PSI - Issue 13

Michihiro Kunigita et al. / Procedia Structural Integrity 13 (2018) 198–203 Kunigita / Structural Integrity Procedia 00 (2018) 000 – 000 Value of was treated as an adjustable parameter in the present study. The local fracture stress, , can be evaluated numerically as a probability distribution from the microstructural parameters including the probability distributions of MA particle thickness, inter-MA particle distance and GB ferrite thickness, together with the number density of MA particles, volume fraction of GB ferrite and yield strength. 2.2. Stress strain analysis It is necessary to evaluate the history of stress and strain exerted on each volume element for calculating the probability of fracture. For this purpose, dynamic elastic-plastic finite-element analysis was conducted using a commercial software, ABAQUS 6.14-1. Because strain-rate near the notch-tip of a Charpy impact specimen reaches as high as about 10 3 s -1 , change in yield strength with strain-rate and temperature was formulated using the temperature strain-rate parameter, R = T ln(A/ ̇ ), where T [K] is temperature and ̇[ −1 ] is strain-rate. Temperature rise by deformation heating was also considered. Figure 4 shows an example of the distribution of maximum principal stress ahead of the Charpy specimen notch-tip. 201 4

Fig. 4. Maximum principal stress distribution ahead of notch-tip, normalized by yield strength (429MPa). Numbers indicate absorbed energy.

3. Experiment

Charpy impact tests were conducted to validate the model proposed by the present study. Tested material was laboratory melt steel having chemical composition of 0.07%C-0.29%Si-1.31%Mn-0.20%Cu-0.59%Ni-0.21%Cr 0.040%V-0.058%Al-0.003%N (mass%). Samples were subjected to simulated HAZ thermal cycle using an induction heating machine. Peak temperature was 1,400 ℃ and cooling rate from 1,000 ℃ down to 100 ℃ (CR) was 1, 3, 10 and 30 ℃ /s. The simulated HAZ samples were subjected to Charpy impact test. Microstructures of the polished and etched (2% natal) samples were observed by SEM, see Fig.5 for CR: 1 ℃ /s. Statistical distributions of MA particle thickness, inter-MA particle distance and GB ferrite thickness (CR: 1 ℃ /s only) were obtained. Volume fraction of GB ferrite (CR: 1 ℃ /s) was calculated as 2.1%. 4. Model prediction Figure 6 shows predicted Charpy impact absorbed energy (50J) transition curves for the samples CR: 1, 3 and 10 ℃ /s, together with the experimental values. Volume element was assumed as a 50 μm cube and total number of the volume elements was 32,000. Because the present model is stochastic, the absorbed energy is calculated as probability distribution. Therefore, the transition curves are displayed as 5%, 50% and 95% cumulative probability curves. In the calculation, = 0.01 was assumed. As an order of magnitude, this value agrees with the experimentally determined ones (Kawata et al., 2016). Also, value of , corresponding to the effective surface energy of 40J/m 2 at 0 ℃ , was assumed. It is noted that the experimentally determined value of the effective surface energy at 0 ℃ was 50J/m 2 (Kawata et al, 2018). It is clearly seen that the predicted transition temperatures agree well with the experiment. Interesting point in the prediction is that the same values for and can be used for all the samples although they have different strengths and microstructures. Additional calculation was conducted for the sample CR: 1 ℃ /s, in which GB ferrite was ignored. As shown by the green plot in Fig.6, the transition temperature shifted to lower temperature