EMI-based Rebar Localization using Neural Network Regression Tool

Rebar is the primary support in lateral infrastructures such as a bridge superstructure (bridge deck) and substructures such as an abutment and pier where geometric measurements are essential elements. Human inspectors mainly utilize hand-held and non-destructive measurement devices such as radar to measure various parameters. However, due to radar's dependency on concrete dielectric properties, electromagnetic induction (EMI) can be an alternative for measurements on highly moisture structures, especially bridge piers. In this study, an EMI-based localization device is developed that can be used for aerial infrastructure inspection. Likewise, a neural network model was developed, and gathered data were trained using a regression tool to characterize the system’s behavior. Specifically, this study applied the circular mapping concept that makes it significant and suitable for pier rebar scanning baseline; a direct measurement approach was used for B-scan. The system applied the time-domain method for the self-induced magnetic signal with the HMC5883L magnetometer that produced EMI data coherently combined with a known parameter such as rebar diameter and depth. Using the Levenberg-Marquardt algorithm for neural network training, the system shows a high correlation between the input and output with an overall regression R equivalent to 0.9964.