Study on Coal Gangue Identification Based on Vibration and Its Adaptability to Coal-Caving Process Parameters

,


Introduction
Coal is an important energy pillar in China, accounting for more than half of the national energy consumption, and China's coal consumption accounted for 50.72% of the world's total in 2017 [1,2].Te thick coal seam refers to the coal seam whose thickness is greater than 3.5 m.China has rich reserves of the thick coal seam.Fully mechanized cave mining technology has been widely used in many mining areas due to its high yield and high efciency [3,4].How to control the closing of the coal opening is one of the key links in the process of top coal cave mining.Te earlier closing of the coal opening will make part of the coal not to be released thus leading to wastage, and the later closing will make the gangue to be excessively released, thereby reducing the quality of the coal and increasing the cost of late coal preparation.At present, the closing time in coal mines still depends mainly on the manual observation and hearing.However, this method is easily afected by the experience of the operators, and the operating environment is also dangerous, which is not in line with the developmental trend of unmanned coal mines in the future.
Because of the current situation of low automation of the coal caving process, many scholars have carried out research studies on the coal gangue identifcation.Zhang et al. [5] proposed that the diference in natural gamma rays between coal and gangue can be used to identify the gangue content in a coal gangue mixture.Wang et al. [6] proposed image recognition intelligent coal caving technology, to improve the automation level of the coal caving process.Hou et al. [7] established a coal gangue sorting system based on the artifcial neural network and image feature extraction by using the diference in surface texture and the gray level of coal and gangue.Wang and Zhang [8] proposed a method to obtain volume data and to further calculate the density by using three-dimensional laser scanning technology, which efectively solved the problem of separating coal and gangue.Guo et al. [9] proposed an identifcation method based on the diference in dielectric properties of coal gangue and obtained the relationship curve of the dielectric constant of coal and gangue with frequency and voltage.
In summary, while the current top coal caving working face is still dominated by manual control, scholars have also explored various intelligent coal caving control technologies.However, the harsh environment of the top coal caving feld, such as noise, dust, and low visibility, seriously afects the application efect of many coal gangue identifcation technologies [10][11][12].
Te basic principle of the vibration-based coal gangue identifcation method is to use the obvious diference in the physical properties of coal and gangue, resulting in diferent vibration signals of the tail beam after being impacted by coal and gangue.Tis method can efectively resist the infuence of various noises and dust, and since its characteristics are based on the vibration sensors, it also makes the installation of equipment more convenient, so it has a good application prospect.However, at present, the research on coal gangue identifcation based on vibration mainly focuses on the impact process of a small number of rock particles and signal processing [13][14][15].Tere are a few studies on the coal gangue identifcation efect in the whole process of coal caving, while no study has yet focused on how the application efect of the vibration identifcation method is afected by the numerous process parameters of the coal release process and environmental conditions.
DEM as a powerful tool to study particle movement has been widely studied by Zheng et al. [16].Ketterhagen et al. [17] used DEM to analyze the hopper powder fow, and DEM simulation results and design charts are very consistent.Coetzee et al. [18] proposed a calibration process of the DEM method to calibrate the parameter value of broken rock particles with a size of 40 mm.Su and Ali Akcin [19] used DEM to propose a numerical model that can predict the tool force in the process of rock cutting, and the results of this model have a strong correlation with the results of theoretical and experimental studies.Hang et al. [20] used the DEM method to study the micromotion and macrodisturbance behavior of the soil under the combined action of two subsoils.
Mastering the variation law of the application efect of the vibration-based coal gangue identifcation method under various working conditions can better guide the formulation of the coal caving process in the future.Terefore, based on the coupling technology of DEM and FEM in LS-DYNA, this article established a three-dimensional numerical model of the top coal caving in fully mechanized cave mining.Te diference of vibration signals of the tail beam in diferent stages of coal caving is studied, and the infuence of coal caving step distance, mining and caving ratio, and the pitching angle of the mining feld on the application efect of the vibration identifcation method is discussed.Tis article analyzes how to control the coal caving process to better obtain a better recognition efect from the perspective of vibration-based coal gangue identifcation technology.Te conclusion can provide guidance for better application of the vibration identifcation method by adjusting the coal caving process.

