Long-Term Monitoring and Early Warning of Coal Mine Underground Reservoirs—A Case Study in Shigetai Coal Mine

Abstract

Coal mining is often associated with groundwater pollution and loss, and coal–water conflicts are becoming increasingly prominent in Western China. As a new way to protect mine water, coal mine underground reservoirs (CMURs) have effectively alleviated the water shortage problem in Western China. The CMURs have been in existence for 25 years, but field-monitoring studies on their long-term stability are rare. In this paper, we take the Shigetai coal mine in the Shendong mining area as the research background. Long-term observation of stress, deformation, seepage pressure, and other parameters of 22 dams in five goaves (31201–31205) of the Shigetai coal mine for the whole year of 2022 has been pioneeringly carried out. A novel early-warning model, which incorporates expert evaluations and real-time indicator fluctuations, is proposed to assess the stability of CMURs. The stability characteristics of coal pillar dams (CPDs) and artificial dams (ADs) of CMURs are evaluated by this model, proving the validity and applicability of this model. This model provides theoretical and methodological guidance for long-term monitoring and early-warning systems for CMURs in the Shendong mining area.

1. Introduction

Currently, coal remains the primary source of energy production and consumption in China [1]. Coal mining activities disturb the coal–rock mass, causing redistribution of in situ stress and inducing secondary fractures. The groundwater originally stored in the pores and fractures of coal and rocks would enter the roadways during mining and be discharged to the surface, causing pollution and the loss of groundwater bodies. This may further lead to problems, such as surface subsidence. In Western China, coal mining has further exacerbated water scarcity.

To address these challenges, Chinese academician Gu Dazhao proposed CMUR technology, focusing on ‘water conduction, storage, and utilization’. This includes design, construction, operation, and safety monitoring technologies [2,3,4,5]. The CMURs are constructed by utilizing the goaf as a water storage space and connecting the discontinuous CPDs with concrete ADs to form a composite dam to construct a groundwater reservoir. This represents a technical method to synergistically utilize coal mining and protect water resources [4]. At present, more than thirty underground water reservoirs have been constructed in the Shendong mining area that effectively utilize the underground space of the mines. This has resulted in the full realization of the resource potential of mine water, the alleviation of the problem of water shortage and water supply in the western region, and the provision of an important reference for the transformation and upgrading of the closed mines in a wide range of China.

The first CMUR in China was constructed in 1998 at the Daliuta coal mine in the Shendong mining area [6]. CMURs use voids within coal–rock formations in mined areas for water storage, but the soft texture of coal and its fracturing can lead to large deformations and water bursts over time. These issues, caused by in situ stress and water softening, threaten the safety of coal mine operations, making safety monitoring and early-warning technologies a priority. Wen et al. [7] derived a mathematical expression for the internal force distribution of the coal–rock mass, and determined the water storage capacity of the CMUR in the Shendong mining area. Yao et al. [8] conducted an edge-shear test of coal rocks with different water content in CMURs, thereby providing a novel understanding of the crack extension and strength deterioration of coal rock in coal mine underground reservoirs. Wang et al. [9] analyzed the relationship between the internal composition, structure, and external deformation of coal rocks in CMURs under long-term stress–seepage interactions and proposed the equation of the coal and rock expansion coefficient against axial stress.

In CMUR construction, CPDs, which are reinforced post-mining, and ADs, which fill gaps between CPDs with concrete, are the main water-blocking infrastructures that are critical for CMUR stability. The time-dependent damage deterioration of CPDs and ADs may induce the leakage of the reservoir. In the case of CPDs, Gu et al. [10] conducted experimental studies on the dynamic damage of CMURs under different seismic intensities, proposed the concept of the safety coefficient for CMUR dams, and compared the safety of CMUR dams with ground reservoir dams under similar conditions. By analyzing the deformation and stress state of CPDs, Zhang et al. [11] determined the optimal width of CPDs and deduced the characteristics of crack development evolution and sliding instability of CPDs. Fan et al. [12] proposed a numerical simulation approach that considered the unsaturated seepage characteristics and the weakening effect of the CPDs and studied the stability of CMURs. In the case of ADs, Fang [13] constructed a theoretical model of the stress distribution on the ADs of CMURs and obtained the surface deformation and damage of the AD body. Kong et al. [14] explored the effect of slot width on the stability of ADs and gave the distribution of plastic zones on the surface of ADs. Ma et al. [15] conducted hydraulic–mechanical coupling numerical calculations of ADs in CMURs and analyzed the stability characteristics of ADs.

The aforementioned studies demonstrate that CMUR technology has gained significant attention and widespread adoption in China [16,17]. Many scholars have also investigated the safety of CMURs. However, most of the current research focuses on laboratory experiments and theoretical explorations of CMUR safety, with a notable absence of on-site monitoring studies on the long-term stability of existing CMURs. This paper presents a year-long observation (2022) of 22 dams in five goaves (31201–31205) at the Shigetai coal mine, Shendong mining area, focusing on monitoring parameters such as stress, deformation, and seepage pressure. The long-time stability of CPDs and ADs of the CMUR under in situ stress and external disturbance is analyzed. A dynamic monitoring and early-warning model that integrates expert scoring with real-time indicator fluctuations is proposed. This model is adopted to analyze the stability of the Shigetai coal mine, thus further proving the validity of the model.

2. Field-Monitoring Information of CMUR in Shigetai Coal Mine

2.1. Geological Conditions for Shigetai CMUR

The Shigetai CMUR is located in goaves 31201–31205 and is the only CMUR in the Shigetai coal mine. The average depth of the working face is 120 m, with a coal seam dip angle of 1–3° and a design mining height of 3.4 m.

The Shigetai coal mine is generally confined by CPDs and ADs (Figure 1). The average width of the CPDs is 20 m, with an elasticity modulus of 7.5 GPa. The average compressive strength of the CPDs is 15.9 MPa, and the tensile strength is 1.2 MPa. On-site research identified leakage points in some areas of the CPD, with water accumulation observed nearby. There are no historical records of dynamic load factors, such as mining earthquakes. However, mining activities occur in the adjacent area but not in the lower coal seams.

The ADs in the Shigetai CMUR are set in the contact lane between the main and auxiliary retraction passages of the mining faces. The concrete grade is C25, and the dam body height is 3.9 m. The embedding depth of the ADs is 0.5 m, with the sides and top fixed by anchor rods.

2.2. Monitoring Principle

Existing studies [18,19] indicate that the most vulnerable areas of the dams in the CMUR are the connecting edges between CPDs, ADs, roofs, and floors, as well as the middle and bottom areas of CPDs and the top and bottom edges of ADs. These areas are of particular concern with regard to monitoring the stability of the CMUR.

In addition, there are significant differences in the mechanical properties and internal structural characteristics between the CPD and AD. The CPD has low strength, significant fracture propagation, and significant stress concentration. Therefore, stress is the main monitoring parameter for CPD stability analysis. The AD has high strength, fewer fractures, and smaller internal stress distribution. Therefore, the stability state of the AD is mainly judged by the surface micro-deformation.

