PSI - Issue 42

Chahboub Yassine et al. / Procedia Structural Integrity 42 (2022) 1025–1032 Author name / Structural Integrity Procedia 00 (2019) 000 – 000

1029

5

Fig. 2. Through wall fracture

Only a quarter of the pipe was mesh to reduce the computation time required due to the high number of elements. Fig.3 (a)

a

b

Fig. 3 (a) Modelling of the mock-up FP (b) mesh size near the crack tip

The GTN model is only considered in a condensed region for time consumption Fig. 3 (b). The dimensions of the Mock-up FP1 are mentioned in the study published by Moinereau.D et al. (2014). We decided to model just a quarter of the mock-up to reduce computation time thanks to the axisymmetry. The FEM model has 104193 nodes and 248038 elements in total. The mesh size in front of the pre-crack tip is (0.125 mm x 0.0625 mm), and the mesh is made up of axisymmetric quadratic elements with eight nodes. We perform a FEM simulation using the GTN parameters only on the crack propagation area, which is the most sensitive part of the model. We decided to start using the direct method, which combines the experiment and finite element results. Because of the huge number of elements in the model, one simulation took around two days of computing. So to get the correct set of GTN parameters, we had to do around 15 simulations; with basic math calculation, it took around 30 days of computation. The ferritic mock-up is modeled to have elastic behavior (E=203 GPa). In Fig.4 (Force vs. Crack opening displacement), the GTN model predicts the pipeline's failure very well. It leads us to the right prediction of the crack propagation, and the maximum initiation load is also well predicted. We can conclude from the curve analysis that the onset of ductile tearing is correctly predicted at 2255 kN (simulation) versus the 2240 kN reported in the experimental data. The maximum projected load is 2741.624 kN, which will occur at a CMOD of 22 mm. The experiment's highest force is 2808 kN, which occurs at a CMOD of 23.42 mm. After using the direct method, we can notice that the GTN model was a strong tool and very practical. However, it took us 30 days to find the right set of parameters, hence the need to find an alternative, complementary way to predict the failure of pipes in a relatively short time. For this reason, we have integrated artificial intelligence into the simulation process, and we were able to reduce the duration of the 30 days to 6 hours. The next section will describe in detail the work done and the optimization process used to predict the failure of the pipe in a short time.