About this Abstract |
| Meeting |
2026 TMS Annual Meeting & Exhibition
|
| Symposium
|
Novel Strategies for Rapid Acquisition and Processing of Large Datasets From Advanced Characterization Techniques
|
| Presentation Title |
Machine Learning Assisted Structure-Property Relationships by Nanoindentation |
| Author(s) |
Eric D. Hintsala, Bernard Becker, Benjamin Stadnick, Kevin Schmalbach, Ude Hangen, Douglas Stauffer |
| On-Site Speaker (Planned) |
Eric D. Hintsala |
| Abstract Scope |
Nanoindentation can give a highly localized fingerprint of the materials elastic and plastic properties via the measured reduced modulus and hardness, respectively. Many thousands of indents can be done in a reasonable amount of time with modern instrumentation which can cover the sub-micron to mm-scale, allowing for structure-property relationships to be determined in complex heterogeneous materials. Machine learning can assist in this process in numerous ways, which will be discussed here. First, automatically identifying phases as regions of similar properties through clustering will be presented alongside a method to evaluate the uncertainty and bias of this approach. Secondly, Bayesian optimization will also be employed to improve instrument efficiency in terms of placing indents in the most needed areas. Lastly, workflow improvements for the correlation of the indentation properties to co-located structural data will also be detailed. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Mechanical Properties, Machine Learning, Characterization |