Thus, stress is the primary monitoring parameter for CPD stability analysis. ADs, having higher strength, fewer fractures, and smaller internal stress distribution, are mainly evaluated for stability through surface micro-deformation. When conducting on-site monitoring, it is essential to focus on the stress concentration areas of the dams. Additionally, employing diverse monitoring techniques is crucial due to the differences between CPDs and ADs.

2.3. General Installation Information

The CPDs and ADs of the CMUR in the Shigetai coal mine were the focus of the research. The monitoring system was installed, and field monitoring was subsequently carried out (

Table 1. Area and quantity of monitoring device installation.

No.	Device/Method Name	Installation Area and Device Name	Parameters	Accuracy
1	Strain Gauge	4 on each AD	Measurement Range: 0–3000 με	Resolution: ±0.1 με
2	Osmometer	2 on each CPD	Measurement Range: 0–0.5 MPa	Accuracy: ±0.5% F.S
3	Stress Gauge	2 on each CPD	Measurement Range: 0–30 MPa	Accuracy: ±0.5% F.S
4	Fiber Optic Pressure Sensor	1 on the AD in each goaf	Measurement Range: 0–0.5 MPa	Accuracy: ±0.5% F.S
5	Micro-seismic Sensor	1 on the AD in each goaf	Frequency Range: 0.5–200 Hz	Sensitivity: 10−3 m/s2 @ 30 Hz
6	Water-Measuring Weir Meter	1 on the AD in each goaf	Flow Rate Range: 0–2 m3/s	Accuracy: ±2%
7	Direct current (DC) detection	Each CPD in 31203 goaf	Depth Detection: up to 50 m	Accuracy: ±5%

). The monitoring content primarily includes stress monitoring, deformation monitoring, reservoir water level and seepage monitoring, and vibration monitoring. Additionally, due to the field environmental situation, direct current (DC) detection was also conducted on the dams in the 31203 goaf as a supplement. The general installation information of the long-term monitoring devices is shown in Figure 2.

3. Field-Monitoring Results and Analysis

3.1. Overview of Field Monitoring

Based on the field geological data of the Shigetai coal mine, the water level contour of CMUR is illustrated in Figure 3. According to the water level contours shown in the figure, goaves 31201 to 31203 are located in areas of low terrain, with goaf 31201 remaining below the water level line throughout the year. The water levels in these areas are higher, warranting closer attention. Additionally, field observations indicate that goaf 31203 has a potential risk of dam leakage. Therefore, this paper primarily focuses on and discusses the field-monitoring results from goaves 31201 and 31203.

3.2. CPD Field-Monitoring Results Analysis

3.2.1. Goaf 31201

As a zone characterized by low topography and prolonged water immersion, the long-term monitoring results of the dam body in goaf 31201 are particularly critical. Consequently, dam-monitoring equipment was installed in key areas of potential damage within the dam, and the relevant field-monitoring data were analyzed.

The seepage pressure monitoring results of the CPD are shown in Figure 4. It is observed that, in 2022, the seepage pressure of the CPD at goaf 31201 fluctuated significantly, with two periods of continuous increase occurring from January to April and from June to August. Notably, the seepage pressure rose rapidly from 0.01 MPa to 0.06 MPa in August.

When comparing the continuous water level-monitoring data for goaf 31201 and 31202-1 (Figure 5), it is apparent that, while the water level in goaf 31201 remained relatively stable in August, the adjacent goaf 31202-1 (Figure 6) experienced large fluctuations in its water levels during August and December, with a maximum fluctuation of up to 1 m/day. Related research indicates that [20] an increase in seepage pressure can initiate cracks in the CPD and increase the flow rate of water through the dam, further inducing the softening of the coal body. This suggests that not only does a high water level affect the stability of the CMUR but significant fluctuations in the water level can also impact stability, especially when the groundwater level is relatively high. Therefore, it is crucial to control both the water level and its fluctuations during the long-term operation of the CMUR.

Figure 7 shows the borehole stress distribution on CPD 31201-2#. The results indicate that borehole stress increased rapidly from 0.1 MPa to 0.4 MPa in August, followed by a further increase in December. An analysis suggests that this rise in borehole pressure may be linked to the water level fluctuations in the adjacent goaf 31202-1-2.

These fluctuations, observed during August and December, likely disturbed the mechanical equilibrium of the CPD, contributing to changes in the local stress field. This phenomenon suggests that rapid water level fluctuations, often induced by frequent water pumping, can influence the stress redistribution in CPDs. Although the measured stress levels alone may not directly weaken the CPD, the stress field alterations and damage development could exacerbate pre-existing fractures or initiate localized damage over time.

To mitigate the impact of such fluctuations, maintaining water level stability through a scientifically arranged pipeline network and optimized groundwater utilization strategy is recommended within the CMUR. Future studies should explore the relationship between stress field redistribution, microcrack evolution, and long-term CPD stability under fluctuating water levels to better understand the underlying mechanisms.

3.2.2. Goaf 31203

In response to the potential leakage area in goaf 31203, a DC detection test was conducted, resulting in a high-resolution electrical apparent resistivity contour (Figure 8). In the figure, if the coal seam is unaffected by water-rich zones or water-conducting structures, the apparent resistivity of the coal seam changes in an orderly manner, and the contour lines exhibit a stable, layered distribution. However, when a low-resistance water-rich area or a water-conducting structure is present, the resistivity value at the anomaly decreases, and the contour lines exhibit distortion, deformation, or a dense strip pattern.

In the contour cross-section result map, the green areas indicate low-resistance zones of apparent resistivity (water-rich zones or fractured areas), while the red areas represent high-resistance zones (less water, coal, or CPDs). Other colors indicate transitional areas.

In Figure 8, a discontinuity is observed between the high-resistance and low-resistance areas. By comparing the distribution of apparent resistivity in the transverse direction, four low-resistance anomalous zones are identified within the detected area. These are designated as anomalous zones No. 1, No. 2, No. 3, and No. 4, corresponding to the ranges of 10–45 m, 75–115 m, 150–170 m, and 200–240 m, respectively. Field observations revealed leakage points in CPD 1 and 2, which coincide with the DC detection results. This correlation indicates that the long-term erosion of CPDs by pressurized water has softened the internal structure of the dam body, leading to the formation of seepage channels. These channels manifest as more pronounced seepage points, which should be key areas for further monitoring and reinforcement.

Building on this, long-term stability monitoring of the CPD in goaf 31203 was conducted. Significant stress and seepage pressure concentrations were detected in CPD 31203-1# (Figure 9 and Figure 10), which are markedly higher than those in the CPD of goaf 31201. During field investigations, a potential leakage point was identified in goaf 31203.

