Research Investigation on the Movement and Fracture Characteristics of an Extra-Thick Hard Roof during Longwall Panel Extraction in the Yima Mining Area, China

As an extra-thick hard roof is a signi ﬁ cant contributing factor to frequently induced sudden roof collapse accidents and coal bursts, this study investigates the relationship between extra-thick hard roof movement and mining-induced stress using physical experiments and numerical simulation methods based on mining activities in a longwall panel in the Yima mining area, Henan province, China. The results suggested that the movement and failure processes of the extra-thick roof could be divided into three main periods: the undisturbed, movement stabilization, and sudden collapse periods. The roof displacement remained essentially unchanged during the undisturbed period. During the movement stabilization period, the displacement gradually increased into the upper roof. However, the extra-thick main roof remained undisturbed until the immediate roof experienced its fourth periodic caving in the physical model. Consequently, the displacement expanded rapidly into the extra-thick main roof during the sudden collapse period and the strain energy was violently released when it accumulated in the extra-thick main roof. Additionally, the mining-induced stress was characterized by a sudden decrease in the gradual increase trend when the extra-thick roof instantly collapsed. The deformation and fracture of the extra-thick roof could cause a sudden decrease in the mining-induced stress and lead to continuous and unstable subsidence pressure exerted on the mining panel and roadway. This signi ﬁ cantly contributes to the occurrence of coal bursts.


Introduction
An extra-thick hard roof is the strata with large thickness and high strength occurring above a coal seam or a thin immediate roof. Large area movement and unstable fracture of the extra-thick hard roof are significant hazards threatening safe production in coal mines. It will cause extensive damage to the entire stope and is an important factor of inducing typical dynamic disasters such as coal bursts [1][2][3]. In general, a sudden and violent collapse of the extra-thick hard roof can release masses containing deformation energy and cause coal bursts during longwall panel advancement where machinery and workers assemble.
The frequent extra-thick roof collapse events caused by mining activities have made an important influence on the safety of coal mine production for several years. In November 2011, a severe coal burst resulted in heavy casualties and damaged the roadway section at the 21-221 mining face of the Qianqiu mine in the Yima mining area [4,5]. Although the accident occurrence was related to the F 16 thrust fault near the mining panel, the presence of a 550 m thick and hard conglomerate rock above the coal seam was also a key disaster factor. Therefore, dynamic movement evolution of the extra-thick roof is of primary interest in this study.
The deformation and instability of the extra-thick roof are a source of dynamic disasters. Scholars have performed analyses from different perspectives and obtained fruitful results on the extra-thick roof deformation [6][7][8][9]. Gao et al. [10] studied the mechanism of ground pressure induced by hard roof instability through physical experiment and field measurement. Zhao et al. [11] introduced the large-sized cantilever beam theory to calculate the caving step distance of the thick and hard roof. Ning et al. [12] investigated dynamic movement and fracturing of the thick and hard roof by microseismic monitoring technology. Shen et al. [13] found the strong influence of the fracture position of a thick hard roof on the roadway. Singh and Singh [14] assessed the effect of basic roof thickness on the strata caving behavior and stated that the caving span of basic roof fall increased linearly with the increase in basic roof thickness. Zhou et al. [15] stated that the solid backfill body is an effective way to control the deformation of hard roof and the hard roof's subsidence decreased as solid backfilling ratio increased. Wang et al. [16] found that overburden pressure induced a fracture in a thick hard roof after initial and periodic instability failure. In addition, extensive field investigation results have been obtained regarding the correlation between the movement of the thick hard roof and roadway failure [17,18].
Due to the complexity, periodicity, and suddenness of the extra-thick roof collapses, the movement of the extra-thick roof is a significant determinant for underground coal mining activities. To acquire further knowledge of the dynamic movement and fracture of the extra-thick roof, integrated physical experiments and numerical simulations were conducted to study the relationship between displacement and mining-induced stress of the extra-thick roof. Moreover, the mechanism contributing to the sudden occurrence of coal bursts induced by the extra-thick roof collapse in the Yima mining area is introduced.

Site Descriptions
The Yima mining area contains several faults and synclines, as shown in Figure 1, and was selected as the principal geological background for this study. The Yima mining area is located south of Yima city, China, and covers five majorproducing coal mines, that is, the Gengcun, Yuejin, Yangcun, Changcun, and Qianqiu mines. The number of coal bursts occurring in the last 10 years is also presented in Figure 1.

