Improved GPR-based Condition Assessment of Reinforced Concrete Bridge Decks Using Artificial Neural Network

Accounting for the effect of rebar depth variation is one of the most important and challenging tasks to accurately assess the condition of reinforced concrete elements using ground penetrating radar (GPR) technique. In current practices, this task is performed on the individual basis, as for each bridge deck a unique depth correction model is derived from the GPR data collected on it. It is found that such a practice has led to a limited capability of GPR in assessing the condition of highly-deteriorated concrete. Therefore, a generic model to account for the depth-amplitude effect is proposed in this research. Using artificial neural network (ANN) modeling, a model for depth correction was calibrated from extensive data collected for a group of bare concrete bridge decks. The obtained ANN for depth correction was then used to assess the condition of a bridge deck, and the attenuation map was compared with those using a traditional depth correction technique. Whereas the conventional approach only detected the relative difference in condition between local deck areas, the outputs using the proposed methodology clearly indicated its capability to assess deck deterioration in absolute terms.