Strategies for computing second-order derivatives in CFD design problems

Massimiliano Martinelli Alain Dervieux Laurent Hascoët

Proceedings of West-East High Speed Flow Field Conference, Moscow, Russia, nov. 19-22 2007 (4 pages)

Abstract: Important advances in CFD design methods has been obtained using first-order derivatives and adjoints. However, in order to take into account uncertainties, reduced models are built with second-order derivatives. New functionals involving first-order derivatives are also used for robust design so their gradients will involve second-order derivatives. For these developments, building second-order derivatives is a mandatory task. We analyze strategies for computing second-order derivatives using Automatic Differentiation tools. The target is a solver for shape design with supersonic flow.

Keywords: Second-Order Derivatives, Robust Design, Uncertainties, Reduced Model, Automatic Differentiation, Gradient, Hessian

Full text (pdf)


@inproceedings{martinelliWEHSFF2007,
  author = {Martinelli, M. and Dervieux, A. and Hasco{\"e}t, L.},
  title = {Strategies for computing second-order derivatives in {CFD} design problems},
  booktitle = {{P}roc. of {W}est-{E}ast {H}igh {S}peed {F}low {F}ield {C}onference},
  month = {nov. 19-22},
  address = {Moscow, Russia},
  year = 2007
}