Active data collection for inadequate models

Obtaining informative measurements is a fundamental problem when inadequate models are used to guide the design of experiments. A comprehensive approach to experimental design for inadequate physics-based models is proposed by focusing on the coupling between the structural uncertainty modeling and the adaptive data collection process. First, by taking advantage of the structure of physics-based models, unlike current approaches, rigorous structural uncertainty models are created to yield solutions, which satisfy physical constraints such as conservation of mass. Second, new adaptive data collection strategies are proposed by combining two current approaches, model driven and model free experimental design, to optimally trade off between model exploitation and design space exploration. The applicability and feasibility of these new ideas will be demonstrated on dispersion models, which are widely used in practice from regulatory applications to emergency response in chemical, nuclear, biological and radiological releases. These dispersion models are polluted by structural errors due to various assumptions (e.g. diffusion coefficients) that can only be informed using limited experimental data.