An analysis indicates that these stress and seepage pressure concentrations could contribute to localized changes in the mechanical behavior of the CPD. The elevated in situ stress, combined with seepage pressure, may accelerate the evolution of pre-existing fractures or enhance permeability in the coal pillar, potentially leading to increased permeability and localized damage. However, it is important to note that the observed stress levels (at a depth of approximately 120 m) are relatively low and, on their own, are unlikely to cause significant structural damage to the CPD.

The identification of a potential leakage point underscores the importance of continuous monitoring and further investigation to determine whether fracture development or weakening of the CPD is occurring. Future work could focus on detailed microstructural analysis, numerical modeling, and laboratory tests to establish the relationship between stress, seepage, and fracture evolution in CPDs under similar conditions.

3.3. AD Field-Monitoring Results Analysis

3.3.1. Goaf 31201

The surface strain evolution of the AD on goaf 31201 is shown in Figure 11. The results indicate that the deformation of AD 31201-2# exhibited an overall negative trend throughout 2022, suggesting a continuous expansion. Moreover, the deformation followed a two-stage trend, with a notable change occurring around August 22, indicating a significant alteration in the stress state of the AD during this period.

When correlating these findings with the borehole water level monitoring for this goaf (Figure 12), it is observed that, during the rapid increase in water level (around early August), the original equilibrium state of the dam’s deformation was disrupted, leading to a rapid progression to the next stage of deformation. This suggests that rapid fluctuations in the water level not only affect the CPD but also trigger a secondary phase of AD deformation, which is detrimental to the long-term stable operation of the AD.

3.3.2. Goaf 31203

The monitoring results of the AD in goaf 31203 are presented in Figure 13. The data show that the overall strain value of the AD in goaf 31203 is greater than that in goaf 31201, with the strain evolution also displaying a two-stage growth trend. This behavior can also be attributed to fluctuations in the water level. Although the terrain of goaf 31203 is higher, resulting in a lower absolute water level and, thus, lower water pressure compared to goaf 31201, Figure 14 indicates that the regional water level in goaf 31203 is higher than in goaf 31201. This higher water level induces greater deformation and increases the risk of dam leakage.

This phenomenon can be explained by the uneven floor of the CMUR, a condition exacerbated by mining activities. Over the long term, water storage can cause the roof of the upper coal seam to further collapse, worsening the unevenness of the CMUR. As a result, the mechanical behavior of the dam is primarily influenced by the local water storage height.

4. Early-Warning Model of CPD and AD

The CMUR construction technology was pioneered in the Shendong mining area, but an early-warning system for CMUR safety is still in its early stages. During the long-term operation of the reservoir, the safety status of the CMUR is primarily determined by real-time monitoring data from CPDs and ADs, which are critical for stability analysis. To date, no established early-warning models specifically address the unique conditions of CMURs. In this chapter, we propose a suitable early-warning model based on real-time monitoring data, drawing on methodologies used for groundwater reservoirs [21,22]. This model provides a theoretical framework for the automatic detection of abnormal conditions in the long-term safety analysis of CMURs.

4.1. Basic Introduction and Theory for the Early-Warning Model of CPD and AD

The early-warning model for CPDs and ADs includes the following steps:

1. Determine the influencing factors: Identify the factors affecting the long-term safety of CPDs and ADs, such as construction methods, environmental parameters, and monitoring data. Classify the hazard level of each factor within different monitoring ranges, resulting in an evaluation matrix I;

2. Calculate subjective weights θ_i: Use the analytic hierarchy process (AHP) to calculate the subjective weights θ_i for each evaluation factor, based on expert scoring. Once determined, these weights remain constant. The detailed process of calculating subjective weights is discussed in Section 4.1.1;

3. Standardize the evaluation matrix I*: Standardize the evaluation matrix I to convert the monitoring data into positive values, ensuring that each indicator value becomes more dangerous as it increases. The standardization process is as follows. Assume the monitoring value of indicator i is x_i, i = 1, 2, …, m, and there are c levels of the evaluation criteria $I=\{\left[a_{ih},b_{ih}\right]\},h=1,2,\dots ,c$. The standardized evaluation matrix I* can be formulated as:

$$\{\begin{array}{c}{x}_i^{*}=\left|x_i-a_{i1}\right|\\ a_{ih}^{*}=\left|a_{ih}-a_{i1}\right|\\ b_{ih}^{*}=\left|b_{ih}-a_{i1}\right|\end{array}$$

(1)

4. Calculate the objective weights λ_i and total weights w_i: Use a relevant function method to calculate the objective weights λ_i and total weights w_i for each evaluation factor. These weights will vary as the monitoring data change. Details of the objective weights and the total weights calculation are discussed in Section 4.1.2;

5. Determine the membership function: adopt the membership function from Zhang, Zhang, Sa, Du, and Xue [22], which will be introduced in Section 4.1.3;

6. Evaluate the stability: assess the stability of the target CPDs and ADs based on the combined analysis of the above steps.

4.1.1. Analytic Hierarchy Process (AHP) Method Introduction

The analytic hierarchy process was first proposed in the early 1970s by Thomas Sethi, an operations researcher at the University of Pittsburgh. AHP is a hierarchical weighted decision analysis method that applies network systems theory and multi-objective comprehensive evaluation techniques.

The AHP analysis typically breaks down decision objectives into three levels, namely the objective level, the criterion level, and the program level. The process generally consists of the following four steps.

S1: Scale Determination and Judgment Matrix Construction

In this step, the factors influencing the stability of the CMUR are identified. Experts assess the importance of each factor using a 1–5 scale (with 1 being least important and 5 being most important). This results in the construction of a judgment matrix.

S2: Eigenvectors, Eigenroot Calculation, and Weight Calculation

This step involves calculating the weight values for each factor. First, the eigenvector values are determined, followed by the calculation of the maximum eigenvalue (consistency index, CI), which will be used in the next step for the consistency test.

S3: Consistency Analysis

When constructing a judgment matrix, logical inconsistencies may arise. To address this, a consistency analysis is performed by calculating the consistency ratio (CR). A CR value of less than 0.1 indicates that the judgment matrix passes the consistency test.

S4: Calculate the Weights and Final Target Score

In the final step, the weights are calculated, and the overall target score is determined.

