Simulation of Asymmetric Destabilization of Mine-void Rock Masses Using a Large 3D Physical Model

The Weihuliang coal mine at Urumchi coal field serves as a case study to illustrate the asymmetric destabilization characteristics of rock masses from mine voids (Fig. 2). The Weihuliang coal mine is located in the pre-Cenozoic strata, approximately centered at the city of Urumchi. In-situ geological and physical settings of the case study are totally distinct (Lai et al. 2014). Fierce crushing stress is a crucial characteristic of tectonic stress and the seismic mechanism is mainly influenced by fracture of a thrust fault. Crustal movement and tectonic activity frequently occur and result in diastrophism velocity reaching up to 10 mm per year. Diastrophism is an elastico-plastic behavior, whereby plastic-shear failure occurs in the faults with strong mobility. Seismic loading would induce rock mass slide and collapse in large underground chambers when coal is extracted near past fault or well-developed joint fissure zones.

The Xishan–Wangjiagou fault group of the Urumchi coal field is located in a conversion region between a fault-fold structure of northern Tianshan Mountain and a nappe structure of Bogda Mountain. The fault-fold structure of northern Tianshan Mountain is composed of several reverse faults whose spatial span is about 30 km. The basin with a southward nappe is formed by the Xishan fault group where the depth of decollement is greater in the north and shallower in the south but generally is 10–15 km. All faults in the Xishan fault group converge to the decollement whose cover depth is about 11 km. Urumchi coal field is composed of the Badaowan syncline and Qidaowan anticline. Weihuliang coal mine is located in the Badaowan syncline with other coal mines. The total thickness of the Weihuliang coal mine is 513.77–902.90 m with 40 m of coal-bearing strata above the mine and a total thickness of approximately 117.05–175.45 m. Strike and dip angles of minable seams are 322°–335° and 46°–67°, respectively. The seam angles vary from 63° to 88° in ESTCS. B₁ and B₂ seams are extremely thick and available for practical excavation in the J₂x₁ strata. The mechanical condition of the B₁₊₂ seams is relatively poor, because the incompact roof and floor do not meet prerequisite characteristics that are suitable for large-level mechanized SSTCC. However, large-scale “V” collapse grooves are formed in the surface after top coal caving and would allow precipitation to pour into the collapse grooves and induce destabilization of roof surrounding rock. Vast noxious gases assembled in the mine voids have been developed by irregular mining and continual excavation disturbance. Usually, the extra-high section height of the top coal does not hinder abrupt large-scale collapse of rock masses that may be induced by local variations of horizontal stress, methane accumulation, mine flood, or fire and coal dust with recovery process; hence, it is necessary to develop an effective methodology for assessing destabilization of rock masses with potential hazards.

The working face of the mechanized SSTCC could be horizontal to the direction of an ESTCS with short actual length for all staff operation. A finite-size excavation disturbed zone (EDZ) after coal mining would be regarded as a neo-active-like fault structure. Deformation of coal-rock masses is intensively concentrated on a narrow zone called an asymmetric destabilization zone under high strain rate loading. Theoretically, we have a stability analysis method for mechanical conditions and stress–strain relationships of plastic destabilization for the asymmetric destabilization of surrounding rock with abundant numerical simulations, which provides a profound understanding of asymmetric destabilization characteristics.

We pay little attention to micro-structures, which play a key role in prediction and control of the asymmetric destabilization. Because in-situ monitoring methods are limited for micro-structure evolution mechanisms on dynamic mechanical response parameters of asymmetric destabilization, we have seldom effectively uncovered the destabilization mechanism. Also, we should analyze the asymmetric destabilization characteristics with hybrid microscopic and macroscopic methods. Currently, we have developed a series of basic experimental materials and structures.

However, a specific large-scale experiment (meter size) should be built to scientifically realize mechanical conditions, evolution mechanisms, and macro-structure characteristics of the asymmetric destabilization. Particularly, asymmetric destabilization of coal-rock masses should be accurately predicted and used for optimization of operational facilities.

In time t, V _i (x,y,t) was the ith weight of V(x,y,t) in the ith sub-domain, and N-dimensional vector \(\overrightarrow {V} (t)\) has been established as Eq. (1).

Difference value of N-dimensional vector \(\overrightarrow {V} (t)\) would be another N-dimensional vector in adjacent two time t − Δt and t as Eq. (2).

