Comparison of air-coupled GPR data analysis results determined by multiple analysts

Abstract

Current bridge deck condition assessments using ground penetrating radar (GPR) requires a trained analyst to manually interpret substructure layering information from B-scan images in order to proceed with an intended analysis (pavement thickness, concrete cover, effects of rebar corrosion, etc.) For example, a recently developed method to rapidly and accurately analyze air-coupled GPR data based on the effects of rebar corrosion, requires that a user “picks” a layer of rebar reflections in each B-scan image collected along the length of the deck. These “picks” have information like signal amplitude and two way travel time. When a deck is new, or has little rebar corrosion, the resulting layer of rebar reflections is readily evident and there is little room for subjectivity. However, when a deck is severely deteriorated, the rebar layer may be difficult to identify, and different analysts may make different interpretations of the appropriate layer to analyze. One highly corroded bridge deck, was assessed with a number of nondestructive evaluation techniques including 2GHz air-coupled GPR. Two trained analysts separately selected the rebar layer in each B-scan image, choosing as much information as possible, even in areas of significant deterioration. The post processing of the selected data points was then completed and the results from each analyst were contour plotted to observe any discrepancies. The paper describes the differences between ground coupled and air-coupled GPR systems, the data collection and analysis methods used by two different analysts for one case study, and the results of the two different analyses.