Model Description.
In this article, a numerical model is established as shown in Figure 1.Te left side of the fgure is the complete model, and the bafe on the top is the boundary of the coal gangue particles fowing around, and there is an opening at the bottom of the bafe to simulate the coal opening.A funnel is arranged below the coal opening to load the particles released from the coal opening.Te short side width of the funnel is 500 mm, which is slightly larger than the maximum diameter of the particles, and it is convenient to count the gangue content in the top coal released.Te model simplifes the hydraulic support, and the complete hydraulic support is shown in the fgure.Considering the calculation time and the motion state of the components in the coal caving process, only the canopy, shield beam, and tail beam are left.At the same time, the canopy and shield beam are simplifed as fat plates.Te degrees of freedom of bafes, canopy, and shield beams are fully constrained, and gravity is applied to the model, g � 9.81 m/s 2 .

Discrete Element Model of Coal Gangue Particles.
Te DEM is proposed by Cundall in the 1970s [21]; it is a numerical simulation method for solving discontinuous media problems.In LS-DYNA, the interaction between particles is defned by the keyword * CON-TROL_DISCRETE_ELEMENT, including the normal damping coefcient, tangential damping coefcient, sliding friction coefcient, and rolling friction coefcient between particles.Te infuence of the normal damping coefcient and tangential damping coefcient on particle motion is very important.When two particles collide, a damping coefcient of 0 is a completely elastic collision.When the damping coefcient is 1, the two particles are no longer separated after the collision, and the normal and tangential damping coefcients are set to 0.5.Te CAP parameter in this keyword can control whether it is dry or wet particles.If CAP � 0, it is dry particles.Coal gangue particles can be regarded as an ideal loose medium.Tere is no adhesion between the particles in this kind of a medium.Terefore, this parameter is set to 0.
Te interaction between particles is processed by the algorithm based on the penalty function [22].Figure 2 shows the normal contact force and tangential force between particles [23].
Some basic assumptions are needed to establish the numerical model of the coal gangue particles during coal caving and they are as follows: 2 Shock and Vibration (1) Te support is only afected by the coal gangue particles from above (2) Te boundary of the broken top coal above the support is located in half of the top beam (3) Te broken top coal is loose, and the pressure from the surrounding particles in the process of coal caving is not huge, so the crushing of coal gangue particles is not considered (4) Te shape of the broken coal gangue particles is spherical LS-PrePost software is used for the creation of particles.In LS-PrePost, the boundary of the discrete element particles must consist of shell elements.After the boundary is specifed, software generates loose particles randomly according to the radius range and gap entered.Te radius range of the particles is 150 mm∼200 mm.Tere is a certain gap between the particles in the initial state, and the particles gradually form the compaction state under the action of gravity after the beginning of the simulation.Te sliding friction coefcient of the normal damping coefcient between particles is 0.65, and the rolling friction coefcient is 0.1.
Because particles are regarded as rigid bodies in the DEM algorithm, * MAT_ELASTIC is selected as the material of coal and gangue.Te material model parameters of coal and gangue are shown in Table 1 [24], where ρ 0 is density, E is Young's modulus, and PR is Poisson's ratio.
Te broken top coal and gangue before coal caving will form a clear boundary between coal and gangue, which can generally be described by a parabola, as shown in Figure 3 [25].Te shape of the coal gangue boundary is related to the coal caving step distance, top coal thickness, and threedimensional size of the support.According to Wang's research [25], the numerical model of the coal gangue interface settings is shown in Figure 4.

Finite Element Model of the Tail Beam of Hydraulic
Support.Te hydraulic support used in this article is ZF4800-17-32 top coal-caving hydraulic support.Te maximum and minimum work height of the support is 3.2 m and 1.7 m, respectively, and the working resistance is 4800 kN.Te structural composition and fnite element model of the support are shown in Figure 5. Te tail beam and the shield beam are parallel before the coal caving, and the coal gangue particles are accumulated above the hydraulic support.After the coal caving begins, the tail beam swings downward and forms the coal opening.Te coal gangue particles fow out along the coal opening and are transported to the scraper conveyor below.In previous studies, we found that the vibration signal on the tail beam is signifcantly stronger than that on the shield beam because the state of the tail beam was closer to the cantilever beam structure.Due to a large number of discrete element particles in the numerical model, it is found that the calculation time is not acceptable after trying to use the complete hydraulic support model.Terefore, only the tail beam structure is retained, and the rigid plate is used to replace the canopy and the shield beam, as a part of the fow boundary of coal gangue particles.Te degrees of freedom of the position      shown in Figure 6 are all constrained to simulate the fxed mode of the tail beam in the hydraulic support.

