High rock stress and ground temperature pose great threats to the routine production of longwall top coal caving (LTCC) panels. In this risky condition, the width of the chain pillar is considered a factor adjustable for controlling coal burst and goaf ignition hazards. However, a contradiction, as suggested by longwall experience, is that narrowing the pillar helps coal burst prevention but negatively leads to higher self-ignition potentials, while widening the pillar restrains goaf ignition but increases the likelihood of coal burst. This paper conducted a case study on a longwall panel from Tangkou Mine, China. The paper first analysed stress, elastic strain energy, and goaf temperature variation with varying pillar widths, by which the coal burst risk index δr and goaf ignition risk index Qs were defined and correlated to pillar width D. Further, a pillar width determination method considering coal burst and goaf ignition dual-hazard management was developed by means of the operating point principle. By this method, a reasonable width range was defined by plotting both correlation curves δr=fD and Qs=gD on a chart, followed by optimal width determination according to the intersection of both curves and further verification via a field trial.

Coal burst and goaf ignition are two frequent risks in underground coal mining, potentially threatening mine safety and routine production [14]. The potentials can be very high as the mine depth exceeds 1000 m due to great rock stress and high ground temperature [5]. In this environment, compared to a fully mechanised mining method, the longwall top coal caving (LTCC) method has greater coal burst and goaf ignition potentials, as LTCC leads to significant ground pressurisation due to the large mining height and leaves residual coals of about 15% of the total reserve in goaf voids. As indicated by LTCC experience, adjusting the chain pillar width can help control the coal burst and goaf ignition considering its effect on ground stress and goaf-to-goaf ventilation. However, determining the optimal width is rather challenging because of a contradiction that wide pillars accumulate more elastic strain energy and simultaneously result in coal recovery ratio decline [68], whereas narrow pillars are less capable of hindering air leakage via fractures into adjacent goafs and causing goaf ignition in the high ground temperature environment, as shown in Figure 1. Therefore, developing a method to determine the chain pillar width is of great significance for mine safety and production.

A mass of studies has been conducted to understand coal burst mechanisms and safety control [912]. For the contributing factors, Bukowska [13] and Dyke et al. [14] identified that the depth of cover, geological structures, coal measure rock mechanics, coal seam thickness, and basement thickness have varying influences on the coal burst behaviour. Mazaira and Konicek [15] and Pu et al. [16] proposed that, in addition to coal measure geology, coal burst is also correlated to panel configuration pattern, panel geometry, and mining methods. Kaiser and Cai [17] suggested the correlation between coal burst and ground support parameters, further developing an interactive ground support design tool taking account of coal burst management. Dou et al. [18] identified the impact of overburden structure variation on coal burst and discussed the mechanism from an energy variation perspective. Saharan and Mitri [19] found that destressing blasting can be an effective technology to decrease the coal burst potential, to which destressing drilling, hydraulic softening, backfilling, ground support reinforcement, and nonpillar mining have similar functionality. Huang et al. [20] tested the effectiveness of hydraulic fracturing in restraining coal burst hazards, of which the fundamentals involved coal and rock mass strength reduction. Konicek et al. [21] proposed that drillhole blasting can be applied to control coal burst hazards in deep mines. Besides geological factors, chain pillar failure is likely to induce coal burst [22]. From the energy accumulation and dissipation perspective, Vardar et al. [23] and Xue et al. [24] analysed pillar energy transfer due to structural failure, unraveling the mechanism of pillar failure inducing coal burst [23, 24]. Wang et al. verified that an inappropriate pillar width can lead to severe mine seismicity and coal burst, based on which a pillar width design method was developed via numerical simulations [25]. These studies focused on the pillar width design considering coal burst management, ignoring the pillar width impact on goaf ignition.

Goaf ignition is another safety threat and has attracted extensive studies especially about the ignition inducements and control techniques [2629]. Yuan and Smith [30] investigated the impact of the surface area of coals and reaction heat. Taraba et al. [31] analysed the impact of coal particle diameter on the oxidation heating zone and gas. Mu [32] identified the longwall retreat rate as a key factor contributing to residual coal ignition in previous panel goafs. Liu et al. [33] and Liu and Qin [34] proposed that increasing the retreat rate and decreasing residual coals and ventilation rate can help restrain the goaf ignition risk, further determining the minimum retreat rate for goaf ignition control according to the critical temperature of coal self-ignition. Currently, available technologies and materials for coal self-ignition control include grouting, fire retardant, gel materials, and inert gases [35, 36]. Colaizzi [37] developed a porous foam cement material to prevent coal ignition. Qin et al. [38] developed a FA-based three-phase foam material to prevent and eliminate goaf ignition potentials. These technologies have proved effective in goaf ignition control but inevitably increase economic inputs and decrease production efficiency.

