Automatic Differentiation for Optimum Design, Applied to Sonic Boom Reduction

Laurent Hascoët
Alain Dervieux
(INRIA, BP93, 06902 Sophia-Antipolis, France)
Mariano Vázquez
(GridSystems, Mallorca, Spain)

Proceedings of the ICCSA'03 conference, Montreal, Canada, 2003 (10 pages)

Abstract: We propose a methodology for adjoint-based optimum design that combines hand-coding with Automatic Differentiation in the reverse mode, therefore managing to keep the memory cost acceptable. This methodology also involves improvements to the reverse mode differentiation, based on dataflow information on the original program. We validate the method on a real-size 3D optimum design problem: reducing the sonic boom under a supersonic aircraft.

Keywords: Optimum Design, Gradient, Adjoint, Automatic Differentiation, Reverse mode, Data-Flow analysis

Full text (pdf)

  author = {Hasco\"et, L. and V\'azquez, M. and Dervieux, A.},
  title = {Automatic Differentiation for Optimum Design,
           Applied to Sonic Boom Reduction},
  booktitle = {Proceedings of the International Conference on
               Computational Science and its Applications, ICCSA'03, Montreal, Canada},
  editor = {{V.Kumar et al.}},
  publisher = {LNCS 2668, Springer},
  pages = {85-94},
  year = 2003