Close

A note on cookies

We use cookies to improve your experience of our website. If you want to find out more see our Privacy Policy

Menu

Industrial Optimal Design using Adjoint CFD

People menu

Research Fellows

Ilias Vasilopoulos

Early Stage Researcher 11 at Rolls-Royce Deutschland

 

The second CAD-based approach evaluated in this work is the AD approach from UPB, which uses the software tool ADOL-C. This method targets to compute exact geometric derivatives by differentiating the CAD kernel, but it can only be applied in cases where there is access to the CAD software source code (as in OpenCASCADE). During the Secondment of ESR 12 in RRD, part of the in-house aerofoil design tool Parablading was differentiated in forward vector mode using the operator overloading technique. As a result, sensitivities w.r.t. compressor design parameters (e.g. leading and trailing edge radii, inlet and exit blade angles, maximum thickness, axial and circumferential shifts of design sections) can be currently computed using AD. This methodology was validated using central (2nd order) FD and it was integrated into the improved automated CAD-based optimization workflow illustrated in the figure below.

The above process chain was first demonstrated in [4], using exclusively FD to compute the geometric sensitivities (increased efficiency for the CAD part compared to the inner re-meshing loop of the previous workflow). Sensitivities using FD were computed inside Parablading by exploiting the NURBS analytical description and keeping the node parametric coordinates fixed. Here, the same process chain was used to optimize again the TU Berlin TurboLab Stator, with the difference being that sensitivities w.r.t. the compressor design parameters were computed by AD. The exact AD is currently less time (but more memory) consuming than the central FD. Future work should focus on extending the differentiation to include more parameters, and also introduce the reverse mode to further increase the code’s efficiency.

^ Back to Top