Therefore, determining the chain pillar width of LTCC panels with high coal burst and goaf ignition potentials should consider the width impact on pillar behaviour and residual coal ignition in adjacent goaf voids. This research can help control mine hazards without additional economic input. In this context, this paper selected an LTCC panel LW6304 from Tangkou Mine in Shandong Province, China, as a case, of which the negative factors include the great depth of cover and high tendency in coal burst and goaf ignition. The paper studied pillar pressurisation, pillar energy variation, and oxidation heating zone variation under different pillar width conditions. On this basis, a method to determine the optimal pillar width with considerations of coal burst and goaf ignition control was developed and further proved effective by a field trial in the mine.

LW6304 is 960 to 985 m deep, 980 m on average, working a mining height of 10 m and dip angle of 3°. In situ observation indicated that the coal has a high ignition tendency, with a self-ignition period of 21 days as the shortest. The average gas content of the coal seam is 2.54 m3/t. The geological structure is simple, characterising a minor impact on longwall activities. Table 1 lists the coal measure lithology and parameters. LW6304 adopted the LTCC mining method, 4 m cutting height, and 85% recovery ratio. The panel was divided into two sections, incorporating a 610 m long and 182 m wide section in the inner position and a 1065 m long and 60 m wide section in the outer position. LW6304 adjoined LW6305 goaf within the 345 to 610 m range from the LW6304 set-up position; the panel width of LW6305 was 55 m. Figure 2 exhibits the panel configuration.

The great depth of cover and large mining height resulted in significant frontal abutment pressurisation, coupled with significant dynamic loading due to the large space for rock block rotation. To prevent coal burst, the panel adopted goaf-side entry tunnelling to leave a 5 m wide pillar. In this narrow pillar condition, the panel experienced localised pillar failure, and the residual coals in LW6305 goaf tended to ignite, as indicated by (i) significant CO concentration increases, greater than 50 ppm many times and sometimes up to 75 ppm (parts per million), and (ii) frequent out-of-limit alarms of coal bed methane. Because of the problems, LW6304 production was sometimes suspended, during which coal ash solutions and nitrogen were injected into LW6305 goaf to prevent residual coal ignition. The risk itself and corresponding measures delayed the routine production, simultaneously increasing additional inputs.

Numerical simulation is frequently used to study geotechnical problems in underground mining, tunnelling, and slope stability [39]. FLAC3D (Fast Lagrangian Analysis of Continua in 3 Dimensions) software is a numerical modeling software for geotechnical analyses, which is widely used in the design and analysis of engineering in civil, mining, and geotechnical excavations, etc. In this paper, a continuum-based model using FLAC3D was established to simulate the maximum principal stress regime and elastic strain energy variation with varying pillar widths, thus assessing the impact of chain pillar width on coal burst behaviour.

3.1. Model Configuration

The model adopted the parameters summarised in Table 1 and 7 m wide chain pillar, totally 400 m long, 300 m wide, and 47 m high, as shown in Figure 3. LW6304 panel length was 182 m, and the roadway width was 5 m. The model adopted a fixed-displacement boundary for the bottom and roller boundary for the four lateral edges, applying 23.25 MPa stress vertically on the top to simulate the loading of unmodeled overlying strata. The rock behaviour obeyed the Mohr-Coulomb criterion and was calibrated using measured results (see [4042] for model calibration procedures).

3.2. Pillar Stress and Energy Variation

The coal body around LW6304 experienced stress redistribution and significant concentration in localised areas. The ground pressurisation led to elastic strain energy accumulation in the chain pillar and thus high potentials of coal burst, implying that the stress profile and energy variation can be critical indexes to quantify the coal burst tendency under different pillar width conditions. Therefore, in the model, the initial pillar width of 7 m was changed to 3, 5, 9, and 11 m to obtain the corresponding stress and energy responses.

