The effect of index parameters on the static properties of limestone in dry and saturated conditions using artificial neural network

Previous studies have shown that moisture has a special effect on the static properties (uniaxial compressive strength (UCS) and elastic modulus (Es) of the rock. In this study, thin section, X-ray diffraction (XRD), porosity, UCS and Es, point load index, and Brazilian tensile strength of the limestone specimens were determined in Khersan 2 dam site, in south west of Iran. Then, using artificial neural network and simple regression, the effect of dry point load index, dry and saturated tensile strength, and porosity on UCS, Es were assessed. Microscopic studies of the samples showed that calcite is the main mineral and samples classified from the Mudstone to the Grainstone. The effect of water on the static properties showed that prediction models in dry conditions are more accurate. Calibration of the relationships presented by previous researchers based on the experimental results of this study and using the criteria of coefficient of determination and root mean square error (RMSE) showed that most of the relationships can be used to estimate the properties of Asmari limestone. Also, investigation of heteroscedasticity graphs of residual variance at predicted levels, determination coefficient and RMSE of the methods showed that the neural network has higher accuracy than simple regression. As compared to the regression method, the neural network is conservative in estimating these properties.