Adjoint based optimization connected with the CAD

Application of adjoint method in computational fluid dynamics makes the calculation of gradients of objective functions with respect to design parameters computationally affordable. By applying traditional methods, e.g methods based on finite differences, the number of simulations required for the gradient calculations increases drastically with an increasing number of design parameters. Through the use of adjoint method is it possible to calculate all gradients in “one shot”.

In general, adjoint-based optimization is a parameter-free procedure, which permits to morph shapes freely. Thereby, the problem arise due to regenerate the geometry in CAD. Through a conjunction of adjoint sensitivities and CAD parameters, this situation can be solved. This connection, once established, will stay during the whole process.

Furthermore, optimization approaches in industrial cases are often gradient based.

The multiobjective optimization makes it necessary to rate appropriate gradients.

The aim of this work is to create a method, which satisfied all of the requirements above. (Connect CAD-Parameter and adjoint-sensitivities, create gradient-based optimization cycle and rate appropriate gradients in multiobjective optimization.) The verification of this method is done by typically academical and industrially relevant geometries.

Keywords: computational fluid dynamics, optimization, sensitivities, gradient-based, multiobjective, discrete adjoint method, CAD, automatic processes

Presenter(s): 
Porsche AG
Christian Boehmer
Session Time Slot(s): 
Time: 
07/03/2017 - 2:00pm-07/03/2017 - 2:25pm
Room: 
Convention Hall 1C
Presentation Image: 
Session Track: 

Session Tracks (STAR Global Conference 2017)

Co-Author(s): 
First Name: 
Josef
Last Name: 
Dubsky
Company / Institution: 
Porsche AG