|About this Abstract
||MS&T23: Materials Science & Technology
||Leveraging Integrated Computational Materials Engineering for High-fidelity Physics-based and Machine Learning Models
||Hybrid Simulation Method Based on Molecular Dynamics and Machine Learning to Improve Property Prediction with Lower Computational Cost in Complex System
||Owais Ahmad, Somnath Bhowmick
|On-Site Speaker (Planned)
In materials science, simulations of molecular dynamics are frequently used to examine the atomic and molecular properties of complex systems.
Yet, these simulations are computationally intensive, especially for m ulticomponent and high-entropy systems with significantly increased interactions and degrees of freedom. In recent years, machine learning algorithms have proven tremendous potential for predicting molecular dynamics simulation properties and reducing their computing cost.
In this study machine learning is used to simulate the molecular dynamics of multicomponent, high-entropy systems. We describe the Molecular Dynamics-Machine Learning Hybrid Simulation technique for forecasting atomic locations and velocities, which is then used for property computation to reduce processing costs. Overall, we conclude that machine learning may be an effective strategy for predicting features and reducing the computational cost particularly for complex systems such as multicomponent and high entropy materials.