About this Abstract |
| Meeting |
2027 TMS Annual Meeting & Exhibition
|
| Symposium
|
Algorithms Development in Materials Science and Engineering
|
| Presentation Title |
Quantification of microstructure variability with sparse data |
| Author(s) |
Arulmurugan Senthilnathan, Sankaran Mahadevan |
| On-Site Speaker (Planned) |
Arulmurugan Senthilnathan |
| Abstract Scope |
Variability in the manufacturing process propagates to the microstructural features that influence macro-scale mechanical properties and performance of a material. Microstructures are experimentally observed using scanning techniques. However, scanning multiple microstructures to quantify feature variability is expensive, thus leading to sparse scans. This work proposes a novel Bayesian approach to quantify feature variability in the presence of sparse scans. First, the variability within each scan is characterized using probability distribution (random variable) and 2-point correlation (random field) function that quantify the probability of finding two points within the same phase. A prior distribution is then assumed for various probabilities, and the prior distributions are updated to estimate the posteriors, given the sparse scans. At each feature, the estimated posterior distributions are aggregated to quantify the overall feature variability. Quantification of feature variability using the proposed approach helps in generating microstructure input for the micromechanical model and validating the process model. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Computational Materials Science & Engineering, ICME, Characterization |