آرتیمس قاسمی

Accurate determination of geo-mechanical properties of rocks such as uniaxial compressive strength (UCS) and modulus of elasticity (E) using standard laboratory tests is a difficult, expensive and time consuming task. This is particularly true for anisotropic, thinly bedded, highly fractured, porous and weak rocks. In order to overcome these difficulties, the development of predictive models for the estimation of these properties seems to be essential in rock engineering projects. The main purpose of this study is prediction of UCS and E of sandstones using artificial neural network (ANN) and multiple regression analysis (MLR). For this, a database of laboratory tests (including 130 sandstone samples) was prepared, which includes porosity, P-wave velocity, dry density, slake durability index, and water absorption as input parameters and UCS and E as output parameter. The performance of the MLR and ANN models are evaluated by comparing statistic parameters, including correlation coefficient (r), root mean square error (RMSE), and variance account for (VAF). Comparison of the multiple linear regressions and ANNs results indicated that respective ANN models were more acceptable for predicting UCS and E than the other.