Deep learning for estimating low-range concrete sub-surface boundary depths using ground penetrating radar signals

This letter proposes a deep learning approach for nondestructively detecting concrete subsurface boundaries between corroded and noncorroded layers using ground penetrating radar (GPR). We utilize a finite difference time domain technique to simulate GPR electromagnetic wave propagation on various concrete models mimicking corrosion situations. Following that, a deep learning method based on convolutional neural networks is utilized to estimate the bulk relative permittivity of the compound concrete structure, as well as a multilayer perceptron-based method for clutter removal through surface wave prediction. By estimating relative permittivity and removing clutter in GPR signals, the proposed approach can reliably detect the subsurface boundaries, which is demonstrated by the evaluation results.