3.2.1. Stress Profile

Along the longwall retreat direction, vertical sections were extracted via the peak maximum principal stress point, as shown in Figure 4.

As can be seen from Figure 4, the maximum principal stress regime varies a lot from the 3 m to 11 m wide pillar conditions, featuring an increase trend with the pillar width. In narrow pillar (3 and 5 m width) configurations, the peak maximum principal stress is smaller in the pillar than in the panel abutment (solid coal body) where stress concentrates. The chain pillar shows significant squeezing deformation. As the pillar width is greater than 7 m, the stress concentrates towards the pillar, varying from 44.3 MPa in the 5 m width condition to 76.1 MPa in the 11 m width condition, as shown in Figure 5. Fitting the data obtained the correlation of pillar stress with width, y=100.54112.02e0.14xR2=0.98, a negative exponential relation.

3.2.2. Elastic Strain Energy Variation

The simulation obtained the elastic strain energy variation with varying pillar widths, as exhibited in Figure 6.

Figure 6 shows that the pillar strain energy has a rapid increase and decrease, ultimately levelling off with minor fluctuations, characterising an evident energy accumulation and release. The energy increase rate is related to the pillar width: the wider the pillar, the greater the increase rate and peak value. Taking the 3 m width as a reference, the maximum elastic strain energy becomes 1.12, 1.17, 1.31, and 1.42 times in 5, 7, 9, and 11 m width conditions. Another phenomenon is that the times of pillar energy release increase with pillar width. In 9 and 11 m width conditions, the maximum energy release is 1.31×105 J and 1.91×105 J, respectively. Narrowing the pillar width to 7 m can reduce the energy release to be lower than 1×105 J.

3.3. Coal Burst Risk Assessment

Based on the maximum principal stress profiles and elastic strain energy variation in varying pillar widths, the paper further assessed coal burst potentials and obtained a reasonable pillar width range considering coal burst management.

Coal burst is a frequent mine hazard and has attracted a mass of studies [4345]. Dou et al. proposed a representative method to quantify the coal burst tendency by defining an index referred to as the stress concentration coefficient, δi. The index measures the product of all components determined as the ratio of the maximum principal stress to the self-weight stress in localised areas. The coal burst potential can be obtained by comparing δi and the stress threshold of coal burst. This method also includes coal burst potential classification [43].

The factors contributing to coal burst problems in underground coal mining include geological structures, abutment pressurisation, overburden strata movement, and coal measure lithology. For mine stress, the paper considered frontal abutment pressurisation and side abutment pressurisation induced by LW6305 mining operation, defining the ratio of the maximum principal stress to the self-weight stress as a coal burst risk index, δr. For LW6304, the self-weight stress (σ0) is 24.5 MPa, and the uniaxial compressive strength (RC) of coals was tested to be 13 MPa, cooperatively yielding the coal burst intensity gradings as listed in Table 2 [43]. Further, according to the maximum principal stress obtained in the numerical simulation, LW6304 coal burst potentials in varying pillar width conditions were calculated and summarised in Table 3.

Table 3 shows that narrow pillars (3, 5, and 7 m wide) have an index lower than 2.86, subject to a weak coal burst intensity according to Table 2. Widening the pillar to 9 m increases the intensity to a medium level. By comparison, the LW6304 chain pillar width should be smaller than 9 m, by which the coal burst potentials can be controlled to some extent.

The impact of chain pillar width on goaf ignition was studied using an Ansys Fluent numerical model. Ansys Fluent software is a computational fluid dynamics software to solve the sophisticated models for multiphase flows, chemical reaction, and combustion, which is widely used in the study of goaf ignition (https://www.ansys.com/products/fluids/ansys-fluent). In the model, the pillar width had the same configuration as the previous model, by which the oxidation heating zone variation was obtained and analysed.

4.1. Model Design and Configuration

The model was established in Ansys Fluent software, geometrically 275 m long, 264 m wide, and 139.5 m high, as shown in Figure 7. Combined with on-site drilling observation and FLAC3D numerical simulation, the height of the caving zone was 23 m, and the fractured zone was 116.5 m.