Coupling Algorithm.
Tere are two ways to deal with the contact between the discrete element and the fnite element in LS-DYNA.One is using * CON-TACT_AUTOMATIC_NODE_TO_SURFACE.Although static friction and dynamic friction coefcient can be defned in this classical way, it is impossible to apply rolling friction, and the friction force acts on the center position of the discrete element particles.Te other is using * DEFIN-E_DE_TO_SURFACE_COUPLING.Tis method is a special keyword defned in the LS-DYNA.It defned the contact characteristics between particles and structures through friction coefcient, rolling friction coefcient, and damping coefcient.It takes the outer surface of the particles as the physical boundary, and the force acts on the periphery of the particles rather than the center of mass, so it is closer to the real situation.Terefore, this article chooses this method to defne the contact between the coal gangue particles and the tail beam.Te friction coefcient between coal and gangue and each part of hydraulic support is 0.3, the rolling friction coefcient is 0.1, and the damping coefcient is 0.6.

Signal Analysis at Different Stages of the Whole Process of Coal Caving
3.1.Signals at Diferent Stages of Coal Caving.In order to quickly determine the time when the coal-caving state is coal gangue mixing, the tail beam in the numerical simulation is defned as the rigid body, and the numerical simulation time is 15 s.In this case, the simulation time will be greatly reduced.After determining the time of coal gangue mixing, the rigid-fexible transformation method in LS-DYNA is used in the original model, and the material transformation of the tail beam is realized by the keyword * DEFORMA-BLE_TO_RIGID_AUTOMATIC.Before reaching the set time, the tail beam is rigid, and after reaching the set time, the tail beam is automatically transformed into a fexible body.Te space occupied by the released top coal before they are released is called the released body.In order to observe the shape of the released body, we fnd out the number of the released coal gangue particles and delete them one by one in the initial model, and the hole is the shape of the released body.Figure 7 shows the shape of the released body at diferent times in the process of coal caving.From Figure 7, it can be seen that the shape of the released body is basically elliptical and conforms to the rule summarized by Wang through experiments and simulations [25].Terefore, the setting of the numerical model is reliable.
During the development of the released body, the long axis and the short axis of the ellipse increase continuously, and the increase rate of the long axis is signifcantly higher than that of the short axis under the action of gravity.When the released body develops in the coal, then the release medium is pure coal; when the boundary of the released body is tangent to the coal gangue boundary, the gangue in the released body will be released.Figure 8 shows the shape of the coal gangue boundary at diferent times.After the beginning of the coal caving, the bottom of the coal gangue boundary began to shrink in the direction of the hydraulic support.Te friction coefcient between the coal gangue and the steel plate is about 0.3, and the friction coefcient between the coal gangue is about 0.65.Te fow velocity of the particles above the tail beam and the shield beam is faster than that on the right side of the coal caving port.At 13 s, there are both coal and gangue in the particles contacting the tail beam, and the coal accumulated at the bottom belongs to the part that cannot be released.
Te images at 12 s, 13 s, 14 s, and 15 s were intercepted, respectively, and Image-Pro Plus software is used to calculate the gangue content of the particles left in the funnel.Te gangue content was calculated by counting the number of coal and gangue pixels in the image.Te statistical process is shown in Figure 9, and the red pixels are the statistical objects.Table 2 shows the statistical results.Te gangue content less than 20% is acceptable in actual production.
In order to compare the diference in vibration signal of the tail beam between the coal-caving stage and coal gangue where X rms is the valid value, x i is the value of the sampling point, and n is the number of sampling points.Te time-domain characteristics of the three signals are shown in Figure 13.It is obvious that, in the coal caving process, the three characteristics are signifcantly increased.In the coal gangue mixing stage, the gangue is continuously contacted with the tail beam and released; because the density of the gangue is signifcantly higher than that of the coal, it is more likely to transfer higher energy to the tail beam, so the acceleration signal will be stronger.In the frst stage of the coal-caving process, the particles arched in a short time, resulting in less coal caving, so its eigenvalues are slightly less than the eigenvalues of the second stage of the coal-caving process.
Te vibration signal of the tail beam is a random signal, which cannot be reproduced by a specifc time function, so it needs to be analyzed by statistical methods.Empirical mode decomposition (EMD) was proposed by Norden E. Huang et al. in 1998.Tey believed that any complex signal was composed of diferent modes, and the superposition of these    Shock and Vibration modes formed a nonstationary signal.Te EMD method is used to decompose the abovementioned three signals.
Figure 14 shows the EMD results of the acceleration signals in the two-stagecoal-caving process, and Figure 15 shows the EMD results of the acceleration signals in the coal gangue mixing stage.Te original signal is decomposed into nine IMF components, which contain diferent frequency components from high to low.Te frst four IMF components have higher frequency and amplitude, indicating that possible important information is included in the frst four IMF components.
In order to represent the diference between signals more intuitively, the decomposed components are further processed to extract feature parameters.
Te mean of the absolute value is used to represent the average level of the signal.Te expression is where Mean is the mean of absolute value.Te peak value is used to represent the range of IMF component amplitude.Te formula of the peak value is where Peak is the peak value.Te peak factor is used to represent the fuctuation of the signal, and its value is the ratio of the peak value to the mean value of the absolute value.Te calculation formula is as follows: where PeakFactor is the Peak factor.Te comparison of each eigenvalue of each IMF component of the three acceleration signals is shown in Figure 16.In general, the eigenvalue of the IMF component decreases with the decrease in frequency.For a single eigenvalue, the diference between the frst two IMF components is the most stable, and the eigenvalue of the coal gangue mixing stage is signifcantly larger than that of the coal caving stage.With the decrease in the frequency of the component, the eigenvalue of the coal-caving stage is higher   Shock and Vibration than that of the coal gangue mixing stage in some lowfrequency components.Terefore, when the gangue is released, the eigenvalue of the high-frequency component of the vibration signal will be signifcantly enhanced.
In the numerical simulation, the gangue content of the fnal released coal gangue particles is less than 10%.By analyzing the time-domain characteristics of the acceleration on the tail beam and the eigenvalues of the IMF components after EMD decomposition, it is found that the vibration signal characteristics of the coal gangue mixing stage and the vibration signal characteristics of the coal caving stage have an obvious upward trend.Terefore, it can be considered that when the gangue content increases, the nature of the medium in the discharge body can be judged by monitoring the vibration signal on the tail beam.

