Utilization of the point load and block punch strengths to predict the mechanical properties of several rock samples using regression analysis methods

In the current research, laboratory investigations including mineralogical studies, X-ray diffraction analyses, physical properties, Schmidt rebound hardness, P-wave velocity, point load, block punch, Brazilian tensile and uniaxial compressive strengths have been carried out on fifteen rock samples to predict their mechanical properties from the point load and block punch strengths. The samples were four types of various sedimentary rocks including sandstone, limestone, travertine and conglomerate collected from the northern and northwestern parts of Damghan, north of Iran. Empirical relationships between the properties were determined using simple and multiple regression analysis methods. The simple regression analyses show reliable direct equations with high determination coefficients between the parameters. This is confirmed by three variables contour maps which indicated the mechanical properties were increased by increasing the other mechanical properties. Calculated values of the mechanical properties from the regression model were compared with experimental values of the parameters obtained from different laboratory tests. Also, strong relations were found between the values of point load index, block punch strength, Brazilian tensile strength and uniaxial compressive strength of the tested rocks from multiple regression analyses. Validating the empirical correlations was shown good agreement between experimental data and predicted results. Some statistical indices (e.g., RMSE, VAF and MAPE) were calculated for assessing the performance degree of models presented by multiple regression analyses. It was found that the models provide acceptable equations for predicting the mechanical properties.