Filters








11 Hits in 5.7 sec

Airfoil Design Parameterization and Optimization using Bézier Generative Adversarial Networks [article]

Wei Chen, Kevin Chiu, Mark Fuge
2020 arXiv   pre-print
We use the airfoil design as an example to demonstrate the idea and analyze B\'ezier-GAN's representation capacity and compactness.  ...  We propose a deep generative model, B\'ezier-GAN, to parameterize aerodynamic designs by learning from shape variations in an existing database.  ...  Meanwhile, machine learning researchers have conducted a vast amount of DR research using deep neural networks such as variational autoencoders (VAEs) [22] and generative adversarial networks (GANs)  ... 
arXiv:2006.12496v2 fatcat:fsugudugazg6lgitfocrkiqd5u

BézierGAN: Automatic Generation of Smooth Curves from Interpretable Low-Dimensional Parameters [article]

Wei Chen, Mark Fuge
2021 arXiv   pre-print
Results show that our method can generate diverse and realistic curves, while preserving consistent shape variation in the latent space, which is favorable for latent space design optimization or design  ...  Many real-world objects are designed by smooth curves, especially in the domain of aerospace and ship, where aerodynamic shapes (e.g., airfoils) and hydrodynamic shapes (e.g., hulls) are designed.  ...  Generative Adversarial Networks A generative adversarial network [4] consists of a generative model (generator G) and a discriminative model (discriminator D).  ... 
arXiv:1808.08871v2 fatcat:oth2fvgp4fhgvm5feztf6adyj4

Airfoil GAN: Encoding and Synthesizing Airfoils forAerodynamic-aware Shape Optimization [article]

Yuyang Wang, Kenji Shimada, Amir Barati Farimani
2021 arXiv   pre-print
Our model is built upon VAEGAN, a neural network that combines Variational Autoencoder with Generative Adversarial Network and is trained by the gradient-based technique.  ...  In this work, we propose a data-driven shape encoding and generating method, which automatically learns representations from existing airfoils and uses the learned representations to generate new airfoils  ...  Generative Adversarial Network (GAN) [8] pushes learning from self-supervision even further via a min-max game between a generator and a discriminator.  ... 
arXiv:2101.04757v1 fatcat:ffm2a46uwvasjfpuskfp4fttiy

MO-PaDGAN: Reparameterizing Engineering Designs for Augmented Multi-objective Optimization [article]

Wei Chen, Faez Ahmed
2021 arXiv   pre-print
To address these challenges, we propose MO-PaDGAN, which adds a Determinantal Point Processes based loss function to the generative adversarial network to simultaneously model diversity and (multi-variate  ...  generator ignores design performance, and 3)~the new parameterization is unable to represent designs beyond training data.  ...  Generative Adversarial Nets Generative Adversarial Networks [6] model a game between a generative model (generator ) and a discriminative model (discriminator ).  ... 
arXiv:2009.07110v2 fatcat:syf6dvubg5anxgge3ilw3zuu7y

Soft and Hard Constrained Parametric Generative Schemes for Encoding and Synthesizing Airfoils [article]

Hairun Xie, Jing Wang, Miao Zhang
2022 arXiv   pre-print
Soft-constrained scheme: The CVAE-based model trains geometric constraints as part of the network and can provide constrained airfoil synthesis; 2.  ...  In this paper, two parametric generative schemes based on deep learning methods are proposed to represent the complicate design space under specific constraints. 1.  ...  The variation In the case of optimization design based on existing airfoil, the soft and hard constrained schemes can be used to provide geometry parametric variables to generate many new airfoils while  ... 
arXiv:2205.02458v1 fatcat:ejt5wo5kw5fodkvqepqogfexui

A Manifold-based Airfoil Geometric-feature Extraction and Discrepant Data Fusion Learning Method [article]

