Since 2022, 3D generation AI (text-to-3D models) has emerged one after another, including Dream Fusion (Google), Magic 3D (NVIDIA), and Point-E (OpenAI). However, compared to natural language, images, and videos, there are fewer datasets for 3D shapes, and no datasets that include structural mechanical information have been proposed.

Nagoya University is currently conducting research 3D generation AI (parameter-to-3D model) based on data sets created through ultra-multiple case elastic-plastic analysis using the supercomputer "Fugaku," with the aim of applying this technology to gigacast structures, whose structural design is difficult to optimize.

In this keynote speech, the speaker discussed about the current status and future developments of AI for Science, as well as the latest trends in AI for Science, such as the development of AI-based models for scientific research, which is becoming active in Japan, the United States, and Europe, and the advancement of computational technology that combines simulation and AI.

・Mr. Nishiguchi, Graduate School of Engineering at Nagoya University

In recent years, the trend towards faster cycle times in production has resulted in increased loads on press molds, resulting in frequent damage to the outer periphery rib structure.

In this presentation, ADVENTURECluster was used to establish an impact analysis method to solve the damage to the PAD outer periphery rib structure.

In addition, the knowledge gained from this experience was used in designing new vehicle models and the repair of production mold damage, and the speaker also presented those cases.

・Mr. Hayashi, MITSUBISHI MOTORS CORPORATION

Nowadays, there is a demand for weight reduction to combat the increased vehicle weight caused by the installation of batteries in PHV, EV, as well as to improve the fuel efficiency of gasoline vehicles.

Nifco has attempted to reduce the weight of various parts, but since there is a trade-off with stiffness (hardness), they were searching for a method to reduce weight while maintaining stiffness.

In this presentation, Nifco presented a case that propose a shape achieving both high stiffness and light weight by using topology optimization.

・Mr. Akamatsu, Nifco Inc.

By using ADVENTURECluster, it is now possible to create a large-scale finite element model with tens of millions of elements from X-ray CT images and calculate the diffusion coefficient of the filler-containing resin based on the assumption of the resin basement.

Because this method can take into account for structures on the order of μm, so that in the future, it is expected to be used to explore the mechanical properties of materials with structures that are difficult to create mathematical model, such as long-fiber-containing resins.

・Mr. Aoki,Polyplastics Co., Ltd.

The Earthquake Disaster Mitigation Experimental Research Division of the National Research Institute for Earth Science and Disaster Resilience is developing elastic-plastic damage constitutive laws for materials such as steel and concrete and is conducting earthquake response analyses of architectural structures in order to predict damage to structures and develop methods for evaluating earthquake resistance.

In this presentation, the pre-processing module for creating detailed models of architectural structures, the reproducibility analysis of a full-scale building experiment at the full-scale 3D earthquake destruction test facility (E-Defense), and the results of an earthquake response analysis using observed waves from the Noto Peninsula earthquake were presented.

・Mr. Yamashita, National Research Institute for Earth Science and Disaster Resilience