AD-based perturbation methods for uncertainties and errors

Massimiliano Martinelli Alain Dervieux Laurent Hascoët Valérie Pascual Anca Belme (INRIA Sophia-Antipolis, BP93, 06902 Sophia-Antipolis, France)

International Journal Engineering Systems Modelling and Simulation, 2010 (10 pages)

Abstract: The progress of automatic differentiation (AD) and its impact on perturbation methods is the object of this paper. AD studies show an important activity for developing methods addressing the management of modern CFD kernels, taking into account the language evolution and intensive parallel computing. The evaluation of a posteriori error analysis and of resulting correctors will be addressed. Recent works in the AD-based construction of second-derivatives for building reduced-order models based on a Taylor formula will be presented on the test case of a steady compressible flow around an aircraft.

Keywords: Automatic Differentiation, AD, second-order derivatives, uncertainty, error correction, design, CFD

Full text (pdf)
Presentation slides at the 44th AAAF, Nantes, France, 2009 (pdf)


@article{BMHPD09,
  author = {Martinelli, M. and Dervieux, A. and Hasco\"et, L. and Pascual, V. and Belme, A.},
  title = {{AD}-based perturbation methods for uncertainties and errors},
  journal = {Int. J. Engineering Systems Modelling and Simulation},
  volume = 2,
  number = "1/2",
  pages = "65-74",
  year = 2010
}