In terms of the key problems we are concerned with in this study, a series of numerical simulations will be conducted. The numerical simulations in this study are based on a numerical code known as RFPA, which is based on the finite-element method (FEM) and can simulate the failure process of quasi-brittle materials such as rock. Since a description of the RFPA has been previously presented in detail [
25], this section will give no further detailed introduction.
3.1. The Failure Characteristics of Rock Samples with Different Brittle Properties
Two numerical rock samples (diameter: 50 mm; height: 100 mm) are established for uniaxial compressive and tensile tests, as shown in
Figure 2. Each sample is discretized into 480,000 finite elements. Sample A contains five layers of identical size. The layers are interbedded with two kinds of rock materials with different levels of brittleness. One material represents a brittle rock and the other represents a relatively ductile rock. Sample B contains two axial halves and each half represents a brittle or ductile material, as in sample A. Both two samples are fixed on the bottom and loaded on the top with axial displacement −5 × 10
−3 mm/step in uniaxial compressive tests and 1 × 10
−3 mm/step in uniaxial tensile tests. Numerical calculation will be performed until rock failures crush down the samples. Since each test involves a sample containing rock materials with different levels of brittleness, contrastive analyses will be conducted based on the rock failure characteristics from the simulation results.
Although there is not one singular definition of rock brittleness and related expressions about rock brittleness have made it more confusing to understand what rock brittleness really is, brittle rock exhibits certain common behaviors that have been recognized by many studies [
26], such as (1) higher Young’s modulus and lower Poisson’s ratio values, (2) low elongation upon load application, (3) a higher ratio of compressive strength to tensile strength, (4) higher internal friction angles, and (5) a large gap between the peak strength and residual strength occurring with failure. In this study, the mechanical properties of materials in the two samples reference these widely recognized behaviors and these mechanical parameters are listed in
Table 1.
Figure 3 shows a typical flash of samples in the model shown in
Figure 2 during the loading steps. It correlates with the uniaxial compressive tests of sample A and B. a and b are, respectively, the Young’s modulus field and the rock damage field of sample A at loading step 40. For the Young’s modulus field, a cooler color represents a higher value of Young’s modulus and a warmer color represents a lower value. Elements in pure red represent those that have been totally damaged at this step. Therefore, it can be clearly seen in
Figure 3a that a series of elements have already been damaged at this step and it is also obvious that most of these elements are located in the brittle material sections. Rock damage field in
Figure 3b shows a similar result.
Figure 3c,d show the Young’s modulus field and rock damage field of sample B at step 45. We can see from this figure that damaged elements appear at two locations in this sample. One is at the interface of the brittle half and ductile half and the other is in the brittle half. Damage in the interface occurs from the local stress concentration due to incoordination in the deformation of the two materials. Damage in the brittle half demonstrates that brittle rocks are easier to damage than ductile rocks under compressive loads. This is consistent with the results of the test on sample A.
Typical flashes of samples A and B in uniaxial tensile tests are shown in
Figure 4. Similarly, it is easy to discover from the Young’s modulus fields and rock damage fields that the two samples undergo failure in the brittle parts because most element damage occurs in the brittle material. Therefore, combined with the uniaxial compressive tests, it is reasonable to conclude from the numerical simulation that brittle rock is easier to destroy than ductile rock under compressive or tensile loads. Since it is obvious that a fractured rock floor could transmit more water than intact rock, the brittle floor is more inclined to induce denser fractures under tectonic stress that serve as better conduits. Moreover, the larger frangibility of a brittle floor may exert an effect on the depth of the floor damage zone I, and bring inspiration to the study of rock damage in coal floors that contain multiple layers with different levels of brittleness.
3.2. Fracture Development in Layered Rock with Different Levels of Brittleness
Due to relatively low strength and high conductivity, natural fractures, including natural fissures and large fractures such as faults, could bring different levels of reduction to the aquiclude zone thickness
h2 during mining activities, and even directly lead to water inrush disasters in some cases. To tentatively explore the development of natural fractures in floor layers under geological stress, the fracture development characteristics in layered rock with different levels of brittleness should be investigated in advance. A two-dimensional numerical model is established here, as shown in
Figure 5. The model is 600 mm in the x-direction and 200 mm in the y-direction, containing four layers of identical size. A brittle layer composed of brittle material and a ductile layer composed of ductile material are sandwiched between the top layer and the bottom layer. The mechanical parameters of the materials in each layer are listed in
Table 1. The left boundary of the model is fixed in the x-direction. The right boundary is applied with constant displacement 2 × 10
−3 mm/step in the x-direction to create a tensile load on the model. It is important to state that although this model is not exact enough to reflect the long-term action of quasi-static or dynamic geological loads, it is also meaningful to conduct such a simple simulation to explore the basic differences between brittle and ductile rock damage in layered rock.