To describe the \(\overrightarrow {V} (t)\) change, a new scalar value, module of N-dimensional vector \(\overrightarrow {V} (t)\), M ₀ was applied to state general level of the whole physical field as Eq. (3).

We defined angle (Φ) as being between \(\vec{V}_{t - \Delta t}\) and \(\vec{V}_{t}\) Another module of \(\left| {\vec{V}_{t - \Delta t,t} } \right|\), ΔM ₀ stated physical field variation in the other view. The value of M ₀ and ΔM ₀ explained coal-rock masses damage situation. They would increase distinctly with internal damage increased in the model, which can be described by AE total event and energy rate.

A particular model frame is required for developing a specific large-scale experiment. The model device must initially be stiff enough to ensure the stability of the model in the experiment. With current requirements in a simulation experiment, five kinds of rock masses were considered including carbon mudstone, silty mudstone, malmstone, B₁, and B₂ seam. Based on the geological section of the coal seam, the angle of the coal seam is over 64° and falls into the extremely steep and thick coal seams. We determined that, in the simulation scope, the excavation level of the Weihuliang mine had already dropped to +574 m with the mechanized sub-horizontal section top coal caving from +625 m for the whole excavation level. So, approximately 49.00 m was considered in the vertical direction. According to the geological section, the thickness of the simulation seam was 110.00 m, where the thicknesses of the B₁ and B₂ seams were, respectively, 16.90 and 14.00 m. For the actual simulation of the excavation in the trend, 72.50 m was used as the width for the simulation because it represented the advance distance (per month) of the working face. Finally, we determined that constant of geometric similarity (C _l ) with comprehensive consideration of all experimental conditions using:

In which, m represents the model parameter, and p is the parameter of in-situ rock masses. According to the properties of the simulation material and the ratio of the rock masses, the constant of density similarity (C _ρ ) is represented by Eq. (5).

Using the similarity principle and dimensional analysis, the constant of geometric similarity (C _l ), the constant of density similarity (C _ρ ), the constant of stress similarity (C _σ ), and the constant of time similarity (C _t ) are related by Eq. (6).

The constant of stress similarity is 0.0272, and the constant of time similarity is 0.2. Considering the depositional pattern of the original strata, we established the large 3D model with 0.01 m in length direction. Due to the complex geological settings, an extensive number of faults were distributed throughout the research area. The model was unable to simulate all geological structural planes; thus, we emphasized the crucial fault that closely neighbored the B₂ seam. Furthermore, a projected coordinate system was used to locate all parts accurately in the model.

The model device not only applied geostatic stress as an analog ground stress, but also co-located loading equipment in the horizontal direction simulating in-situ tectonic stress as an auxiliary stress and as a dynamic load for the model. Ultimately, large underground structure experiment system (LUSE) was modeled as (length × width × height) 4.42 m × 2.90 m × 1.95 m in the Xi’an University of Science and Technology (XUST), China. The greatest sizes of the system is 8.80 m × 5.60 m × 12.00 m.

The large underground structure experiment system, to date is the largest device of its kind in Asia, and has several favorable aspects including loading ability, geometric size, and monitoring methods.

Core samples from geo-climatic boreholes in the experimental site were classified using the rock quality designation (RQD) system. The designation system provides a quantitative method of categorizing the engineering character of rock masses based on core hole information. A summary of the RQD for the boreholes is presented in Table 1, which shows, in general, that the RQD of rock masses of the extremely steep and thick coal seams at Weihuliang coal mine was low. Locally, mudstone and malmstone were broken and had extremely low RQD values.

The physico-mechanical properties of the main sedimentary rock masses used in the large physical simulation were based on evaluation of exploratory drilling and laboratory test results. In addition, the data utilized in the experiment were finished at Key Laboratory of the Ministry of Education of China for Western Mines and Hazard Prevention at Xi’an University of Science and Technology. The main properties used in the physical simulation were summarized in Table 2.

Much research work has been done on model material (Lai et al. 2012, 2013; Li et al. 2011; Pan et al. 1997). In the present work, we prepared a new kind of modeling material for establishing the specific three-dimensional model. Porous rock-like composite material (PRCM) adopted bank sand as the aggregate, with 3.9 % in the 0.50–1.00 mm size range, 29.1 % in the 0.25–0.50 mm size range, and the remaining portion in the 0.10–0.25 mm size range. We used bank sand and coal powder as the aggregate for the coal seam material being simulated. Gypsum and flour were both grouting agents and water was used as the plastic impact agent, which largely influenced the accuracy of the material proportion. The porous rock-like composite material was light material whose massive density was 1700 kg/m³. Pebbles, clay, and mica were added separately to represent the fault and distinguish between different strata. Pebbles and clay were used to simulate the fault and improve the brittleness of the model material.