Analysis of the Vibration Identification Effect under Different Coal Mining Technology and Environments
In the process of top coal-caving mining, there will be a variety of diferent conditions, such as diferent coal seam thickness and pitching angle in the mining feld.In the previous study of coal gangue identifcation based on vibration, researchers focus on the infuence of the change of the particle itself on the vibration signal and have not yet analyzed the infuence of environmental factors in the whole coal-caving process.In order to better apply the method of coal gangue identifcation based on vibration in the future, it is necessary to analyze the environmental factors involved in the coal mining process.Tis section analyzes the coalcaving step distance, the mining and caving ratio, and the pitching angle in the mining feld.

Efect of Coal-Caving
Step Distance.Te coal-caving step distance refers to the distance with which the fully mechanized mining work face moves forward between the two coal-caving processes.Te coal-caving step distance is an important factor afecting the recovery rate and the gangue content of the working face.In order to analyze the feasibility of applying the vibration-based coal gangue identifcation method under diferent coal-caving step distances, the distance between the tail beam and the bottom of the coal gangue interface is adjusted to 2.8 m, 3.5 m, and 4 m, respectively.Te coal gangue interface before coal caving is shown in Figure 17.We observed the coal-caving process in three cases, focusing on the contact time of gangue and tail beam.It should be emphasized that based on the work in Section 3, we have proved that when the particles contacted with the tail beam that contained some gangue particles, some characteristic values of the acceleration signal will change signifcantly.In order to reduce the repetitive work and to avoid excessive length, the vibration signal is no longer analyzed in this section, but the gangue content in the particles' contact with the tail beam is directly observed.When the gangue particles occur, we believe that the characteristics of the vibration signal are enhanced.
Figure 18 shows the process of coal caving when the distance is 2.8 m.After the beginning of coal caving, the vertical part of the coal gangue boundary gradually moves to Shock and Vibration 7 the coal-caving mouth.Te gangue begins to be released between 6 s and 8 s.However, there is no gangue in the particles contacting with the tail beam at 10 s.Te gangue began to contact the tail beam at around 11 s, and the gangue content was about 20.2%.It can be seen that when the coal caving step is small, the gangue will be released in advance from the side of the coal-caving mouth away from the hydraulic support, resulting in a high gangue content in the coal gangue particles released when the gangue is in contact with the tail beam.Terefore, when the vibration-based coal gangue identifcation method is used in the small step of coal caving, the identifable signal will lag behind the change of the gangue content.
Figure 19 shows the process of coal caving with a distance of 3.5 m.At about 12 s, the gangue is released from the coal opening, and at about 13.4 s, the gangue particles contact the tail beam.Te time interval from the beginning of gangue release to the beginning of gangue contact with the tail beam is far less than that of the distance of 2.8 m.Te gangue content is about 4.67% at 13.4 s, which fully meets the requirements of mining.
Figure 20 shows the process of coal caving with a distance of 4 m, due to the increase in the distance, the coalcaving time needs to be extended.At about 11 s, the gangue began to discharge, and at about 12.8 s, the gangue began to contact the tail beam.At this time, the gangue content was about 5.1%.Since the fuidity of the bottom is worse than that in the vertical direction, more coal will be cut of and will be accumulated at the bottom in the case of a large step distance.However, according to the requirement of gangue content, these coal particles accumulated at the bottom cannot be released, resulting in waste.
Due to the efect of gravity, the horizontal movement speed of the bottom particles is much smaller than that in the vertical direction.Trough the analysis of the coal-caving process under three diferent coal-caving step distances, it is 8 Shock and Vibration found that the particles that fnally contacted with the tail beam are mainly from the gangue at the high position.Terefore, when the gangue mentioned above reaches the tail beam, the movement of the vertical coal gangue boundary has a great infuence on the application efect of the vibration-based coal gangue identifcation method.When the step distance is small, the change of vibration signal on the tail beam seriously lags behind the change of the gangue content, and the application efect of the coal gangue identifcation method based on the vibration is poor.When the step distance increases properly, the change in the gangue content can be refected in the vibration signal of the tail beam in time.However, when the step distance is too large, some coal will accumulate at the bottom and will cause loss.Terefore, in the future, when using vibration-based coal gangue identifcation, the caving step should be reasonably adjusted to achieve better identifcation results.

