About this Abstract |
| Meeting |
MS&T26: Materials Science & Technology
|
| Symposium
|
Additive Manufacturing Modeling, Simulation, and Machine Learning: Microstructure, Mechanics, and Process
|
| Presentation Title |
Quantitative Validation Methodologies for Physics-Based Microstructure Prediction Models in Metal Additive Manufacturing |
| Author(s) |
Arulmurugan Senthilnathan, Sankaran Mahadevan |
| On-Site Speaker (Planned) |
Arulmurugan Senthilnathan |
| Abstract Scope |
Mechanical properties and performance of a material are predicted using micromechanical models that require microstructure information. In metal additive manufacturing (MAM), experimental identification of process conditions that provide the desired microstructure is expensive; therefore, physics-based MAM process and post-process models are developed to predict the microstructure. However, the assumptions and approximations in the physics-based model form propagate to the predicted microstructures. Currently, quantitative validation methods for physics-based process and post-process models in MAM consider only accuracy in a few microstructural features. This work proposes two quantitative validation methods that estimate both accuracy and precision of physics-based process and post-process models, considering the MAM process variability in the observed microstructure. In the proposed validation methods, probabilistic metrics are defined to compare the predicted and observed microstructural feature variability. Quantitative validation methods using probabilistic metrics pave the way for systematic uncertainty aggregation and support model-assisted qualification and certification in MAM. |