A Bayesian approach to evaluating the uncertainty of thermodynamic data and phase diagrams

Abstract A heuristic optimization methodology based on Bayesian statistics is presented. The goal is to help researchers decide on the optimal set of thermodynamic data to use. This approach accounts for the errors associated with reported data and how reliable the researcher believes the model to be. The optimization is conducted with a multi-objective genetic algorithm (GA) coupled with Bayesian statistics to more accurately link the limited and uncertain experimental thermodynamic data to thermodynamic models of interest. The computer program provides guidance as to which experiments are needed to enhance the reliability of the dataset and is ideally suited for parameter optimization and sensitivity analysis. Applications include the UO 2 –PuO 2 and UO 2 –BeO systems.