4.1.2. Objective Weight and Total Weight Calculation

The objective weight λ_i is determined by the related function method. Assuming that the total range of index i varies with the range $V_{ip}=\langle a_{i1},b_{ic}\rangle$ and at its h level, the range is $V_{ih}=\langle a_{ih},b_{ih}\rangle ,i=1,2,\dots ,m;h=1,2,\dots ,c.$ Define

$$r_{ih}\left(x_i,V_{ih}\right)=\{\begin{array}{c}{\displaystyle \frac{x_i-a_{ih}}{b_{ih}-a_{ih}}},x_i\le b_{ih}\\ {\displaystyle \frac{b_{ih}-x_i}{b_{ih}-a_{ih}}},x_i>b_{ih}\end{array}$$

(2)

When $x_i\in V_{ip}$, then:

$$r_{i{h}^{*}}\left(x_i,V_{ih}\right)=\underset{h=1}{\overset{c}{\max }}\left\{r_{ih}\left(x_i,V_{ih}\right)\right\}$$

(3)

The larger the value of i, the higher the risk. So, more weight should be assigned:

$$r_i=h^{*}+r_{i{h}^{*}}\left(x_i,V_{ih}\right)$$

(4)

The objective weight of index i can be calculated as:

$${\lambda }_i=r_i/{\displaystyle \sum _{i=1}^m{r}_i}$$

(5)

Finally, the total weight w_i can be calculated by combining the subjective weight θ_i and the objective weight λ_i:

$$w_i={\lambda }_i{\theta }_i/{\displaystyle \sum _{i=1}^m{\lambda }_i{\theta }_i}$$

(6)

4.1.3. Membership Function Method [22]

Assume the evaluation target is $u=\{x_i\},i=1,2,\dots ,m$. The set of possible evaluations can be divided into c levels, with intervals represented by $I=\{\left[a_{ih},b_{ih}\right]\},h=1,2,\dots ,c$. Here, $a_{ih},b_{ih}$ represent the lower and upper boundary of index i for level h. Assume that k_ih has a relative membership of one to level h within the interval $\left[a_{ih},b_{ih}\right]$. The calculation model of k_ih is:

$$\{\begin{array}{c}{k}_{i1}=a_{i1}\\ k_{ih}={\displaystyle \frac{a_{ih}+b_{ih}}{2}},h=2,3,\dots ,c-1\\ k_{ic}=b_{ic}\end{array}$$

(7)

The standard matrix can be expressed as K = (k_ih). The membership of x_ij between intervals h and h + 1 can be expressed as:

$${\mu }_{ih}(u)=\{\begin{array}{c}0.5(1+{\displaystyle \frac{b_{ih}-x_{ij}}{b_{ih}-k_{ij}}}),x_{ij}\in \left[k_{ih},b_{ih}\right]\\ 0.5(1-{\displaystyle \frac{b_{ih}-x_{ij}}{b_{ih}-k_{i(h+1)}}}),x_{ij}\in \left[b_{ih},k_{i(h+1)}\right]\end{array}$$

(8)

The comprehensive membership of object u to level h can be expressed as:

$$v_h(u)={\left\{1+{\left[{\displaystyle \frac{{\displaystyle \sum _{i=1}^m{\left[w_i\left(1-{\mu }_{ih}\left(u_j\right)\right)\right]}^p}}{{\displaystyle \sum _{i=1}^m{\left[w_i{\mu }_{ih}\left(u_j\right)\right]}^p}}}\right]}^{{\displaystyle \frac{\alpha }{p}}}\right\}}^{-1}$$

(9)

where w_i is the total weight of index i, and ${\displaystyle \sum _{i=1}^m{w}_i=1}$. α is the optimization criterion parameter. When α = 1 is the linear criterion, α = 2 is the least square criterion. p is the distance parameter. When p = 1 is the Hamming distance, p = 2 is the Euclidean distance.

The eigenvalue of evaluation target u is:

$$H\left(u\right)={\displaystyle \sum _{h=1}^c{v}_h\left(u\right)\times h}$$

(10)

4.2. Early Warning of CPD in Shigetai CMUR

4.2.1. Single-Parameter Warning Values

Based on existing numerical calculations and long-term field monitoring results, it has been determined that stress, fracture network connectivity, seepage field evolution, and underground water levels have significant impacts on the safety of the CPD in CMUR.

Since there is no unified standard for CMUR evaluation, the stability of the CPD is classified into four early-warning levels using a fuzzy evaluation method, namely safe (I), relatively safe (II), mildly hazardous (III), and hazardous (IV). The single-parameter warning thresholds and the intervals of the evaluation set are established through a combination of expert scoring, numerical simulation, laboratory tests, theoretical analysis, and on-site measurements, as summarized in

Table 2. Single-parameter hierarchical graded indicators of CPD: a case in Shigetai CMUR.

Evaluation Set u			Standardized Marking Criteria														Setting Accordance
			0	0.1		0.2	0.3	0.4	0.5		0.6	0.7	0.8		0.9	1.0
			Graded Indicators for CPD
Goal	Criteria	Alternatives	Safe		Relatively Safe					Mildly Hazardous				Hazardous
A1 Early Warning of CPD in Shigetai CMUR	B1 Design Parameters	x1 CPD width/m	≥30		20~30					10~20				≤10			Numerical Simulation
		x2 CPD height/m	≤3.5		3.5~5.0					5.0~7.0				≥7.0			Numerical Simulation
		x3 CMUR burial depth/m	≤100		100~200					200~500				≥500			Numerical Simulation
	B2 Occurrence Conditions	x4 Elastic Modulus/GPa	≥40		8~40					1~8				≤1			Lab test
		x5 Compressive Strength/MPa	≥40		15~40					10~15				≤10			Lab test
		x6 Dip Angle of Coal Seam/°	0~3		3~18					18~36				36~90			Similar Simulation
		x7 Hydraulic Conductivity/cm·s−1	<10−10		10−10~10−8					10−8~10−4				≥10−4			Similar Simulation
		x8 Seismicity and Mining-induced response	No mining activity, no mine earthquake		No mining activity, rare mine earthquake					Mining activity exists, medium mine earthquake frequency				Constant mining activity, frequent mine earthquake			Field Monitoring
	B3 Deformation Parameters	x9 Borehole Pressure/MPa	≤5		5~10					10~20				≥20			Field Monitoring
		x10 Vibration speed/mm·s−1	≤1.5		1.5~3					3~6				≥6			Small Probability Event
	B4 Seepage Parameters	x11 Seepage Pressure/MPa	≤0.2		0.2~0.4					0.4~1.0				≥1.0			Field Monitoring
		x12 Underground Water Level/m	≤3		3~6					6~10				≥10			Field Monitoring
		x13 Daily Fluctuations of Water Level/m·d−1	≤0.6		0.6~1.2					1.2~2.4				≥2.4			Small Probability Event
		x14 Borehole Water Level/mm	≤300		300~600					600~900				≥900			Confidence Interval
		x15 Leakage Points within single CPD	≤1		1~3					3~5				≥5			Confidence Interval

.

It is important to note that Table 2 serves as a comprehensive framework designed to encompass a wide range of conditions that may arise in coal mine underground reservoirs (CMURs). While the methodology has been specifically applied to the Shigetai CMUR in this study, the data presented in Table 2 reflect a broader categorization to provide flexibility for application in different contexts. For the Shigetai CMUR, the included thresholds align with the specific geological and operational conditions of this mine, where surveyed CPDs and ADs are situated in a relatively small region with minimal variability in the geological conditions.

For applications beyond the Shigetai CMUR, the thresholds and intervals in Table 2 should be adjusted to account for site-specific factors, such as variations in coal seam thickness, burial depth, and material strength. This ensures that the framework remains adaptable and accurate for evaluating CMUR stability across diverse geological settings.

