All Issue

2026 Vol.39, Issue 3 Preview Page

Research Paper

Attention Mechanism-Based TabNet for Predicting Seismic Response Strain of Buried Pipelines

어텐션 메커니즘 기반 TabNet을 활용한 매설 배관의 지진 응답 변형률 예측 모델 개발

Yangrok Choi1
,
Youngjae Bae2
,
Seungchan Oh3
,
Kyungrok Kwon4
,
Jung Sik Kong5*
최 양록1
,
배 영재2
,
오 승찬3
,
권 경록4
,
공 정식5*
1Postdoctoral Researcher, Future and Fusion Lab of Architectural, Civil and Environmental Engineering, Korea University, Seoul, 02841, Korea
2Ph.D. Student, School of Civil, Environmental and Architectural Engineering, Korea University, Seoul, 02841, Korea
3Junior Engineer, Structural Engineering Department, BANDI Consultants, Seoul, 31075, Korea
4Postdoctoral Researcher, Division of Ocean Space Development & Energy Research Department, Institute of Ocean Science and Technology, Busan, 49111, Korea
5Professor, School of Civil, Environmental and Architectural Engineering, Korea University, Seoul, 02841, Korea
1고려대학교 미래건설환경융합연구소 박사후연구원
2고려대학교 건축사회환경공학과 박사과정
3반디컨설턴트 구조부 연구원
4해양과학기술원 해양공간개발·에너지연구부 박사후연구원
5고려대학교 사회환경시스템공학과 교수
*Corresponding Author
30 June 2026. pp. 153-161
PDF XML Citation
Abstract

Three-dimensional finite element analysis (FEA), which is widely used for the seismic response analysis of buried pipelines, has high computational costs. Furthermore, conventional response surface methodology (RSM) is limited in its ability to accurately simulate complex nonlinear relationships. To address these issues, this study created a database of 2,000 finite element analyses of buried pipelines and proposes a strain prediction model based on TabNet, a deep learning architecture designed specifically for tabular data. The proposed approach outperformed conventional RSM prediction accuracy and stability for both maximum tensile and compressive strains within a single model. In addition, by quantifying feature importance using an attention mechanism, the model overcomes the inherent "black-box" limitations of artificial intelligence (AI) and secures structural interpretability. These findings will significantly contribute to the efficient seismic performance evaluation of buried pipelines and the development of optimal design guidelines in the future.

Keywords
buried pipeline
seismic response
finite element analysis
TabNet
attention mechanism

매설 배관의 지진 응답 분석에 주로 쓰이는 3차원 유한요소해석은 막대한 연산 비용을 수반하며, 기존 반응표면법(RSM)은 복잡한 비선형 관계 모사에 한계가 있다. 이를 극복하기 위해 본 연구는 2,000건의 매설 배관 유한요소해석 데이터베이스를 구축하고, 정형 데이터에 특화된 딥러닝 아키텍처인 TabNet 기반의 변형률 예측 모델을 제안한다. 제안된 모델은 단일 모델로 최대 인장 및 압축 변형률 모두에 대해 기존 RSM보다 높은 예측 정확도와 안정성을 입증했다. 또한, 어텐션 메커니즘을 통해 변수 영향도를 정량화함으로써 AI의 블랙박스 한계를 극복하고 구조적 해석력을 확보하였다. 이는 향후 매설 배관의 효율적인 내진 성능 평가 및 최적화된 설계 가이드라인 수립에 기여할 것이다.

키워드
매설 배관
지진 응답
유한요소해석
TabNet
어텐션 메커니즘
References
1

American Lifelines Alliance (2001) Guidelines for the Design of Buried Steel Pipe, American Society of Civil Engineers (ASCE), Reston, VA, USA.

2

Arik, S.Ö., Pfister, T. (2021) TabNet: Attentive Interpretable Tabular Learning, In Proceedings of the AAAI Conference on Artificial Intelligence, 35(8), pp.6679~6687.

10.1609/aaai.v35i8.16826
3

Choi, Y., Ann, H., Kwon, K., Jang, Y.C., Kim, Y. P., Kong, J.S. (2020) Seismic Structural Analysis of Buried Pipelines and Safety Evaluation Model based on Response Surface Method, J. Korean Soc. Hazard Mitig., 20(3), pp.225~236.

10.9798/KOSHAM.2020.20.3.225
4

Khuri, A.I., Mukhopadhyay, S. (2010) Response Surface Methodology, Wiley Interdiscip. Rev.: Comput. Stat., 2(2), pp.128~149.

10.1002/wics.73
5

Oh, S., Kwon, K., Choi, Y., Bae, Y., Kong, J.S. (2026) Displacement-Based Seismic Analysis of Underground Lifelines Using Response Surface-Based Strain Predictions, KSCE J. Civil Eng., 30(6), 100529.

10.1016/j.kscej.2026.100529
6

O'Rourke, M.J., Hmadi, K.E. (1988) Analysis of Continuous Buried Pipelines for Seismic Wave Effects, Earthq. Eng. & Struct. Dyn., 16(6), pp.917~929.

10.1002/eqe.4290160611
7

O'Rourke, M.J., Liu, X. (1999) Response of Buried Pipelines subject to Earthquake Effects, New York: Mceer, p.260.

8

Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A.N., Kaiser, Ł. Polosukhin, I. (2017) Attention is All You Need, Advances in Neural Information Processing Systems, 30.pp.5998~6008.

9

Wang, L.R.L. (1980) Seismic Vulnerability, Behavior and Design of Underground Piping Systems: Static and Quasi-static Seismic Analyses of Buried Pipelines, Technical Report No. 11, Department of Civil Engineering, Rensselaer Polytechnic Institute, Troy, New York.

10

Wang, L.R.L., Cheng, K.M. (1979) Seismic Response behavior of Buried Pipelines, J. Press. Vessel Technol., 101(1), pp.21~30.

10.1115/1.3454594
11

Wong, K.C., Datta, S.K., Shah, A.H. (1986) Three-Dimensional Motion of Buried Pipeline. I: Analysis, J. Eng. Mech., 112(12), pp.1319~1337.

10.1061/(ASCE)0733-9399(1986)112:12(1319)
12

Xie, X., Symans, M.D., O’Rourke, M.J., Abdoun, T.H., O’Rourke, T.D., Palmer, M.C., Stewart, H.E. (2013) Numerical Modeling of Buried HDPE Pipelines subjected to Normal Faulting: A Case Study, Earthq. Spectra, 29(2), pp.609~632.

10.1193/1.4000137
13

Xu, R., Jiang, R., Qu, T.J. (2021) Review of Dynamic Response of Buried Pipelines, J. Pipeline Syst. Eng. & Pract., 12(2), 03120003.

10.1061/(ASCE)PS.1949-1204.0000527
Information
  • Publisher :Computational Structural Engineering Institute of Korea
  • Publisher(Ko) :한국전산구조공학회
  • Journal Title :Journal of the Computational Structural Engineering Institute of Korea
  • Journal Title(Ko) :한국전산구조공학회 논문집
  • Volume : 39
  • No :3
  • Pages :153-161
  • Received Date : 2026-03-18
  • Revised Date : 2026-05-19
  • Accepted Date : 2026-05-20
  • DOI :https://doi.org/10.7734/COSEIK.2026.39.3.153
Journal Informaiton Journal of the Computational Structural Engineering Institute of Korea Journal of the Computational Structural Engineering Institute of Korea
Online Submission
  • scopus
  • NRF
  • KOFST
  • crossref crossmark
  • crossref cited-by
  • crosscheck
  • orcid
  • open access
  • ccl
Journal Informaiton Journal Informaiton - close