Numerische Optimierungsverfahren für die Parameterschätzung und den Entwurf optimaler Experimente unter Berücksichtigung von Unsicherheiten für die Modellvalidierung verfahrenstechnischer Prozesse der Chemie und Biotechnologie


It is the goal to investigate how the evaluation of derivatives in nonlinear optimum experimental design can be improved in terms of efficiency and applicability to more general problem formulations. The research includes the development of tailored Newton- and Quasi-Newton-methods and the usage resp. development of suitable techniques from algorithmic differentiation.

Furthermore, the methods should be generalized to be useful for models of partial differential equations. The methods and the implemented software are supposed to meet the requirements of the partners from industry, namely BASF and Knauer. The project is primarily concerned with Newton- and Quasi-Newton-methods, the combination of direct and adjoint differentiation in optimum experimental design, derivative generation of model equations, optimum experimental design for iterative state solvers and optimum experimental design for the SMB process. The methods are implementd in a way such that they can be used by the industrial partners and other interested users. The advances of this project make it possible to investigate larger and more accurate models and treat new problem classes.

Principal investigators
Griewank, Andreas Prof. Dr. (Details) (Non-linear Optimization)

Financer
BMBF

Duration of project
Start date: 07/2007
End date: 12/2010

Last updated on 2022-07-09 at 21:08