Yu Xiang, Guangbo Zhang, Liwei Hu, Jun Zhang, Wenyong Wang
2022 arXiv   pre-print
Experimental results show that our method could extract geometric-features of airfoils more accurately compared with existing methods, that the average MSE of re-built airfoils is reduced by 56.33%, and  ...  Motivated by the advantages of manifold theory and multi-task learning, we propose a manifold-based airfoil geometric-feature extraction and discrepant data fusion learning method (MDF) to extract geometric-features  ...  Yanqing Cheng, Associate Research Fellow, and Dr. Weiqi Qian, Research Fellow, both from China Aerodynamics Research and Development Center for their valuable suggestions with this paper.  ... 
arXiv:2206.12254v1 fatcat:kbwqshfw4zg2nkwqonkrxerlqa

Deep Generative Models in Engineering Design: A Review [article]

Lyle Regenwetter, Amin Heyrani Nobari, Faez Ahmed
2022 arXiv   pre-print
Recently, DGMs such as feedforward Neural Networks (NNs), Generative Adversarial Networks (GANs), Variational Autoencoders (VAEs), and certain Deep Reinforcement Learning (DRL) frameworks have shown promising  ...  Deep Generative Models (DGMs) typically leverage deep networks to learn from an input dataset and synthesize new designs.  ...  [55] use a neural network to generate optimized topologies from loading conditions.  ... 
arXiv:2110.10863v4 fatcat:zc4mo4nwzjdlne5jbvaetyugxy

Inverse design optimization framework via a two-step deep learning approach: application to a wind turbine airfoil [article]

Sunwoong Yang, Sanga Lee, Kwanjung Yee
2021 arXiv   pre-print
A variational autoencoder and multi-layer perceptron are used to generate a realistic target distribution and predict the quantities of interest and shape parameters from the generated distribution, respectively  ...  Finally, the framework is validated through aerodynamic shape optimizations of the wind turbine airfoil.  ...  Acknowledgments This work was supported by the National Research Foundation of Korea (NRF) Grant funded by the Ministry of Science and ICT (NRF-2017R1A5A1015311).  ... 
arXiv:2108.08500v2 fatcat:5bwu32pdkzfbnf4xabr6d25nni

Engineering Sketch Generation for Computer-Aided Design [article]

Karl D.D. Willis, Pradeep Kumar Jayaraman, Joseph G. Lambourne, Hang Chu, Yewen Pu
2021 arXiv   pre-print
Both models generate curve primitives without the need for a sketch constraint solver and explicitly consider topology for downstream use with constraints and 3D CAD modeling operations.  ...  We propose two generative models, CurveGen and TurtleGen, for engineering sketch generation.  ...  BézierGAN [4] synthesizes smooth curves using a generative adversarial network [12] and applies it to 2D airfoil profiles.  ... 
arXiv:2104.09621v1 fatcat:xxm2z53t2nh7tfbryfa5kusyma

Flow predictions with deep neural networks [article]

Ioannis Baklagis, National Technological University Of Athens
2021
A Generative Adversarial Network (GAN) combined with CNNs, called ffs-GAN, predicted transonic flow field profiles of parameterized supercritical airfoils, [16] .  ...  Training Dataset Creation and Network Training In order to create the training dataset, the airfoil shape was parameterized using two Bezier curves, with 6 CPs each, for the pressure and the suction side  ...  Statistical methods can be used to optimize the network input consisting of sequential data and, with the addition of architecture optimization, the training cost can be further decreased. 3.  ... 
doi:10.26240/heal.ntua.21425 fatcat:dckjq3ytrzcl7a4pkjoienoz4y

Dagstuhl Reports, Volume 9, Issue 12, December 2019, Complete Issue

2020
or adversarial games; with full or partial observation, etc.  ...  A work is progress is using evolutionary computation to generate playable game boards of different difficulty levels and using search algorithms to solve and evaluate the designed games.  ...  We also need statistically accurate physical simulators that allow us to optimize designs for manipulation.  ... 
doi:10.4230/dagrep.9.12 fatcat:hebigxkvinhjdb6qlg3j5hw25u