|About this Abstract
||2018 TMS Annual Meeting & Exhibition
||Building an ICME Infrastructure: Developing Tools that Integrate Across Length and Time Scales to Accelerate Materials Design
||Uncertainty Quantification and Propagation through CALPHAD Thermodynamics and Integrated Computational Materials Engineering (ICME)
||Jeff W. Doak, Abhinav Saboo, Dana Frankel, Nick Hatcher, James Saal, Greg Olson
|On-Site Speaker (Planned)
||Jeff W. Doak
ICME models rely heavily on CALPHAD thermodynamics fit to experimental and ab-initio data. Uncertainty in this underlying data can drive uncertainty in the CALPHAD and ICME models. However, traditionally, once a CALPHAD database has been fit, the experimental data and statistics of the fit are discarded and the CALPHAD database is taken as ground truth. By discarding this information, considerable value to the material design process is lost.
At QuesTek Innovations, we are developing tools to incorporate Bayesian inference into CALPHAD thermodynamics and ICME modeling. Bayesian inference simultaneously fits thermodynamic models to data and quantifies the uncertainty in the resulting models due to this fitting. Because this uncertainty comes in the form of probability distributions over model parameters, the uncertainty in CALPHAD thermodynamic models can be straightforwardly propagated through ICME models. We will discuss the development of tools for Bayesian inference and their application to thermoelectrics and high-entropy alloys.
||Planned: Supplemental Proceedings volume