| About this Abstract |
| Meeting |
MS&T23: Materials Science & Technology
|
| Symposium
|
Additive Manufacturing Modeling, Simulation, and Machine Learning: Microstructure, Mechanics, and Process
|
| Presentation Title |
Physics-constrained, Inverse Design of High-temperature Strength Printable Aluminum Alloys with Low Cost and CO2 Emissions for High Demand Industries |
| Author(s) |
Benjamin M. Glaser, S. Mohadeseh Taheri-Mousavi |
| On-Site Speaker (Planned) |
Benjamin M. Glaser |
| Abstract Scope |
Printable high-temperature aluminum alloys can be used in industries such as automotive, provided that their cost and sustainability metrics match the requirements of large-scale production. These metrics add additional constraints besides printability for the design of these alloys beyond typical mechanical performance. We recently designed a record high-temperature strength printable Al alloy from the Al-Ni-Er-Zr-Y-Yb system and validated its performance in experiments. To adapt this design for high-demand industries, on data from CALPHAD-based ICME calculations, we applied various unsupervised machine learning techniques and Bayesian optimization to efficiently explore high dimensional compositional space and optimize the complex objective functions. We discovered that changing the composition enables cost reductions of 30-50 percent while sacrificing as little as 5-10 percent of the performance. This is accomplished by reducing Er concentrations, due to its prohibitively high cost and CO2 emissions, and balancing with increased concentrations of lower cost elements. Several scenarios will be discussed. |