4.2.2. Subjective Weights θ_i Determination

According to the content of hierarchical grading results in Table 2 and the expert scoring results, the subjective weights are calculated by the AHP method. Furthermore, the consistency of each matrix is examined, and the subjective weight of CPD θ_i is calculated in

Table 3. Subjective weight of CPD.

Goal	Criteria	Weight	Alternatives	Weight	Subjective Weight θi
A1 Early Warning of CPD in Shigetai CMUR	B1 Design Parameters	0.0516	x1 CPD width	0.1429	0.0074
			x2 CPD height	0.1429	0.0074
			x3 CMUR burial depth	0.7143	0.0369
	B2 Occurrence Conditions	0.5289	x4 Elastic Modulus	0.3438	0.1818
			x5 Compressive Strength	0.3438	0.1818
			x6 Dip Angle of Coal Seam	0.1287	0.0681
			x7 Hydraulic Conductivity	0.1287	0.0681
			x8 Seismicity and Mining-induced response	0.0550	0.0291
	B3 Deformation Parameters	0.2097	x9 Borehole Pressure	0.8333	0.1747
			x10 Vibration speed	0.1667	0.0350
	B4 Seepage Parameters	0.2097	x11 Seepage Pressure	0.0800	0.0168
			x12 Underground Water Level	0.3525	0.0739
			x13 Daily Fluctuations of Water Level	0.1676	0.0351
			x14 Borehole Water Level	0.0474	0.0099
			x15 Leakage Points within single CPD	0.3525	0.0739

.

4.2.3. Standard Evaluation Matrix Establishment k_ih

First, the lower and upper bounds of the standard value interval matrix are determined by Equation (7). For indicators whose values are safer with larger values (such as the elastic modulus), use Equation (1) for the subsequent calculation convenience, and for the indicators that need qualitative evaluation (such as construction quality), the standardized marking criteria in Table 2 are used to quantitatively process the results, as shown in

Table 4. Standard evaluation matrix of CPD k_ih.

Alternatives	Safe		Relatively Safe		Mildly Hazardous		Hazardous
	ai1	bi1	ai2	bi2	ai3	bi3	ai4	bi4
x1 CPD width index	0	10		20		30		40
x2 CPD height	0	3.5		5.0		7.0		10
x3 CMUR burial depth	0	100		200		500		100
x4 Elastic Modulus index	0	20		52		59		60
x5 Compressive Strength index	0	20		45		50		60
x6 Dip Angle of Coal Seam	0	3		18		36		90
x7 Hydraulic Conductivity	0	10−10		10−8		10−4		1
x8 Seismicity and Mining-induced response	0	0.25		0.5		0.75		1
x9 Borehole Pressure	0	5		10		20		30
x10 Vibration speed	0	1.5		3		6		9
x11 Seepage Pressure	0	0.2		0.4		1.0		2.0
x12 Underground Water Level	0	3		6		10		15
x13 Daily Fluctuations of Water Level	0	0.6		1.2		2.4		3.6
x14 Borehole Water Level	0	300		600		900		1500
x15 Leakage Points within single CPD	0	1		3		5		10

.

4.2.4. Objective Weight Determination and Field Verification

(1)

Calculation parameter selection for the Shigetai CMUR;

As the monitoring results always change for the Shigetai CMUR, we selected the monitoring data of the CPDs in each goaf in 31201–31205 on 31 December 2022 as an example, and evaluated the real-time safety of the CPD in the Shigetai CMUR (

Table 5. Parameter selection of CPD.

	31201-4#	31202-4#	31203-1#	31204-3#	31205-1#
x1 CPD width index	20
x2 CPD height	3.4
x3 CMUR burial depth	120
x4 Elastic Modulus index	52.5
x5 Compressive Strength index	44.1
x6 Dip Angle of Coal Seam	3
x7 Hydraulic Conductivity	10−9
x8 Seismicity and Mining-induced response	Mining activity exists, no mining earthquake = 0.6
x9 Borehole Pressure	0	0	6.1	0	0.1
x10 Vibration speed	0	2.06	0	0	0
x11 Seepage Pressure	0.04	0.08	0.1	0.12	0.19
x12 Underground Water Level	0.78	4.97	4.00	2.90	4.04
x13 Daily Fluctuations of Water Level	0	0.20	0.90	0.34	0.1
x14 Borehole Water Level	119	550.7	374.8	423.5	300
x15 Leakage Points within single CPD	0	1	3	0	1

). Each CPD is considered the most dangerous CPD in its goaf, and the parameters of each CPD are determined as follows:

(2)

Total weight w_i calculation;

According to the parameters in Table 5, the total weight of the CPDs in goaves 31201–31205 can be calculated by Equations (2)–(6), which is shown in

Table 6. Total weight of CPDs in Shigetai CMUR.

	31201-4#	31202-4#	31203-1#	31204-3#	31205-1#
x1 CPD width index	0.0107	0.0096	0.0086	0.0103	0.0099
x2 CPD height	0.0071	0.0063	0.0057	0.0068	0.0065
x3 CMUR burial depth	0.0393	0.0350	0.0316	0.0377	0.0361
x4 Elastic Modulus index	0.2702	0.2406	0.2170	0.2591	0.2483
x5 Compressive Strength index	0.2608	0.2322	0.2094	0.2500	0.2396
x6 Dip Angle of Coal Seam	0.0659	0.0587	0.0529	0.0632	0.0606
x7 Hydraulic Conductivity	0.0824	0.0734	0.0662	0.0790	0.0757
x8 Seismicity and Mining-induced response	0.0479	0.0426	0.0385	0.0459	0.0440
x9 Borehole Pressure	0.0845	0.0753	0.1507	0.0811	0.0792
x10 Vibration speed	0.0169	0.0358	0.0136	0.0162	0.0156
x11 Seepage Pressure	0.0098	0.0101	0.0098	0.0125	0.0146
x12 Underground Water Level	0.0451	0.0846	0.0670	0.0674	0.0771
x13 Daily Fluctuations of Water Level	0.0170	0.0202	0.0341	0.0255	0.0182
x14 Borehole Water Level	0.0067	0.0121	0.0087	0.0111	0.0088
x15 Leakage Points within single CPD	0.0358	0.0637	0.0862	0.0343	0.0657

.

(3)

Membership function calculation and eigenvalue determination.

According to Equation (8), the membership function and affiliation of each CPD could be calculated. Taking CPD 31201-4# as an example, according to the judgment conditions, all of the monitoring indicators of this coal pillar dam body belong to classes I–II. The affiliation of each factor of CPD 31201-4# for each safety level can be calculated in

Table 7. Affiliation of 31201-4# CPD in Shigetai CMUR.

