An indirect design representation for topology optimization using variational autoencoder and style transfer

Tinghao Guo, Danny J. Lohan, James T. Allison, Ruijin Cang, Yi Ren

Research output: Chapter in Book/Report/Conference proceedingConference contribution

7 Citations (Scopus)

Abstract

In this paper we propose an indirect low-dimension design representation to enhance topology optimization capabilities. Established topology optimization methods, such as the Solid Isotropic Material with Penalization (SIMP) method, can solve large-scale topology optimization problems efficiently, but only for certain problem formulation types (e.g., those that are amenable to efficient sensitivity calculations). The aim of the study presented in this paper is to overcome some of these challenges by taking a complementary approach: achieving efficient solution via targeted design representation dimension reduction, enabling the tractable solution of a wider range of problems (e.g., those where sensitivities are expensive or unavailable). A new data-driven design representation is proposed that uses an augmented Variational Autoencoder (VAE) to encode 2. D topologies into a lower-dimensional latent space, and to decode samples from this space back into 2. D topologies. Optimization is then performed in the latent space as opposed to the image space. Established topology optimization methods are used here as a tool to generate a data set for training by changing problem conditions systematically. The data is generated using problem formulations that are solvable by SIMP, and are related to (but distinct from) the desired design problem. We further introduce augmentations to the VAE formulation to reduce unrealistic scattering of small material clusters during topology generation, while ensuring diversity of the generated topologies. We compare computational expense for solving a heat conduction design problem (with respect to the latent design variables) using different optimization algorithms. The new non-dominated points obtained via the VAE representation were found and compared with the known attainable set, indicating that use of this new design representation can simultaneously improve computational efficiency and solution quality.

Original languageEnglish (US)
Title of host publicationAIAA/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials
PublisherAmerican Institute of Aeronautics and Astronautics Inc, AIAA
Edition210049
ISBN (Print)9781624105326
DOIs
StatePublished - Jan 1 2018
EventAIAA/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials Conference, 2018 - Kissimmee, United States
Duration: Jan 8 2018Jan 12 2018

Other

OtherAIAA/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials Conference, 2018
CountryUnited States
CityKissimmee
Period1/8/181/12/18

Fingerprint

Shape optimization
Topology
Computational efficiency
Heat conduction
Scattering

ASJC Scopus subject areas

  • Civil and Structural Engineering
  • Building and Construction
  • Mechanics of Materials
  • Architecture

Cite this

Guo, T., Lohan, D. J., Allison, J. T., Cang, R., & Ren, Y. (2018). An indirect design representation for topology optimization using variational autoencoder and style transfer. In AIAA/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials (210049 ed.). American Institute of Aeronautics and Astronautics Inc, AIAA. https://doi.org/10.2514/6.2018-0804

An indirect design representation for topology optimization using variational autoencoder and style transfer. / Guo, Tinghao; Lohan, Danny J.; Allison, James T.; Cang, Ruijin; Ren, Yi.

AIAA/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials. 210049. ed. American Institute of Aeronautics and Astronautics Inc, AIAA, 2018.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Guo, T, Lohan, DJ, Allison, JT, Cang, R & Ren, Y 2018, An indirect design representation for topology optimization using variational autoencoder and style transfer. in AIAA/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials. 210049 edn, American Institute of Aeronautics and Astronautics Inc, AIAA, AIAA/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials Conference, 2018, Kissimmee, United States, 1/8/18. https://doi.org/10.2514/6.2018-0804
Guo T, Lohan DJ, Allison JT, Cang R, Ren Y. An indirect design representation for topology optimization using variational autoencoder and style transfer. In AIAA/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials. 210049 ed. American Institute of Aeronautics and Astronautics Inc, AIAA. 2018 https://doi.org/10.2514/6.2018-0804
Guo, Tinghao ; Lohan, Danny J. ; Allison, James T. ; Cang, Ruijin ; Ren, Yi. / An indirect design representation for topology optimization using variational autoencoder and style transfer. AIAA/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials. 210049. ed. American Institute of Aeronautics and Astronautics Inc, AIAA, 2018.
@inproceedings{51b3af6a99be47ea83b0cbf45cc5772b,
title = "An indirect design representation for topology optimization using variational autoencoder and style transfer",
abstract = "In this paper we propose an indirect low-dimension design representation to enhance topology optimization capabilities. Established topology optimization methods, such as the Solid Isotropic Material with Penalization (SIMP) method, can solve large-scale topology optimization problems efficiently, but only for certain problem formulation types (e.g., those that are amenable to efficient sensitivity calculations). The aim of the study presented in this paper is to overcome some of these challenges by taking a complementary approach: achieving efficient solution via targeted design representation dimension reduction, enabling the tractable solution of a wider range of problems (e.g., those where sensitivities are expensive or unavailable). A new data-driven design representation is proposed that uses an augmented Variational Autoencoder (VAE) to encode 2. D topologies into a lower-dimensional latent space, and to decode samples from this space back into 2. D topologies. Optimization is then performed in the latent space as opposed to the image space. Established topology optimization methods are used here as a tool to generate a data set for training by changing problem conditions systematically. The data is generated using problem formulations that are solvable by SIMP, and are related to (but distinct from) the desired design problem. We further introduce augmentations to the VAE formulation to reduce unrealistic scattering of small material clusters during topology generation, while ensuring diversity of the generated topologies. We compare computational expense for solving a heat conduction design problem (with respect to the latent design variables) using different optimization algorithms. The new non-dominated points obtained via the VAE representation were found and compared with the known attainable set, indicating that use of this new design representation can simultaneously improve computational efficiency and solution quality.",
author = "Tinghao Guo and Lohan, {Danny J.} and Allison, {James T.} and Ruijin Cang and Yi Ren",
year = "2018",
month = "1",
day = "1",
doi = "10.2514/6.2018-0804",
language = "English (US)",
isbn = "9781624105326",
booktitle = "AIAA/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials",
publisher = "American Institute of Aeronautics and Astronautics Inc, AIAA",
edition = "210049",

