東華大學圖書館 |

Artificial Neural Network-Based Surrogate Modelling Methods in Accelerating Topology Optimization Process for Large-Scale Designs.

紀錄類型:	書目-電子資源 : Monograph/item
正題名/作者:	Artificial Neural Network-Based Surrogate Modelling Methods in Accelerating Topology Optimization Process for Large-Scale Designs./
作者:	Ren, Tan Kai.
面頁冊數:	1 online resource (117 pages)
附註:	Source: Dissertations Abstracts International, Volume: 84-09, Section: A.
Contained By:	Dissertations Abstracts International84-09A.
標題:	Load. -
電子資源:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=30276679click for full text (PQDT)
ISBN:	9798374421026

Artificial Neural Network-Based Surrogate Modelling Methods in Accelerating Topology Optimization Process for Large-Scale Designs.
Ren, Tan Kai.

Artificial Neural Network-Based Surrogate Modelling Methods in Accelerating Topology Optimization Process for Large-Scale Designs. - 1 online resource (117 pages)

Source: Dissertations Abstracts International, Volume: 84-09, Section: A.

Thesis (Ph.D.)--Hong Kong University of Science and Technology (Hong Kong), 2022.

Includes bibliographical references

Topology design which optimizes the materials or components distributions in a selected domain toachieve certain objectives has been widely applied in engineering field such as transportation vehiclesor architectural designs. In structural mechanics, topology optimization (TO) provides systematicapproach to optimize the topology of the structure so that desired material properties are fulfilled.However the method requires the iterative calculation of objective function, which involvescomputationally expensive numerical simulation such as finite element method (FEM) calculation. Asthe computational cost for the process scales exponentially with the domain or mesh sizes, constantdevelopment is performed to improve the efficiency of the method. In recent years, various deeplearning-based methods have been developed to accelerate the design process. Although the existingapplications of deep learning models could accelerate the design process through different approaches,most of them suffer from the issue of transferability and the requirement of large amount of trainingdata. Since the training data is required to be generated through numerical simulation itself, the overallefficiency of those methods is reduced.In this thesis, three major ANN-based methods are developed with the ultimate objective of acceleratingthe traditional topology design process, while also taking into account the transferability of the networksand the training data generation process. In the first method, an inverse design model is built with acombination of a surrogate model based on Convolutional Neural Network (CNN), and a generativemodel, Deep Convolutional Generative Adversarial Network (DCGAN). The method is developed tosolve inverse design problem in a short time, by generating designs that satisfy desired mechanicalproperties, while also subjected to prescribed geometrical constraint. The design of microstructuralmaterials to achieve specified effective compliance tensor is used as the demonstration for effectivenessof the developed model.In the second method, a deep learning model known as Mapping Network is developed to reduce thetime taken for training data generation of neural network-based surrogate model. Instead of generatingall the training data for the surrogate model by performing FEM calculation on the field of interest inthe full-scale mesh, a large portion of the data are instead generated in much coarser mesh. The coarse-scalefield is then mapped back to the original scale by MapNet. Since the simulation time in coarse-scalemesh is much faster, and the prediction time of Mapping Network is also relatively short, theoverall time required during the data generation process can then be greatly reduced. The applicationof surrogate model in TO process for structural design problem is used to demonstrate the time savingthat could be achieved by using the proposed method as compared with the traditional method oftraining data generation.Next, using the insights gained from the second method, the idea of Mapping Network in mapping thefield of interest from coarse to fine scale is further developed and improved upon. Since the trainednetwork provides a great transferability to different design problems, the idea is integrated to develop ahighly scalable surrogate modelling method used for accelerating the most expensive part of TO process,the FEM calculation of objective values. In the developed method, the FEM calculations during eachTO step are performed at coarse scale mesh, which are then mapped back to the fine scale using a deeplearning model known as MapNet. Combining with the fine scale structure available from each iterationof TO, and fragmentation technique which crops a domain into many smaller subdomains, the trainedMapNet has great transferability and can be easily used for different design problems. The developedframework is demonstrated to be easily applied across TO processes for various design problemsincluding structural and thermal problem, while achieving high efficiency and massive time saving.

Electronic reproduction.
Ann Arbor, Mich. :
ProQuest,
2023

Mode of access: World Wide Web

ISBN: 9798374421026Subjects--Topical Terms:

