About this Abstract |
Meeting |
2023 TMS Annual Meeting & Exhibition
|
Symposium
|
Frontiers in Solidification: An MPMD Symposium Honoring Jonathan A. Dantzig
|
Presentation Title |
An Integrated Machine Learning and Phase-field Approach for Accurate Prediction of Dendritic Arm Spacing |
Author(s) |
Sepideh Kavousi, Mohsen Asle Zaeem |
On-Site Speaker (Planned) |
Mohsen Asle Zaeem |
Abstract Scope |
This study describes the exploration of data science techniques to predict primary dendritic arm spacing (PDAS) for solidification of different material systems. The theoretical models, which relate the PDAS to the material properties and solidification conditions—pulling velocity and temperature gradient—do not match the relationships obtained from experimental and computational studies. We present a machine learning model for predicting PDAS via performing atomistic-informed phase field modeling of solidification for various compositions of Al-Cu, Ti-Ni, and Mg-Al alloys. A three-layer hidden feedforward artificial neural network (ANN) model is developed and trained to mimic the arm spacing associated with the training data set. To produce a new global relation for PDAS, we used material systems with different crystal structures (FCC, BCC, and HCP) and accounted for slow, moderate, and rapid solidification rates. |
Proceedings Inclusion? |
Planned: |
Keywords |
Solidification, Computational Materials Science & Engineering, Machine Learning |