Data-Driven Prediction of Crack Evolution in Regional Concrete Bridges: Comparison of Multiple Machine Learning Models

The maximum crack width of the main girder of concrete girder bridges is one of the key indicators for evaluating bridge service performance and formulating maintenance decisions. At present, crack information in engineering practice is obtained mainly through manual inspection, which involves a certain lag and makes it difficult to meet the requirements of modern bridges for preventive and predictive maintenance. This study compared 12 regression models with GridSearch and with the Tree-structured Parzen Estimator (TPE) in Optuna, producing 24 model-optimization combinations. The dataset contains 669 bridge inspection records of 170 expressway bridges from 2019 to 2022, covering 12 structural, environmental, traffic, and maintenance features. Before preprocessing, the records were assigned to a training set and a test set containing 535 and 134 samples, respectively. Both optimization strategies used the same bridge-grouped five-fold cross-validation split and maximized the mean validation R2, with preprocessing fitted independently within each training fold. CatBoost–Optuna achieved the highest observed test R2 of 0.9208, with RMSE = 0.0484 and MAE = 0.0387 on the standardized MCW scale. Gradient boosting models performed best in this comparison. SHapley Additive exPlanations (SHAP) further identified bridge age, heavy-vehicle traffic, and cumulative vehicle load as the main contributors to the model predictions. The results provide a regional benchmark for MCW estimation. In future work, covariate scenarios can be supplied to the model to explore maintenance demand.

Authors

Institutions

Publication Details

Journal
Infrastructures
Published
2026-11-18
DOI
https://doi.org/10.3390/infrastructures11110427
Primary Topic
Infrastructure Maintenance and Monitoring
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Data-Driven Prediction of Crack Evolution in Regional Concrete Bridges: Comparison of Multiple Machine Learning Models

Dong Liang, Haibin Huang, Jing Shi, Ning Liang et al.
1 links
Infrastructures
Infrastructure Maintenance and Monitoring
article

Data-Driven Prediction of Crack Evolution in Regional Concrete Bridges: Comparison of Multiple Machine Learning Models

Dong Liang, Haibin Huang, Jing Shi, Ning Liang, Ding-Yu Shen, Zhi-Yi Zhang
article en

Abstract

The maximum crack width of the main girder of concrete girder bridges is one of the key indicators for evaluating bridge service performance and formulating maintenance decisions. At present, crack information in engineering practice is obtained mainly through manual inspection, which involves a certain lag and makes it difficult to meet the requirements of modern bridges for preventive and predictive maintenance. This study compared 12 regression models with GridSearch and with the Tree-structured Parzen Estimator (TPE) in Optuna, producing 24 model-optimization combinations. The dataset contains 669 bridge inspection records of 170 expressway bridges from 2019 to 2022, covering 12 structural, environmental, traffic, and maintenance features. Before preprocessing, the records were assigned to a training set and a test set containing 535 and 134 samples, respectively. Both optimization strategies used the same bridge-grouped five-fold cross-validation split and maximized the mean validation R2, with preprocessing fitted independently within each training fold. CatBoost–Optuna achieved the highest observed test R2 of 0.9208, with RMSE = 0.0484 and MAE = 0.0387 on the standardized MCW scale. Gradient boosting models performed best in this comparison. SHapley Additive exPlanations (SHAP) further identified bridge age, heavy-vehicle traffic, and cumulative vehicle load as the main contributors to the model predictions. The results provide a regional benchmark for MCW estimation. In future work, covariate scenarios can be supplied to the model to explore maintenance demand.

InfrastructuresVol. 11(11)
Hebei University of Technology (CN), China Railway Shanghai Design Institute Group (China) (CN), China Railway Major Bridge Reconnaissance & Design Institute (China) (CN), China Railway Construction Machinery Research & Design Institute (CN)
National Natural Science Foundation of China
Industry, innovation and infrastructure, Sustainable cities and communities
Infrastructure Maintenance and Monitoring
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon