Prediction of rock loading stages using average infrared radiation temperature under shear and uniaxial loading

Abstract

Listen

Translate article

Accurately predicting geologically rock loading phases under varying water levels is crucial to prevent rock engineering operations from severe geological hazards. Acoustic emission (AE) traditionally determines these phases, but machinery in rock projects can disrupt AE sensors, leading to imprecise data and potential accidents. To address this, our study tested a novel approaches using AE, Average Infrared Radiation Temperature (AIRT), Critical Slow Down Theory (CSDT), and Convolutional Neural Networks (CNNs) to predict rock loading phases accurately under different water conditions. The research findings are: (1) AE analysis divides the stress–strain curve into four phases: crack closure, elastic deformation, and stable and unstable crack propagations. Each phase exhibits distinct characteristics in terms of AE signal behavior. (2) AIRT analysis reveals specific patterns during different loading phases. The AIRT and cumulative AIRT have distinct trends in different phases. (3) The AIRT sequence's fractal dimension shows distinct trends during loading phases: it initially rises during the elastic phase and then declines. Fractal dimensions for crack closure, elastic deformation, and stable crack propagation increase, while that for unstable crack propagation decreases with water content. 4) The stress–strain curve phases were successfully predicted by using AIRT and CSDT, demonstrating a strong correlation with actual stress levels. Notably, this marks the first-time prediction of phases through non-destructive tests (IR and CSDT), where CSDT fluctuations at each stage can serve as early indicators of rock failure.5) The proposed CNNR model achieves high accuracy (98%) in predicting different phases of the stress curve using AIRT. It particularly performs well in the elastic deformation and stable crack propagation phases, with only a few instances of false negative predictions. The model shows promising potential for effective and safe geological projects, especially under varying water conditions. The findings contribute to the understanding of rock behavior and offer insights for the design of safer and more efficient rock engineering projects.