Adjoint-Based Error Estimation and Mesh Adaptation for Problems with Output Constraints
Speaker: Marco Ceze, NASA Ames Research Center
Presentation
Abstract
In this talk we extend adjoint-based output error estimation and mesh adaptation to computational fluid dynamics simulations that involve trim constraints—i.e. constraints on certain outputs. These constraints may involve outputs other than those directly targeted for error estimation and adaptation. A representative example used in this work is the reduction of numerical error in drag under fixed-lift conditions. The lift output comes in only through the constraint, but numerical errors in the lift will indirectly affect the drag calculation. We present a method for including this effect in the output error calculation, and for adapting the discretization to better predict both the target output and the trim constraint. The method relies on a discrete adjoint solution, which is also used to efficiently trim the simulations. We demonstrate the method on a discontinuous finite element discretization of the compressible RANS equations.
Biography
Marco Ceze is a NASA postdoctoral fellow under the supervision of Dr. Scott Murman. Before NASA, Marco was a postdoctoral fellow at the University of Michigan where he earned his doctorate under the supervision of Prof. Krzyzstof Fisdkowski. His research interests are efficient and robust nonlinear solvers, hp adaptivity, and optmization methods.