Date Approved
6-9-2025
Embargo Period
6-9-2025
Document Type
Thesis
Degree Name
M.S. Civil Engineering
Department
Civil engineering
College
Henry M. Rowan College of Engineering
Advisor
Adriana Trias Blanco, Ph.D.
Committee Member 1
Gilson Lomboy, Ph.D.
Committee Member 2
Kaz Tabrizi, Ph.D.
Keywords
Bridge Engineering;Corrosion in Bridges;Damage detection;Ground Penetrating Radar;Signal Processing;Structure health monitoring
Disciplines
Civil and Environmental Engineering | Civil Engineering | Engineering
Abstract
This research addresses the critical issue of deteriorating civil infrastructure by utilizing GPR integrated with advanced signal processing and machine learning. Employing the Conquest 100 system, four concrete slab configurations were tested: multi-layer rebar with defects, single- and double-layer rebar setups, varied rebar spacing, and pre-corroded rebars (0–30% mass loss). GPR B-scans collected over 3–60 days demonstrated that detection accuracy varied significantly with curing time, structural complexity, and moisture levels. A custom-developed Python script surpassed EKKO_Project software in rebar localization, achieving lower position errors (0.4–0.7 inches compared to 0.5–1.25 inches). Notably, the amplitude of GPR signals unexpectedly increased with corrosion progression (from 9.667 mV at 20% mass loss on Day 3 to 16.000 mV on Day 28), challenging conventional understanding by suggesting rust-induced scattering enhances reflections. A semi-supervised deep learning technique—Bootstrap Your Whole Latent (BYWL)—trained on 1,543 synthetic images and fine-tuned on 179 real GPR scans, demonstrated superior rebar localization accuracy (RMSE: 19.05 pixels; errors down to 2%), outperforming baseline models. This methodology minimizes manual interpretation and enhances inspection reliability under challenging conditions and holds significant promise for infrastructure assessment and non-destructive evaluation.
Recommended Citation
Chowdhury, Priyam, "ENHANCING GPR RELIABILITY FOR CONCRETE INFRASTRUCTURE: INTEGRATING SIGNAL PROCESSING AND DEEP LEARNING FOR DEFECT AND CORROSION ASSESSMENT" (2025). Theses and Dissertations. 3384.
https://rdw.rowan.edu/etd/3384