PSI - Issue 75

Thomas Constant et al. / Procedia Structural Integrity 75 (2025) 660–676 Author name / Structural Integrity Procedia 00 (2025) 000–000

670

11

Fig. 4. Illustration of a realization of the excitation profile for each load configuration.

The folded bracket area is then considered as the critical region of the system, i.e. the location where cracks are most likely to initiate. The random variables related to the system here include the stress concentration factor k t , the fatigue material parameters σ ′ f and ϵ ′ f , and the 0.2% yield strength R p 0 . 2 . These variables are grouped into the random vector X B whose properties are summarized in table 2. All random variables are assumed to be mutually independent. Their probabilistic distributions come either from Zhang et al. (2024) or industrial feedback.

Table 2. Probabilistic setting of X B . Random Variable

Distribution

Mean

Standard Deviation

R p 0 . 2

Log-normal Log-normal Log-normal

253

5

ϵ ′ f

0 . 976 1601

0 . 0488

σ ′ f

80

Normal

3

0 . 3

k t

The probability of failure is now computed with d th arbitrarily fixed at d th = 0 . 0025. Table 3 compares three Active Kriging approaches, briefly described below. The first method enriches all points selected by the U -learning function, see equation 16, requiring n c calls per iteration. This strategy, AK-MCS with Multi points Enrichment (AK-MCS-ME), serves as the baseline for comparison. The second method applies the critical configuration approach but does not use the AK-SM imputation criterion. It works as the AK-MCS method, described in Appendix B. This approach is termed AK-MCS with Critical Configuration Approach (AK-MCS-CCA). Finally, the third method is the proposed AK-SM-CCA.