Aquiclude Stability Evaluation and Significance Analysis of Influencing Factors of Close-Distance Coal Seams: A Case Study of the Yili No. 4 Coal Mine in Xinjiang, China

,


Introduction
Mining of shallow coal seams in areas with scarce water resources and environmental problems inevitably causes the overlying rock subsidence and groundwater level drop, jeopardizing the fragile ecological balance [1][2][3]. In particular, this refers to the arid and semiarid regions of western China, for which the water-preserving mining seems to be the most lucrative solution [4][5][6][7]. Water-preserving mining is concerned with the protection of water resources throughout the process of mining. This is critical for rationally realizing green and climate-smart mining [8]. Its realization depends, in turn, on the aquiclude stability control and its robust substantiation. Therefore, the quantitative evaluation of aquiclude stability has important guiding significance for water-preserving mining [9][10][11][12].
The impact of coal seam mining on shallow water resources can be reduced to the following three cases: (1) The aquifer water level remains unchanged. (2) The aquifer water level decreases. (3) The aquifer water level restores after decreasing. Numerous experts and scholars have studied the aquifer water level behavior using different test methods and criteria. Thus, several authors reported that the relationship between the development height of the fracture zone controlled by the key stratum and the waterresisting key strata and the aquifer determined the variation behavior of the aquifer's water level [13][14][15][16][17]. When the aquifer is above the fracture zone, the water level does not change; otherwise, the water level drops. Alternatively, other researchers reported that the degree of development of cracks in aquicludes determines the variation behavior of the aquifer level [18][19][20]. When the upward and downward cracks penetrate through the aquiclude, the water level drops; otherwise, the water level does not change. Other researchers studied the tensile, shear, and bending failure criteria of aquiclude stability through laboratory experiments and theoretical analysis [21][22][23]. Other researchers explored the water level changes before and after mining through on-site observation of the aquifer water level [24][25][26]. It was found that the water level gradually recovered after the drop. The above findings were mainly related to cases where the fracture zone reached the aquiclude or the aquiclude deformation reached its limiting value. They provided no quantitative assessment of the aquiclude stability or considered the seepage effect. In coal-bearing strata of western China, which have a short diagenesis time, aquicludes are rich in clay minerals, such as montmorillonite, illite, and kaolin, and are prone to swelling and sealing fractures when exposed to water [27,28]. As a result, such aquicludes are not necessarily completely destabilized after cracks are generated, or the deformation reaches the limit value. Therefore, the stability assessment of such aquicludes from the aspects of fracture development and deformation does not conform to the real situation. Therefore, it is necessary to develop a new aquiclude stability assessment method.
The aquiclude stability is mainly affected by geological factors (topography, geological structure, hydrogeology, stratum lithology, etc.) and mining factors (mining height, mining speed, working face size, working face layout, etc.) [29,30]. Among them, geological factors are the unchangeable and artificial uncontrollable factors formed during geological tectonic movement. In contrast, mining factors are artificially controllable factors that have a significant effect on aquiclude stability. Multiple researchers have studied the effect of different factors on aquiclude stability for close-distance coal seams mining. Thus, some authors explored the development of water-flowing fracture zone under the conditions of close-distance coal seams through physical simulation and field measurement and analyzed the feasibility of close-distance coal seams mining with water preservation [31,32]. Other authors investigated the influence of different stratum structures of coal seams on the development of overlying stratum cracks and classified the close-distance coal seams into three types: no inferior key stratum between coal seams, single inferior key stratum between coal seams, and two or more inferior key strata between coal seams [33,34]. Other authors studied the influence of mining staggered distance of the upper and lower coal seam on aquiclude stability through physical analog model and numerical simulation, as well as substantiated the optimal mining staggered distance [35]. Other authors investigated the effect of bedrock thickness and coal seam spacing on the aquiclude stability [36,37]. They reported that with the bedrock thickness increase, the development degree of water-flowing fracture zone decreased, improving the aquiclude stability. With an increase in the interlayer spacing, the development degree of secondary cracks dropped, and the aquiclude stability improved. In summary, previous studies provided a single-factor analysis of multiple factors on aquiclude stability, without their ranking or interaction account. Therefore, it is necessary to comprehensively consider various influencing factors and study their significance to substantiate the optimal water-preserving mining conditions in close-distance coal seams.
Based on the geological conditions, subsidence, seepage, and deformation perspectives of the close-distance coal seams in the Yili No. 4 Coal Mine located in the Xinjiang Uygur Autonomous Region of China, this paper established an aquiclude stability evaluation system with three evaluation factors (subsidence, seepage, and deformation), three subfactors (subsidence gradient, seepage rate, and horizontal deformation), and four evaluation criteria (unstable, weakly stable, medium stable, and stable). Besides, a multifactor analysis of aquiclude stability was carried out to determine the significance of the upper coal seam mining height, seam spacing, and the lower coal seam mining height on aquiclude stability. It provides a reference for the design of waterpreserving mining parameters for close-distance coal seams in the Yili No. 4 Coal Mine and similar ones.