Physical Experiment of the Movement
Characteristics of the Extra-Thick Hard Roof 3.1. Physical Model Construction. In recent years, many investigators have conducted extensive physical geological engineering experiments to summarize the characteristics of ground subsidence, overburden rock displacement behavior, distribution characteristics of mining-induced stress, and even the fault activation process in mining engineering [19][20][21][22][23][24][25][26][27][28][29][30][31]. To reproduce the process of extra-thick roof movement through physical experiments, the relationship between the physical model and field prototype must satisfy the similarity law [32][33][34]. In this study, three principal types of similarity law parameters were considered: the material strength, material geometry, and material density. The similarity parameters C σ , C L , and C ρ are the ratios of material strength, material geometry, and material density between the field prototype and physical model, respectively. In addition, all of these should satisfy the equation C σ /C ρ ⋅ C L = 1. For the similar simulation experiment of mining engineering, when the qualitative analysis model is needed, the geometric similarity ratio of the model should be taken in the range of 50 to 200. As the main purpose of this paper is to explore the deformation and failure mechanism of an extra-thick hard roof, the physical model design should mainly meet the requirements of strength similarity theory. According to the geological conditions of the extra-thick hard roof and the objective conditions of the experiment system, the optimal geometric similarity ratio is determined as 100. Meanwhile according to the maximum loading limit of the physical model and the mechanical parameters of rock strata, the strength similarity ratio is determined as 160. Therefore, the similarity coefficients C σ , C L , and C ρ are considered as 160, 100, and 1.6, respectively. The similarity materials used to simulate rock strata in field prototypes contain fine sand, gypsum, lime, and water, which are widely used as binder and aggregate materials in physical experiments [19,[32][33][34][35][36]. Figure 2 shows that the physical model built on the experimental platform (GDSTM) has dimensions of 150 × 90 × 10 cm. Therefore, the similarity thickness of siltstone, that of coal seam, that of sandstone mudstone, and that of conglomerate rock in this physical model are 5, 5, 10, 24, and 46 cm, respectively. Because the full height mining scheme is adopted in the similar simulation experiment, this paper mainly studies the characteristics of extra-thick hard roof deformation and collapse under the influence of mining. In order to lay a higher hard roof as far as possible in the physical model, the thickness of coal seam is reduced to only 5 cm. Five layers of rock strata with different similarity materials in physical experiment are made with five mixed proportions, as shown in Table 2. The humidity change will cause the strength change of a similar material in physical experiment, resulting in the similarity error of mechanical conditions between the model and the prototype.

Monitoring Plan and Analysis.
Eleven stress sensors were positioned on the extra-thick conglomerate rock, 34 cm away from the coal seam, to monitor the mining-induced stress evolution during the longwall mining process. The horizontal space between the sensors was 10 cm. The digital speckle image correlation technique was utilized to study the displacement evolution of overburden rock during continuous coal seam mining [37][38][39]. The technique can determine material displacement by following a specific speckle to calculate the movement path between two different pictures [40][41][42][43]. The black speckles with a diameter of 1 cm in this study were arranged in an orderly manner on the rock strata surface to monitor the movement characteristics of the extrathick roof, as shown in Figure 2