Goaf ignition potentials depend on pillar porosity. Considering that the chain pillar is subject to compressive stress during the longwall retreat process, the pillar porosity can be determined via Equation (1) ([46]):
where η is the pillar porosity, η0 is the initial porosity, εv is volumetric strain, σ0 is the virgin in situ stress, calculated by σ0=σ1+σ2+σ3/3, and K is the bulk modulus. Incorporating the laboratory test and numerical simulation results obtained the pillar porosity, respectively, 0.21, 0.19, 0.15, 0.11, and 0.09 with pillar width increased from 3, 5, 7, and 9 to 11 m. The goaf porosity was assumed to be 0.3 (LW6304) and 0.25 (LW6305) according to the compaction degree of collapsed rock. The belt roadway worked for air-in ventilation, providing fresh air with O2 of 12% and N2 of 77%. Other gas components of air have little influence on the results due to their small content, which is not considered in the simulation. According to the geological conditions of LW6304, the gas content of the coal seam had little influence on the goaf ignition. Therefore, the influence of coal seam gas content on goaf ignition was not considered here. The model adopted the standard Kε turbulent model for iterative calculation. The permeability governing equation used in the Ansys Fluent model is derived from the Carman formula of porous media [47]. The expression is
where K¯ is the permeability in goaf, Dm¯ is the average particle size of fractured rock mass in goaf, and Kp¯ is the average coefficient of bulk increase of fractured rock mass in goaf.

Dm¯ and Kp¯ at different areas of goal in the longwall top coal caving coalface are different. Their relationship is shown in Table 4 [47].

4.2. Simulation Results

The oxidation heating zone profiles were obtained via the goaf ignition model. By assuming 10% O2 concentration as the threshold of the oxidation heating zone against the asphyxiation zone [48], the oxidation heating zone of LW6305 goaf was determined, coupled with the zone edge distance away from LW6304 (Figure 8).

As indicated by Figure 8, the maximum distance of the LW6305 heating zone edge to the LW6304 face has a negative correlation to pillar width. In detail, the maximum distance decreases from 128.1 to 50.4 m as the pillar width increases from 3 to 11 m. Taking the 11 m width condition as a reference, the maximum distance is 2.54 times in the 3 m pillar width condition, gradually declining to 2.25 times (5 m width), 1.90 times (7 m width), and 1.48 times (9 m width). In addition, the air-in ventilation roadway has a generally greater oxygen concentration than the air-out ventilation roadway.

4.3. Goaf Ignition Risk Assessment

Goaf ignition is likely to occur in the oxidation heating zone. In mining practice, the goaf ignition risk is frequently assessed by comparing the longwall retreat distance in a certain period (L0) with the maximum distance of the heating zone to the current coalface position (Smax). In cases of SmaxL0, the previous panel goaf tends to ignite spontaneously; in Smax<L0 cases, the goaf ignition potential is considered lower, meaning a safe environment for longwall mining.

The mine data indicates that the shortest self-ignition period of LW6304 and LW6305 coals should be 21 days, and the retreat rate is 5 m/d. Therefore, LW6304 coalface marches a distance (L0) of about 105 m in the potential goaf ignition period. Further define a goaf ignition risk index Qs as the ratio of Smax to L0. The potential of LW6305 goaf ignition in response to different pillar widths was quantified as summarised in Table 5.

Table 5 indicates that, as the chain pillar width is not greater than 5 m, residual coals in the adjacent goaf tend to ignite, with the goaf ignition risk index (Qs) higher than 1. Widening the pillar to 7 m or larger can decrease Qs to be smaller than 1, a low ignition tendency. Therefore, LW6304 chain pillar width should be more than 5 m considering goaf ignition control.

5.1. The Optimal Width Determination

The above analyses revealed that widening the chain pillar helps decline the coal burst potentials but negatively increases the possibility of goaf ignition, a contradictive requirement of controlling both mine hazards on the pillar width. However, there may exist an optimal pillar width that allows both hazards to be balanced, for which the optimal value can be determined according to the operating point principle. The operating point selection has been widely applied to obtain the optimal working condition of ventilation and water pump facilities [49, 50]. This paper also used the operating point selection for determining the chain pillar width, which includes the following steps:

  • (i)

    Establish the correlation δr=fD between the coal burst risk index δr and pillar width D and correlation Qs=gD between the goaf ignition risk index Qs and D