Te Infuence of Mining and Caving
Ratio.In a certain direction, the thickness of the coal seam in the mining area often changes.For example, the thickness diference between the east and west of the coal seam in Jiulishan Mine is large,

Shock and Vibration
and the maximum thickness is 7 m, which becomes thin from east to west.According to the change in coal seam thickness, it is necessary to determine a reasonable mining and caving ratio to achieve a high resource recovery rate to ensure the safety and stability of the production.In order to study the efect of using the vibration identifcation method under diferent mining and caving ratios, this section conducts a numerical simulation of the coal-caving process under three diferent mining and caving ratios.Figure 21 shows three diferent mining and caving ratios, which are 1 : 1, 1 : 1.43, and 1 : 2.67, respectively.Te distance from the bottom of the coal gangue interface to the tail beam is the same.
Figure 22 shows the coal caving process under the condition of mining and caving ratio of 1 : 1. Te frst time of gangue discharge is 9.3 s, and the time of gangue beginning to contact with the tail beam is 9.9 s.At this time, the gangue content of the coal gangue particles in the funnel is 2.9%. Figure 23 shows the coal caving process under the condition of mining and caving ratio of 1 : 1.43; the time of the frst release of the gangue is 11.7 s, and the time when the gangue begins to contact the tail beam is 12.5 s.At this time, the gangue content of the coal gangue particles in the funnel is 1.9%.Figure 24 shows the coal-caving process under the condition of mining and caving ratio of 1 : 2.67, the time of the frst release of the gangue is 15.1 s, and the time of the within the scope of meeting the production requirements, and the efect of vibration identifcation method is better.

Te Infuence of Pitching Angle of Mining Field.
In some areas, due to the complex geological conditions, coal seam inclination is often encountered in the process of advancing the coal mining face.In this case, it is necessary to combine the method of upward mining and downward mining to complete the mining work.Te infuence of upward mining and downward mining on top coal-caving law is diferent.In this section, the discrete element numerical models at different angles are established to analyze the infuence of the angle of upward and downward mining on the identifcation efect of the vibration-based coal gangue identifcation method.Figure 25 shows the numerical models of the coal drawing process at diferent angles established in this section.α � −20 °, −10 °, 0 °, 10 °, and 20 °. α are the angles between the ground and the forward direction, which stipulates that the upward mining is positive and the downward mining is negative.
Figure 26 shows the coal-caving process under the conditions of −20 °and −10 °.Diferent from the horizontal mining, under the action of gravity, the lower part of the coal gangue boundary moves signifcantly faster to the coal opening.When the angle is −10 °, the frst time of the gangue discharge is 7.2 s, and when the angle is −20 °, this time was advanced to 6 s.In contrast, in downward mining, the frst time the gangue particles discharge from the coal-caving mouth under the horizontal mining with the same coal gangue interface is 11.7 s.It can be seen that the time of gangue discharge is greatly advanced under the condition of downward mining, and the larger the angle is, the earlier the time is.When the gangue is released, there is still a large number of coal particles at the top of the coal opening, which will keep the gangue particles from contacting the tail beam.When the gangue particles located at the top of the coal gangue boundary in the initial state start to contact with the tail beam, the gangue located at the top of the coal gangue boundary will also be released in large quantities.In the case of −20 °, the contact time between the gangue and the tail beam is 13.7 s, and the gangue content of the released coal gangue particles is 33.3%.When the angle is −10 °, the contact time between the gangue and the tail beam is 11.6 s, and the gangue content is 13.3%.By contrast, when α � 0, the contact time between the gangue and tail beam is 12.5 s, and the gangue content is 1.9%.Terefore, when the working face is undermined, the signal will lag behind the change of gangue content.
Te caving process is shown in Figure 27 when the upward mining angles are 10 °and 20 °, respectively.Unlike