The nonlinear coupling effect of porous rock-like composite material is remarkable because of the plastic impact agent (Huang 2009; Lin 1984; Lai et al. 2013a, b). Because the porous rock-like composite material consisted of loose particles, their mechanical properties are anisotropic. Comprehensive strength of porous rock-like composite material depends on nonlinear coupling effect with water. In some cases, various ingredients would blend well with water, whereas, in other cases, the alternation effect of water on the material plays a key role in deciding strength of the modeling materials. Therefore, modeling of these effects must consider a comprehensive set of properties including static pressure (μ) and hydrodynamic pressure (τ _d ). The effective stress of porous rock-like composite material (σ _a ) could be divided into three stages (Lai et al. 2013a, b).

Stage I Modeling material is unsaturated. In such a situation, under an external loading, the pore water would not provide enough internal stress (σ) to create internal micro-cracks and dilatancy in the material and decrease the strength of the material. In this stage, along with increase of water, the unprofitability of static pressure would be gradually decreased and strength of the porous rock-like composite material is steadily increased while the effective stress of the material is increased, as expressed by Eq. (7).

Stage II Modeling material is saturated. Under an external loading, drainage equilibrium process occurred among internal pores with the structure of the material stabilized, which would result in the skeleton and pore water bearing the external loading. The beneficial static pressure for the skeleton reaches peak value and the effective stress of the porous rock-like composite material reaches the maximum, as expressed by Eq. (8).

Stage III Modeling material is oversaturated. Under such a condition, all pores are filled with flowing water that would add an extra hydrodynamic pressure and shear stress that results in incremental and tangential deformation in the skeleton. Micro-cracks are evolved into macro-cracks that may cause dynamic destabilization and damage the whole structure. Under these conditions, the mechanical properties of the porous rock-like composite material are less stable. Meanwhile, the effective pressure is sharply decreased, as expressed by Eq. (9).

In which, R and J are coefficient of hydrodynamic pressure and hydraulic slope angle, respectively. When flowing water streams in the porous rock-like composite material, macro-cracks would propagate remarkably forming the macro-crack plane, and the extra hydrodynamic pressure is parallel with the plane and internal static stress is perpendicular to the plane at the same time. So, R would be regarded as br/2. Here, b is weighted width of the macro-crack plane, and r is unit weight of the water in the experiment. Besides, J meets the Darcy’s law.

The strength of the porous rock-like composite material was adjusted by changing contents of aggregate and grouting agents. Mainly, the compressive strength and bulk modulus of the porous rock-like composite material were analyzed. Standard coal-rock specimens (¢50 × 100 mm) with different ingredients have been made and each group had nine specimens with the same ingredients avoiding such errors. Plastic impact agent for the specimens in the experiment was water whose amount just occupied one in every ten units of the aggregate and grouting agents. Figure 3 showed stress–strain relationship of coal-rock specimens with various ingredients being adopted in the simulation.

Proportions (coal:sand:gypsum:flour) comprising the coal seam were 30:30:4:6 and 30:30:3:7, in the experiment. Maximum axial deformation of specimens with 30:30:4:6 was 1.6 %, and had an elastic deformation stage that was lower than the uni-axial stress, peak values of compressive strength and bulk modulus were 13.7 MPa and 5.3 GPa, respectively. While the proportion was altered to 30:30:3:7 with a reduction in the amount of gypsum, specimens obtained peak values of compressive strength and plastic stage. However, compressive strength and bulk modulus decreased to 10.6 MPa and 5.1 GPa, respectively. Compared with mechanical parameters of the prototype, we finally decided the model proportions of various rock masses. Tables 3 and 4 separately indicated the main mechanical parameters of the porous rock-like composite material and the state of model placement and the amount for different materials.

When the model was completed, we began to set all monitoring facilities and dry the model out for at least 2 months so the strength of the model were able to reach an ideal state. The mining and observation of the model complied with parameter of the constant of time similarity, and as a matter of experience and custom, we used 1 h to simulate one equivalent day. The roof collapsed method was adopted in the model and, first, caved the upper level seam at the +579 m level in Weihuliang mine to simulate the actual mine voids, as shown in Fig. 4a.