3562902
Load.
Index Terms--Genre/Form:

542853
Electronic books.

Artificial Neural Network-Based Surrogate Modelling Methods in Accelerating Topology Optimization Process for Large-Scale Designs.
LDR:05511nmm a2200397K 4500 001 2356181
005 20230612071837.5
006 m o d
007 cr mn ---uuuuu
008 241011s2022 xx obm 000 0 eng d
020 $a 9798374421026
035 $a (MiAaPQ)AAI30276679
035 $a (MiAaPQ)HongKongSciTech_991013060729103412
035 $a AAI30276679
040 $a MiAaPQ $b eng $c MiAaPQ $d NTU
100 1 $a Ren, Tan Kai. $3 3696662
245 1 0 $a Artificial Neural Network-Based Surrogate Modelling Methods in Accelerating Topology Optimization Process for Large-Scale Designs.
264 0 $c 2022
300 $a 1 online resource (117 pages)
336 $a text $b txt $2 rdacontent
337 $a computer $b c $2 rdamedia
338 $a online resource $b cr $2 rdacarrier
500 $a Source: Dissertations Abstracts International, Volume: 84-09, Section: A.
500 $a Advisor: Ye, Wenjing.
502 $a Thesis (Ph.D.)--Hong Kong University of Science and Technology (Hong Kong), 2022.
504 $a Includes bibliographical references
520 $a Topology design which optimizes the materials or components distributions in a selected domain toachieve certain objectives has been widely applied in engineering field such as transportation vehiclesor architectural designs. In structural mechanics, topology optimization (TO) provides systematicapproach to optimize the topology of the structure so that desired material properties are fulfilled.However the method requires the iterative calculation of objective function, which involvescomputationally expensive numerical simulation such as finite element method (FEM) calculation. Asthe computational cost for the process scales exponentially with the domain or mesh sizes, constantdevelopment is performed to improve the efficiency of the method. In recent years, various deeplearning-based methods have been developed to accelerate the design process. Although the existingapplications of deep learning models could accelerate the design process through different approaches,most of them suffer from the issue of transferability and the requirement of large amount of trainingdata. Since the training data is required to be generated through numerical simulation itself, the overallefficiency of those methods is reduced.In this thesis, three major ANN-based methods are developed with the ultimate objective of acceleratingthe traditional topology design process, while also taking into account the transferability of the networksand the training data generation process. In the first method, an inverse design model is built with acombination of a surrogate model based on Convolutional Neural Network (CNN), and a generativemodel, Deep Convolutional Generative Adversarial Network (DCGAN). The method is developed tosolve inverse design problem in a short time, by generating designs that satisfy desired mechanicalproperties, while also subjected to prescribed geometrical constraint. The design of microstructuralmaterials to achieve specified effective compliance tensor is used as the demonstration for effectivenessof the developed model.In the second method, a deep learning model known as Mapping Network is developed to reduce thetime taken for training data generation of neural network-based surrogate model. Instead of generatingall the training data for the surrogate model by performing FEM calculation on the field of interest inthe full-scale mesh, a large portion of the data are instead generated in much coarser mesh. The coarse-scalefield is then mapped back to the original scale by MapNet. Since the simulation time in coarse-scalemesh is much faster, and the prediction time of Mapping Network is also relatively short, theoverall time required during the data generation process can then be greatly reduced. The applicationof surrogate model in TO process for structural design problem is used to demonstrate the time savingthat could be achieved by using the proposed method as compared with the traditional method oftraining data generation.Next, using the insights gained from the second method, the idea of Mapping Network in mapping thefield of interest from coarse to fine scale is further developed and improved upon. Since the trainednetwork provides a great transferability to different design problems, the idea is integrated to develop ahighly scalable surrogate modelling method used for accelerating the most expensive part of TO process,the FEM calculation of objective values. In the developed method, the FEM calculations during eachTO step are performed at coarse scale mesh, which are then mapped back to the fine scale using a deeplearning model known as MapNet. Combining with the fine scale structure available from each iterationof TO, and fragmentation technique which crops a domain into many smaller subdomains, the trainedMapNet has great transferability and can be easily used for different design problems. The developedframework is demonstrated to be easily applied across TO processes for various design problemsincluding structural and thermal problem, while achieving high efficiency and massive time saving.
533 $a Electronic reproduction. $b Ann Arbor, Mich. : $c ProQuest, $d 2023
538 $a Mode of access: World Wide Web
650 4 $a Load. $3 3562902
650 4 $a Deep learning. $3 3554982
650 4 $a Optimization techniques. $3 3681622
650 4 $a Neural networks. $3 677449
650 4 $a Engineering. $3 586835
650 4 $a Energy. $3 876794
650 4 $a Algorithms. $3 536374
650 4 $a Microstructure. $3 555179
650 4 $a Bridges. $3 868720
650 4 $a Design techniques. $3 3561498
650 4 $a Composite materials. $3 654082
650 4 $a Design. $3 518875
650 4 $a Finance. $3 542899
650 4 $a Industrial engineering. $3 526216
650 4 $a Materials science. $3 543314
655 7 $a Electronic books. $2 lcsh $3 542853
690 $a 0791
690 $a 0537
690 $a 0800
690 $a 0389
690 $a 0508
690 $a 0546
690 $a 0794
710 2 $a ProQuest Information and Learning Co. $3 783688
710 2 $a Hong Kong University of Science and Technology (Hong Kong). $3 1022235
773 0 $t Dissertations Abstracts International $g 84-09A.
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=30276679 $z click for full text (PQDT)