We combine structural testing, field inspection, computer vision, and machine learning to identify damage and measure its progression. The goal is to make condition assessment more consistent, quantitative, and useful for maintenance and repair decisions.
Wang, J., Kim, Y.J., and Liu, C. · ACI Structural Journal, 120(3) (2023), 3–15.View paper ↗
A self-built convolutional neural network processes thousands of experimental images to identify cracking in UHPC, while agent-based modeling separates cracks from surface noise and quantifies damage. The combined workflow turns visual observations into repeatable measurements of crack width, area, and post-peak deterioration.
Kim, Y.J. and Wang, J. · ACI Structural Journal, 120(2) (2023), 3–17.View paper ↗
This case study evaluates five asphalt-overlaid concrete bridge decks in Colorado using visual inspection, chain dragging, rebound hammer testing, and ground-penetrating radar (GPR). It examines how GPR scanning rates affect the estimated deterioration area and develops analytical and agent-based models that separate corrosion initiation and progression from concrete delamination, improving agreement with actual repair areas.
Ji, Y., Wang, D., and Wang, J. (2024). Study of Recycled Concrete Properties and Prediction Using Machine Learning Methods. Journal of Building Engineering, 94, 110067.
Wang, J., Kim, Y.J., and Liu, C. (2023). Deep learning for detection and characterization of cracking in UHPC. ACI Structural Journal, 120(3), 3–15.
Kim, Y.J. and Wang, J. (2023). Nondestructive testing of bridge decks: a case study and suggestions. ACI Structural Journal, 120(2), 3–17.