Artificial Intelligence Forecasting Models of Uniaxial Compressive Strength

Abstract The uniaxial compressive strength (UCS) is a vital rock geomechanical parameter widely used in rock engineering projects such as tunnels, dams, and rock slope stability. Since the acquisition of high-quality core samples is not always possible, researchers often indirectly estimate these parameters. The main objective of the present study is to evaluate the performance of the long short term memory (LSTM), deep neural networks (DNN), K-nearest neighbor (KNN), Gaussian process regression (GPR), support vector regression (SVR), and decision tree (DT) to predict the UCS of different rock types of Claystone, Granite, Schist and Sandstone, Travertine, Limestone, Slate, Dolomite and Marl acquired from almost all quarry locations of Iran. 170 data sets, including porosity (n), Schmidt hammer (SH), P-wave velocity (Vp), and point load index (Is(50)) were applied in the methods. Finally, a comparison was made between the results made by the prediction methods. To assess the performance ability of the applied methods, the 5-fold cross-validation (CV) was considered. The results proved that computational intelligence approaches are capable of predicting UCS. On the whole, the GPR with a correlation coefficient (R2) of 0.9955 and a route mean square error (RMSE) of 0.52169, performs best. Lastly, the UCS prediction intelligence methods were ordered as GPR, DT, SVR, LSTM, DNN and KNN, respectively.