Rock damage is shown from the evolution of the model’s minimum principal stress field in
Figure 6. The model stress field is displayed in different loading steps, in which a series of fractures, depicted in pure black, initiate and propagate. As shown in
Figure 6a, during loading step 30, a vertical fracture (No. I) forms in the brittle layer. At this time, fracture I is exactly sandwiched between the top layer and the ductile layer. As the loading continues, the second fracture (No. II) forms in step 35 on the right of fracture I, as shown in
Figure 6b. It is worth noting that fracture II also initiates in the brittle layer. Afterwards, fracture III forms and then fracture IV forms, as shown in
Figure 6c,d. Both of these two fractures initiate in the brittle layer.
During this simulation, it is noteworthy that once the loading increases to a certain degree, rock damage will appear and then a fracture will initiate. However, these fractures will only form in the brittle layer, rather than in the ductile layer. Moreover, when these fractures extend to an interface, for example, the interface between the brittle layer and ductile layer, they prefer to stop their propagation to induce a new parallel fracture in another position of the brittle layer, rather than cutting through the interface and extending to form a fracture with a larger height.
It is necessary to explain that the layered rock is under a uniaxial tensile load and that the numerically obtained fractures begin in the brittle layer under tensile stress. Obviously, this fracture formation is a result of fissure propagation under external energy, which is in accord with the development of Griffith fissures. No new fractures develop in the brittle layer after four fractures, i.e., after the number of fractures reaches saturation, and this phenomenon is similar to the equally spaced cracks in layer materials [
27]. This numerical investigation has further confirmed the distinct difference of natural fracture development in layered rock and illustrates that brittle layers are more likely to develop denser natural fractures than ductile layers.
3.3. Damage Zone of Rock Floor with Different Levels of Brittleness under Mining
To investigate the damage zone of floor with different levels of brittleness during mining disturbances, a numerical model measuring 200 m in the x-direction and 150 m in y-direction is established as shown in
Figure 7. This model consists of five rock layers, in which a working face is preset in the coal seam to simulate the mining activities. This model is constrained on the left and right side in the horizontal direction (x-direction), and is under 10 MPa pressure in the vertical direction to simulate the overburden pressure. The working face keeps moving forward in the x-direction at 5 m/step until excavating 80 m. All of the rock mechanical parameters for this model are listed in
Table 2. Similarly, for comparison, two numerical cases are established based on the difference in floor layer brittleness.
In the case of the brittle floor, the evolution of the model’s AE fields is shown in
Figure 8. AE events are the emission of the elastic wave generated in rock-like materials during loading, which originates from the damage of microstructures. AE events provide instant damage information and they can be treated as real-time indicators of the progressive failure of rock. In RFPA, when damage is applied to the overstressed elements, it can bring about reductions in element stiffness and strength, as well as an increase in permeability. Microcracks are considered to consist of totally damaged elements and macrocracks can form where they propagate and coalesce. Therefore, AE fields can be used in this study to describe the floor damage zone and help to quantify the floor damage depth
h1. It is noted that model AE events can be found in
Figure 8 in red, white, or black. The red represents the areas damaged due to tensile stress in the current step, the white represents those damaged due to shear stress in the current step, and the black represents all of those damaged in previous steps. The mining distance
Md (10 m, 30 m, 60 m, 80 m) represents the advance of the working face.
Figure 8a provides the AE field of the brittle case when the mining distance (
Md) is 10 m, with a detailed view concerning the coal floor attached. In this step, sparse AE events appear in the floor, which indicates the floor damage zone coming into being. As the working face moves forward, more AE events are discovered in the floor when the mining distance (
Md) reaches 30 m, as shown in
Figure 8b. In this step, the maximum floor damage depth that the AE field exhibits grows to 7.0 m. It is worth noting that the newly formed AE events in the floor are mostly in red, which demonstrates that rock failure in the floor under mining disturbance is due to tensile damage. As the working face continues to move forward at a mining distance of 60 m, the floor damage zone expands both in length and depth, as shown in
Figure 8c. In this step, the maximum floor damage depth increases to 17.0 m and is located 20 m in front of the cut in the horizontal direction. When the working face finally moves forward at a mining distance of 80 m, as shown in
Figure 8d, the floor damage zone continues to expand and the maximum floor damage depth increases to 22.0 m.
Figure 9 depicts the maximum floor damage depth at different mining distances in all loading steps. It clearly shows the positive variation in the maximum floor damage depth along with the mining distance. However, it is important to state that the location in which the deepest rock damage appears is not constant. It varies as the working face moves. In this case, the final deepest floor damage appears 40 m in front of the cut, as shown in
Figure 8d, which is twice the distance of that when the mining distance is 60 m. Moreover, this simulation indicates that during most of the mining process, the floor damage depth along the moving direction of the working face is not consistent. The deepest site where the rock damage appears at present means the floor aquiclude zone here is the weakest and is the site most likely to connect to the upward flow of water and contribute to the formation of a water inrush pathway.