Aquiclude Stability Evaluation System and Evaluation
Criteria. Aquiclude stability is the result of the joint effect of many factors, which may be coupled (interrelated). It is not rigorous to evaluate the stability of the aquiclude from singlefactor analysis. Referencing to the existing literature, it was discovered that subsidence, seepage, and deformation are the key aspects affecting the stability of aquiclude, which have not yet been comprehensively incorporated in the evaluation system. Among the three aspects, subsidence gradient of aquiclude can effectively represent the longitudinal shear action, seepage velocity can effectively represent the erosion effect on aquiclude, and horizontal deformation can effectively represent the transverse tensile action. In this paper, starting from the properties of aquiclude subsidence, seepage, and deformation, three influencing factors, namely, (i) the aquiclude subsidence gradient, (ii) seepage rate, and (iii) horizontal deformation, were selected to evaluate the stability of aquiclude comprehensively.
For the evaluation of aquiclude stability, a fuzzy comprehensive evaluation method was applied. The fuzzy comprehensive evaluation method is based on the fuzzy mathematical correlation theory. It is a mathematical model used to comprehensively evaluate the systems that evaluations are based on complexity and fuzziness. Different from the absolute membership in conventional sets, where a factor belongs to a set or not (0 or 1), the fuzzy set assesses the membership degree of a factor in the set as any number in the closed interval [0, 1], replacing absolutely belonging (yes or no) with relative belonging.
For the evaluation criteria, referring to the mining water barrier classification of aquicludes by Yu [38], the aquiclude stability was classified by the following four levels: unstable, 2 Geofluids weakly stable, medium stable, and stable. The interval from 0 to 1 was split into four equal segments to characterize the stability of aquiclude (Sta). The specific evaluation levels V and evaluation values Sta are shown in Table 1.
In summary, based on the above influencing factors, evaluation method, and evaluation criteria, referring to the methods of establishing evaluation index system of other scholars [8,39,40], the following evaluation index system is established, as shown in Figure 1.

Membership Function Determination
(1) Aquiclude subsidence gradient The subsidence gradient represents the change of subsidence per unit length. Its impact on the aquiclude is mainly manifested in the longitudinal shear effect, which produces shear cracks and affects the stability of the aquiclude structure, causing water loss. Based on the geological conditions of the Yili No. 4 Coal Mine, semitrapezoidal and trapezoidal distributions were selected to determine relevant parameters, and the subsidence gradient was determined as the membership function of the evaluation standard: where i (m m -1 ) is the subsidence gradient.
(2) Aquiclude seepage rate Aquiclude seepage rate has a great impact on the stability of the aquiclude. It is mainly manifested in the erosion effect on the aquiclude, which creates water channels and affects the seepage stability of the aquiclude, causing the loss of water resources. Similar to previous subsection, semitrapezoidal and trapezoidal distributions were used to determine the relevant parameters, and the membership function with the seepage rate as the evaluation standard was determined: where v (m s -1 ) is the seepage rate. Target layer Figure 1: The evaluation index system.   The horizontal deformation of the aquiclude has a significant impact on the stability of the aquiclude. It is mainly manifested in the lateral stretching effect on the aquiclude, resulting in tensile cracks and affecting the aquiclude structure's stability, causing water loss. For semitrapezoidal and trapezoidal distributions of relevant parameters, the membership function with the horizontal deformation as the evaluation standard was derived as follows: where ε (m m -1 ) is the horizontal deformation.

Weight
Calculation of Impact Factors. The evaluation index system of aquiclude stability is complex and comprehensive, and different factors are interrelated. For instance, aquiclude horizontal deformation affects the width of cracks and then the seepage rate. On the contrary, the seepage rate affects the expansion characteristics of the aquiclude and then the cracks width. Therefore, how to accurately determine the weight of each factor is significant for the evaluation results. In this paper, using the "1-9" scaling method, the scale value was evaluated based on the relative magnitude of the influence of each impact factor on the aquiclude stability evaluation, and the judgment matrix was established by pairwise comparison, as shown in Table 2. Then, the weight of each influencing factor was calculated, which finally yielded the evaluation index. The steps for obtaining the weight of each factor were as follows: (1) Normalization of the column vector of the judgment matrix: (2) Matrix summation by the rows of A ij : (3) W ij was normalized to get the sort vector: W = ðw 1 , w 2 , ⋯, w n Þ T After calculation, we got the following:

