PSI - Issue 57

Yuki Ono et al. / Procedia Structural Integrity 57 (2024) 290–297 Author name / Structural Integrity Procedia 00 (2019) 000 – 000

293

4

kinematic hardening parameters, so-called Voce- Chaboche’s (VC) parameters, were employed. Table 2 provides the VC parameters for each material section. These parameters were obtained by calibrating experimental cyclic hysteresis loops with a unique optimization algorithm called RESSPyLab [de Castro e Sousa et al. (2020)]. Fatigue test data from Mikkola et al. (2016) were utilized to calibrate VC parameters. The calibrations considered the first five hys teresis loops up to reaching stability. Young’s modulus was pre -defined from the first tensile load of the test data [Mikkola et al. 2016] and not considered an optimization parameter, which better estimates crack initiation life for high-strength steel welded joints when little plasticity is involved [Petry et al. 2022]. Fig. 4 shows two types of HFMI geometry models. This study chose a representative HFMI geometry close to the average value among the measured profiles in Section 2.2. The simplified geometry model has the smooth geometries of the HFMI groove, which are 1.91 mm- radius (ρ), 0.15 mm -depth (d), and 2.8 mm-width (w). In contrast, the actual geometry model directly incorporates the corresponding measured profile with surface imperfection. The weld size of both models was the weld leg length of 4.75 mm for the base plate side ( l x ), 6.55 mm for the gusset plate side ( l y ), and the weld angle ( θ ) of 36 degrees. This study investigated the stress-strain response at/near three target locations at weld toe, A, B, and C. A is the reference location corresponding to the HFMI groove bottom. B and C are the locations that can be possible crack initiation sites, according to the study by Ono et al. (2022), and preliminary elastic stress analysis shows higher stress concentrations compared to other surfaces. To model local stress-strain behaviour sufficiently in local geometry imperfection, the element siz e around three locations was set to about 0.5 μm for the actual geometry models and then the size gradually increased towards the other global parts. In the case on simplified model with smoother geometry, the element size was about 10 μm based on the resu lts convergence study. Initial residual stress distributions implemented in the FE model are shown in Fig. 5. These residual stress distributions were based on the experimental measurement for the stress states after welding and HFMI treatment and represented by means of a predefined temperature field in Abaqus; see Ono et al. (2022). The changes in temperature through the plate thickness were applied at the nodes for three paths at x = -3.5 mm, 0 mm, and 3.5 mm, which created

Ϭ 0.0 0.2 0.4 0.6 0.8 1.0 -0.2 -1.0 -0.8 -0.6 -0.4 0 Ͳϭ ͲϬ͘ϴ ͲϬ͘ϲ ͲϬ͘ϰ ͲϬ͘Ϯ Ϭ Ϭ͘Ϯ Ϭ͘ϰ Ϭ͘ϲ Ϭ͘ϴ ϭ Residual stress σ RS / f y

(a) Simplified geometry model ρ = 1.91 mm, = 36 deg, w = 2.80 mm d = 0.15 mm, l y = 4.75 mm, l x = 6.55 mm

Ϭ͘ ϱ

0.5

Stress in x -direction (MPa)

1.0

ϭ

0.5 mm

− 500 − 310 − 120 − 405 − 215 355 165 70 − 2.5 450 260

1.5

Depth y (mm)

ϭ͘ ϱ

2.0

(0,0,0)

Ϯ

z

x

2.5

Ϯ͘ ϱ

y

(b) Actual geometry model

Simplified geometry model Actual geometry model

3.0

3.0 mm

ϯ

Residual stress σ RS / f y

Ͳϭ -1.0

ͲϬ͘ϴ -0.8

ͲϬ͘ϲ -0.6 -0.4 ͲϬ͘ϰ

ͲϬ͘Ϯ -0.2

Ϭ 0.0

0.5 mm

0

Ϭ

Ϭ͘ ϬϮ 0.1 0.08 0.06 0.04 0.02 Ϭ͘ Ϭϰ Ϭ͘ Ϭϲ Ϭ͘ Ϭϴ Ϭ͘ ϭ

Depth y (mm)

: Target location

Fig. 5 Implemented initial residual stress distribution in FE model

Fig. 4 HFMI geometry models

S , Nominal stress

S , Nominal stress

S , Nominal stress

・・・ ・・・

・・・ ・・・

・・・ ・・・

Single high-peak load cycle

20 load cycles

Single high-peak load cycle

20 load cycles

20 load cycles

(b) Loading case 2

(c) Loading case 3

(a) Loading case 1

Fig. 6 Applied load cycles for residual stress relaxation assessment