PSI - Issue 38

Jinchao Zhu et al. / Procedia Structural Integrity 38 (2022) 621–630 Author name / Structural Integrity Procedia 00 (2021) 000 – 000

625

5

average finite-element edge length is 0.5 mm outside of the weld toe region. A mesh refinement is performed at the weld toe with a mesh convergence study resulting in an average element edge length of 0.1 mm, see Fig. 3. The SCF is obtained by applying a nominal stress of 1 MPa.

Fig. 3. Two-dimensional plane strain finite-element model with mesh refinement at weld toe.

2.5. MC simulation to include influence of geometrical variabilities MC simulation is used to study the influence of stochastic variations of the weld geometry on the probability distribution of fatigue life. A total of 2000 data sets are randomly generated from lognormal distributions of leg length, toe angle and toe radius with mean values and standard deviations given in Section 2.1. The lognormal distribution is motivated by the positive sign of the three geometrical parameters. In the NS method with r ref = 1 mm and FAT 225, only the probability distributions of the leg length and toe angle are used since the toe radius is fixed to be 1 mm, see Section 2.2. For each of the 2000 randomly generated set of geometrical parameters, a 2D FEA is performed to compute the probability distribution of K t , see Section 1.1. Based on Eq. (4), the probability distribution of fatigue life is thereafter computed. It is noted that, since the number of MC simulations is very large, the width of the 75% confidence interval typically used in IIW recommendations is negligible. An overview of the MC simulation is shown in the flowchart in Fig. 4.

a

b

Fig. 4. Flowchart of MC simulation (a) NS method with r ref = 1 mm and FAT 225; (b) NS method with r ref = r actual + 1 mm and FAT 200

3. Results and discussions 3.1. Influence of weld convexity and concavity

In the NS methods, straight lines are often used to model the idealized weld geometry. However, concave and convex shape of welds are typically observed in cruciform joints (Mansour et al. (2019), Kainuma et al. (2006)). In