${\mathit{\mu }}_{\mathit{i}\mathit{h}}(\mathit{u})$	h = 1	h = 2	h = 3	h = 4
x1 CPD width index	0.000	0.500	0.000	0.000
x2 CPD height	0.514	0.000	0.000	0.000
x3 CMUR burial depth	0.300	0.000	0.000	0.000
x4 Elastic Modulus index	0.000	0.429	0.000	0.000
x5 Compressive Strength index	0.000	0.536	0.000	0.000
x6 Dip Angle of Coal Seam	0.500	0.000	0.000	0.000
x7 Hydraulic Conductivity	0.000	1.000	0.000	0.000
x8 Seismicity and Mining-induced response	0.000	0.100	0.000	0.000
x9 Borehole Pressure	1.000	0.000	0.000	0.000
x10 Vibration speed	1.000	0.000	0.000	0.000
x11 Seepage Pressure	0.900	0.000	0.000	0.000
x12 Underground Water Level	0.870	0.000	0.000	0.000
x13 Daily Fluctuations of Water Level	1.000	0.000	0.000	0.000
x14 Borehole Water Level	0.802	0.000	0.000	0.000
x15 Leakage Points within single CPD	1.000	0.000	0.000	0.000

.

Referring to the above method, the affiliation values for the remaining dam bodies 31201–31205 can be calculated. Using Equation (9), the integrated relative affiliation function of each CPD can be calculated, and the eigenvalues of the CPD can be further obtained. By changing the values of α and p, the eigenvalues of each dam body under different optimization criteria and distance parameters are obtained (α = 1, p = 1: linear model, α = 2, p = 1: Sigmoid model, α = 1, p = 2: Topsis model, α = 2, p = 2 multi-objective fuzzy preference model), and the parameter with the largest eigenvalue of the computing model is selected as the indicator of the monitoring and warning level. The calculation results are shown in

Table 8. Integrated relative affiliation and eigenvalues of CPDs in 31201–31205.

31201-4# Integrated Relative Affiliation					31201-4# Eigenvalue H(u)	Warning Level
h	1	2	3	4
α = 1, p = 1	0.256	0.348	0.000	0.000	0.952	I
α = 2, p = 1	0.106	0.222	0.000	0.000	0.550	I
α = 1, p = 2	0.218	0.454	0.000	0.000	1.127	II
α = 2, p = 2	0.072	0.409	0.000	0.000	0.891	I
31202-4# integrated relative affiliation					31202-4# eigenvalue H(u)	Warning Level
h	1	2	3	4
α = 1, p = 1	0.175	0.391	0.000	0.000	0.956	I
α = 2, p = 1	0.043	0.291	0.000	0.000	0.625	I
α = 1, p = 2	0.194	0.463	0.000	0.000	1.120	II
α = 2, p = 2	0.055	0.427	0.000	0.000	0.908	I
31203-1# integrated relative affiliation					31203-1# eigenvalue H(u)	Warning Level
h	1	2	3	4
α = 1, p = 1	0.105	0.383	0.000	0.000	0.871	I
α = 2, p = 1	0.014	0.278	0.000	0.000	0.570	I
α = 1, p = 2	0.105	0.459	0.000	0.000	1.023	II
α = 2, p = 2	0.014	0.418	0.000	0.000	0.850	I
31204-3# integrated relative affiliation					31204-3# eigenvalue H(u)	Warning Level
h	1	2	3	4
α = 1, p = 1	0.228	0.348	0.000	0.000	0.925	I
α = 2, p = 1	0.081	0.222	0.000	0.000	0.524	I
α = 1, p = 2	0.209	0.454	0.000	0.000	1.118	II
α = 2, p = 2	0.065	0.409	0.000	0.000	0.884	I
31205-1# integrated relative affiliation					31205-1# eigenvalue H(u)	Warning Level
h	1	2	3	4
α = 1, p = 1	0.198	0.348	0.000	0.000	0.894	I
α = 2, p = 1	0.057	0.222	0.000	0.000	0.501	I
α = 1, p = 2	0.196	0.454	0.000	0.000	1.105	II
α = 2, p = 2	0.056	0.409	0.000	0.000	0.874	I

:

The results of the safety assessment for CPDs in the Shigetai CMUR as of 31 December 2022 indicate that all CPDs are rated as “relatively safe” or above. However, the relative safety ranking of the CPDs across the five mining areas varies. The ranking from the most hazardous to the safest is as follows: 31201-4# > 31202-4# > 31204-3# > 31205-1# > 31203-1#. This suggests that the safety of the CPDs in Shigetai is closely linked to the regional geological conditions.

For goaves with lower terrain, such as 31201 and 31202, the early-warning model indicates that their CPDs are among the most hazardous of the five goaves. It is important to note that the CPDs selected for this analysis have shown the most significant deformation and the highest number of seepage points within their respective goaves, resulting in a relatively low safety level. However, other CPDs that exhibit less deformation and no obvious seepage are rated as “safe”, and their safety can be ensured with real-time monitoring.

4.3. Early Warning of AD in Shegetai CMUR

The steps for early warning of the ADs in the Shigetai CMUR are close to the CPDs, as detailed below:

4.3.1. Single-Parameter Warning Values

Based on existing numerical calculations and long-term field-monitoring results, it has been determined that the ADs, as one of the main water-retaining structures of the CMUR, generally do not bear significant loads, such as ground stress or mining stress on their surfaces. Instead, strain is typically used to characterize the deformation of the ADs in practical measurements.

The stability evaluation of ADs in the CMUR is categorized into four levels of early warning, namely safe (I), relatively safe (II), mildly hazardous (III), and hazardous (IV). This classification is based on the fuzzy evaluation method. The single-parameter warning thresholds and the intervals of the evaluation set for ADs are presented in

Table 9. Single-parameter hierarchical graded indicators of AD: a case in Shigetai CMUR.

Evaluation Set u			Standardized Marking Criteria												Setting Accordance
			0	0.1	0.2	0.3	0.4	0.5		0.6	0.7	0.8	0.9	1
			Graded Indicators for AD
Goal	Criteria	Goal	Safe			Relatively Safe			Mildly Hazardous			Hazardous
A2 Early Warning of AD in Shigetai CMUR	B1 management system	x1 Construction Quality	Good construction quality, complete documentation			Satisfactory construction quality, generally complete documentation			Average construction quality, loss of documentation			Bad construction quality, no documentation			Expert advice
		x2 Operation and Maintenance System	Well-established daily inspections system			Basically completed daily inspection system			Daily inspection system to be improved			No daily inspection system			Expert advice
	B2 Parameters of the dam body	x3 Dam concrete marking	≥C25			C20~ C25			C15~ C20			≤C15			Similar Simulation
		x4 Dam Height/m	≤4			4~6			6~8			≥8			Numerical Simulation
		x5 Dam Embedding Depth/m	≥1.0			0.5~1.0			0.1~0.5			≤0.1			Numerical Simulation
		x6 CMUR burial depth/m	≤100			100~200			200~500			≥500			Numerical Simulation
		x7 Seismicity and Mining-induced response	No mining activity, no mine earthquake			No mining activity, rare mine earthquake			Mining activity exists, medium mine earthquake frequency			Constant mining activity, frequent mine earthquake			Expert advice
	B3 Deformation Parameters	x8 Tensile Strain/μ	≤80			80~160			160~320			≥320			Field Monitoring
		x9 Compressive Strain/μ	≤100			100~200			200~400			>400			Field Monitoring
	B4 Seepage Parameters	x10 Seepage Pressure/MPa	≤0.2			0.2~0.4			0.4~1.0			≥1.0			Field Monitoring
		x11 Underground Water Level/m	≤3			3~6			6~10			≥10			Theoretical Calculation
		x12 Daily Fluctuations of Water Level/m·d−1	≤0.6			0.6~1.2			1.2~2.4			≥2.4			Small Probability Event
		x13 Borehole Water Level/mm	≤300			300~600			600~900			≥900			Confidence Interval
		x14 Leakage Points within single CPD	≤1			1~3			3~5			≥5			Confidence Interval

.

