All Issue

2025 Vol.3, Issue 2 Preview Page

Research Article

Data Augmentation and Interpretable Deep Learning for Remaining Life Prediction of Highway Bridges

고속도로 교량 결함도지수의 장기 예측을 위한 데이터 증강 및 Deep Lattice Network 모델 개발

Chaemin Lee1
,
Yangrok Choi2
,
Jung Sik Kong3*
이 채민1
,
최 양록2
,
공 정식3*
1Graduate Student, Department of Civil, Environmental and Architectural Engineering, Korea University, Seoul 02841, Korea
2Graduate Student, Department of Civil, Environmental and Architectural Engineering, Korea University, Seoul 02841, Korea
3Professor, Department of Civil, Environmental and Architectural Engineering, Korea University, Seoul 02841, Korea
1고려대학교 건축사회환경공학부 석사과정
2고려대학교 건축사회환경공학부 박사과정
3고려대학교 건축사회환경공학부 교수
*Corresponding Author
31 December 2025. pp. 99-114
PDF XML Citation
Abstract

Machine learning models are commonly used for long-term prediction of bridge deterioration. However, tree-based models such as XGBoost exhibit fundamental extrapolation limitations, where predictions converge to constant values beyond the training data range, failing to capture continued deterioration. This study develops a Deep Lattice Network (DLN) model with monotonicity constraints to overcome this limitation. Monotonicity constraints were applied to key variables including service years, chloride, and traffic volume to ensure physically valid predictions in extrapolation regions. To address the scarcity of high-defect data, balanced data augmentation was implemented to expand the training dataset. The DLN model successfully predicted that the vast majority of bridges would reach D-grade, while XGBoost predicted only a negligible fraction due to prediction saturation. Data augmentation significantly improved prediction performance by reducing prediction errors and substantially increasing the number of test samples in the rare high-defect region. The results demonstrate that combining monotonicity constraints with balanced data augmentation effectively resolves both extrapolation and imbalanced data problems in bridge deterioration prediction, demonstrating the potential for realistic long-term maintenance planning.

Keywords
bridge deterioration prediction
Deep Lattice Network
monotonicity constraints
data augmentation
extrapolation
machine learning

교량 열화의 장기 예측에는 기계학습 모델이 널리 사용된다. 그러나 XGBoost와 같은 트리 기반 모델은 학습 데이터 범위를 벗어나면 예측값이 일정 수준에서 수렴하여 지속적인 열화를 포착하지 못하는 근본적인 외삽 한계를 보인다. 본 연구에서는 이러한 한계를 극복하기 위해 단조성 제약을 적용한 Deep Lattice Network(DLN) 모델을 개발하였다. 공용년수, 염화물, 교통량 등 주요 변수에 단조성 제약을 적용하여 외삽 구간에서도 물리적으로 타당한 예측을 보장하였다. 고결함도 데이터의 희소성 문제를 해결하기 위해 균형잡힌 데이터 증강을 적용하여 학습 데이터를 확대하였다. DLN 모델은 대다수의 교량이 D등급에 도달할 것으로 예측한 반면, XGBoost는 예측값 포화로 인해 극히 일부만 도달할 것으로 예측하였다. 데이터 증강은 예측 오차를 크게 감소시키고 희소 고결함도 구간의 테스트 샘플 수를 대폭 증가시켜 예측 성능을 현저히 개선하였다. 본 연구는 단조성 제약과 균형잡힌 데이터 증강의 결합이 교량 열화 예측에서 외삽 문제와 불균형 데이터 문제를 효과적으로 해결하여 현실적인 장기 유지관리 계획 수립의 가능성을 제시하였다.

키워드
교량 열화 예측
Deep Lattice Network
단조성 제약
데이터 증강
외삽
기계학습
References
1

Aggarwal, C. C. (2018) Neural Networks and Deep Learning. Vol. 10. No. 978. Cham: Springer.

2

Alam, K. M. R., Siddique, N., and Adeli, H. (2020) A Dynamic Ensemble Learning Algorithm for Neural Networks. Neural Computing and Applications 32(12), 8675-8690.

