PSI - Issue 28

Adrian Loghin et al. / Procedia Structural Integrity 28 (2020) 2304–2311

2307

4

A. Loghin et al. / Structural Integrity Procedia 00 (2020) 000–000 Table 1: Crack front mesh density and advancement size used in crack path sensitivity study.

Mesh ID

Number of corner nodes along crack front

Advancement size (mm)

Model#1 Model#2 Model#3 Model#4

30 40 50 90

0.15

0.2 0.1 0.1

Fig. 3: Comparison between experimental data and two 3D predictions using a di ff erent method to apply loading at the pin holes. Same mesh design was used in the two simulations and same advancement increment of 0.1 mm. Crack path is extracted along the front surface.

Fig. 4: Mesh sensitivity study using models defined in Table 1. Crack paths associated with Model#1, #2 and #4 overlap.

2.2. Response Surface Modeling

Response surface modeling (AKA surrogate modeling) is widely used in engineering applications to accelerate the model run-time speed within a subset of the problem domain. With the recent advance in machine learning al gorithms, various supervised learning models have emerged that tackle diverse sets of problems, most of which are freely accessible to the modeling community. In this study, a simple Radial Basis Function (RBF) based interpolation