Predictive Models for Strength Properties in Sandstone and Quartzite Using Soft Computing Techniques: Multiple Linear Regression, Artificial Neural Network, and eXtreme Gradient Boosting

ABSTRACT Geomechanical parameters like elastic modulus, uniaxial compression, and tensile strength have a substantial role in excavation development in underground and surface mining infrastructures. The catastrophic failure behaviour of rock presents a direct risk to people as well as an indirect risk by impacting the infrastructure development, stability, and economics, especially in mine transportation operations. However, testing these geomechanical parameters requires an experimental facility to conclude the results. The authors discuss the application of artificial intelligence techniques to predict the geomechanical parameters (elastic modulus, uniaxial compression, and tensile strengths) using the physical and index properties of the rock. Sandstone and quartzite rock samples are tested in this study to determine the physical (density, porosity, P-wave, and S-wave velocities), mechanical, and index properties (Schmidt rebound hammer and CERCHAR abrasivity index). Furthermore, the database is pre-processed, and a linear correlation between the input and output parameters is recognised. Techniques like Multiple Linear Regression, Artificial Neural Network, and eXtreme Gradient Boosting are used to predict the output parameters. The evaluation parameters, including coefficient of determination (R2), adjusted R2, Root Mean Square Error (RMSE), Mean Adjusted Error (MAE), and Residual Sum of Squares (RSE), are observed, and comprehensively, the use of Artificial Intelligence Techniques in geomechanics is discussed.