Master Final Defense Seminar: Levi Dickson

Tuesday, December 1, 2026 at 1:30-3:30 PM

Calibration and Uncertainty Quantification of a Multicomponent Benchmark Solidification Trial

Event Details

Event Description

"Calibration and Uncertainty Quantification of a Multicomponent Benchmark Solidification Trial" 

Levi Dickson, MSE MS Candidate 

Advisor: Professor Matthew Krane

ABSTRACT

Benchmark static castings allow numerical solidification programs to be validated for accuracy by examining a simplified casting scenario, ensuring reliable estimations of microstructure and macrosegregation in multicomponent alloy castings. This and subsequent uncertainty quantification allow modelers to assess uncertainty in output of interest predictions. However, when casting boundary conditions are unknown, validation becomes difficult because of the epistemic uncertainty in the physics included and aleatoric uncertainty of associated input values. This work examined a Bayesian-inference-based approach to predict boundary conditions in multiple nickel-based benchmark castings based on thermal histories from experimental results, revealing which parameters were most physically likely during the experiment. After calibrating the model to individual experiments, model inputs related to boundary conditions, thermophysical quantities, and experimentally measured values have their uncertainties propagated through, revealing the highest sensitivity input parameters and providing probability distributions of selected outputs of interest.