Assessment of physio-mechanical properties of rocks using ultrasonic pulse velocity tests- a case study

Abstract

Abstract Determining the physio-mechanical properties of rocks is essential for the safe and effective design of civil and mining engineering projects. Traditional laboratory-based testing methods, although reliable, are often labour-intensive, time-consuming, and costly, primarily due to the need for precise core specimen preparation. As a more practical and efficient alternative, non-destructive techniques particularly ultrasonic pulse velocity (UPV) testing can be employed to estimate key rock properties without damaging the samples. In this study, rock samples comprising sandstone, shale, and coal were collected from three boreholes: two located in the Piparwar region of North Karanpura, Jharkhand, and one from the Narankuri region in Raniganj, West Bengal, India. The samples were subjected to laboratory testing under both dry and saturated conditions to determine a range of physio-mechanical properties, including uniaxial compressive strength (UCS), modulus of elasticity (E), porosity (n), P-wave velocity (V P ), and S-wave velocity (V S ). Multivariate regression analysis revealed strong correlations between UCS and E with porosity, V P , and V S . To further enhance the predictive modelling, two machine learning approaches vis Back Propagation Neural Network (BPNN) and Generalized Regression Neural Network (GRNN) were applied. The novelty of this work lies in the comparative evaluation of GRNN and BPNN for predicting rock strength parameters using non-destructive inputs. The results demonstrated that GRNN outperformed BPNN in all cases, offering higher accuracy and generalization. These findings underline the potential of machine learning based non-destructive testing as a rapid, cost-effective solution for evaluating rock strength properties in engineering applications.