The uniaxial compressive strength (UCS) of the rock is one of the most important design parameters in various engineering applications. Therefore, the UCS requires to be either preciously measured through extensive field and laboratory studies or could be estimated by employing machine learning techniques and several other measured physical and mechanical explanatory rock parameters. This study is proposed to estimate the UCS of the evaporitic rocks by using a simple, measured point load index (PLI) and Schmidt Hammer (SHVRB) test rock blocks of evaporitic rocks. Finite mixture regression model (FMR), hybrid fuzzy inference systems model (HYFIS), multiple regression model (MLR), and locally weighted regression (LWR) are employed to predict the UCS. Different algorithms are implemented, including expectation-maximization (EM) algorithm, Mamdani fuzzy rule structures, Gradient descent-based learning algorithm with multilayer perceptron (MLP), and the least squares. Coefficient of Determination (R2), Root Mean Square Error (RMSE), Mean Absolute Error (MAE) and A20-index accuracy measures are used to compare the performances of the competing models. Based on all the above measures, LWR outperformed with the other models whereas the HYFIS model has a slight advantage over the other two models.
© 2023. Springer Nature Limited.