Quantitative analysis of rockburst in the surrounding rock masses around deep tunnels

Highlights

• Rockbursts in deep tunnels subjected to high crustal stress were investigated using a general particle dynamics code.

• The numerical results showed that the GPD code can satisfactorily simulate the phenomenon of rockburst.

• The GPD code can efficiently reproduce the 2D and 3D crack growth in the surrounding rock.

Abstract

Rockbursts in deep tunnels subjected to high crustal stress were investigated using a general particle dynamics (GPD) code with the Holmquist–Johnson–Cook damage model (HJC). Under dynamic excavation disturbance, several numerical tests were carried out to investigate the dynamic failure patterns of surrounding rock masses with different preexisting cracks around deep tunnels. The numerical results showed that the GPD code can satisfactorily simulate the phenomenon of rockburst caused by crack evolution, including predicting the ejection velocity, rockburst location and ejected rock block volume. Moreover, the analysis of the circumferential stress in the surrounding rock of deep tunnels at the Jinping II Hydropower Station shows the efficiency of the GPD code. In addition, four typical rockburst, which occurred under different geological conditions at the Jinping II Hydropower Station, were analyzed by using the GPD code. The numerical simulations suggested that the GPD code can efficiently reproduce the 2D and 3D crack growth in the surrounding rock, which eventually leads to the occurrence of rockburst.

Introduction

Rockburst is a type of geological disaster that often occurs in brittle rocks around deep underground engineering subjected to high crustal stress (Zhou et al., 2018). The occurrence of rockbursts in deep rock masses proves that the release of elastic strain energy can cause the dynamic failure of rock masses. The rockburst disasters cause considerable casualties and equipment losses, leading to the increase in construction costs or even the abandonment of engineering projects.

In recent decades, many attempts have been made to reveal the mechanism of rockbursts from different points of view (Zhang et al., 2012a; Xiao et al., 2016; Keneti and Sainsbury, 2018). Rockbursts initiate deep in rock masses, and no early warning is given before it occurs. Therefore, the entire process of rockburst from initiation to ejection is difficult to observe. Based on model tests and rockburst field monitoring, some theoretical models of rockburst processes were proposed. However, understanding, predicting and controlling rockbursts are still considerable challenges in deep underground engineering.

In general, rockbursts may occur in deep rock masses subjected to high crustal stress and are triggered by dynamic excavation disturbances that arise from blasting or boring (Zhang et al., 2012b; Cheng et al., 2013; Xiao et al., 2016; Li et al., 2017; Liang et al., 2019; Zhang et al., 2019). Rock masses subjected to either static stress or dynamic loading may exhibit different characteristics (Lee and Pietruszczak, 2008). Therefore, it is of great significance to consider a high static stress and a dynamic disturbance as two key factors that trigger rockbursts in deep underground engineering (Li et al., 2012). To reveal the rockburst mechanism, the finite element method is often applied to simulate the rockburst phenomenon. However, rockburst is always related to the initiation, propagation and coalescence of microcracks and the formation of macrocracks (Zhou et al., 2018). Therefore, it is very difficult to capture the crack evolution and to simulate the discontinuous deformation of rockburst ejection using the finite element method.

The stability of large-scale rock engineering under complex stress conditions has always been studied. Rockburst is a type of geological disaster that is very difficult to numerically reproduce. Compared with mesh-based numerical methods, the meshless numerical method, which has the advantage of simulating large deformation and crack initiation, propagation and coalescence in rock materials, can be used to assess the stability of large-scale rock engineering under complex stress conditions. For example, the propagation of cracks can be simulated by smooth particle hydrodynamics (SPH) (Monaghan and Gingold, 1983) and pseudo-spring SPH (Chakraborty and Shaw, 2013; Islam and Shaw, 2020). However, tensile instability problems are encountered in SPH (Monaghan, 1988; Libersky and Petsehek, 1990) and pseudo-spring SPH (Chakraborty and Shaw, 2013; Islam and Shaw, 2020). Molecular dynamics can be applied to model the propagation of cracks (Izumi and Katake, 1993). However, molecular dynamics has some shortcomings, such as a longer computational time and a lower computational efficiency. To overcome the aforementioned shortcomings, peridynamic theory (PD) (Silling, 2000), which is a meshless numerical method based on the nonlocal concept, was introduced to model crack problems. It is assumed that particles in a continuum interact with each other across a finite distance; thus, the formulation of this theory is by integral equations rather than partial differential equations (Silling, 2000; Silling and Bobaru, 2005). Therefore, the peridynamic method (PD) can be applied to model the problems of continuous or discontinuous displacements. Moreover, it is not affected by the singularity problem of crack tips; similar to meshless methods and molecular dynamics methods, the PD avoids the limitations of the calculation dimension of the molecular dynamics method. However, the bond-based peridynamics (BB-PD) are based on the assumption of pairwise interactions of the same magnitude, which leads to a restriction of the Poisson's ratio to 1/4 (3D) or 1/3 (2D) for isotropic problems (Silling, 2000; Zhou et al., 2015). Moreover, an unstressed configuration is used in bond-based peridynamics, which does not result in the application of the stress-based strength criterion to peridynamic theory (Zhou and Wang, 2016). To overcome the above disadvantages of BB-PD and SPH, the general particle dynamics (GPD) code, which was developed from smooth particle hydrodynamics (Monaghan, 1988; Libersky and Petsehek, 1990; Chakraborty and Shaw, 2013), was developed to simulate crack problems in rocks (Zhou and Bi, 2018). During the simulation process, a given particle interacts with its neighboring particles by virtual bonds except if a crack exists between neighboring particles. If particles exist within the influence domain of a given particle, interactions among the particles are established through bonds. Then, when the stresses of the virtual bonds satisfy the stress-based strength criterion, the virtual bonds are broken. Finally, the sequence of failure of such neighboring bonds can be easily captured to trace the initiation and propagation of cracks. Therefore, GPD code is more in line with the physical phenomenon than SPH, which overcomes the disadvantages of SPH (Monaghan, 1988; Libersky and Petsehek, 1990; Chakraborty and Shaw, 2013). Moreover, in GPD, Poisson's ratio is not limited to 1/4 (3D) or 1/3 (2D), and the stress-based strength criterion can be applied to define the initiation and growth of cracks, which overcomes the disadvantages of bond-based peridynamics (BB-PD).

In this paper, the rockburst ejection process is simulated under conditions of dynamic excavation disturbance by using GPD code. In the GPD code, rock masses are discretized into particles, and these particles together with the contact surfaces between them are considered the numerical model (Bi et al., 2016, Bi et al., 2017; Bi and Zhou, 2017a, Bi and Zhou, 2017b; Zhou et al., 2020; Zhou and Zhang, 2017). These discretized particles and additional parameters, such as ejection speed and kinetic energy, compose the basis of the precise simulation of the rockburst process under dynamic excavation disturbance. To understand the rockburst mechanism around deep tunnels, rock masses containing different preexisting cracks around the deep tunnel are investigated using the general particle dynamics (GPD) code with the Holmquist–Johnson–Cook damage model. The numerical results show that GPD code can satisfactorily simulate the phenomenon of rockburst induced by crack evolution, which implies that the GPD code can provide a new way to evaluate rockbursts around deep tunnels.