Geofluids
Next, a consistency test on the value of CR = CI/RI was conducted, where CI = ðλ max − nÞ/ðn − 1Þ. The largest eigenvalue of the matrix was derived as λ max = 1/n∑ n i=1 ðAWÞ i /w i = 3:0037, and the value of RI was based on the Saaty average random consistency index, as shown in Table 3.
If CR < 0:1, the judgment matrix passed the consistency test; otherwise, it had to be adjusted. After calculation, we got CI = 0:0019 and the above matrix's CR = 0:0033. Since the latter was less than 0.1, it passed the consistency test.        Then, the FLAC3D numerical model with dimensions (length × width × height = 600 × 490 × 165 m) wasconstructed. The upper and lower working face length was 240 m, the staggered distance was 50 m, and the continuous advancing length was 400 m. Monitoring points were set at 5 m intervals in the aquiclude position in the middle of the advancing direction of the working face, as shown in Figure 6. The model adopted the Mohr-Coulomb criterion, with fixed constraints on the surrounding and lower boundaries. The aquifer maintained a water pressure boundary condition of 0.05 MPa, and the physical and mechanical parameters of each rock layer were adopted, as shown in Table 4.

Case Study Analysis and Verification
Referring to the multifield coupling model established by Fan et al. [41][42][43], the loading and unloading stages were specified by the self-developed subprograms written in Fish language and incorporated into the FLAC3D software package. Then, the corresponding permeability-axial strain relationship was selected to simulate the permeability evolution of the overlying rock under the disturbance of shortdistance coal mining. The specific determination process is shown in Figure 7. For the number of loading and unloading times F = 1, the permeability values of the loading and unloading stages were k 1 and k 2 , respectively. At F = 2, the respective values were k 3 and k 4 . The aquiclude subsidence gradient, the aquiclude horizontal deformation, the seepage rate of overburden, and the permeability variation behavior after mining of the upper and lower coal seams were numerically simulated, as shown in Figure 8.
It can be seen in Figure 8 that after mining of the upper and lower working faces, the maximum values of the subsidence gradient, seepage rate, and horizontal deformation of aquiclude were 0.142 m m -1 , 0:379 × 10 −7 m s −1 , and 0.0075 m m -1 , respectively. According to the above simulation results, substituting the corresponding membership function and matching it with the corresponding membership degree yielded the Sta evaluation index of 0.8096, corresponding to the stable level of aquiclude.
The water level variation in the observation hole during the mining of the actual working face is shown in Figure 9, with an initial water level of 777 m.
As the working face advanced, the water level continued to drop. When the working face was pushed through the observation hole by 40 m, the water level dropped to the lowest point, reaching 772.2 m. The maximum water level drop was 4.8 m. As the working face continued to advance, the water level began to rise gradually. At 145 m from the observation hole, the water level finally stabilized at 777.8 m, showing a steady-decreasing-rising-steady variation pattern. After mining, the water level of the working face has not dropped significantly and even increased by 0.8 m from the initial water level. This implies that mining at the working face did not significantly impact aquiclude, verifying the rationality of the adopted evaluation method.

Test Plan Design.
To study the influence of the three factors (namely, the thickness values of the upper and lower coal seams and coal seam spacing) on the groundwater flow field, the coal seam occurrence parameters were generalized based on the comprehensive subdivision of the coal rock structure. The distance between the upper coal seam and the aquifer was kept at 85 m, while other parameters are listed in Table 5.
To improve the multifactor analysis efficiency and accuracy, the orthogonal method was used to design the experimental plan. The orthogonal array table was designed as L 9 ð3 4 Þ, and the specific simulation scheme was adopted, as shown in Table 6.
Model dimensions were as follows: length × width = 600 × 440 m, while the model's height was automatically adjusted according to the height and spacing of the coal seams. As the influence of the staggered distance on the experimental results was not considered, zero staggered distances of the upper and lower working faces were assumed. The length of the working face was 240 m, and the continuous advancing length was 400 m. Monitoring points were set with 5 m intervals in the aquiclude position in the middle of the advancing direction of the working face, as shown in Figure 10. The corresponding mechanical parameters and seepage parameters of rocks were the same as those described in Section 3.

