Computation and verification of steel constructions using physics-informed artificial intelligence

被引：8

作者：

Kraus, Michael A. ^{[1
]}

Taras, Andreas ^{[2
]}

机构：

[1] Stanford Univ, CEE, Y2E2,473 Via Ortega, Stanford, CA 94305 USA

[2] Swiss Fed Inst Technol, Dept BAUG, Inst Baustat & Konstrukt IBK, Stahlbau & Verbundbau, Stefano Franscini Pl 5, CH-8093 Zurich, Switzerland

来源：

STAHLBAU | 2020年 / 89卷 / 10期

关键词：

physics-informed artificial intelligence; digital twin; machine learning; deep learning; design; structural verification; THEORY-GUIDED DATA; SCIENCE; DESIGN;

D O I：

10.1002/stab.202000074

中图分类号：

TU [建筑科学];

学科分类号：

0813 ;

摘要：

Currently the technology of artificial intelligence (AI) spreads into research and industry practice of all branches in diverse forms. Given that situation, this article serves to introduce the reader to theoretical background on AI in general as well as to the specific case of physics-informed AI (PIKI). Selected examples from design and verification practice of steel construction then illustrate the application of physics-informed neural nets (PINN) methods, where the specific requirements for the formulation of the learning problem are highlighted. PINN serves as an alternative to established computational methods for design of steel structures and its components using available experimental and simulation data. This enables the interpretation and use of PINNs being the digital twin of a steel structure over its lifecycle. PIKI as presented here does not per se cause a "big data" situation and is therefore interesting for engineering research and practice. This paper eventually gives perspectives on future applications of AI for steel construction.

引用

页码：824 / 832

页数：9

共 28 条

[1]

Bishop CM., 2006, Pattern Recognition and Machine Learning

[2]

Brownlee J., 2016, XGBoost With python: Gradient boosted trees with XGBoost and scikit-learn

[3] Dimensioning of silicone adhesive joints: Eurocode-compliant, mesh-independent approach using the FEM [J].

Drass, Micheal ;

Kraus, Michael A. .

GLASS STRUCTURES & ENGINEERING, 2020, 5 (03) :349-369

[4] Theory-Guided Data Science for Climate Change [J].

Faghmous, James H. ;

Banerjee, Arindam ;

Shekhar, Shashi ;

Steinbach, Michael ;

Kumar, Vipin ;

Ganguly, Auroop R. ;

Samatova, Nagiza .

COMPUTER, 2014, 47 (11) :74-78

[5]

Frochte J, 2019, Maschinelles Lernen: Grundlagen und Algorithmen in Python

[6]

Goodfellow I, 2016, ADAPT COMPUT MACH LE, P1

[7]

Goulet J.-A., 2020, Probabilistic machine learning for civil engineers

[8]

Heaton J, 2018, Genet Program Evolvable Mach, V19, P305, DOI [DOI 10.1007/S10710-017-9314-Z, 10.1007/s10710-017-9314-z.pdf, DOI 10.1007/S10710-017-9314-Z.PDF, 10.1007/s10710-017-9314-z]

[9] MULTILAYER FEEDFORWARD NETWORKS ARE UNIVERSAL APPROXIMATORS [J].

HORNIK, K ;

STINCHCOMBE, M ;

WHITE, H .

NEURAL NETWORKS, 1989, 2 (05) :359-366

[10]

Karpatne A., 2017, ARXIV

← 1 2 3 →