Optimal Design of Validation Experiments Based on Informatic Entropy

This electronic document focuses on how optimal design of validation experiment to allow conclusive comparison of model prediction with experimental data in model assessment. Our scheme promises the model assessment to be implemented at highest or lowest similarity between the distribution of the experimental outcome and model prediction through minimizing or maximizing the corresponding expected relative-entropy. Besides, Bayesian theory is incorporated into updating the distribution of experimental outcome based on new experimental data. Moreover, fuzzy inference is used to catch the optimal values in our methodology. Our scheme is illustrated for the optimal design of Sandia's Thermal problem validation experiment.