Skip to main navigation Skip to search Skip to main content

GNN-Based Structural Dynamics Simulation for Modular Buildings

  • Jun Zhang
  • , Tong Zhang
  • , Ying Wang*
  • *Corresponding author for this work
  • Harbin Institute of Technology Shenzhen
  • Peng Cheng Laboratory

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Modular buildings are made up of standardized building sections manufactured in a controlled environment. Their advantages include speedy construction process, cost-effectiveness, and higher quality. Further, it provides an opportunity to perform data-driven numerical simulations through a standardized process. Due to its capability of topological generalization, a Graph Neural Network (GNN) based approach is proposed in this study as an innovative structural dynamics simulation tool. The proposed approach can predict the dynamic response of structures with different topologies by changing the relationship matrix. To demonstrate its effectiveness, three spring-mass systems, with 3-DOF, 6-DOF, and 10-DOF, are used as examples of modular buildings. The dynamic response data of the 3-DOF system are used to train a GNN model. After necessary topological adaptation, the model is then used to predict the dynamic response of the 6-DOF system and the 10-DOF system, without any new training process. The results show that the proposed approach can predict the dynamic responses of structures with different topologies with a very low peak mean square error (PMSE), which is less than 0.01. It has the potential to enhance the generalization capabilities of data-driven numerical simulation methods.

Original languageEnglish
Title of host publicationPattern Recognition and Computer Vision - 5th Chinese Conference, PRCV 2022, Proceedings
EditorsShiqi Yu, Jianguo Zhang, Zhaoxiang Zhang, Tieniu Tan, Pong C. Yuen, Yike Guo, Junwei Han, Jianhuang Lai
PublisherSpringer Science and Business Media Deutschland GmbH
Pages245-258
Number of pages14
ISBN (Print)9783031189128
DOIs
StatePublished - 2022
Externally publishedYes
Event5th Chinese Conference on Pattern Recognition and Computer Vision, PRCV 2022 - Shenzhen, China
Duration: 4 Nov 20227 Nov 2022

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume13536 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference5th Chinese Conference on Pattern Recognition and Computer Vision, PRCV 2022
Country/TerritoryChina
CityShenzhen
Period4/11/227/11/22

Keywords

  • AI for Science
  • Graph neural networks
  • Modular construction
  • Structural dynamic response prediction
  • Structural health monitoring

Fingerprint

Dive into the research topics of 'GNN-Based Structural Dynamics Simulation for Modular Buildings'. Together they form a unique fingerprint.

Cite this