Estimating Geotechnical Properties of Sedimentary Rocks Based on Physical Parameters and Ultrasonic P-Wave Velocity Using Statistical Methods and Soft Computing Approaches

In this research, the north and northwest parts of Damghan in the northeast of Iran were selected as the study areas, and different sedimentary rocks including, sandstone, limestone, travertine, and conglomerate, were collected. Laboratory investigations including petrography study, X-ray diffraction, dry density, porosity, point load strength (PLS), Brazilian tensile strength (BTS), block punch strength (BPS), uniaxial compressive strength (UCS), and ultrasonic P-wave velocity (VP) were determined. The main aim of this study is to establish predictive models to estimate the PLS, BTS, BPS, and UCS of the studied rocks based on P-wave velocity. Four experimental models were developed using multivariate regression analysis (MRA), artificial neural network (ANN), and adaptive neuro-fuzzy inference system (ANFIS). Statistical parameters including, R, RMSE, VAF, MAPE, and PI, were calculated and compared to assess the performance of MRA, ANN, and ANFIS models. Correlation coefficient values were obtained from 0.73 to 0.85, 0.96 to 0.99, and 0.99 for the MRA, ANN, and ANFIS models, respectively. A good RMSE value equal to 0.11 was obtained for ANFIS when using VP for predicting block punch strength. Calculating residual error and correlation between experimental and predicted values indicated that the ANFIS models have the best coefficients. Also, the results of this research demonstrated that the ANN approach is more efficient than MRA in predicting the mechanical properties of the studied rocks.