All Issue

2026 Vol.42, Issue 3 Preview Page
Analysis of the Effects of Training Conditions on Deep Learning-Based Prediction of Liquefaction-Induced Damage to Caisson-Type Quay Walls

액상화에 의한 딥러닝 기반 케이슨식 안벽 피해 예측을 위한 학습 조건의 영향 분석

Hyeonsu Yun1
,
Seong-Kyu Yun2
,
Jiseong Kim3
,
Gichun Kang4*
윤 현수1
,
윤 성규2
,
김 지성3
,
강 기천4*
1Member, Ph.D Candidate, Department of Civil Engineering, College of Engineering, Gyeongsang National Univ.
2Member, Ph.D, Department of Civil Engineering, College of Engineering, Gyeongsang National Univ.
3Member, Assistant Prof., Department of Land Information and Construction, Vision College of Jeonju
4Member, Associate Prof., Department of Civil Engineering, College of Engineering, Gyeongsang National Univ.
1정회원, 경상국립대학교 공과대학 토목공학과 박사과정
2정회원, 경상국립대학교 공과대학 토목공학과 공학박사
3정회원, 전주비전대학교 국토정보건설과 조교수
4정회원, 경상국립대학교 공과대학 토목공학과 부교수
*Corresponding Author
30 June 2026. pp. 113-124
PDF XML Citation
Abstract

Recent studies in geotechnical engineering have increasingly focused on deep learning-based approaches for the rapid prediction of structural responses under seismic loading. In particular, caisson-type quay walls constructed on liquefiable ground exhibit time-dependent nonlinear behavior, creating a growing demand for data-driven prediction methods that can replace repetitive numerical analyses. However, existing studies have primarily focused on improving the prediction accuracy of deep learning models, whereas systematic investigations of how training conditions affect prediction stability and reliability remain limited. Therefore, this study constructed time-series datasets of settlement and horizontal displacement using FLIP numerical analyses under various ground conditions (SPT N-values) and seismic intensities (0.1–2.0 m/s2). Based on these datasets, the effects of training conditions on the prediction performance and stability of deep learning models were evaluated. The results demonstrated stable prediction performance and reproducibility for batch sizes of 8–32, whereas the batch size of 64 resulted in increased prediction variability.

Keywords
Batch size
Caisson-type quay wall
Deep learning
Earthquake
Liquefaction

최근 지반공학 분야에서는 지진 시 구조물 응답을 신속하게 예측하기 위한 딥러닝 기반 연구가 활발히 수행되고 있다. 특히 액상화 지반에서의 케이슨식 안벽과 같은 항만 구조물은 시간 의존적 비선형 거동을 보이기 때문에, 반복적인 수치해석을 대체할 수 있는 데이터 기반 예측 기법의 필요성이 증가하고 있다. 그러나 기존 연구는 주로 딥러닝 모델의 예측 정확도 향상에 초점을 두고 있으며, 학습 조건에 따라 예측 결과의 안정성과 신뢰성이 어떻게 변화하는지에 대한 체계적인 분석은 부족한 실정이다. 이에 본 연구에서는 FLIP 수치해석을 통해 다양한 지반 조건(N치)과 지진동 세기(0.1~2.0 m/s2)에 따른 침하량 및 수평변위의 시계열 데이터를 구축하고, 이를 기반으로 딥러닝 모델의 학습 조건이 예측 성능 및 결과 안정성에 미치는 영향을 분석하였다. 분석결과 batch size 8~32 범위에서는 안정적인 예측 성능과 재현성이 확보된 반면, batch size 64 조건에서는 변동성이 증가하는 경향이 나타났다.

References
1

Cho, K., van Merriënboer, B., Gulcehre, C., Bahdanau, D., Bougares, F., Schwenk, H., and Bengio, Y. (2014), “Learning Phrase Representations using RNN Encoder–Decoder for Statistical Machine Translation”, arXiv preprint, arXiv:1406.1078.

10.3115/v1/D14-1179
2

Han, S.-J., Choo, Y. W., Yoon, S., Han, S.-H., and Kim, J.-E. (2025), “Study on Deep Learning-Based Image Analysis for Structural Vibration in Dynamic Centrifuge Modeling”, Journal of the Korean Geotechnical Society, Vol.41, No.4, pp.53-64. https://doi.org/10.7843/kgs.2025.41.4.53.

10.7843/kgs.2025.41.4.53
3

Hochreiter, S., and Schmidhuber, J. (1997), “Long Short-Term Memory”, Neural Computation, Vol.9, No.8, pp.1735-1780.