Shock and Vibration 13
with the release of gangue, and the gangue content in the released gangue particles will be very low.

Conclusions
In this article, the discrete element and fnite element coupling method is used to study the process of coal caving by monitoring the vibration signal of the tail beam to determine whether there are gangue particles released from the coal opening.After verifying the feasibility of the method, the infuence of coal-caving step distance, mining and caving ratio, and pitching angle of the mining feld on the recognition efect is studied.
(1) In the process of coal caving, when the gangue contacts with the tail beam, the time-domain characteristics of the acceleration signal on the tail beam and some characteristics of the IMF component after EMD will be signifcantly higher than those in the pure coal-caving stage, indicating that the acceleration signal can be used as a basis for judging whether the gangue particles will be released.(2) Te coal-caving step distance has an important infuence on the recognition efect.When the coalcaving step distance is small, the gangue particles will be released in advance from the side away from the hydraulic support of the coal opening, resulting in a large number of gangue particles being released when the gangue particles from the top contact with the tail beam, and the gangue content is unqualifed.When the coal-caving step is large, although when the gangue particles contact the tail beam, the gangue content in the released particles is qualifed, some coal particles at the bottom will pile up and cannot be released, resulting in coal waste.(3) Te application efect of the vibration identifcation method in diferent mining and caving ratios is ideal.With the thickness of the coal seam, the time from the gangue particles arrival at the coal opening to the gangue particles' contact with the tail beam increases gradually, and the discharge of the gangue also increases.However, because the number of coal particles also increases at the same time, the gangue content is always maintained in an acceptable range.(4) Te caving process of upward mining and downward mining is obviously afected by gravity.In the case of upward mining, the time from the gangue particles' arrival at the coal opening to the gangue particles' contact with the tail beam is very short.Te application of the vibration identifcation method has a good efect, which can ensure that the mined coal has a low gangue content, but it will cause the coal particles at the bottom to be difcult to be discharged leading to wastage.Downward mining will greatly advance the time of gangue being released, and the top coal will continue to hinder the contact between gangue particles and the tail beam, resulting in the change of signal lagging behind the change of gangue content.Te efect of the vibration identifcation method is not good.
Te conclusions of the article can provide a certain reference for the popularization and application of the vibration-based coal gangue identifcation method.

Figure 1 :
Figure 1: Te numerical model of drawing coal process.

Figure 3 :
Figure 3: Te coal gangue boundary line before and after coal caving.

Figure 4 :Figure 5 :
Figure 4: Te coal gangue boundary in the numerical model.

3. 2 .
Signal Processing and Feature Extraction.In order to show the diferences among the three signals more intuitively, the maximum, minimum, and root mean square of the signals are extracted, respectively.Te root means square is defned as follows:

Figure 6 :
Figure 6: Te fnite element model of the tail beam.

Figure 7 :Figure 8 :Figure 9 :
Figure 7: Forms of the released body in diferent stages of coal caving.

Figure 10 :
Figure 10: Vibration signal of the tail beam in 3∼5 s stage.

Figure 14 :
Figure 14: EMD results of acceleration signals in coal-caving stage.

Figure 15 :
Figure 15: EMD results of vibration signals in coal gangue mixing stage.

Figure 16 :
Figure 16: Eigenvalues of the IMF components of acceleration signals at diferent stages.

Figure 23 :
Figure 23: Te coal-caving process when the mining and caving ratio is 1 : 1.43.

Table 1 :
Material model parameters for coal and gangue.

Table 2 :
Gangue content of released body at diferent time periods.