Influence of preprocessing on the fatigue life prediction of scanned weld topologies

• 1089 fatigue life predictions were calculated to quantify the influence of NURBS quantity and smoothening in preprocessing on fatigue life prediction using six different FDP. • Deviation of fatigue life prediction reduced from 4 down to 2 and standard deviation was decreased by 50 %. • Gradient-based FDP are less affected by the preprocessing process compared to common FDP. • Radii > 0.5 mm could only be achieved by manually adjusting the NURBS nodes and patches. This work quantifies the influence of preprocessing scanned weld topologies on the accuracy and scatter of fatigue life prediction. Based on 13 welded fatigue specimens, 99 weld topologies were created using different settings in preprocessing. Mesh convergence was achieved in FEA using elastic plastic material models. Common fatigue damage parameters (FDP), such as von Mises (vM), Smith-Watson-Topper (SWT), Fatemi-Socie (FS), and their gradient-based extensions were used to create 1089 fatigue life predictions for comparison with experimental fatigue life. The deviation of fatigue life prediction is defined as the logarithmic fraction of experimental and predicted fatigue life. This paper examines the influence of different approaches when scan data is transformed to a solid in FEA. Increasing the number of Non-Uniform Rational B-Splines (NURBS) patches in preprocessing to 250 resulted in smaller radii r (in mm) {r| 0.05 ≤ r ≤ 0.19}. Reducing this number to 10 increased the interval to {r| 0.07 ≤ r ≤ 0.44}. With a reduction of NURBS nodes the deviation of common FDP of ca. 3 decreased by a third. Gradient based FDP were not affected to a comparable magnitude with a deviation closer to null. By manual displacement of NURBS nodes the radii increased to {r| 0.76 ≤ r ≤ 2.24}. Here, deviation of common FDP decreased over 50 % while gradient based extension of SWT resulted in the best deviation of 0.042. It is evident that the user-influence on the preprocessing stage, whether conscious or unconscious, has a substantial impact on the fatigue life prediction.