Test Results.
In order to avoid extra-thick hard roof sudden collapse in the preliminary mining process, a coal pillar with a length of 20 m is reserved on the boundary of the on-site 21211 mining face. So, a coal pillar with a length of 20 cm was reserved on the right boundary of the model to reduce the boundary effect. The advancing length of the onsite 21221 mining face is 3 m to 8 m each step according to the field data. The average advancing length is 5 m each step under normal mining conditions. The coal seam was mined 5 cm each step from the right boundary to the left boundary of the physical model. Figure 3 presented the mining process and roof caving during mining in the test. During the entire mining process in this test, the roof experienced five periodic caving, which showed distinct continuous failure with caving heights of 2, 6, 6, 14, and 36 cm above the goaf. The roof movement process could be divided into three main periods: the undisturbed, movement stabilization, and sudden collapse periods. When the mining face was mined at 35 cm, the roof did not change significantly within a stable state. This stage was the undisturbed period, as shown in Figures 3(a) and 3(b). With the development and connection of cracks in the roof strata, the roof separation phenomenon became increasingly conspicuous. However, the extra-thick roof was essential in the movement stabilization period, as shown in Figures 3(c)-3(g). Meanwhile, the roof progressively sank repeatedly and the roof caving height increased steadily from 2 to 14 cm during this period. The reason for the roof  3 Geofluids suspension for a long time is that the extra-thick hard roof has great strength and stiffness, which can bear large deformation. Deformation energy is gradually accumulated in the extra-thick hard roof during coal seam mining. The bending and subsidence of hard rock lead to forming the large separation area of the extra-thick hard roof. When the mining face was mined at 110 cm, the roof collapse suddenly expanded into the extra-thick conglomerate in a large area, as shown in Figure 3(h). The roof movement showed an unstable dynamic change in behavior in the sudden collapse period. In this physical experiment, the extra-thick main roof collapse was only observed once when periodic roof caving occurred in the immediate roof. The structural instability of the extra-thick roof was a significant feature of large deformation.
In order to further analyze the spatial structure characteristics of the extra-thick roof in five times the caving state, Figure 4 showed spatial structure distribution of roof caving during advancement of the mining face. The roof caving areas at different stages were distinguished by different color boxes during the experiment process. As the mining face continued advancing and goaf area gradually increased, the caving space was gradually expanded. The development of roof caving was a dynamic process. The characteristic param-eters of fractured strata in caving areas were usually the caving height, high-level fractured rock span, low-level fractured rock span, and caving angle, further analyzed in Figure 5.
Low-level and high-level fracture rocks refer to the lowest and the uppermost fracture rock in the roof collapse area, respectively. They can generally describe the spatial range of the roof collapse area. The span of high-level fracture strata is obviously larger than that of low-level fracture strata in the roof collapse area. That is because the roof collapse area expands upward in the mining process of the coal seam. At the first caving, the fracture spans of the low-level and high-level strata were 35 and 32 cm, respectively. The average fracture span of the low-level strata was 20 cm within the next four periodic roof caving. The fracture span of the high-level strata increased steadily except for the third periodic caving, as shown in Figure 5(a). Figure 5(b) shows the minimum and maximum collapse angles of 50°and 56°, respectively. The caving angle showed little fluctuation during the five periodic caving. When the rock strata occur shear failure model, the caving angle between the failure surface and horizontal direction is approximately same under the same lithology condition. The caving angle is closely related to the internal friction angle of the rock stratum. With the advancement of mining faces, the extra-thick roof always    5 Geofluids experienced change of state alternating between stability and instability after the first roof weighting, and roof pressure release occurred periodically. Because the lithology of low-level strata was mudstone, the breaking span of lowlevel strata remained stable. However, the lithology of high-level strata was composed of mudstone and extrathick conglomerates, and the breaking span of high-level strata changed frequently.
The influence of mining disturbance on the extra-thick roof was relatively small in the undisturbed period, and the entire model vertical displacement was zero, as shown in Figure 6(a). During the movement stabilization period, the disturbance effect of continuous mining gradually became obvious. The extra-thick roof experienced five caving in succession and collapsed rock fills in the goaf. The caving strata displacement in the center of the goaf increased to 5.5 cm due to repeated compression. Because of the rock volume expansion, the strata displacement far away from the center of the mining face gradually decreased. In addition, the displacement gradually increased into the upper roof and could be observed only in the immediate roof and the extra-thick roof was not disturbed by the mining activities, as shown in Figures 6(b)-6(e). However, the mining face was mined to a distance of 110 cm in the sudden collapse period and Figure 6(f) shows that the displacement expanded rapidly into the extra-thick roof and the energy accumulated in the hard conglomerate was suddenly released.
With the mining face advancement, the stress balance of the extra-thick roof was broken due to roof collapse. The stress was redistributed to form a new overburden equilibrium structure. Five stress monitoring points marked as #1, #4, #6, #8, and #11 in the extra-thick conglomerate were selected to analyze the characteristics of stress variation, as shown in Figure 7. When the mining face was advanced to a distance from 0 cm to 70 cm, the stress of all monitoring points gradually increased. And the stress of each monitoring point had slightly fluctuated at the same mining distance. When the mining face was advanced to a distance from 75 cm to 110 cm, the stress of 8# monitoring point in extrathick roof would go through two stages, including sharp increase stage and sharp decrease stage. Therefore, the stress variation and the movement state of the extra-thick roof are combined for integrated analysis. During the undisturbed period, the mining disturbance influence on the extra-thick conglomerate increased. The extra-thick roof stress gradually increased, and the deformation energy slowly accumulated in the extra-thick conglomerate. In the movement stabilization period, the extra-thick conglomerate stress increased to the peak stress. In the sudden collapse period, the extra-thick conglomerate stress dropped suddenly. Because of the extensive deformation energy of the extra-thick conglomerate, the deformation energy was violently released in a sudden manner when the extra-thick conglomerate collapsed over a large area. Therefore, the mining-induced stress was characterized by a sudden decrease in the gradually increasing trend when the extra-thick roof suddenly collapsed.  6 Geofluids movement and large deformation of a massive system under static or dynamic loading of discontinuous medium (such as jointed rock mass). It is suitable for applying to simulate the movement characteristics of an extra-thick roof under the influence of mining disturbance. The material in the 3DEC model was divided into two parts: the rock block as the continuity and the structural plane between rock blocks as the discontinuity. Considering the same physical model dimensions, a geometrical dimension of 150 × 10 × 90 m was established in the 3DEC model, as shown in Figure 8. Combined with the specific characteristics of each rock stratum, the relevant mechanical parameters of the strata and structure plane between blocks in the numerical mode were shown in Tables 3 and 4 As the mining face advanced to 20 m, there was no obvious extra-thick roof movement but the immediate roof had a triangular shape displacement, as shown in Figure 9(a). As the mining face advanced to 40 m, the immediate roof collapsed under the action of gravity. The main roof, 24 m above the coal seam, exhibited a distinct separation phenomenon, shown in Figure 9  7 Geofluids separation area expanded further, as shown in Figure 9(e). The extra-thick roof moved downward, and the roof displacement increased gradually as the mining face advanced to 110 m, as shown in Figure 9(f). The extra-thick roof collapsed in a large area, which could easily cause coal bursts. Clearly, the broken strata of the extra-thick roof successively progressed through separation, bending subsidence, closing separation, and caving compaction processes.

