About this Abstract |
Meeting |
2024 TMS Annual Meeting & Exhibition
|
Symposium
|
Algorithm Development in Materials Science and Engineering
|
Presentation Title |
Machine Learning-guided MEAM Interatomic Potential Development for Predicting Melting Point Properties |
Author(s) |
Sepideh Kavousi, Mohsen Asle Zaeem |
On-Site Speaker (Planned) |
Sepideh Kavousi |
Abstract Scope |
Due to the computational complexities of lengthy simulations and the necessity of feedback loops to optimize the solid-liquid properties, the decisions on modifying interatomic potentials are made based on the developer’s prior experience and knowledge. Our preliminary investigations have identified four parameters in MEAM interatomic potential that have the highest impact on the melting point of pure elements. We will perform molecular dynamics simulations to investigate the interplay of these parameters on the melting point, cohesive energy, and equilibrium crystal structure of several elements (available on the NIST repository). This data will be used as a training/testing dataset for a machine learning (ML) model. The artificial neural network (ANN) regression model will correlate the variations of the target potential parameters to minimize the loss function. This ML framework will change the manual potential parameter tuning to an accelerated ML-guided framework for development of high-temperature- interatomic potentials. |
Proceedings Inclusion? |
Planned: |
Keywords |
Machine Learning, Solidification, Computational Materials Science & Engineering |