It also should be noted that Table 9 only reflects the specific geological and operational conditions of the Shigetai CMUR, where surveyed ADs are situated in a relatively small region with minimal variability in geological conditions. While the methodology can be extended to other regions, the thresholds and intervals in Table 9 must be adapted to account for site-specific factors, such as variations in concrete marking, and in situ stress condition changes.

4.3.2. Subjective Weights θ_i Determination

According to the content of the hierarchical grading results and the expert scoring results in Table 9, the subjective weights of the ADs are also determined, as shown in

Table 10. Subjective weight of AD.

Goal	Criteria	Weight	Alternatives	Weight	Subjective Weight θi
A2 Early Warning of AD in Shigetai CMUR	B1 management system	0.0597	x1 Construction Quality	0.8333	0.0497
			x2 Operation and Maintenance System	0.1667	0.0100
	B2 Parameters of the dam body	0.1749	x3 Dam concrete marking	0.4440	0.0777
			x4 Dam Height/m	0.1653	0.0289
			x5 Dam Embedding Depth/m	0.2596	0.0454
			x6 CMUR burial depth/m	0.0847	0.0148
			x7 Seismicity and Mining-induced response	0.0464	0.0081
	B3 Deformation Parameters	0.3827	x8 Tensile Strain/μ	0.5000	0.1914
			x9 Compressive Strain/μ	0.5000	0.1914
	B4 Seepage Parameters	0.3827	x10 Seepage Pressure/MPa	0.0800	0.0306
			x11 Underground Water Level/m	0.3525	0.1349
			x12 Daily Fluctuations of Water Level/m·d−1	0.1676	0.0641
			x13 Borehole Water Level/mm	0.0474	0.0181
			x14 Leakage Points within single CPD	0.3525	0.1349

.

4.3.3. Objective Weight Determination and Field Verification

(1)

Calculation parameter selection for the Shigetai CMUR;

By referring to the standard evaluation matrix in the CPD, the standard evaluation matrix in the AD can be established. For AD stability analysis in the Shigetai CMUR, we also selected the monitoring data of the ADs in each goaf in 31201–31205 on 31 December 2022 as an example and evaluated the real-time safety of the AD in the Shigetai CMUR. For the three indicators of qualitative evaluation, namely construction quality, operation and maintenance system, and mining-induced response, the standardized marking criteria are used for assessment. The evaluation parameters of the ADs in each goaf in the CMUR are shown in the following table.

(2)

Total weight w_i calculation;

According to the parameters in

Table 11. Parameter selection of CPD.

	31201-4#	31202-4#	31203-1#	31204-3#	31205-1#
x1 Construction Quality	Satisfactory construction quality, complete documentation = 0.2
x2 Operation and Maintenance System	Completed daily inspections system = 0.2
x3 Dam concrete marking	10
x4 Dam Height/m	3.9
x5 Dam Embedding Depth/m	1.5
x6 CMUR burial depth/m	120
x7 Seismicity and Mining-induced response	Mining activity exists, no mining earthquake = 0.6
x8 Tensile Strain/μ	150	30	105	11	8
x9 Compressive Strain/μ	48	0	55	17	38
x10 Seepage Pressure/MPa	0.04	0.08	0.1	0.12	0.19
x11 Underground Water Level/m	0.78	4.97	4.00	2.90	4.04
x12 Daily Fluctuations of Water Level/m·d−1	0	0.20	0.90	0.34	0.1
x13 Borehole Water Level/mm	119	550.7	374.8	423.5	300
x14 Leakage Points within single CPD	0	1	3	0	1

, the total weight of the CPDs in goaves 31201–31205 can be calculated by Equations (2)–(6), which is shown in

Table 12. Total weight of CPDs in Shigetai CMUR.

	31201-4#	31202-4#	31203-1#	31204-3#	31205-1#
x1 Construction Quality	0.0673	0.0455	0.0427	0.0600	0.0539
x2 Operation and Maintenance System	0.0135	0.0091	0.0085	0.0120	0.0108
x3 Dam concrete marking	0.1168	0.0788	0.0740	0.1041	0.0936
x4 Dam Height/m	0.0429	0.0290	0.0272	0.0383	0.0344
x5 Dam Embedding Depth/m	0.1024	0.0692	0.0649	0.0913	0.0821
x6 CMUR burial depth/m	0.0245	0.0165	0.0155	0.0218	0.0196
x7 Seismicity and Mining-induced response	0.0207	0.0140	0.0131	0.0185	0.0166
x8 Tensile Strain/μ	0.1439	0.0971	0.2024	0.1282	0.1176
x9 Compressive Strain/μ	0.1439	0.2306	0.0912	0.1282	0.1153
x10 Seepage Pressure/MPa	0.0276	0.0218	0.0219	0.0328	0.0360
x11 Underground Water Level/m	0.1278	0.1819	0.1500	0.1778	0.1907
x12 Daily Fluctuations of Water Level/m·d−1	0.0482	0.0434	0.0764	0.0673	0.0451
x13 Borehole Water Level/mm	0.0190	0.0261	0.0194	0.0293	0.0219
x14 Leakage Points within single CPD	0.1014	0.1370	0.1928	0.0904	0.1625

.

(3)

Membership function calculation and eigenvalue determination.

By using Equation (8), the relative affiliation function and value of each AD can be calculated. Taking 31201-4# AD as an example, the relative affiliation value of the 31201-4# AD is shown in

Table 13. Affiliation of 31201-4# AD in Shigetai CMUR.

