Skip to Main content Skip to Navigation

Optimisation Différentiable en Mécanique des Fluides Numérique

Francois Courty 1
1 TROPICS - Program transformations for scientific computing
CRISAM - Inria Sophia Antipolis - Méditerranée
Abstract : Our contribution concerns the following three complementary domains: Automatic Differentiation, optimal shape design for large systems, mesh adaption. In the chapter 1 of the part 1, we expose a method to compute gradients using Automatic Differentiation for a classical optimal shape design problem. We explain how to deduce an exact gradient based on an adjoint state without storing explicitly the Jacobian matrix. In the chapter 2 of the part 2, we propose a SQP-like method to solve a class of optimization problems with equality constraints. The new algorithm enables to solve simultaneously the optimality system. In the chapter 3 of the part 2, we study a new preconditioning strategy for optimal shape design. We build an additive multilevel preconditioning starting from the classical Bramble-Pasciak-Xu principle and from the agglomeration principle. In the chapter 1 of the part 3, we study the problem of the best adapted mesh for a pure interpolation problem. The optimality system solution gives a completely explicite expression of the optimal metric as a function of the function to adapt. In the chapter 2 of the part 3, we extend the method of the previous chapter to the problem of mesh adaption for P.D.E. We obtain an optimal control formulation with an adjoint state.
Document type :
Complete list of metadatas
Contributor : Francois Courty <>
Submitted on : Tuesday, January 27, 2004 - 5:57:43 PM
Last modification on : Saturday, January 27, 2018 - 1:30:54 AM
Long-term archiving on: : Friday, April 2, 2010 - 8:06:15 PM


  • HAL Id : tel-00004344, version 1



Francois Courty. Optimisation Différentiable en Mécanique des Fluides Numérique. Mathématiques [math]. Université Paris Sud - Paris XI, 2003. Français. ⟨tel-00004344⟩



Record views


Files downloads