Estimation of punch strength index and static properties of sedimentary rocks using neural networks in south west of Iran

•Engineering properties of 142 specimens of sedimentary rocks were investigated.•Petrographic studies showed that the problem of swelling is not expected.•Vp, rock type, point load index, porosity, and moisture were selected as inputs.•Using ANN, UCS, Es and punch strength index were predicted based on inputs.•ANN results show high accuracy for predicting of punch strength index, UCS and Es. Shear strength and static parameters of intact rock are of the most important properties which are vitally required for rock mechanics studies in different engineering projects as the basic data of the study. Different rock mechanical tests are performed in order to determine such characteristics. Considering the difficulty of conducting tests on rocks specifically on weak rocks and high expenses of these tests, it is feasible to have an acceptable estimation of such mechanical and physical parameters via geological properties of the field and developing proper relations and depending the importance of the project, reducing the number of the required tests for characterization of intact rock. In this study, physical tests such as porosity, moisture, and density of rocks, and mechanical tests such as uniaxial compressive strength (UCS), point load index, elastic modulus (Es), punch strength index and compressional wave velocities were performed on 142 specimens from limestone, shale, marl, and mudrock. These specimens were prepared from the cores taken from the drilled boreholes located in three important project sites such as Bazoft dam, Godarkhosh dam, and Konjancham tunnel. Then, using the neural networks, some relationships were developed and presented in order to estimate uniaxial compressive strength (UCS), elastic modulus (Es), and punch strength index of these rocks based on the corresponding compressional wave velocity, rock type, point load index, and physical properties. The results obtained from neural network simulations showed that all the three parameters such as UCS, Es, and punch strength index are having significant correlation with physical parameters, point load index, and compressional wave velocity of rock in such a way that the correlation of the punch strength index is higher than that of Es and UCS. The correlation coefficients of punch strength index, UCS, and Es with investigated parameters are 0.99, 0.99, and 0.97, respectively.