An Acoustic Emission Data-Driven Model to Simulate Rock Failure Process

Abstract

Numerical simulation is a commonly used method for investigating rock failure. However, the numerical model is usually insufficient to predict real rock damage and failure because of rock microstructural heterogeneity. In fact, rock damage can be quantified using acoustic emission (AE) data. The aim of this study is to simulate and predict the failure of Brazilian and uniaxial compression specimens using AE data recorded during experiments. An AE data-driven model, in which cracks are assumed to be tensile in nature, is developed. AE data recorded from the test start up to a fraction of the peak stress (e.g., 20%, 40%, and 60%) are input into the data-driven model to predict the evolution of failure pattern beyond that stress level up to failure. First, we quantified stress-induced rock damage with AE data based on the tensile model. The results indicate that most of damage source radii are less than one millimeter, and the corresponding damage degree is close to one. Then, the inversed damage is input as the initial conditions for the numerical simulation to predict the future damage and failure of rock. With the increase of damage elements driven by AE data, the inversed damage zone develops from diffuse to localized, and the dominant factor for rock failure transits from microstructural heterogeneity into stress-induced rock damage. The damage and failure pattern of rock is well predicted when sufficient AE data are taken into account as known conditions.