Modeling Atomistic Dynamic Fracture Mechanisms Using a Progressive Transformer Diffusion Model.

Area of Science:

• Materials Science
• Computational Materials Science
• Mechanical Engineering

Background:

• Dynamic fracture is crucial for understanding material failure over time at the atomic scale.
• Predicting brittle material failure requires sophisticated analysis of crack dynamics and initiation.

Purpose of the Study:

• To develop and validate a machine learning model for describing dynamic fracture in brittle materials.
• To assess the model's ability to capture crack dynamics, instabilities, and initiation mechanisms.

Main Methods:

• Utilized an atomistically derived progressive transformer diffusion machine learning model.
• Trained the model on a limited dataset of atomistic simulations.
• Validated the model on diverse cases, including complex geometries and generative neural network-designed microstructures.

Main Results:

• The machine learning model effectively described dynamic fracture, including crack dynamics and instabilities.
• The model demonstrated strong generalization capabilities beyond the training data.
• Accurate predictions were achieved for validation cases with distinct geometric features and microstructures.

Conclusions:

• The developed machine learning model provides a rapid and effective tool for assessing dynamic fracture mechanisms.
• The model's generalization allows for analysis of complex geometries and novel material designs.
• This approach advances the understanding and prediction of material failure.