Validation Methods

Validation of mathematical models of physical phenomena is critical to developing reliable predictive simulations. However, robust and reliable validation processes are not well established. One of the central goals of the PECOS Center is to develop and deploy such validation cycles. We refer here to a “validation cycle” in recognition of the fact that validation is iterative, with results of a validation exercise commonly leading to refinement of either the model or its calibration. The validation process is made more subtle and complicated by the need to account for uncertainty and the fact that one generally needs to validate the model’s ability to predict particular quantities of interest (QoI’s), and these quantities are usually not accessible experimentally. Among the research and development issues being pursued in the development of validation methods are:

  • Procedures for determining surrogate quantities of interest
  • Acceptance measures and criterion
  • Selection of tolerances for decision making
  • Parameter identification by Bayesian methods
  • Procedures for selecting relevant observables
  • Adaptive validation methods
  • Validation of data reduction models in laboratory experiments