Multi-mean timely track monitoring including acoustic emission (AE) sensors, crack optical acquirement (COA), roof separation observation (RSO), and close-field photogrammetry (CFP) was applied. Data from the buried pressure sensor (BPS) was linked with the stress–strain relationship of the stope in the extremely steep and thick coal seams. Simulated hydraulic support (SHS) was simulated for parameter optimization of hydraulic supports in the extremely steep and thick coal seams. We measured the initial conditions of all facilities before coal mining. The mechanisms of asymmetric destabilization characteristics of rock mass from mine voids were clearly detected and analyzed, as shown in Fig. 4b.

Hydro-cylinders located on top and on the lateral side of the large underground structure experiment system exerted the asymmetric static–dynamic loading for the experimental purpose, which simulated the tectonic stress derived from fault-fold structures of the northern Tianshan Mountain and nappe structure of Bogda Mountain. The maximum loading of the top and lateral hydro-cylinders were 10 and 12 MPa, respectively.

Asymmetric destabilization characteristics would induce local deformation and failure of coal-rock masses, which are predisposed to mine hazards. Stress around the working face must be redistributed after coal mining by using abundant AE data on failure of the surrounding rock. We set up AE sensors in different locations to achieve information induced by rock failure in four stages, which were (I) excavation of working face, (II) first caving of top coal, (III) coal caving, and (IV) local roof caving. Figure 6 showed the mechanism of asymmetric destabilization-AE in various stages.

Initially, the model was excavated; the AE signal was less with seldom total and big events due to good condition of coal-rock masses (Fig. 5a), when the working face just advanced 0.5 m. With the working face advancing continuously, integrality of coal-rock masses was beginning to be broken and cracks were propagating. During first caving of the top coal, the AE signal waves appeared to skip. The crest value of the AE event occurred at the No. 33 monitoring serial number. The signal from the No. 5 AE sensor was weaker without severe rock mass failure than others, which indicated that coal excavation and first caving of top coal seldom affected the upper zone of the model monitored by the No. 5 AE sensor.

Figure 5b indicated that, after first caving of the roof, several small-scale fluctuations of AE signals from the top coal emerged and cracks gradually stretched. Energy rate of coal-rock masses released abruptly up to 600, and then values of all parameters declined and tended toward a gentle state.

Figure 5c shows the parameter traits after coal caving. A large-scale roof collapsed suddenly and the AE signal surged indicating that added cracks propagated. The coal caving was a dynamic wave process where the total and big event increased to 700 and 650, respectively. The peak value of energy rate occurred at the No. 37 monitoring serial number. Thereafter, the energy rate was reduced to the point that caving no longer influenced the coal-rock masses.

The roof located upon the top coal started caving in a large scale, with Fig. 5d showing AE characteristics in the process. The AE signal rose and fluctuated repeatedly, with damage increasing to the coal-rock masses. The integrality of 3D internal part model continued degradation. According to AE trait analysis of the 3D model, the common mechanism of AE traits was summarized with parameters of the AE signal emerging as a commutative wave in time, which indicated that cracks in the coal-rock masses were propagating steadily after structural planes alternatively opened and closed. Large and small strain zones continuously alternated with time variation.

Researches on strata behavior in the extremely steep and thick coal seams indicate that parameter optimization would be needed for in-situ support (Lai et al. 2009a, b, 2014). According to various field loadings of surrounding rock, the support should be determined by comparing the theoretical analysis and numerical simulation results (Lai et al. 2009a, b). Yield loading of the in-situ supports has been set at 4500 kN by simulation theories and setting loading must be 4200 kN. The simulated hydraulic support (Fig. 9a) has been installed in the working face with static-resistance strain equipment adopted for loading observation. The strain values fluctuated in the experiment and the coal caving process had a major impact on the support (Fig. 9b). A large-scale roof caving would increase the amount of shock to the support. Based on statistical results, the setting loading of the hydraulic support was between 4000 and 4200 kN, and also the phenomena that occurred after first roof caving in the experiment were similar with the actual situation of roof failure. Particularly, the individual hydraulic support had such a low setting loading that they could not adequately maintain contact with the roof. When the roof caved in a large-scale zone, the peak value of the support loading was up to 4200 kN; however, the simulated hydraulic support with 4200 kN would ensure demand of pressure control.