| About this Abstract |
| Meeting |
2024 TMS Annual Meeting & Exhibition
|
| Symposium
|
2024 Technical Division Student Poster Contest
|
| Presentation Title |
SPG-26: A Comparison of the Generalizability of Machine Learning and Constitutive Modeling Approaches for the Prediction of Flow Stress |
| Author(s) |
Thomas McCarthy, Jubert Pasco, Clodualdo Aranas |
| On-Site Speaker (Planned) |
Thomas McCarthy |
| Abstract Scope |
To achieve an efficient manufacturing process, a complete understanding of the material's deformation behavior must be achieved. Once achieved, processing maps and computational simulations can be performed to determine suitable processing conditions for both formative and subtractive processes. Traditionally, this has been accomplished using phenomenological- and physical-based constitutive modeling; however, recently machine learning methods have demonstrated exceptional accuracy in predicting the material's deformation behavior. Despite this, limited research has compared the ability of constitutive modeling and machine learning approaches to estimate the flow stress under unknown conditions. Furthermore, many investigations applying machine learning solely focus on Artificial Neural Networks (ANN), while limited research explores alternative machine learning models. To address these limitations, the hot deformation behavior of a high entropy alloy (CoCrFeMnNi) is modeled using ANN, Gaussian Process Regression, modified Johnson-Cook, and modified Hensel-Spittel models. The generalizability of these models is evaluated and compared using a temperature-based cross-validation strategy. |
| Proceedings Inclusion? |
Undecided |
| Keywords |
Machine Learning, Modeling and Simulation, High-Entropy Alloys |