Skip to Main content Skip to Navigation

Calcul de gradient sur des paramètres CAO pour l’optimisation de forme

Abstract : In this manuscript, we present a shape optimization method based on CAD parameters such as lengths, angles, etc. We rely on gradient-based optimization techniques. The sensitivity of the objective function, with respect to the mesh nodes position, is provided by an adjoint solver considered here as a black box. To optimize with respect to CAD parameters, we focus on computing the sensitivity of the nodes positions with respect to these parameters. Thus, we propose two approaches based on finite differences. The first method uses a harmonic projection to compare in the same space the initial mesh and the one obtained after a change of the set of CAD parameters. The developments presented in this manuscript open up new doors like the application to shapes with multiple borders such as exhaust manifolds. We also developed an interpolation method suitable for this comparison. The entire process is automated, and we demonstrate the entire effectiveness on internal aerodynamics industrial applications. The second method is directly based on the CAD geometries to assess this sensitivity. To perform this comparison, we use the intrinsic definition of the patches in the parametric space (u;v). Through the use of the exact coordinates at any point on the surface provided by the CAD, we avoid using an interpolation to get the best calculation accuracy possible. However, unlike the first method, it requires to identify the correspondence between patches from one shape to another. An application on an external aerodynamics academic case was made. The relevance of the first method is demonstrated on a representative multi-objective case, which facilitate its deployment use in an industrial environment. Regarding the second method, we showed its great potential. However, further developments are needed to handle more advanced cases. Because they are independent of the mechanical solver and the number of parameters, these methods significantly reduce product development time, particularly by allowing large and multiphysics optimization.
Document type :
Complete list of metadata

Cited literature [62 references]  Display  Hide  Download
Contributor : ABES STAR :  Contact
Submitted on : Thursday, April 6, 2017 - 4:50:13 PM
Last modification on : Wednesday, July 21, 2021 - 2:16:02 PM
Long-term archiving on: : Friday, July 7, 2017 - 3:26:28 PM


Version validated by the jury (STAR)


  • HAL Id : tel-01503189, version 1


Timothée Leblond. Calcul de gradient sur des paramètres CAO pour l’optimisation de forme. Autre. Université Paris-Saclay, 2017. Français. ⟨NNT : 2017SACLC017⟩. ⟨tel-01503189⟩



Record views


Files downloads