ISAE-SUPAERO Conference paper
The 1st International Conference on Cognitive Aircraft
Systems – ICCAS
March 18-19, 2020
https://events.isae-supaero.fr/event/2
Scientific Committee
•
Mickaël Causse, ISAE-SUPAERO
•
Caroline Chanel, ISAE-SUPAERO
•
Jean-Charles Chaudemar, ISAE-SUPAERO
•
Stéphane Durand, Dassault Aviation
•
Bruno Patin, Dassault Aviation
•
Nicolas Devaux, Dassault Aviation
•
Jean-Louis Gueneau, Dassault Aviation
•
Claudine Mélan, Université Toulouse Jean-Jaurès
•
Jean-Paul Imbert, ENAC
Permanent link :
https://doi.org/10.34849/cfsb-t270
Rights / License:
ICCAS 2020 Geodesic Convolutional Neural Ne …
Geodesic Convolutional Neural Network for 3D
Deep-Learning based Surrogate Modeling and
Optimization.
Content
The role of numerical simulation in product development has shifted from being a validation tool of mature designs into a means of exploration of product design space. Yet, the time required to run a simulation is, most of the time, a bottleneck in the engineer’s optimisation loop and for larger design spaces it can result in automated shape optimization being simply intractable. This needs to be addressed on the way to better simulation-driven design. Surrogate models are used in CFD simulations, and other computationally intensive simulations, as a cheaper data-driven substitute for the full-fledged numerical simulator. Most of the existing surrogate modeling approaches rely on Gaussian Process regressors (Kriging) and are thus limited to predicting the performance of shapes with a fixed low-dimensional parameterization. On top of that, kriging methods are meant for predicting global scalar values but they are not capable of predicting fields (e.g. velocity or pressure values at every point of the shape).
In this abstract, we give an account of Neural Concept Shape (NCS), a surrogate modelling software that proposes using geodesic deep neural networks as regressors to overcome the aforementioned limitations of Gaussian Process regressors 1. NCS builds on top of state-of-the-art 3D geometric learning techniques [2, 3] to accelerate 3D simulations and automate shape design optimization. In particular, we demonstrate the case of a fixed-wing drone shape optimisation. We train a deep Geometric Convolutional Neural Network (GCNN) to reproduce the flow and pressure field, nor-mally obtained using a CFD solver, directly as a Neural Network prediction. Importantly, the input of our Neural Network can be a mesh representation of the shape since we rely on spatial mesh convolutions as described in [3]. In our software, we use geodesic coordinates for convolutions on the surface of the objects, and Euclidean coordinates for convolutions in the bulk of the domain. As opposed to existing methods, we do not use interpolation on a regular grid or an image, and do not require prior re-meshing of the shape.
The reported approach is beneficial on many critical aspects: first, it allows us to compute approx-imate solutions orders of magnitude faster than the typical numerical simulators (tens of millisec-onds instead of multiple hours); and, second, it allows to compute the gradients of the objective (e.g. aerodynamic efficiency) with respect to the input shape. This ultimately allows us to run op-timization loops in reasonable time and makes it possible to use first-order opop-timization methods. References
1 Pierre Baque, Edoardo Remelli, Francois Fleuret, and Pascal Fua. Geodesic convolutional shape optimization. In International Conference on Machine Learning, pages 481–490, 2018.
[2] Michael M Bronstein, Joan Bruna, Yann LeCun, Arthur Szlam, and Pierre Vandergheynst. Geo-metric deep learning: going beyond euclidean data. IEEE Signal Processing Magazine, 34(4):18–42, 2017.
[3] Federico Monti, Davide Boscaini, Jonathan Masci, Emanuele Rodola, Jan Svoboda, and Michael M Bronstein. Geometric deep learning on graphs and manifolds using mixture model cnns. In Pro-ceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 5115–5124, 2017.
Dr BAQUÉ, Pierre (Neural Concept) ; Mr ALLARD, Théophile (Neural Concept); Dr BASET, Selena (Neural Concept); Mr VONT TSCHAMMER, Thomas (Neural Concept); Mr ZAMPIERI, Luca (Neural Concept); Prof. FUA, Pascal (EPFL)