Numerical Simulation
Results. The aquiclude subsidence gradient, horizontal deformation, seepage rate, and permeability variation behaviors in nine adopted schemes were obtained through numerical simulation. Scheme 1 (upper coal seam mining height 3 m × lower coal seam mining height 5 m × coal seam spacing 10 m) was chosen for detailed analysis. The variation behavior of each parameter is shown in Figure 11.
It can be seen in Figure 11(a) that after the double coal seam was mined, the aquiclude tilted and sank to the goaf side. The subsidence gradient varied with horizontal positions, and its positive and negative values corresponded to different tilt and subsidence directions. A positive (negative) value implied that the direction of tilting and sinking coincided with (was opposite to) that of the established 7 Geofluids coordinate axis, respectively. The maximum subsidence gradient was 0.0923 m m -1 , which was located 30 m inside the working face end. The subsidence gradient within a certain range in the middle of the working face was zero.
It can be seen from Figure 11(b) that after the double coal seam was mined, the aquiclude horizontal deformation above different working face positions varied and could be split into the following four areas. Area 1 was the original strain area located 70 m outside the working face end. The aquiclude's horizontal deformation in this area was very small. Area 2 was the tensile deformation area, ranging from 70 m outside the working face end to 30 m inside it. In this area, the horizontal deformation value first increased and then decreased, reaching a maximum of 0.0088 at 40 m outside the working face end. Area 3 was the compression deformation area, ranging from 30 m inside the working face end to 100 m inside it. In this area, the horizontal deformation first increased and then decreased, reaching its maximum of -0.0086 m m -1 at 50 m inside the working face end. Area 4 was located 100 m from the inside of the working face end to its middle, its horizontal deformation was zero, and thus, it was the deformation recovery area.
According to Figure 11(c), during mining of the double coal seam, the aquifer water seeped in the goaf direction. The arrows at the working face end were denser, which indicates that the seepage rate increased greatly. The highest seep-age rate of 4:02 × 10 −7 m s −1 was observed at the end of the working face between the upper and lower coal seams. The maximum seepage rate of aquiclude was 0:45 × 10 −7 m s −1 . Figure 11(d) shows that after the double coal seam mining, the overlying rock permeability exhibited an overall increase. Within a certain range at both ends of the working face, the permeability of the overlying rock increased greatly. For the mudstone layer between double coal seams, the permeability of the lower coal seam increased significantly at the initial stage of mining. In the later stage of mining, due to the compaction effect of the mined-out area and the swelling and plugging of the mudstone after encountering water, the crackinduced permeability began to decrease again. In contrast, near the ends of the working face, the permeability increased due to the presence of open large cracks. For the aquiclude below the aquifer, despite the increasing trend of permeability after the double coal seam mining, the permeability increment was smaller than that in other rock formations. Figure 12 shows the numerically predicted changes in four aquiclude parameters for nine different schemes. The derived numerical values are listed in Table 7.
A comprehensive analysis of nine sets of test data revealed the following trends.
(1) The maximum subsidence gradient interval of aquiclude was 0.0923-0.1932 m m -1 . When the upper and

Aquiclude Stability Evaluation and
Significance Analysis of Influencing Factors 5.1. Aquiclude Stability Evaluation. Based on the maximum subsidence gradient, seepage rate, and horizontal deformation parameters of aquiclude of different schemes, the corresponding membership functions were substituted and matched with corresponding parameter weight, yielding the aquiclude stability evaluation indices for nine different schemes adopted in this study. The performed calculation procedure was described below for Scheme 1.

Significance Analysis of Influencing Factors.
Based on the aquiclude stability evaluation indices of the above nine schemes, range and variance analyses were carried out to study the influence of different factors on the aquiclude stability.
(1) Range analysis The range analysis results are summarized in Table 9.
The range analysis revealed the following decreasing order: η Upper coal seam mining height > η Lower coal seam mining height > η Seam spacing . This indicates that the decreasing order of the significance of the influence on the evaluation index of aquiclude stability was as follows: upper coal seam mining height, lower coal seam mining height, and seam spacing. The variation of the evaluation index with the level of each factor is shown in Figure 13, where one can observe that the evaluation index negatively correlated with the height of the upper and lower coal seams and positively correlated with coal seam spacing.
Although the range analysis is simple and easy to implement, it cannot distinguish data fluctuations caused by changes in factor levels from those caused by test errors nor can it give a quantitative estimate of the significance of the influence of factors. To make up for the range analysis deficiencies, the following variance analysis of the experimental results was performed.     12 Geofluids (2) Variance analysis The variance analysis was performed based on numerical simulation results, and the test level α was given. The critical value F α ð f x , f e Þ was obtained from the F distribution table, where f x and f e were the degrees of freedom (DoF) of the influencing factors and the error, respectively. The calculated F value was compared with the critical value. If F > F α ðf x , f e Þ, this factor had a significant influence on the test results. The greater the difference, the greater the significance of the impact. The results of detailed variance analysis are shown in Table 10. If F > F 0:01 ðf x, f e Þ, the influence of this factor was considered highly significant and designated by " * * " sign in Table 10. If F 0:01 ðf x, f e Þ > F > F 0:05 ðf x, f e Þ, the influence of this factor was considered significant and represented by a single " * " sign. If F < F 0:01 ð f x, f e Þ, the influence of this factor was insignificant.
It can be seen from Table 10 that such factors as upper and lower coal seam mining heights had a significant influence, while the influence of seam spacing was not significant. The decreasing order of significance of each factor's influence on the aquiclude horizontal deformation was as follows: upper coal seam mining height, lower coal seam mining height, and seam spacing. Since this order was identical with the range analysis results, this verified the rationality of the proposed approach.

