Robust Data-Driven Prediction Of Sandstone Resistivity Using UCS, Porosity, And P-Wave Velocity Through ANN Modelling

Abstract

Electrical resistivity is a fundamental geophysical property widely utilised in subsurface investigations, offering insights into lithology, fluid saturation, and pore structure in geological formations. Traditional methods for determining resistivity often require direct measurement, which can be time-consuming, equipment-intensive, and impractical in remote or inaccessible environments. In this context, indirect predictive models based on measurable geotechnical parameters offer a promising alternative. This study presents a robust, data-driven approach for predicting the electrical resistivity of sandstone using Artificial Neural Networks (ANNs) with three key input variables: Uniaxial Compressive Strength (UCS), porosity, and P-wave velocity. A comprehensive experimental dataset comprising 500 sandstone samples was used to train and validate the model. Laboratory testing was conducted according to ASTM and ISRM standards to ensure accuracy and consistency. The ANN architecture, developed, demonstrated strong predictive performance with an R² value of 0.7892 and a Mean Absolute Error of 23.84 Ohm-m. Sensitivity analysis revealed porosity as the most influential factor, followed by UCS and P-wave velocity. The results confirm the feasibility of using ANN-based models for reliable and non-invasive resistivity prediction in geotechnical and hydrogeological applications, enabling improved site characterization and decision-making in resource exploration, construction, and environmental monitoring.