PSI - Issue 28

Adrian Loghin et al. / Procedia Structural Integrity 28 (2020) 2304–2311 A. Loghin et al. / Structural Integrity Procedia 00 (2020) 000–000

2310

7

(a) DOE28

(b) DOE33

Fig. 8: Surrogate model verification: response surface model crack path prediction against 3D explicit FEA.

The runtime savings are very significant when the surrogate models are employed. For the example provided herein, the 3D explicit deterministic simulation is performed in about 90 minutes while for the three surrogate models the cumulative runtime is 0.14 seconds. These verified surrogate models for the crack path and the K I values can be conveniently combined to set up a crack propagation model that can account for hole location uncertainly as well as material fatigue crack growth rate properties. This additional e ff ort will be a topic of a future study.

3. Conclusions

RBF response surface modeling trained on accurate 3D FEA representations provides an e ffi cient simulation pro cess to perform probabilistic fatigue crack growth assessments. An experimental measurement which serves also as a validation reference, is used as a demonstration for the development. Deviation from nominal geometry is considered as the source of crack path variability and a space filling algorithm (LHS) is used to identify o ff -nominal representa tions. 3D fatigue crack growth simulations are performed for all instances to provide accurate crack path and mode I stress intensity factor solutions for calibrating and verifying three response surface models. The coupled modeling procedure, 3D FEA and RBF response surface, provides an accuracy-runtime trade-o ff that extends its application to component level probabilistic life assessment.