Discussion
A single-factor variance analysis was carried out on the evaluation factors of subsidence gradient, horizontal deformation, and seepage rate. The results are summarized in Tables 11-13. It can be seen from Tables 11-13 that the significant results obtained by the variance analysis of different factors were different. The variance analysis of aquiclude subsidence gradient revealed that the decreasing order of significance was as follows: upper coal seam mining height, lower coal seam mining height, and seam spacing. The variance analysis of aquiclude horizontal deformation derived the following order of significance: upper coal seam mining height > lower coal seam mining height > seam spacing. The variance analysis of aquiclude seepage rate implied the order of significance as follows: lower coal seam mining height > upper coal seam mining height > seam spacing. The main   13 Geofluids reason for such discrepancies is that the stability of the aquiclude is affected by factors such as subsidence, seepage, and deformation. Different factors contribute in controlling the stability of aquicludes, which cannot be assessed by singlefactor analysis. Multiple factors need to be considered simul-taneously. Based on the established aquiclude stability evaluation system and evaluation criteria, a more reliable order of significance was obtained using the comprehensive evaluation index Sta: upper coal seam mining height > lower coal seam mining height > seam spacing.

Conclusions
Aquiclude stability is the result of the joint effect of many factors, which may be coupled (interrelated). It is not rigorous to evaluate the stability of the aquiclude from singlefactor analysis. In this paper, a new evaluation method for the stability of aquiclude was proposed, which adopting fuzzy comprehensive evaluation and taking subsidence gradient, seepage rate, and horizontal deformation of the aquiclude as evaluation factors. Then, the stability of weakly cemented aquiclude under different mining disturbance conditions of the close-distance coal seams in the Yili No. 4 Coal Mine was evaluated. Besides, the significance of factors affecting the aquiclude stability under the disturbance of close-distance coal seams mining was analyzed. The following conclusions were reached: (1) The fuzzy comprehensive evaluation system and standard for the aquiclude stability of the Yili No. 4 Coal Mine were elaborated. The "1-9 scale method" was used to construct the judgment matrix, and the weight factors of aquiclude subsidence gradient, seepage rate, and horizontal deformation were determined to be 0.1095, 0.3090, and 0.5815, respectively, through normalization and consistency check (2) The relationship between permeability and axial strain of mudstone and sandstone in different cyclic loading and unloading stages was obtained through laboratory tests. By establishing the FLAC3D model and self-developed code compilation in Fish language, various evaluation factors after mining at 21105 and 23213 working faces were studied, and the comprehensive evaluation index Sta was determined to be 0.8096, which corresponded to the stable level. After the actual coal seam was mined, the water level increased by 0.8 m compared with the initial value, indicating the stability of the aquiclude and verifying the feasibility of the adopted evaluation system (3) Based on the comprehensive evaluation index Sta, the range and variance analysis were carried out. It was determined that the aquiclude stability evaluation index of the Yili No. 4 Coal Mine negatively correlated with the mining heights and positively correlated with the seam spacing. The decreasing order of the significance of influencing factors on aquiclude stability was as follows: upper coal seam mining height, lower coal seam mining height, and seam spacing Based on the above results, it can be seen that the fuzzy comprehensive evaluation method is more practical than a single-factor analysis for aquiclude stability evaluation. This paper studied the influence significance of the mining height of the upper and lower coal seams and seam spacing on the aquiclude stability, which has important guidance meaning for the water conservation mining of close-distance coal seams. However, the influence of working face length, layout, and advancing speed on the aquiclude stability is not considered, and further research is still needed.

Data Availability
Data from this research are not publicly available. Interested researchers can contact the corresponding author of this article.

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