Uncertainty reduction by optimal experimental design in the framework of PDE
Monday, August 8, 3PM – 4PM
Thomas Carraro, Institute of Applied Mathematics, University of Heidelberg
Model parameter estimation based on experimental measurements is relevant for many applications. Experimental data are usually affected by uncertainties, which could derive from random measurements errors or random fluctuations in the experimental conditions. When the sources of random perturbations in the measurements are unavoidable the use of optimal experimental design (OED) techniques can significantly increase the accuracy of parameter estimation. We give a rigorous derivation of the OED problem in the framework of partial differential equations (PDE) using the variational formulation, which is appropriate for the numerical solution by the finite element method (FEM). In addition we propose a primal-dual active set method to solve the constrained OED problem. Numerical results are presented including the first steps towards the application of OED in parameter estimation for fuel cell models.
Hosted by Robert Moser