Fatigue assessment in finite element analysis

Abstract

For the fatigue assessment of welded structures, several design methods are described by the design codes of IIW, Eurocode3 and DNV. The relation between the stress range and cycles to failures were derived from experiments: the SN curves. With the increased accessibility of finite element software to engineers, a more flexible approach emerged: the so-called hot spot stress method. Particularly suitable for welded structures, it assesses the influence of the geometric discontinuity on the stress distribution. A stress extrapolation procedure is required to overcome the high stresses computed by FEA at the notch. Normal stresses at the surface, perpendicular to the weld, are to be used. For curved welds and surfaces, e.g. in case of tubular joints, these stresses are complex to determine from FEA output. A method that automates the hot spot stress calculation as a post processor to FEA output would facilitate the process. Such an automated subroutine would further enable a study on the finite element modelling aspects, including the use of shell and solid elements and the inclusion of the weld profile, in relation to conducted fatigue experiments. This report describes the development of a subroutine as post-processor for FEA output to calculate the hot spot stresses. Read out points for stress extrapolation are located independently of the finite element mesh. In addition, for each weld node the local coordinate system is to be determined and subsequently the corresponding stress transformation is to be performed. Equivalent stresses at the read out points are determined by means of interpolation from a triangle-shaped plane formed by three element nodes. Cut-out specimens of an orthotropic bridge deck were tested against fatigue and strain gauges were used to measure the strains. This report shows that FEA stresses corresponded well with the measurements. Inclusion of the weld profile is important; ignoring the weld underestimates the stress levels by 10%. Use of shell elements resulted in 4% lower nominal stresses than solid elements, however in the weld region only the solid element model accurately reproduced the stress distribution. An alternative method for the fatigue classification described in EC3 is proposed, which represents better the observed stress levels for the crack initiation point at the weld root. Finally, this report considers stress concentration factors of tubular joints as recommended by CIDECT design guide. The FEA results are compared to the SCFs from parametric formulae. Good correspondence was found between the FEA and CIDECT results. A validation analysis of the boundary and loading conditions was performed, from comparing the joint model with the entire truss structure for shell elements. A correction function is derived to cover the small observed differences. Additionally, strain gauge measurements from experiments on large scale tubular joints in a marine environment are compared to the FEA results. Good correspondence was found between the measured and numerically computed strains. For solid element models, the weld leg size is found to be important for the hot spot stresses; each millimetre shift of the weld toe affected the stress levels by 3%. A characteristic SN curve was derived from the experiments. Fatigue assessment by means of the FEA hot spot stress in combination with the corresponding DNV SN-curve was found to be more conservative.