Estimating Location and Thickness of a Void in a Concrete Slab using Artificial Neural Network

Abstract

The investigation of voids in concrete slabs is a crucial aspect of concrete diagnostics. We have reported that it is possible to efficiently detect the presence of voids by identifying the scattered waveform through an artificial neural network (ANN). This study serves as a sequel, aiming not only to determine the presence of a void but also to detect its location and thickness. The ANN-based estimation demonstrated good accuracy even when the observed data included some noise or when dealing with models featuring weak conductivity. Additionally, a substantial amount of scattered waveform data is required to train and test the ANN. In this research, we leverage frequency domain analysis and FFT techniques to create the scattered waveform data of a concrete slab considered as a flat layered medium. This approach allows for significant time savings in computation compared to the finite difference time domain (FDTD) method used in our previous studies.