}

TY - GEN

T1 - An indirect design representation for topology optimization using variational autoencoder and style transfer

AU - Guo, Tinghao

AU - Lohan, Danny J.

AU - Allison, James T.

AU - Cang, Ruijin

AU - Ren, Yi

PY - 2018/1/1

Y1 - 2018/1/1

N2 - In this paper we propose an indirect low-dimension design representation to enhance topology optimization capabilities. Established topology optimization methods, such as the Solid Isotropic Material with Penalization (SIMP) method, can solve large-scale topology optimization problems efficiently, but only for certain problem formulation types (e.g., those that are amenable to efficient sensitivity calculations). The aim of the study presented in this paper is to overcome some of these challenges by taking a complementary approach: achieving efficient solution via targeted design representation dimension reduction, enabling the tractable solution of a wider range of problems (e.g., those where sensitivities are expensive or unavailable). A new data-driven design representation is proposed that uses an augmented Variational Autoencoder (VAE) to encode 2. D topologies into a lower-dimensional latent space, and to decode samples from this space back into 2. D topologies. Optimization is then performed in the latent space as opposed to the image space. Established topology optimization methods are used here as a tool to generate a data set for training by changing problem conditions systematically. The data is generated using problem formulations that are solvable by SIMP, and are related to (but distinct from) the desired design problem. We further introduce augmentations to the VAE formulation to reduce unrealistic scattering of small material clusters during topology generation, while ensuring diversity of the generated topologies. We compare computational expense for solving a heat conduction design problem (with respect to the latent design variables) using different optimization algorithms. The new non-dominated points obtained via the VAE representation were found and compared with the known attainable set, indicating that use of this new design representation can simultaneously improve computational efficiency and solution quality.

AB - In this paper we propose an indirect low-dimension design representation to enhance topology optimization capabilities. Established topology optimization methods, such as the Solid Isotropic Material with Penalization (SIMP) method, can solve large-scale topology optimization problems efficiently, but only for certain problem formulation types (e.g., those that are amenable to efficient sensitivity calculations). The aim of the study presented in this paper is to overcome some of these challenges by taking a complementary approach: achieving efficient solution via targeted design representation dimension reduction, enabling the tractable solution of a wider range of problems (e.g., those where sensitivities are expensive or unavailable). A new data-driven design representation is proposed that uses an augmented Variational Autoencoder (VAE) to encode 2. D topologies into a lower-dimensional latent space, and to decode samples from this space back into 2. D topologies. Optimization is then performed in the latent space as opposed to the image space. Established topology optimization methods are used here as a tool to generate a data set for training by changing problem conditions systematically. The data is generated using problem formulations that are solvable by SIMP, and are related to (but distinct from) the desired design problem. We further introduce augmentations to the VAE formulation to reduce unrealistic scattering of small material clusters during topology generation, while ensuring diversity of the generated topologies. We compare computational expense for solving a heat conduction design problem (with respect to the latent design variables) using different optimization algorithms. The new non-dominated points obtained via the VAE representation were found and compared with the known attainable set, indicating that use of this new design representation can simultaneously improve computational efficiency and solution quality.

UR - http://www.scopus.com/inward/record.url?scp=85044612008&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85044612008&partnerID=8YFLogxK

U2 - 10.2514/6.2018-0804

DO - 10.2514/6.2018-0804

M3 - Conference contribution

SN - 9781624105326

BT - AIAA/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials

PB - American Institute of Aeronautics and Astronautics Inc, AIAA

ER -