the effect on SCFs would be greatest in thinner tubes and not main structural members. However, one of the parametric formulations may be used to estimate the variations in hot-spot SCF and expected fatigue life arising from extremes cases within the tolerance band.

The parametric formulation of Kuang, which corresponds closely with the experimental SCF variations, has been used to estimate SCFs and fatigue lives for the extreme wall thickness combinations, and the results are shown below;

Test case T (mm) t (mm) Variation in stress Variation in life Nominal chord / nominal brace 16.0 16.0 0% 0% Thick chord / nominal brace 18.4 16.0 -26% +145% Thin chord / nominal brace 14.0 16.0 +33% -58% Thick chord / thin brace 18.4 14.0 -38% +319% Thin chord / thick brace 14.0 18.4 +60% -76%

The variations in fatigue endurance have been calculated assuming a single segment SN curve with a negative inverse slope of 3. The maximum variation occurs between the thick chord / thin brace and thin chord / thick brace conditions, representing almost an order of magnitude on fatigue life. This variation is greater than the +/- two standard deviations (a factor of approximately 6 on life for 16mm

joints) used to quantify the scatter band of results in the UK design guidance [4.1]. It is perfectly plausible therefore that tubular wall thickness variations could be solely responsible for the scatter in tubular joint fatigue test results in air. More significantly, perhaps, the fatigue life of a thin chord / thick brace combination could be as little as 1/4 of the life of the nominal joint. In any event, it is clear that wall thicknesses are important item of data to be reported in any tubular fatigue test programme.