| About this Abstract |
| Meeting |
Materials Science & Technology 2020
|
| Symposium
|
High Temperature Corrosion and Degradation of Structural Materials
|
| Presentation Title |
Machine Learning and Data Analytics for Accelerating High-temperature, Corrosion-resistant Materials Design |
| Author(s) |
Xuesong Fan, Baldur Steingrimsson, Anand Kulkarni, Peter K. Liaw |
| On-Site Speaker (Planned) |
Peter K. Liaw |
| Abstract Scope |
We will introduce machine learning (ML) models capable of detecting patterns and characteristic trends, and evolving distinguishing characteristics between calcium-magnesium-alumino-silicate (CMAS) and calcium sulfate (CaSO4) hot corrosion attacks, with and without the influence of sea salt, for the purpose of developing coatings resistant to CMAS and calcium sulfate hot corrosion. To this end, we will present a unified feature list, one that captures inputs and outputs describing both CMAS and calcium sulfate attacks, a joint optimization scheme, for identifying a combination of thermal barrier coating and base alloy with good resistance to both corrosion mechanisms, and canonical component analysis, for evolving the distinguishing characteristics. We recommend selecting a ML or data analysis technique suitable for the application at hand and the data available. We recognize close linkage between the feature data extracted and the underlying physics-based models (refer to our patent application “Machine Learning to Accelerate Alloy Design”, No. 16/782,829). |