3-D Phase-Field Simulations to machine-learn 3-D Features from 2-D Microstructure Images

Yuxun Jiang1*, Muhammad Adil Ali1, Irina Roslyakova1*2Ingo Steinbachi

1 Interdisciplinary Centre for Advanced Materials Simulation (ICAMS), Ruhr University
Bochum, 44801 Bochum, GermanyBochum, 44801 Bochum, Germany

2 Matplus GmbH, 42103 Wuppertal, Gennany

EXTENDED ABSTRACT: Estimating 3D stereological information from 2D micrographs has always been a challenge in general. Machine learning has emerged as a promising tool in tackling such complicated tasks in materials science [1]. However, lacking sufficient amount of data impedes the application of machine learning algorithms. The present study aims to establish effective machine learning models driven by Phase-Field simulations, in particular for the description and prediction of creep behavior of Ni-based superalloys. To establish the microstructure feature evaluation, thousands of 2D sections taken from 3D Phase-Field simulations were evaluated and served as training sets to retrieve the 3-D information [2, 3]. In principle, the 2D sections contain full information of characteristic microstructural features as well as information about their evolution in dependence of processing and testing conditions. The trained machine was then applied to extract 3-D information from real 2-D experimental micrographs. The results demonstrate that the experimental images contain a bias, and that this bias can be corrected by our approach. Hence the present work helps to strengthen the analysis of 2-D microstructure images significantly and serves to make predictions on the creep behavior ofNi-based superalloys.

This work has been done in the framework of the Collaborative Research Centre/Transregio Transfer project T2. The 3-D Phase-Field simulations were conducted via OpenPhase [4].

Keywords: 3D Phase-Field simulation; 2D creep microstructure analysis; Machine learning


[1] Sung Wook Kim et al., Scientific reports, 11(1), (2021), 1-14.

⑵ Mansur Ahmed et aL, Modelling Simul. Mater. Sci. Eng., 29, (2021), 055012

[1]   Uchechukwu Nwachukwu et al., Modelling Simul. Mater. Sci. Eng., 30, (2022), 025009

[2]   https://openphase-solutions.com/

Brief Introduction of Speaker
Ingo Steinbachi

Prof Dr. Ingo Steinbach is a pioneer of the phase-field theory. His Multi-Phase-Field approach, with contributions from many other scientists worldwide, forms a standard for multi-phase application in metallurgy and other fields of materials research.
Yuxun Jiang is now a PhD candidate at ICAMS. He is focusing on establishing a creep indicator of Ni-base superalloys which takes microstructure parameters and loading conditions as model inputs, by means of material informatics and physics base modelling.