${\mathit{\mu }}_{\mathit{i}\mathit{h}}(\mathit{u})$	h = 1	h = 2	h = 3	h = 4
x1 Construction Quality	0.600	0.000	0.000	0.000
x2 Operation and Maintenance System	0.600	0.000	0.000	0.000
x3 Dam concrete marking	0.500	0.000	0.000	0.000
x4 Dam Height/m	0.513	0.000	0.000	0.000
x5 Dam Embedding Depth/m	0.750	0.000	0.000	0.000
x6 CMUR burial depth/m	0.300	0.000	0.000	0.000
x7 Seismicity and Mining-induced response	0.000	0.100	0.000	0.000
x8 Tensile Strain/μ	0.000	0.625	0.000	0.000
x9 Compressive Strain/μ	0.760	0.000	0.000	0.000
x10 Seepage Pressure/MPa	0.900	0.000	0.000	0.000
x11 Underground Water Level/m	0.870	0.000	0.000	0.000
x12 Daily Fluctuations of Water Level/m·d−1	1.000	0.000	0.000	0.000
x13 Borehole Water Level/mm	0.802	0.000	0.000	0.000
x14 Leakage Points within single CPD	1.000	0.000	0.000	0.000

.

By using Equation (9), the integrated relative affiliation function $v_h(u)$ of each AD can be calculated, and the eigenvalues of the AD can be further obtained, as shown in

Table 14. Integrated relative affiliation and eigenvalues of ADs in 31201–31205.

31201-4# Integrated Relative Affiliation					31201-4# Eigenvalue H(u)	Warning Level
h	1	2	3	4
α = 1, p = 1	0.623	0.092	0.000	0.000	0.807	I
α = 2, p = 1	0.733	0.010	0.000	0.000	0.753	I
α = 1, p = 2	0.570	0.235	0.000	0.000	1.040	II
α = 2, p = 2	0.636	0.087	0.000	0.000	0.810	I
31202-4# integrated relative affiliation					31202-4# eigenvalue H(u)	Warning Level
h	1	2	3	4
α = 1, p = 1	0.575	0.172	0.000	0.000	0.920	I
α = 2, p = 1	0.647	0.041	0.000	0.000	0.730	I
α = 1, p = 2	0.566	0.329	0.000	0.000	1.225	II
α = 2, p = 2	0.630	0.194	0.000	0.000	1.019	II
31203-1# integrated relative affiliation					31203-1# eigenvalue H(u)	Warning Level
h	1	2	3	4
α = 1, p = 1	0.285	0.174	0.000	0.000	0.634	I
α = 2, p = 1	0.137	0.043	0.000	0.000	0.223	I
α = 1, p = 2	0.263	0.286	0.000	0.000	0.835	I
α = 2, p = 2	0.113	0.138	0.000	0.000	0.389	I
31204-3# integrated relative affiliation					31204-3# eigenvalue H(u)	Warning Level
h	1	2	3	4
α = 1, p = 1	0.683	0.002	0.000	0.000	0.686	I
α = 2, p = 1	0.822	0.000	0.000	0.000	0.822	I
α = 1, p = 2	0.672	0.006	0.000	0.000	0.683	I
α = 2, p = 2	0.808	0.000	0.000	0.000	0.808	I
31205-1# integrated relative affiliation					31205-1# eigenvalue H(u)	Warning Level
h	1	2	3	4
α = 1, p = 1	0.557	0.002	0.000	0.000	0.561	I
α = 2, p = 1	0.613	0.000	0.000	0.000	0.613	I
α = 1, p = 2	0.503	0.005	0.000	0.000	0.512	I
α = 2, p = 2	0.505	0.000	0.000	0.000	0.505	I

.

The results show that all ADs of the Shigetai CMUR as of 31 December 2022, are classified as either safe (I) or relatively safe (II). For the ADs at goaves 31201 and 31202, the monitoring and warning level is shown as relatively safe (II). Given the higher monitoring and warning levels of the ADs at goaves 31201 and 31202, it is recommended to increase patrol frequency and enhance monitoring systems with advanced real-time data analysis to detect any early signs of instability. Comparison with the contour lines of the CMUR (Figure 2) reveals that goaves 31201 and 31202 are located at the lower levels of the reservoir. Consequently, their actual storage height is higher than that of the remaining three goaves, which explains why the monitoring and warning levels of the ADs at goaves 31201 and 31202 are slightly higher than those of the other ADs. This observation supports the validity and applicability of the model for assessing CMUR stability based on reservoir storage height and geological conditions.

5. Conclusions

This paper focuses on the monitoring and early warning of the long-term stability of coal mine underground reservoirs (CMURs), using the Shigetai CMUR as a typical case study. The long-term monitoring of seepage pressure, borehole stress, and surface strain of the coal pillar dams (CPDs) and artificial dams (ADs) in goaves 31201–31205 was conducted. Using this model, the real-time stability of CPDs and ADs in Shigetai CMUR as of 31 December 2022 was analyzed. The results are consistent with the geological conditions, demonstrating the model’s effectiveness. The main conclusions are listed as follows.

Real-time monitoring equipment was installed on the CPDs and ADs in the Shigetai CMUR to measure stress, deformation, seepage, water level, flow rate, and vibration, building upon existing research. The DC method was applied to detect the seepage situation in the dams. The monitoring results showed that the 31201 and 31202 extraction areas are in the lower area of the whole groundwater reservoir, and the water storage height of the dam body in this area should be higher than other areas, which may have a greater risk in the later stage. At the same time, the rapid fluctuation of the groundwater level will probably lead to the deterioration of the stability of the dam body. For example, the 31202-1-2 joint lane in August near the maximum fluctuation of the water level reached 1 m/day, resulting in the seepage pressure of the CPD 31201-2# in the August period rapidly rising from 0.01 MPa to 0.06 MPa. The fluctuation speed of the groundwater level should also be one of the important elements of the monitoring and early warning of the coal mine underground water reservoir. The deformation of the ADs in the Shigetai coal mine exhibits significant time-dependent deformation.

Using variable set theory, fuzzy mathematics, and the analytic hierarchy process (AHP), this study developed a novel model that classifies the safety levels of CPDs and ADs into four categories. The key parameters that need to be used for long-term stability identification and monitoring and early warning of CMURs are determined. The range of each value for different levels is determined, and a dynamic monitoring and early-warning model, considering real-time parameter fluctuation and expert scoring, is proposed. With the help of this model, the real-time stability of the CPDs and ADs on 2022 December 31 is analyzed. The analysis indicated that, as of 31 December 2022, all CPDs at the Shigetai CMUR were classified as safe. However, continuous monitoring is essential, especially for CPDs experiencing higher stress and seepage. The relatively safe classification of the ADs in goaves 31201 and 31202 suggests a need for increased patrols and enhanced monitoring, especially given their lower elevation and higher water storage levels.

The early-warning model developed in this study may have broader applications in the future. It could be adapted to other coal mine underground reservoirs and even extended to different types of mining environments, including those with varying geological conditions and reservoir characteristics. Abundant experimental data and field analysis are suggested for model parameter determination. Future work could focus on refining the model to accommodate more diverse mining operations and the integration of additional monitoring technologies, such as automated data analysis and real-time sensor networks, to further improve the model’s predictive accuracy and reliability. Future work may also focus on other types of underground reservoirs, such as those in non-coal mining sectors, enhancing the safety and sustainability of mining operations worldwide.