東華大學圖書館 |

語系: 繁體中文

說明(常見問題)

回圖書館首頁

手機版館藏查詢

登入

回首頁

切換: 標籤 | MARC模式 | ISBD

Interpolating Neural Networks for th...

Park, Chanwook,

FindBook

Google Book

Amazon

博客來

Interpolating Neural Networks for the Next-Generation Predictive Scientific Artificial Intelligence /

紀錄類型:	書目-電子資源 : Monograph/item
正題名/作者:	Interpolating Neural Networks for the Next-Generation Predictive Scientific Artificial Intelligence // Chanwook Park.
作者:	Park, Chanwook,
面頁冊數:	1 electronic resource (117 pages)
附註:	Source: Dissertations Abstracts International, Volume: 86-12, Section: B.
Contained By:	Dissertations Abstracts International86-12B.
標題:	Mechanical engineering. -
電子資源:	https://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=32039084
ISBN:	9798315797814

Interpolating Neural Networks for the Next-Generation Predictive Scientific Artificial Intelligence /
Park, Chanwook,

Interpolating Neural Networks for the Next-Generation Predictive Scientific Artificial Intelligence /Chanwook Park. - 1 electronic resource (117 pages)

Source: Dissertations Abstracts International, Volume: 86-12, Section: B.

Predictive scientific Artificial Intelligence (AI) is a subfield of AI that focuses on uncovering relationships between spatial-temporal and parametric inputs and outputs that have physical meaning, aiming to facilitate decision-making in science and engineering. Currently, predictive scientific AI faces several key challenges: generalization capability, scalability, and sustainability. These challenges are closely tied to five performance metrics of an AI model: 1) grid (or mesh) resolution, 2) model accuracy, 3) computational speed, 4) usage of Random Access Memory (RAM), and 5) disk storage requirements. To address these challenges and improve the five performance measures, this dissertation introduces the Interpolating Neural Network (INN), a novel approach that combines interpolation theory, tensor decomposition, and neural networks. Throughout this work, it is demonstrated how INN and its variants-INN-Tensor Decomposition (INN-TD) and immersed INN-TD-can significantly improve these the performance metrics, outperforming traditional numerical methods and existing data-driven models by orders of magnitude.The key findings of this work reveal immense potential for commercialization across various disciplines, particularly in the Computer-Aided Engineering (CAE) industry. For example, the INN solver can be seamlessly integrated within leading CAE software such as Ansys and Abaqus through their user subroutines. When combined with GPU hardware, the INN solver achieves a 2.5 times speedup over standard Finite Element Analysis (FEA) conducted in Ansys, while requiring 100 times fewer mesh nodes, which will substantially alleviate the burden of mesh generation and pre-/post-processing. Additionally, the INN-TD trainer has demonstrated a lightweight structure with fewer trainable parameters and fast convergence compared to state-of-the-art Machine Learning (ML) models. As major CAE companies actively explore the integration of data-driven approaches into their simulation platforms, the INN-TD trainer holds promise as an efficient backend ML model. Moreover, INN-TD and immersed INN-TD solvers have the potential to fundamentally transform how parameterized partial differential equations are solved in real engineering analysis and design. Accordingly, the author is highly motivated to pursue commercialization opportunities for the INN technology introduced in this dissertation.

English

ISBN: 9798315797814Subjects--Topical Terms:

649730
Mechanical engineering.
Subjects--Index Terms:

Computational mechanics

Interpolating Neural Networks for the Next-Generation Predictive Scientific Artificial Intelligence /
LDR:03923nmm a22004453i 4500 001 2435057
005 20250930135711.5
006 m o d
007 cr|nu||||||||
008 261011s2025 miu||||||m |||||||eng d
020 $a 9798315797814
035 $a (MiAaPQD)AAI32039084
035 $a AAI32039084
040 $a MiAaPQD $b eng $c MiAaPQD $e rda
100 1 $a Park, Chanwook, $e author. $0 (orcid)0000-0001-5109-9563 $3 3825240
245 1 0 $a Interpolating Neural Networks for the Next-Generation Predictive Scientific Artificial Intelligence / $c Chanwook Park.
264 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2025
300 $a 1 electronic 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: 86-12, Section: B.
500 $a Advisors: Liu, Wing Kam Committee members: Wagner, Gregory J.; Qian, Dong; Bessa, Miguel A.
502 $b Ph.D. $c Northwestern University $d 2025.
520 $a Predictive scientific Artificial Intelligence (AI) is a subfield of AI that focuses on uncovering relationships between spatial-temporal and parametric inputs and outputs that have physical meaning, aiming to facilitate decision-making in science and engineering. Currently, predictive scientific AI faces several key challenges: generalization capability, scalability, and sustainability. These challenges are closely tied to five performance metrics of an AI model: 1) grid (or mesh) resolution, 2) model accuracy, 3) computational speed, 4) usage of Random Access Memory (RAM), and 5) disk storage requirements. To address these challenges and improve the five performance measures, this dissertation introduces the Interpolating Neural Network (INN), a novel approach that combines interpolation theory, tensor decomposition, and neural networks. Throughout this work, it is demonstrated how INN and its variants-INN-Tensor Decomposition (INN-TD) and immersed INN-TD-can significantly improve these the performance metrics, outperforming traditional numerical methods and existing data-driven models by orders of magnitude.The key findings of this work reveal immense potential for commercialization across various disciplines, particularly in the Computer-Aided Engineering (CAE) industry. For example, the INN solver can be seamlessly integrated within leading CAE software such as Ansys and Abaqus through their user subroutines. When combined with GPU hardware, the INN solver achieves a 2.5 times speedup over standard Finite Element Analysis (FEA) conducted in Ansys, while requiring 100 times fewer mesh nodes, which will substantially alleviate the burden of mesh generation and pre-/post-processing. Additionally, the INN-TD trainer has demonstrated a lightweight structure with fewer trainable parameters and fast convergence compared to state-of-the-art Machine Learning (ML) models. As major CAE companies actively explore the integration of data-driven approaches into their simulation platforms, the INN-TD trainer holds promise as an efficient backend ML model. Moreover, INN-TD and immersed INN-TD solvers have the potential to fundamentally transform how parameterized partial differential equations are solved in real engineering analysis and design. Accordingly, the author is highly motivated to pursue commercialization opportunities for the INN technology introduced in this dissertation.
546 $a English
590 $a School code: 0163
650 4 $a Mechanical engineering. $3 649730
650 4 $a Computer engineering. $3 621879
650 4 $a Mechanics. $3 525881
653 $a Computational mechanics
653 $a Machine Learning
653 $a Solid mechanics
653 $a Finite Element Analysis
653 $a Interpolating Neural Network
690 $a 0548
690 $a 0800
690 $a 0346
690 $a 0464
710 2 $a Northwestern University. $b Mechanical Engineering. $e degree granting institution. $3 3822052
720 1 $a Liu, Wing Kam $e degree supervisor.
773 0 $t Dissertations Abstracts International $g 86-12B.
790 $a 0163
791 $a Ph.D.
792 $a 2025
856 4 0 $u https://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=32039084