View article

This paper deals with model validation of dynamic systems with vehicle systems being of particular interest that have multiple time-dependent output. First, we review several validation methodologies that have been reported in the literature graphical comparison, feature-based techniques, PDFCDF based techniques, Bayesian posterior estimation, classical hypothesis testing and Bayesian hypothesis testing. We discuss their advantages and disadvantages in terms of several attributes applicability to different types of models, need for assumptions computational cost, subjectivity, propensity to type-I or II errors, and others. We then proceed with the most important attribute can the validation method provide a quantitative measure of the goodness of the model We conclude that Bayesian-based model validation frameworks answer this question positively. A bootstrap method is presented that obviates the need to assume a statistical distribution model. The features of the Bayesian validation framework are illustrated using a thermal benchmark problem developed by Sandia National Laboratories and a battery model developed in the Automotive Research Center, a US Army Center of Excellence for modeling and simulation of ground vehicle systems.