Prediction of the tensile strength of sandstones using image processing and GMDH techniques

Abstract The tensile strength of rock plays a crucial role in the planning of tunnels and underground engineering projects. Given the inefficiency of the direct method in measuring this strength, non-destructive testing methods are now being employed to predict the tensile strength (TS) of rocks. Due to the significance of the aforementioned parameter, this paper integrates image processing techniques with the Group Method of Data Handling (GMDH) for the assessment of the tensile strength of sandstones. To this end, 102 disk-shaped sandstone samples were prepared, followed by petrographic analyses and scanning operations for all 102 images. Principal component analysis was conducted for the purpose of feature reduction. Finally, a Group Method of Data Handling (GMDH) model, which utilized input data derived from an image processing technique, was developed to assess the tensile strength of the model. The statistical indices, including a correlation coefficient (R) value of 0.961 for the Brazilian tensile strength (BTS), indicate the workability and feasibility of the proposed model.