87P5-13A Tandem Bayesian Model for Probabilistic Search to Improve Weld Joint Creep Property Hitoshi Izuno1, M. Demura1, M. Yamazaki1, S. Minamoto1, J. Sakurai1, K. Nagata1, Y. Mototake2, D. Abe3,and K. Torigata3. 1 Research Network and Facility Services Division, National Institute for Materials Science (NIMS)2 Hitotsubashi University3 IHI Corporation The creep rupture of welded joints of heat-resistant steel initiated at the fine-grained heat-affected zone (HAZ) and its life is determined by the HAZ shape, which widely varies depending on the thermal history. This work aims at improving the creep rupture life of the weld joint through controlling the HAZ shape with optimizing the welding condition. We first constructed a forward workflow to predict the creep rupture life for arbitrary welding conditions by coupling the heat conduction analysis and subsequent creep damage analysis [1]. Then, using the workflow as a data source, we then developed a tandem Bayesian model that probabilistically connects two surrogate models: one is to predict the HAZ shape factors from the welding conditions; and the other is to predict the creep rupture life from the HAZ shape factors [2]. We finally proposed a framework where the model is sequentially updated by testing the candidates selected based on the probability of updating the longest rupture life obtained so far. The proposed framework has succeeded in improving the rupture life by 12% over the initial one just with the very small number of calculations (137 cases) compared with the huge design space of 78. [1] D. Abe, K. Torigata, H. Izuno, M. Demura, Welding in the World, 68, 1297–1311 (2024) [2] H. Izuno, M. Demura, et al. Welding in the World, 68, 1313-1332 (2024) P5-14First-principles Calculation of Electron-phonon Interaction Atsushi Togo1, Laurent Chaput2, Henrique Miranda3, Manuel Engel3, Martijn Marsman4, and Georg Kresse4 1 Center for Basic Research on Materials, National Institute for Materials Science (NIMS)2 LEMTA, University of Lorraine3 VASP Software GmbH4 Faculty of Physics and Center for Computational Materials Physics, University of Vienna The electron-phonon interaction in condensed matter materials is related to properties such as the thermoelectric effect and conventional superconductivity. These properties can be predicted through first-principles calculations. There are several approaches to computing the electron-phonon interaction from first-principles calculations. We have developed a supercell approach within the projector augmented wave (PAW) method [1, 2]. This approach simplifies the computational procedure and expects to enhance robustness in systematic calculations against a variety of crystals. This method has been implemented in two software packages, the VASP and phelel codes. We overview the theory and methodology for computing the electron-phonon interaction and its related properties. [1] L. Chaput, A. Togo, I. Tanaka., Phys. Rev. B, 100, 174304 (2019). [2] M. Engel, H. Miranda, L. Chaput, A. Togo, C. Verdi, M. Marsman, and G. Kresse, Phys. Rev. B, 106, 094316 (2022)
元のページ ../index.html#87