About this Abstract |
| Meeting |
MS&T25: Materials Science & Technology
|
| Symposium
|
Additive Manufacturing Modeling, Simulation, and Machine Learning: Microstructure, Mechanics, and Process
|
| Presentation Title |
Probabilistic Metrics for Validation of Grain Growth Models |
| Author(s) |
Arulmurugan Senthilnathan, Pranav Karve, Sankaran Mahadevan |
| On-Site Speaker (Planned) |
Arulmurugan Senthilnathan |
| Abstract Scope |
Micromechanical models require microstructure data to predict the performance and properties of a material. Experimentally obtaining the microstructure is expensive; therefore, computational models are developed to predict the microstructure by describing the grain growth phenomenon in the form of governing and constitutive equations. However, uncertainty and numerical errors in the model form propagates to the predicted microstructure features. This work presents probabilistic metrics to validate the grain growth models. Phase-field (PF) and cellular automata (CA) models are used to predict microstructure. Uncertainties in the PF and CA model-predicted microstructural features (e.g., grain shape) are first quantified through descriptors (e.g., sphericity). Various probabilistic validation metrics are then defined to compare the descriptors of predicted and experimental microstructures. Quantitative methods for model validation using probabilistic metrics pave the way for systematic uncertainty aggregation in the prediction of mechanical properties and part performance, and support model-assisted qualification and certification of additively manufactured products. |