[11a-N107-1~11] 23.1 Joint Session N "Informatics"

△ [11a-N107-11] [Highlight]Deep Learning Virtual Experimentations for Materials and Process Informatics of Tangible Materials

〇Shun Muroga¹, Takashi Honda², Hideaki Nakajima¹, Kazufumi Kobashi¹, Taiyo Shimizu¹, Hiroshi Morita¹, Toshiya Okazaki¹, Kenji Hata¹ (1.AIST, 2.ADMAT)

Keywords:Deep Learning, Materials and Process Informatics, Virtual Experimentation

Artificial intelligence is an emerging frontier in material science to discover new materials with targeted properties by an artificial neural network. This approach has not been applicable to tangible materials because the complexities of their structures cannot be easily defined by atoms, chemical bonds, and periodic structures. In this study we propose a deep learning computational framework that can implement virtual experimentations on tangible materials where structural representations of the materials were created by conditional generative adversarial networks and prediction of properties using a convolutional neural network of generated structures. Proposed method was applied to the model tangible material of carbon nanotube film. The data generated by proposed framework can be used as a versatile database for material science, in analogous to databases of molecules and solids used in cheminformatics, as exemplified by investigations of the correlation between the electrical conductivity and specific surface area.