For comparison, the AE fields of the case with the ductile floor when the mining distance reaches 30 m and 80 m are shown in
Figure 10a,b, respectively. It is easily found in these figures that the floor damage depth here is much smaller than that in the brittle case. This comparison directly shows the effect of floor brittleness on the floor damage depth under mining activities. The more brittle the floor rock is, the larger the floor damage depth will be.
3.4. The Risk of Water Inrushes from Floors with Different Levels of Brittleness under Mining
Further investigation on the risk of water inrushes from floors with levels of brittleness under mining will be conducted based on the numerical model shown in
Figure 11, which comes from the model shown in
Figure 7, with the bottom stratum acting as a confined aquifer of 3.0 MPa pressure. A sharply dipping natural fracture develops in the aquifer and extends 27 m upward into the floor strata. O–O’ in the figure is a horizontal dashed line exactly cutting through the uppermost floor elements that are used to monitor the flow rate data in the loading steps. Similarly, two numerical cases are set based on the difference in floor brittleness. All of the mechanical parameters of the rock material in this model are also listed in
Table 2.
The water inrush process in the case of a brittle floor is shown in
Figure 12, in which a, c, and e are the AE fields at the mining distances (
Md) of 5 m, 20 m, and 45 m, respectively, and b, d, and f are the corresponding flow rate fields. To clearly show the detailed information about the floor strata, only part of the model is selected for display.
When the first excavation step is completed, the working face moves forward for 5 m, and the AE field in
Figure 12a shows that there is no damage in the floor. The current flow rate field shown in
Figure 12b indicates that water flow mainly exists in the natural fracture, which has a water conductivity hundreds of times larger than that of the floor rock. Therefore, it is a fact that due to the development of this natural fracture with a strong water conductivity, the effective thickness
h2 of the protective zone in the floor is greatly reduced.
As the working face moves toward the mining distance of 20 m, the AE field in
Figure 12c shows that the floor damage zone has already formed. At this moment, element damage of the natural fracture has occurred and this leads to the enlargement of fracture permeability, directly resulting in the enhancement of the maximum flow rate from 5.31 m
3/min in the initial step to 16.3 m
3/min in the current step. The floor protective zone thickness
h2 is greatly reduced at present due to the combined action of the mining activities from the top and natural fracture from the bottom.
When the working face reaches a mining distance of 45 m, the floor damage zone continues to enlarge, as shown in
Figure 12e,f. Typically, the floor damage zone has reached the depth of the natural fracture and their connection has made the floor ineffective to prevent upward water flow. At this very moment, the water inrush pathway has just formed. The maximum flow rate is thus increased to 37.4 m
3/min.
The flow rates of elements along O–O’ are illustrated in
Figure 13. Three moments are selected to show the data: (1) at the initial excavation step (
Md = 5 m); (2) at the time when the water inrush pathway forms (
Md = 45 m); (3) at the final excavation step (
Md = 80 m). At the initial excavation step, the maximum flow rate of the upper floor occurs near the mine stope with a value of 0.64 m
3/min, and it decreases as the distance from the mine stope to the side decreases.
The mining distance of 45 m is exactly the point at which the water inrush pathway forms. The figure shows there is a sharp jump to 37.4 m3/min in the broken curve at 74.5 m on the x-coordinate, which is the key site for permitting water flow into the mine stope. This key site is 30.5 m behind the working face, which means that the working face is not always the most probable site for inducing a water inrush. The flow damage zone near the geological discontinuities might be the most possible site for inducing a floor water inrush.
At the final excavation step, the working face has moved 80 m and the maximum flow rate has increased to 58.7 m3/min, without the key site changing. This flow rate means a heavy water inrush disaster takes place in the coal mine.
It should be noted that the natural fracture has not been totally damaged during the excavation steps in the numerical simulation. In reality, only the upper part of the natural fracture has damaged and an intact water inrush pathway represented by rock failure has not been discovered. However, the natural fracture in this model is filled with permeable weaklings. Although not totally damaged under mining disturbance, the natural fracture could permit enormous water flow and induce water inrush accidents owing to its original properties. Therefore, natural fractures in the coal floor will always act as a menace to induce water inrushes from underlying aquifers.
In contrast, the AE field in the case of the ductile floor at the mining distance (
Md) of 80 m is shown in
Figure 14. Obviously, the floor damage zone here is not deep enough to connect the natural fracture to create a water inrush pathway. Therefore, water inrush has not occurred in the case with the ductile floor throughout the whole mining process. It can be easily found that a brittle floor is preferable to increase the risk of water inrushes from underlying aquifers in comparison with a ductile floor.