About this Abstract |
| Meeting |
2027 TMS Annual Meeting & Exhibition
|
| Symposium
|
Verification, Calibration, and Validation Approaches for Mechanical Modeling of Metals
|
| Presentation Title |
Multi-fidelity Gaussian process-based calibration of cyclic nitinol crystal plasticity simulations |
| Author(s) |
John A. Moore, Truman Russell, Anahita Alipanahi, Brenden Carty, Jacob Rusch |
| On-Site Speaker (Planned) |
John A. Moore |
| Abstract Scope |
Computational crystal plasticity models are a valuable tool for microstructure-sensitive modeling of metals. However, the micromechanics-based formulation that endows them with strong predictive ability comes with the cost of many calibration parameters. Likewise, as compared to phenomenological measurements for macroscale models, it is difficult in these systems to measure microscale material properties directly and validate with microscale response. This challenge is even larger for transforming materials (e.g., nitinol) that are subject to cyclic loads---as the transformation and cyclic properties need determined. This work presents a novel multi-fidelity Gaussian process (GP)-based approach to determining these calibration properties for nitinol. The multi-fidelity approach uses low-fidelity simulations to find optima cheaply and explores the parameter space using Bayesian optimization. Heteroscedastic noise reduces overfit in the GP. Genetic algorithms are also used for this exploration and high energy diffraction microscopy (HEDM) of an austenitic nitinol is used for crystal scale validation. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Modeling and Simulation, Phase Transformations, ICME |