  • (ii)

    Plot the two correlation curves δr=fD and Qs=gD on a chart

  • (iii)

    Demarcate the threshold regarding δr2.86 (no risk or low coal burst intensity) and Qs1 (no risk or low goaf ignition tendency) and thus obtain a range of pillar width

  • (iv)

    Take the intersection of two curves as the optimal pillar width, as shown in Figure 9 

Figure 9 indicates that both hazards have exponential correlations to pillar width, in detail a negative correlation for coal burst and a positive correlation for goaf ignition. Configuring the threshold suggests a reasonable pillar width from 5.9 to 9.2 m; if smaller than 5.9 m, the pillar has a minor effect on goaf ignition control, while if greater than 9.2 m, the pillar tends to undergo a coal burst problem. The two correlation curves intersect at the point of 6.7 m width, implying that the optimal pillar width should be 6.7 m, rounded to be 7 m for mine operation convenience.

5.2. Field Trial

The LW6304 chain pillar was constructed 7 m wide in practice. When LW6304 coalface approached LW6305 goaf, mine seismicity and gas (CH4 and CO) concentration were monitored along the roadway, as shown in Figure 2. The monitoring lasted for 28 days and collected continuous seismicity and gas concentration variations as shown in Figures 10 and 11, respectively.

The measurement results show that each seismic event had an energy release lower than 1×105 J, practically small energy release events. The events with 0 to 5×103 J energy release dominated, taking 75% of the total amount. No coal burst phenomenon was observed. In the 7 m width condition, the CH4 concentration ranged from 0.28 to 0.46%, not triggering out-of-limit alarming. The CO concentration ranged from 6.22 to 39.78 ppm, with the maximum of 39.78 ppm, far lower than the alarm value of goaf ignition. These results indicate that the 7 m wide pillar can help control the coal burst and goaf ignition risks, verifying the effectiveness of the pillar width determination method.

LTCC panels have coal burst and goaf ignition potentials. Adjusting the chain pillar width can help control both mine hazards, however characterising an opposite influencing mechanism: a narrow pillar prevents excessive energy accumulation and rapid energy release but possibly causes air leakage into goaf voids and hence goaf ignition. Therefore, the pillar width determination should consider the pillar impact on coal burst and goaf ignition.

The pillar strain energy experiences a rapid increase and decrease and ultimately levels off with longwall retreat, behaving significant energy accumulation and release. The pillar width has a positive correlation with the energy accumulation rate, peak energy, and times and amount of energy release. Setting the pillar smaller than 9.2 m can decrease the coal burst risk index lower than the weak coal burst. From the coal burst control, the LW6304 chain pillar should be smaller than 9.2 m. The distance from the LW6305 oxidation heating zone to the LW6304 coalface has a negative correlation with the pillar width. As the pillar width is greater than 5.9 m, the goaf ignition risk index is smaller than the no risk tendency. From the goaf ignition control, the LW6304 chain pillar should be wider than 5.9 m.

The paper develops a pillar width determination method by virtue of operating point selection. The method plots the coal burst risk index (δr) and goaf ignition risk index (Qs) curves against the pillar width on a chart and thus determines a reasonable width range by demarcating δr2.86 and Qs1. The optimal width can be located as the intersection point of both curves δr=fD and Qs=gD. Combined with the on-site construction, the optimal chain pillar width of LW6304 coalface is 7 m. This method has been verified effective in determining the chain pillar width by a field trial.

The datasets analysed during the current study are available from the corresponding author on reasonable request.

The authors declare that they have no conflict of interest.

This work was financially supported by the National Natural Science Foundation of China (grant numbers 52204161 and 51974291), the Fundamental Research Funds for the Central Universities (grant numbers 2022QN1008 and 2021ZDPY0226), the Jiangsu Funding Program for Excellent Postdoctoral Talent (Grant No. 2022ZB511), the Shanxi Province Unveils Bidding Project (grant number 20201101009), the Assistance Program for Future Outstanding Talents of China University of Mining and Technology (grant number 2022WLJCRCZL027), and the Postgraduate Research & Practice Innovation Program of Jiangsu Province (grant number KYCX22_2630).

Exclusive Licensee GeoScienceWorld. Distributed under a Creative Commons Attribution License (CC BY 4.0).