10.1007/s00521-019-04359-7
3

Bektaş, B. A. (2017) Use of Recursive Partitioning to Predict National Bridge Inventory Condition Ratings from National Bridge Elements Condition Data. Transportation Research Record 2612(1), 29-38.

10.3141/2612-04
4

Choi, Y., Bae, Y., Kwon, K., Choi, Y., Sun, J., and Kong, J. S. (2025) Machine Learning-Based Future Performance Prediction Model for Bridge Inspection and Performance Data in South Korea. Advances in Structural Engineering 28(12), 2260-2275.

10.1177/13694332251327835
5

Choi, Y., Kwon, K., and Kong, J. S. (2023) Development of the Deep-Learning Models for Predicting Bridge Performance Considering Time-Series Characteristics. In Proceedings of the Korean Society of Civil Engineers Conference, Jeonnam 19 October 2023. Seoul, Korea: Korean Society of Civil Engineers. (In Korean)

6

Cohen, J. (1988). Statistical Power Analysis for the Behavioral Sciences (2nd ed.). Lawrence Erlbaum Associates.

7

Hong, J., and Jeon, S. (2022) Evaluation of Prediction Performance of the Machine-Learning-Based Models for the Safety Grade Classification of Bridges. In Proceedings of the Korea Concrete Institute Conference, Jeju 2 November 2022. Seoul, Korea: Korea Concrete Institute. (In Korean)

8

Kim, W., Kim, H., Lee, J., and Son, H. (2024) Data Anomaly Detection of Monitoring System. Journal of Korean Society of Disaster and Security 17(4), 117-128.

9

Lee, J., Kim, W., Min, G., and Kim, W. S. (2024) Development of a Predictive Model for Bridge Deck Condition Rating and Defect Index Using Various Machine Learning Algorithms. Journal of the Korea Concrete Institute 36(6), 657-664. (In Korean)

10.4334/JKCI.2024.36.6.657
10

Lyu, C., Lin, S., Lynch, A., Zou, Y., and Liarokapis, M. (2025) UAV-Based Deep Learning Applications for Automated Inspection of Civil Infrastructure. Automation in Construction 177, 106285.

10.1016/j.autcon.2025.106285
11

Manik, M. M., and Sabarethinam, K. (2023) Machine Learning Approach for Predicting Bridge Components' Condition Ratings. Frontiers in Built Environment 9, 1254269.

10.3389/fbuil.2023.1254269
12

Nam, W.-S., Jung, H., Park, K.-H., Kim, C.-M., and Kim, G.-S. (2022) Development of Deep Learning-Based Damage Detection Prototype for Concrete Bridge Condition Evaluation. KSCE Journal of Civil and Environmental Engineering Research 42(1), 107-116. (In Korean)

13

Oh, S. T., Lee, D., and Lee, J. H. (2010) A Condition Rating Method of Bridges Using an Artificial Neural Network Model. Journal of the Korean Society for Railway 13(1), 71-77. (In Korean)

14

Wang, W., and Su, C. (2023) Deep Learning-Based Detection and Condition Classification of Bridge Steel Bearings. Automation in Construction 156, 105085.

10.1016/j.autcon.2023.105085
15

Wellalage, N. K. W., Zhang, T., and Dwight, R. (2020) Bridge Deterioration Modeling by Markov Chain Monte Carlo (MCMC) Simulation Method. Sustainability 12(24), 10625.

Information
  • Publisher :Korean Institute of Bridge and Structural Engineers
  • Publisher(Ko) :한국교량및구조공학회
  • Journal Title :Journal of Structure Research and Practice
  • Journal Title(Ko) :한국교량및구조공학회 논문집
  • Volume : 3
  • No :2
  • Pages :99-114
  • Received Date : 2025-11-25
  • Revised Date : 2025-12-29
  • Accepted Date : 2025-12-29
  • DOI :https://doi.org/10.22725/JSRP.2025.3.2.99
Journal Informaiton Journal of Structure Research and Practice Journal of Structure Research and Practice
Online Submission & Review System
  • KOFST
  • KISTI Current Status
  • KISTI Cited-by
  • KCI 문헌 유사도 검사
  • orcid
  • open access
  • ccl
Journal Informaiton Journal Informaiton - close