Study on precursory recognition and integrated warning modeling to fracture in flawed sandstone under uniaxial compression

Abstract

To explore the precursory information and instability fracture of rocks, we conducted a series of uniaxial compression tests on flawed sandstone. An integrated warning modeling is developed to predict the fracture in flawed sandstone. The results show that both the peak strength and elastic modulus of flawed sandstone demonstrate an “inverted” Gaussian distribution relative to the ligament angle, reaching the minimum values at 60°. The elastic strain energy proportion shows a sharp drop-off, whereas the proportion of dissipative energy increases in steps. The coalescence modes of flawed sandstone change from a mixed tensile-shear failure approximately aligned with the axis to an oblique shear failure. An integrated warning model is developed by integrating a Multi-Output Classifier (MOC) and Grid optimization (GO). The integrated warning model has an accuracy of 97.95%. Additionally, the sensitivity of the model is recorded at 99.26%, confirming its effectiveness in predicting the likelihood of fracturing.