Movement Evolution
The rock strata above the mining face began to move down slowly, and the velocity vector appeared at the immediate roof, as shown in Figure 10(a). The extra-thick roof caving in small areas led to stress redistribution, and the velocity vector of the roof enlarged as the mining face advanced to 40 m, as shown in Figure 10(b). As the mining face further was advancing, the collapsed rock strata far away from the mining face entered the compaction state and the velocity vector was essentially zero, as shown in Figures 10(c) and 10(d). In addition, there was a wide range of unstable roofs near the mining face moving to the goaf, and the roof separation gap enlarged. The extra-thick conglomerate had a downward movement trend with a small velocity, as shown in Figure 10

Mining-Induced Stress Evolution
Characteristics of the Extra-Thick Hard Roof

Numerical Model.
A finite difference program FLAC 3D is applied to simulate the mining-induced stress distribution characteristics of an extra-thick hard roof during the advancement of the mining face. By building the appropriate constitutive relation of geological materials in FLAC 3D , the progressive mining-induced stress evolution of elasticplastic rock mass is effectively tracked. Combined with the results of physical experiment, the stress evolution of the extra-thick roof in the numerical model was comparatively analyzed. A numerical model in FLAC 3D with 12 stress monitoring points numbered #1-#12 located on extra-thick conglomerate rock was established. The location of the stress monitoring points corresponds to the stress sensor distribution in the physical model to monitor mining-induced stress in the extra-thick conglomerate, as shown in Figure 11. Adjacent stress monitoring points were placed 10 m apart. A vertical stress of 20.0 MPa was applied at the top of the numerical model, and a horizontal stress of 23.4 MPa was applied to the two side boundaries. Moreover, the horizontal boundary displacement was limited and the vertical displacement of the numerical model base was fixed. The physical and mechanical parameters of rock strata in the FLAC 3D model after several debugging cycles were used, as shown in Table 5.

Stress Distribution Characteristics in the Extra-Thick
Roof. As the mining face was advancing, the vertical stress in the extra-thick roof was gradually released. The release of vertical stress showed that the absolute value of stress decreased and the release area of vertical stress gradually increased into the upper roof, as shown in Figures 12(a)-12(f). The supportive effect of the original coal seam on the extra-thick roof gradually disappeared. The effect of model top loading and gravity caused bending subsidence of the extra-thick roof. As the mining face advanced to 110 m, the extra-thick roof experienced the largest stress release.
As shown in Figure 13, there were five stages of severe stress change in the extra-thick conglomerate during coal  To confirm the reliability of the numerical simulation results, the mining-induced stress obtained from the physical experiment was compared with that obtained from numerical simulation. In this study, the stress change per    11 Geofluids meter of the #8 monitoring point at the same position as the numerical simulation and physical experiment in the mining face advancement was selected for comparative analysis. Figure 14 showed that the stress change per meter obtained from physical experiments was consistent with that obtained from numerical simulation. In the early stage of coal mining, the stress fluctuation in the extra-thick roof was small. But when the mining face advanced to 110 m, there was a violent fluctuation, indicating that the disturbance of coal mining to extra-thick conglomerate reached its maximum. The fifth caving of the extra-thick roof occurred in the goaf and the stress changed violently; the stress variation behavior at the #8 measuring point at the same position between numerical simulation and physical experiment was basically synchronous.  12 Geofluids

