Details
| Original language | English |
|---|---|
| Article number | 106719 |
| Journal | Computers and Structures |
| Volume | 260 |
| Early online date | 9 Dec 2021 |
| Publication status | Published - Feb 2022 |
Abstract
Surrogate modelling has emerged as a useful technique to study complex physical and engineering systems in various disciplines, especially for engineering analysis. Previous studies mostly focused on developing new surrogate models and/or applying existing surrogate models to practical problems. Despite the computational efficiency, the surrogate for a new task is often built from scratch and the knowledge gained from previous surrogate modelling for similar tasks is neglected. As the need for quickly modifying simulation models to reflect design changes has significantly increased, one potential solution is to utilize prior knowledge from surrogate modelling. In this study, a novel meta-learning-based surrogate modelling framework is presented. The framework includes two phases: a meta-training and a few-shot learning phase. A meta-model that represents a family of tasks and the adaptation of this model to a new task with few data points are the results of the first and second phase, respectively. The study specifies the scope of the framework by classifying similar tasks. Applications of the framework to global sensitivity analysis, optimization, and reliability analysis are also addressed. Four numerical experiments are performed to demonstrate the feasibility and applicability of the framework.
Keywords
- Knowledge transfer, Meta-learning-based surrogate modelling, Model-agnostic meta-learning, Surrogate modelling
ASJC Scopus subject areas
- Engineering(all)
- Civil and Structural Engineering
- Mathematics(all)
- Modelling and Simulation
- Materials Science(all)
- General Materials Science
- Engineering(all)
- Mechanical Engineering
- Computer Science(all)
- Computer Science Applications
Cite this
- Standard
- Harvard
- Apa
- Vancouver
- BibTeX
- RIS
In: Computers and Structures, Vol. 260, 106719, 02.2022.
Research output: Contribution to journal › Article › Research › peer review
}
TY - JOUR
T1 - Transfer prior knowledge from surrogate modelling
T2 - A meta-learning approach
AU - Cheng, Minghui
AU - Dang, Chao
AU - Frangopol, Dan M.
AU - Beer, Michael
AU - Yuan, Xian-Xun
N1 - Funding Information: The second author is grateful for the financial support received from China Scholarship Council (CSC).
PY - 2022/2
Y1 - 2022/2
N2 - Surrogate modelling has emerged as a useful technique to study complex physical and engineering systems in various disciplines, especially for engineering analysis. Previous studies mostly focused on developing new surrogate models and/or applying existing surrogate models to practical problems. Despite the computational efficiency, the surrogate for a new task is often built from scratch and the knowledge gained from previous surrogate modelling for similar tasks is neglected. As the need for quickly modifying simulation models to reflect design changes has significantly increased, one potential solution is to utilize prior knowledge from surrogate modelling. In this study, a novel meta-learning-based surrogate modelling framework is presented. The framework includes two phases: a meta-training and a few-shot learning phase. A meta-model that represents a family of tasks and the adaptation of this model to a new task with few data points are the results of the first and second phase, respectively. The study specifies the scope of the framework by classifying similar tasks. Applications of the framework to global sensitivity analysis, optimization, and reliability analysis are also addressed. Four numerical experiments are performed to demonstrate the feasibility and applicability of the framework.
AB - Surrogate modelling has emerged as a useful technique to study complex physical and engineering systems in various disciplines, especially for engineering analysis. Previous studies mostly focused on developing new surrogate models and/or applying existing surrogate models to practical problems. Despite the computational efficiency, the surrogate for a new task is often built from scratch and the knowledge gained from previous surrogate modelling for similar tasks is neglected. As the need for quickly modifying simulation models to reflect design changes has significantly increased, one potential solution is to utilize prior knowledge from surrogate modelling. In this study, a novel meta-learning-based surrogate modelling framework is presented. The framework includes two phases: a meta-training and a few-shot learning phase. A meta-model that represents a family of tasks and the adaptation of this model to a new task with few data points are the results of the first and second phase, respectively. The study specifies the scope of the framework by classifying similar tasks. Applications of the framework to global sensitivity analysis, optimization, and reliability analysis are also addressed. Four numerical experiments are performed to demonstrate the feasibility and applicability of the framework.
KW - Knowledge transfer
KW - Meta-learning-based surrogate modelling
KW - Model-agnostic meta-learning
KW - Surrogate modelling
UR - http://www.scopus.com/inward/record.url?scp=85120950147&partnerID=8YFLogxK
U2 - 10.1016/j.compstruc.2021.106719
DO - 10.1016/j.compstruc.2021.106719
M3 - Article
AN - SCOPUS:85120950147
VL - 260
JO - Computers and Structures
JF - Computers and Structures
SN - 0045-7949
M1 - 106719
ER -