Effective statistical models for prediction of the tensile strength of rocks

This study investigated the relationships between indirect and direct tensile strength values of rocks using regression analysis. The uniaxial tensile strength (UTS) values of nine distinct rock specimens were ascertained, after which assessments were conducted for Brazilian tensile strength (BTS), flexural strength under concentrated load, and flexural strength under constant moment tests. In order to enhance the accuracy of estimating the UTS, the uniaxial compressive strength, apparent (open) porosity, and ultrasonic pulse velocity (UPV) of the rock specimens were determined, followed by regression analyses. The experimental findings revealed a notable discrepancy between the indirect tensile strength values and those obtained from the UTS. This study introduces a novel statistical approach integrating rock properties and indirect methods for UTS determination, presenting effective regression models applicable to practical scenarios. According to the multiple linear regression results, the BTS was identified as the most suitable indirect testing method, while the UPV was found to be the most relevant rock property for predicting UTS. Furthermore, results from flexural strength tests conducted under various standards advocate a test specimen size of 25∗50∗150mm as optimal for predicting tensile strength. • Tensile strength is a key parameter in mining and civil engineering design. • Various indirect methods were tested for rock tensile strength estimation. • Indirect tensile values were about 2–3 times higher than direct results. • Ultrasonic pulse velocity is the best rock property for tensile prediction. • A regression model using BTS and UPV effectively predicts rock strength.