Experimental design for model discrimination of thermodynamic models

Abstract

Traditionally, the screening of thermodynamic models in phase equilibria is performed only after obtaining a full set of experimental points. The aim of this work is to show that sequential model discrimination procedures may be used successfully to discriminate among rival thermodynamic models and thus, as a guide for more rational data collection. In order to do that, three different sequential experimental design procedures are implemented and used to model the solid–gas equilibrium at high pressures. One of the procedures is based on the Tsallis Statistics and depends on an empirical parameter which measures the risk that the experimenter is resolved to take in the decision-making stage. This procedure has been extended here to include prediction errors. It is shown here that the use of any of the three sequential discrimination procedures allow the selection of adequate models and model parameters with a significant reduction of the experimental effort. The Tsallis parameter is shown to exert a great influence upon the discarding of the models and the sequence of experiments. The prediction error is shown to be of fundamental importance for the proper selection of the next experiment.