Modeling of Electrical Resistivity of Soil Based on Geotechnical Properties

TL;DR

This study compares various models, including neural networks, to predict soil electrical resistivity from geotechnical properties, finding neural networks offer the highest accuracy and establishing systematic analysis in this field.

Contribution

It introduces a comprehensive comparison of linear, non-linear, and neural network models for predicting soil resistivity based on geotechnical data.

Findings

Neural network models provide the most accurate predictions.

Significant exponential negative relationships exist between resistivity and geotechnical properties.

The study offers practical guidelines for modeling soil resistivity.

Abstract

Determining the relationship between the electrical resistivity of soil and its geotechnical properties is an important engineering problem. This study aims to develop methodology for finding the best model that can be used to predict the electrical resistivity of soil, based on knowing its geotechnical properties. The research develops several linear models, three non-linear models, and three artificial neural network models (ANN). These models are applied to the experimental data set comprises 864 observations and five variables. The results show that there are significant exponential negative relationships between the electrical resistivity of soil and its geotechnical properties. The most accurate prediction values are obtained using the ANN model. The cross-validation analysis confirms the high precision of the selected predictive model. This research is the first rigorous systematic analysis and comparison of difference methodologies in ground electrical resistivity studies. It provides practical guidelines and examples of design, development and testing non-linear relationships in engineering intelligent systems and applications.