Reduced basis method for 3D electromagnetic

Thumbnail

Event details

Date 16.09.2009
Hour 16:15
Speaker Dr. Frank Schmidt
Location
MAA110
Category Conferences - Seminars
The reduced basis method enables fast and robust evaluation of input-output relationships described by parametrized partial differential equations. In the present work we apply the technique to geometrically parameterized electromagnetic scattering problems on unbounded domains in 2D and 3D. The main idea of the reduced basis method is to split up the solution process of the parameterized model into an expensive offline and a cheap online part. After constructing the reduced basis offline, the reduced model can be solved online very fast in the order of seconds or below. Error estimators assure the reliability of the reduced basis solution and are used for self adaptive construction of the reduced system. Due to the very high number of terms in the affine expansion of the parametrized system, especially in 3D, currently used online-offline decomposition of the error estimator is practically impossible. Online computation of the residuum becomes too expensive. Here we present a new method for estimation of the residuum in the reduced basis context which is inspired by the sub-domain residual method. The good performance of the new residuum estimator is demonstrated numerically. As application examples we consider scattering problems from 2D and 3D photolithographical masks and an optimization related to optical proximity corrections.

Practical information

  • General public
  • Free

Contact

  • Dr. Imbo Sim

Event broadcasted in

Share