Discussions
According to the displacement and mining-induced stress variation in the extra-thick roof, the following areas should be discussed in greater detail to analyze the dynamic evolution characteristics of deformation and movement of the extra-thick roof. The extraction of a longwall mining face induced mininginduced changes and extra-thick roof movement above the mining face. The movement and failure processes of an extra-thick roof could be divided into three main periods: the undisturbed, movement stabilization, and sudden collapse periods. Because the mining face advancement distance in the physical experiment was short, the extra-thick roof deformation caused by bending and sinking was relatively small. The mining-induced stress of typical stress sensors was small; therefore, this stage could be categorized as belonging to the undisturbed period. With mining face advancement, the suspended roof length gradually increased and the roof bent and sank under the action of gravity. The bending moment and shearing force at both roof ends gradually increased, leading to roof failure. In addition, the horizontal stress was applied to the left and right boundaries of the physical model and the extra-thick roof collapsed slowly under the action of horizontal pressure. Therefore, this stage was part of the movement stabilization period. Because the main roof was composed of an extra-thick conglomerate, it could sustain a large stress level. Although the roof movement was dynamic in coal mining, an extra-thick roof was not disturbed until the roof reached a certain degree of deformation. The UCS of the extra-thick roof was 45 MPa, and its failure could occur only when the effect of the roof subsidence increased the extra-thick roof stress to the ultimate strength. It broke and caused roof caving in a large area and could easily induce coal bursts. This stage corresponded to the sudden collapse period.     14 Geofluids mining-induced stress was characterized by a sudden decrease in the gradually increasing trend when the extrathick roof suddenly collapsed, as shown in Figures 7 and  13. The stress change characteristics of the #8 measuring point at the same position in both the physical experiment and FLAC 3D model were basically synchronous, as shown in Figure 14.
Hence, the deformation and fracture of the extra-thick roof could lead to the continuous and unstable subsidence pressure exerted on the mining face and roadway, which provided a continuous force to the immediate roof, and coal seam and could cause a sudden decrease in the mininginduced stress of the extra-thick roof. This significantly contributed to the occurrence of coal bursts.

Conclusions
Numerical simulations and physical experiments were conducted to study the movement and fracture characteristics of the extra-thick roof during coal seam mining. The detailed conclusions are as follows.
(1) The movement and failure processes of the extrathick roof could be divided into three main periods: the undisturbed, movement stabilization, and sudden collapse periods. The roof displacement essentially did not change during the undisturbed period. During the movement stabilization period, the displacement gradually increased into the upper roof and could be observed only in the immediate roof and the extra-thick roof was not disturbed by the mining activities. However, the displacement expanded rapidly into the extra-thick main roof during the sudden collapse period (2) Although roof movement is a dynamic process during coal seam mining, the extra-thick main roof was undisturbed until the immediate roof experienced the fourth periodic caving. The UCS of the extrathick main roof was 45 MPa, and its failure could occur only when the effect of the roof subsidence increased the extra-thick main roof stress to the ultimate strength. Therefore, the strain energy was violently released when it accumulated in the extrathick roof failure process (3) The extra-thick main roof collapse was only observed once in the physical experiment when a periodic roof caving occurred in the immediate roof. The displacement and velocity vector of the extra-thick main roof increased sharply in this periodic caving, and the collapse height instantly expanded into the extra-thick main roof. Meanwhile, the mining-induced stress was characterized by a sudden decrease in the gradually increasing trend when the extra-thick main roof suddenly collapsed (4) As the mining face advanced, the movement and fracture of the extra-thick roof led to a continuous and unstable subsidence pressure exerted on the mining face and roadway. This pressure exertion provided a continuous force to the immediate roof and coal seam. This significantly contributed to the occurrence of coal bursts

Data Availability
All the original data used to support the findings of this study are available from the corresponding author upon request.

Conflicts of Interest
The authors declare that they have no conflicts of interest.