Inverse analysis of the butterfly effect in dendrite precipitates using machine learning

2020 Fiscal Year Final Research Report

• Project/Area Number: 19K22117
• Research Category: Grant-in-Aid for Challenging Research (Exploratory)
• Allocation Type: Multi-year Fund
• Review Section: Medium-sized Section 28:Nano/micro science and related fields
• Research Institution: Tokyo University of Science
• Principal Investigator: Kotsugi Masato 東京理科大学, 基礎工学部材料工学科, 准教授 (60397990)
• Co-Investigator(Kenkyū-buntansha): 橋爪 洋一郎 東京理科大学, 理学部第一部応用物理学科, 講師 (50711610)
• Project Period (FY): 2019-06-28 – 2021-03-31
• Keywords: パーシステントホモロジー / 機械学習 / 金属組織

Outline of Final Research Achievements

We developed an automated analysis method for predicting physical property parameters from image data of dendrite precipitates and spinodal decomposition using the topological concept of "persistent homology," First, image data of dendrite precipitates and spinodal decomposition were generated using the phase-field method. Next, persistent homology was used to extract the features of the shape of microstructures. Then, principal component analysis was used for dimensionality reduction, and the changes in the data were visualized in low-dimensional space. The results suggest that it is possible to estimate various physical property parameters such as development time, anisotropy parameter, gradient energy coefficient, and total energy in metallographic formation.

Free Research Field

顕微分光解析、マテリアルズインフォマティクス

Academic Significance and Societal Importance of the Research Achievements

パーシステントホモロジーと呼ばれる位相幾何学の概念と教師無し機械学習を組み合わせて、デンドライト組織などの複雑な金属組織から、自動的に物性パラメータ（発展時間、異方性パラメータ、勾配エネルギー係数、全エネルギー）を逆解析するための枠組みを新しく開発した。