Analysing the ability of the UniGrow method to assess the effect of the geometry of the weld excess on the fatigue life of butt-welded plates

Abstract

Fatigue is one of the driving limit states in the design of wind turbines, as they are subject to varying loadsfrom winds, waves and gusts. One such governing area is the butt-welded connection from the shell to thering flange. To optimize the tower design and to help focus future research, more insight on the influence ofmaterial properties and geometrical features on the fatigue life of these butt welds is needed. Due to thecosts associated with testing, SGRE stated the need for a model that could assess the total fatigue life of buttwelds subject to Mode I cracking in a rather quick way. After literature study, it was proposed to use theUniGrow model.The UniGrow method is a total fatigue life model in which the material is considered to consist of elementaryblocks. Fatigue is modelled as a process of continuous crack initiation of the elementary block ahead of thecrack tip. In this model, the plastic compressive residual stress ahead of the crack tip plays an importantrole in the determination of the crack growth speed. These plastic compressive residual stresses ahead ofthe crack tip can either be determined analytically (analytical UniGrow) or by using numerical methodssuch as FEA (numerical UniGrow). The plastic compressive residual stresses ahead of the crack tip areconverted into a plastic residual SIF that is used to reduce the applied residual stress intensity factor (i.e.plasticity decreases crack growth speed). Due to the application of a number of methods such as theCreager-Paris equations, the current UniGrow implementation is solely valid for pure Mode I crack growth.Validation of the UniGrow method in the crack propagation range has been performed with results fromexperiments on CT specimens made of S355 steel. Good correspondence between the experimental resultsand predicted crack growth rates were found for all stress ratios except R = 0.25. Both Analytical andNumerical Unigrow methods have been analysed. Based upon the results of the validation it wasrecommended to use a numerical method (elastoplastic FEA) to determine the plastic residual stressesahead of the crack tip as this provides more physically accurate results. The analytical method can be usedto obtain an initial guess of the elementary block size, as it is a much faster method. Furthermore, using theMorrow method for the determination of the fatigue life of the elementary block ahead of the crack tip wasfound to provide more accurate results than the SWT damage parameter. This is due to the fact that in theelementary block ahead of the crack tip, cyclic stress relaxation occurs and, therefore, mean stress effectsneed not be accounted for. For short cracks, the fatigue life of the elementary block is usually outside theLCF range (Nf > 5*104 cycles). Here, the use of the Morrow method without considering residual stresses isnot recommended since cyclic stress relaxation does not occur due to limited plasticity.The performance of the UniGrow method as total fatigue life was studied using literature research and acomparison to another total fatigue life model: Two-Stage-Model. In the studied literature, researchers tendto use the weight function method to determine the SIF in welded geometries. Analysis comparing theweight function method to FEA results showed that the WF method generally overpredicts the SIFscompared to FEA. For a flat plate, the difference was found to be a maximum of 3.5%, whereas for the weldedgeometry, the difference was up to 12%.From the literature and the comparison of the UniGrow model to the Two-Stage-Model, it was concludedthat the UniGrow model is incapable of predicting proper fatigue crack initiation lives as it is unable tocapture the complexity of short crack growth. The model provides satisfactory results for cracks wherecrack propagation is a dominant part of the fatigue life (such as welded geometries and notched specimensin the LCF regime). It was proposed to use the UniGrow model to determine fatigue lives of notchedspecimens up to 105 cycles. It was found that this limit also depends on the SCF present in the specimen.For welded geometries, available results in literature showed good correspondence. However, the weldsused by SGRE are of better quality (lower SCF) than the ones used in UniGrow research. This means thatrelatively more of the fatigue life will be spent in fatigue crack initiation. More research is therefore neededto examine whether this affects the functioning of the UniGrow method for welds.The UniGrow method was lastly used to determine which of the parameters controlling the geometry of theweld excess (the weld flank angle, weld toe radius and weld excess height) is of greatest influence on thefatigue performance of a butt weld. From this analysis it was concluded that changing the weld excessheight, followed by the weld flank angle has the most beneficial effect on the fatigue life. The results fromthis parametric study should be used with great care: the usage of the UniGrow method combined withneglecting the residual stresses could lead to incorrect predictions of fatigue crack initiation life.Simplifications of the material and crack type could also have had an influence on the results.