Fast geometric factor approximation to evaluate the electrical resistivity of concrete and reinforced concrete specimens using neural network models

The measurement of electrical resistivity has applications in civil engineering, particularly in condition assessment, structural health monitoring, and material characterization. Accurately evaluating material resistivity requires determining a geometric factor, which depends on specimen geometry, electrode positioning, and the presence of reinforcement. Several methods exist for calculating this factor, including numerical modelling and experimental approaches. This study presents an approximation technique using surrogate models to rapidly calculate geometric factors for non-reinforced and reinforced geometries. The studied geometries are prismatic, with the reinforced case involving a single embedded bar, commonly used in laboratory studies. These surrogate models are based on neural networks trained with data generated through finite element modelling. The results are validated against literature data, including studies by Minagawa et al. (2023) [1] and Presuel-Moreno et al. (2013) [2]. This research provides a generalized method for evaluating geometric factors and serves as a proof of concept that this approach can effectively approximate them. It accelerates resistivity data processing and facilitates parametric analyses. Additionally, it complements previous studies, which were limited to the central positioning of measurement devices, by enabling the evaluation of geometric factors across a wider range of specimen dimensions and measurement positions on the surface of the domain. As an initial study, it lays the foundation for applying this method to other geometries in future work.