Graduation Semester and Year
Spring 2026
Language
English
Document Type
Thesis
Degree Name
Master of Science in Civil Engineering
Department
Civil Engineering
First Advisor
Dr. Nur Yazdani
Second Advisor
Dr. Himan Hojat Jalali
Third Advisor
Dr. Panagiotis Danoglidis
Abstract
Corrosion detection remains a crucial factor in the proper maintenance of Reinforced Concrete (RC) structures. While several studies have used Ground Penetrating Radar (GPR) for corrosion detection, quantitative models for prediction are still limited. This study presents an approach for the quantitative prediction of rebar corrosion in concrete using GPR data obtained from a prior experimental program involving accelerated corrosion. A total of 36 RC samples were cast, varying in cover depth, rebar diameter, and concrete porosity. The samples were subjected to an impressed current of 0.65 A in a 5% NaCl solution over three phases: 10, 20, and 30 days. GPR scans were performed at the end of each phase. The samples were then destructively tested to obtain individual rebar mass loss, which served as ground truth. GPR signals were processed to remove outliers, reduce noise, and improve data quality for analysis. Key GPR reflection features, including two-way travel time, amplitude, dielectric constant, and attenuation associated with bar reflections, were extracted and assigned as individual data points, resulting in a dataset of 720 points. Ten Machine Learning (ML) models were trained for the prediction of rebar mass loss and validated using two external samples. Their performance metrics were evaluated and compared, and the influence of the parameters on the best-performing models was discussed. The Artificial Neural Network (ANN) model performed best within the original dataset, achieving an R2 of 0.998, while the Gradient Boosting Machine (GBM) and Bayesian Regression models demonstrated strong generalization on the two external samples, achieving R2 values of 0.833 and 0.799, respectively. Results demonstrate the potential of GPR, combined with ML, for non-destructive, quantitative prediction of corrosion in RC structures.
Keywords
Reinforced Concrete (RC), Non-Destructive Testing (NDT), Ground Penetrating Radar (GPR), Machine Learning (ML), Corrosion
Disciplines
Civil Engineering | Structural Engineering
License
This work is licensed under a Creative Commons Attribution 4.0 International License.
Recommended Citation
Shafiullah, Nafisa, "DATA-DRIVEN PREDICTION OF CONCRETE BRIDGE DECK CORROSION USING GROUND PENETRATING RADAR (GPR)" (2026). Civil Engineering Theses. 4.
https://mavmatrix.uta.edu/civilengineering_theses2/4