Evaluation of engineering characteristics and estimation of static properties of clay-bearing rocks

Estimating the static properties of rocks, especially low-strength rocks, is time-consuming, costly, and in some cases impossible. The current study was carried out to evaluate the petrographic (XRD, thin section, and calcimetry), physical (porosity, absorption, density), mechanical [uniaxial compressive strength (UCS), Young's modulus (E-s), Poisson ratio] and dynamic [compressional wave velocity (V-p), shear wave velocity (V-s), dynamic modulus (E-d)] properties of the Godarkhosh dam site, in western Iran. Then, some relationships were proposed to estimate the mechanical properties using simple regression (SR), multiple linear regression, and artificial neural networks (ANN). The XRD analysis showed that the main clay minerals observed in rocks are Illite, Kaolinite, and Chlorite. Therefore, these clay rocks' swelling potential is low. In addition, due to the high percentage of carbonate minerals in the marl samples, the mechanical and dynamic properties of the marls samples were higher than shale samples. Statistical analysis showed that both UCS and E-s have a significant correlation with physical properties and V-p. The relationship between UCS with these parameters is more than with the Es. Besides, the UCS and Es's relationship with V-p were higher than the physical properties. Presented relationships were compared with previous suggested equations. The UCS and E-s relationship, based on universal average data, showed that there is a moderate correlation (RMSE = 0.30, R = 0.74) between these two variables. The ANN exhibits a higher accuracy than the MLR and SR methods in estimating the E-s and UCS. The neural network is also conservative in estimating the modulus of elasticity of the clay-bearing rocks; however, it is not conservative in predicting the UCS of these rocks.