10.1162/neco.1997.9.8.1735
4

Hoffer, E., Hubara, I., and Soudry, D. (2017), “Train Longer, Generalize Better: Closing the Generalization Gap in Large Batch Training of Neural Networks”, arXiv preprint arXiv:1705.08741, https://arxiv.org/abs/1705.08741.

5

Hong, S., Kwak, T.-Y., Woo, S.-I., and Kim, S.-R. (2024), “Comparison of Deep Learning and Measurement-Based Methods for Settlement Prediction in Soft Ground”, Journal of the Korean Geotechnical Society, Vol.40, No.6, pp.67-78. https://doi.org/10.7843/kgs.2024.40.6.67.

10.7843/kgs.2024.40.6.67
6

Iai, S., Matsunaga, T., and Kameoka, T. (1992a), “Strain Space Plasticity Model for Cyclic Mobility”, Soils and Foundations, Vol.32, No.2, pp.1-15.

10.3208/sandf1972.32.2_1
7

Iai, S., Matsunaga, T., and Kameoka, T. (1992b), “Analysis of Undrained Cyclic behavior of Sand under Anisotropic Consolidation”, Soils and Foundations, Vol.32, No.2, pp.16-20.

10.3208/sandf1972.32.2_16
8

Keskar, N. S., Mudigere, D., Nocedal, J., Smelyanskiy, M., and Tang, P. T. P. (2017), “On Large-Batch Training for Deep Learning: Generalization Gap and Sharp Minima”, Proceedings of the International Conference on Learning Representations (ICLR), https://doi.org/10.48550/arXiv.1609.04836.

10.48550/arXiv.1609.04836
9

Masters, D. and Luschi, C. (2018), “Revisiting Small Batch Training for Deep Neural Networks”, Proceedings of the International Conference on Learning Representations (ICLR), https://doi.org/10.48550/arXiv.1804.07612.

10.48550/arXiv.1804.07612
10

Qian, X. and Klabjan, D. (2020), “The Impact of the Mini-batch Size on the Variance of Gradients in Stochastic Gradient Descent”, arXiv preprint arXiv:2004.13146, https://arxiv.org/abs/2004.13146.

11

Smith, S. L. and Le, Q. V. (2018), “A Bayesian Perspective on Generalization and Stochastic Gradient Descent”, Proceedings of the International Conference on Learning Representations (ICLR), https://doi.org/10.48550/arXiv.1710.06451.

10.48550/arXiv.1710.06451
12

Yun, H.S., Back, M-J., Sun, J., Yun, S-K., and Kang, G-C. (2024), “A Simplified Method for Evaluating Damage of Caisson-Type Quay Wall During Earthquakes”, Journal of the Korean Geosynthetics Society, Vol.23, No.3, pp.1-14. https://doi.org/10.12814/jkgss.2024.23.3.001.

10.12814/jkgss.2024.23.3.001
13

Yun, H.S., Yun, S-K., Lee, S., and Kang, G-C. (2026). “Prediction of Settlement of Caisson-type Quay Walls Using Deep Learning During Earthquakes”, Journal of the Korean Geotechnical Society, Vol.42, No.2, pp.89-99. https://doi.org/10.7843/kgs.2026.42.2.89.

10.7843/kgs.2026.42.2.89
14

Yun, S.K, Kim, S.J., Kim, J.S., and Kang, G-C. (2021), “Dynamic behavior of Quay Walls According to N-values of Backfill and Amplitude of Seismic Accelerations”, J. Korean Soc. Hazard Mitig., Vol.21, No.1, pp.207-217. https://doi.org/10.9798/KOSHAM.2021.21.1.207.

10.9798/KOSHAM.2021.21.1.207
Information
  • Publisher :The Korean Geotechnical Society
  • Publisher(Ko) :한국지반공학회
  • Journal Title :Journal of the Korean Geotechnical Society
  • Journal Title(Ko) :한국지반공학회 논문집
  • Volume : 42
  • No :3
  • Pages :113-124
  • Received Date : 2026-06-02
  • Revised Date : 2026-06-23
  • Accepted Date : 2026-06-23
  • DOI :https://doi.org/10.7843/kgs.2026.42.3.113
Journal Informaiton Journal of the Korean Geotechnical Society Journal of the Korean Geotechnical Society
Archives
: Vol.23~28 (2007~2012)
  • NRF
  • KOFST
  • KISTI Current Status
  • KISTI Cited-by
  • crosscheck
  • orcid
  • open access
  • ccl
Journal Informaiton Journal Informaiton - close