Predictive Modeling of the Uniaxial Compressive Strength of Rocks Using an Artificial Neural Network Approach

Sedimentary rocks provide information on previous environments the surface of Earth. As a result, they are principal narrators former climate, life, and important events The complexity cost direct destructive laboratory tests adversely affect data scarcity problem, making development intelligent indirect methods an integral step in attempts to address problem faced by rock engineering projects. This study established artificial neural network (ANN) approach predict uniaxial compressive strength (UCS) MPa sedimentary using different input parameters; i.e., dry density (ρd) g/cm3, Brazilian tensile (BTS) MPa, wet (ρwet) g/cm3. developed ANN models, M1, M2, M3, were divided as follows: overall dataset, 70% training dataset 30% testing 60% 40% respectively. In addition, multiple linear regression (MLR) was performed for comparison proposed models verify accuracy predicted values. performance indices also calculated estimating models. predictive M2 model terms coefficient determination (R2), root mean squared error (RMSE), variance accounts (VAF), a20-index 0.831, 0.27672, 0.92, 0.80, respectively, revealing ideal results, thus it best-fit prediction UCS at Thar coalfield, Pakistan, among this study. Moreover, performing sensitivity analysis, determined that BTS most influential parameter predicting UCS.