PSI - Issue 19

Théophane Vanlemmens et al. / Procedia Structural Integrity 19 (2019) 610–616

613

Vanlemmens, Elbel, Meneghetti/ Structural Integrity Procedia 00 (2019) 000 – 000

4

The performances have been compared as well, and it was concluded that the overall mean time needed to implement the PSM on those geometries is 3x less than the one needed for the R1 concept. The overall time is composed of the preparation time (geometry + mesh), the meshing time and the computation time.

3.3. Tests on industrial geometries

First the original formulation using hexahedral elements has been applied to the industrial components. The results were good, but for complex geometries, the overall time saving was not as big as what was expected with the experience of the simple geometries. This was due to the preparation time for the meshing. The method requires some rules for the mesh, illustrated at Fig. 4: 4 elements are needed at the root and two elements at the weld toe. It is therefore a little bit tricky and time consuming (several remeshings can be necessary) to achieve it.

Fig. 4: Illustration of the required mesh when using hexahedral elements

The second formulation suggested by Meneghetti in [3] uses tetrahedral elements (solid 187). In this case, no additional rule is necessary to achieve on the mesh. This is the main advantage of using those elements: the meshing time is shorter and no remeshing is needed.

Fig. 5: Example of a tetrahedral mesh for PSM

Two different large structural parts from crawler excavators have been modelled using this approach, an undercarriage and a boom. The whole welds have not been modelled but only some selected welds that need a local stress approach. The results of the comparison between R1 concept and PSM for all the evaluated zones have been put together in the graphic from Fig. 6 . Visually there is a very good correlation between both methods although PSM seems to be systematically more pessimistic than R1 concept. The mean value of the ratio between the number of cycles with PSM and the number of cycles with R1 concept is 0.47. The coefficient of variation is 27%, which confirms the small dispersion of the results. The reason for the low value from this ratio is not clear yet, but since it seems to be systematic, it could be corrected by adapting the FAT class.