Skip to Main content Skip to Navigation

Méthodes d'optimisation pour l'espace et l'environnement

Abstract : This work is composed of two parts, issued from different industrial applications.
The first one is about an active array antenna.
First, we have to calculate the optimal excitations in order to comply with the radiated power requirements. We transform a problem with numerous local minima into a convex optimization one, whose optimum is the global minimum of the initial problem, by using the energy conservation principle.
Then we solve a topological optimization problem: we have to reduce the number of Radiating Elements (RE). We apply a singular value decomposition to all relaxed optimal modulus, and a topological gradient algorithm decides gatherings of elementary RE.

The second part is about a chemical accident black box modelling.
We perform a reliability and sensitivity analysis of a large number of parameters (failure probabilities, design point, and influent parameters). Without the gradient, we use a reduced model.
In a first test case, we compare the neural networks and the Sparse Grids (SG) interpolation method. The SG are an emergent tool: thanks to their hierarchical structure and an adaptive algorithm, they become particularly efficient for real problems (few influent variables).
We apply them to a test case in larger dimension with specific improvements (successive approximations and data thresholds).
In both cases, algorithms have lead to operative software.
Document type :
Complete list of metadatas
Contributor : Thierry Touya <>
Submitted on : Thursday, March 5, 2009 - 11:09:04 PM
Last modification on : Friday, October 23, 2020 - 4:52:38 PM
Long-term archiving on: : Tuesday, June 8, 2010 - 9:20:41 PM


  • HAL Id : tel-00366141, version 1


Thierry Touya. Méthodes d'optimisation pour l'espace et l'environnement. Modélisation et simulation. Université Paul Sabatier - Toulouse III, 2008. Français. ⟨tel-00366141⟩



Record views


Files downloads