WP 3: Return to CAD/transition tools

The transition from topology to shape optimisation. Upgrade techniques which include the smoothing of the surface-outcome of topology optimisation, the CAD-model reconstruction and the subsequent shape optimisation. An integrated/automated kernel is targeted.

Feasibility study w.r.t. the use of Symbolic Regression for CAD (NURBS) generation from point clouds and triangulated surfaces. Survey of alternative mesh-to-CAD methods, including neural networks

Basic implementation of NURBS Symbolic Regression with discontinuity detection and filtering/smoothing of input data.

Compatibility of the hierarchical shape-topology optimisation with CAD software. Linked with RBF-based morphing mainly developed by ESR7 at NTUA

Automatic differentiation of mesh to CAD algorithms to evaluate parameter sensitivity in the symbolic regression.

Objectives:

• Develop topology to shape transition tools.

• Develop tools to recover mesh deformations in CAD format.

• Algorithmic differentiation of "Mesh to CAD" algorithms to optimise the geometric model

Personnel and Project Deliverables

ENGYS - Eugene De Villiers "Return to CAD options"

Marios Damigos (ESR 5) "NURBS by symbolic regression"

Athanasios Liatsikouras (ESR 6) "CAD-based sensitivity option in ESI adjoint kernel"

James Koch (ESR 8) "Transition from topology to shape"

Mladen Banovic (ESR 12) "Differentiated version of mesh to CAD algorithm"