Designing appropriate support for the variable geologic conditions and loading conditions typically encountered in a longwall gateroad is challenging because of the large number of parameters that affect the stability of the excavation. A further complication is the variable duration of longwall-induced loading. In some cases, the supports only need to control the roof for a relatively short period of time. For example, a longwall headgate belt entry may only be subject to the peak loading conditions for less than one day as the longwall face approaches and mines up to the supports. On the other hand, the support system in entries located between the gobs of two longwall panels may be required to control the ground for periods exceeding 2 years, while subject to extreme vertical stress conditions. Support required in a headgate gateroad with short-duration loading is very different from entry support requirements for longer term loading.

At present, gateroad support systems are largely designed based on historical experience in similar geologic and operational settings. Standing support selection can be facilitated by using the NIOSH-developed Support Technology Optimization Program (STOP) [6], which allows various commercially available standing support systems to be compared. The procedures described in this paper represent a further step in the development of gateroad support design procedures that allow designers to simultaneously consider many of the important parameters that affect gateroad stability.

The research in this study was conducted through a combined program of data collection on current gateroad support practices, full-scale testing of gateroad support systems at the NIOSH Mine Roof Simulator; monitoring of longwall-induced ground deformation, stress changes, and support response at longwall operations in the USA; analysis of the results; and development of numerical modeling procedures that allow expected support performance and ground response to be evaluated. The developed numerical models were used to conduct parametric studies in which the important factors affecting entry stability under changing stress were evaluated. The understanding of gateroad stability derived from the results formed the basis of a conceptual model of support requirements in longwall gateroads. Using the conceptual model as a basis, the relationships between important input parameters for gateroad stability were derived through regression analysis of the results of the parametric studies. The derived relationships can be used by support designers to assess support alternatives for local geologic conditions and longwall-induced loads.

In this research, the need for surface control to prevent smaller rock fragments from falling between supports was not evaluated. The focus was on preventing large roof falls that would impede ventilation or safe access to the longwall working face. While the focus of this research was on the impact of longwall-induced stress changes on entry stability, it must be noted that individual geologic structures such as slips or inclined joints and faults in the roof may result in local instability. These structures were not explicitly considered in the analyses nor in the proposed support assessment procedures.

This paper describes the key results of the field-monitoring program and the results of parametric studies that shaped the conceptual model of gateroad stability. The principles of the conceptual model are explained, and calculation procedures are proposed that allow the roof sag and expected support loading to be estimated for given geologic and longwall loading conditions. The results are compared to field-monitoring outcomes, and a case study mine is presented to demonstrate the application of the developed procedures.

The desired set of instruments for each gateroad monitoring site is shown schematically in Fig. 2. It was not possible to install the full set of instruments at all monitoring sites because of practical limitations related to access and timing in high production longwall environments. The instrumentation had the objective to measure stress changes, strata deformations, and support loading induced by longwall extraction. Instrument installation was conducted during development at the time of installation of the supports and was wired-up to dataloggers as soon as through ventilation was achieved.

Three-dimensional stress changes in the roof strata were measured using the CSIRO Hollow Inclusion Cell (HI-cell) [10] typically at locations indicated in Fig. 2. Vertical stress changes in the unmined coal adjacent to the gateroad entry were measured using borehole pressure cells (BPCs) that were originally developed through US Bureau of Mines research and were manufactured locally according to NIOSH specification. The BPCs are fitted with pressure transducers that have an accuracy of 0.25% of range. In this research, BPC results were only used to assess trends in loading, rather than attempting to derive absolute stress-change values.

Roof deformation was measured using multiple point borehole extensometers (MPBX) that were manufactured by NIOSH technicians at the Pittsburgh Mining Research Laboratory. The relative displacement between the anchor and the borehole collar is measured using a linear transducer with 10 cm of travel and resolution of 0.25 mm. Up to eight monitoring wires were installed in the boreholes, with anchor points at selected locations depending on geology and length of roof bolts. Measuring points were located up to 6 m above the entry roof.

Deformation of the coal ribs was measured using multiple point borehole extensometers of various designs. Currently extensometers manufactured by Mine Design Technologies and RST Instruments Ltd. are used that have up to six monitoring points. Deformations are measured using linear potentiometers with up to 500 mm of stroke and ± 2% linearity. Rib deformation anchor points were located between 1.8 and 9.0 m horizontally into the coal bed being mined.

The loading of the cable bolt supports was measured by using cable-bolt load cells (GEOKON Model 3000) installed with 25 mm thick bearing plates at the collar of the cable bolt supports. The load cells use electrical resistance strain gauges with 0.025% full-scale accuracy under laboratory conditions. Four to eight cable bolts were instrumented at each monitoring site.

Standing support loads were measured using a modified Hydrocell (Strata Worldwide) standing support prestressing unit. The unit is a water-filled welded bladder made of expandable sheet-metal that is installed underneath or on top of the support and is equipped with a commercially available pressure transducer that has 1% accuracy. Standing support deformation was measured using various commercially available string potentiometers mounted within about 15 cm of the top and bottom of the standing supports as dictated by local conditions. The potentiometers have an accuracy of 0.25% of full scale.

Instrument calibrations are conducted prior to installation at mine sites. Most instruments such as the HI-cells and cable bolt load cells arrive with manufacturer provided calibration factors, temperature deviation corrections, zero offsets and other instrument specific factors. Calibrations for pressure transducers are conducted internally following ASTM Standard Specifications. The standing support load cells are calibrated using the NIOSH Mine Roof Simulator through the full range of anticipated in-mine loading. The displacement monitoring devices are calibrated to 0.025 mm using a precision caliper.

Dataloggers used for the studies were the Campbell Scientific 21X for the HI-cells and the MIDAS (Golder Associates) originally developed by NIOSH.

Figure 3 summarizes key roof sag monitoring results measured in track and tailgate entries at five of the NIOSH monitored instrumentation sites. It was observed that after development and before the impact of the approaching longwall face becomes evident, the roof is essentially stationary. As the first longwall panel approaches and passes by the monitoring instruments, roof deformations are typically less than 35 mm. The roof sag increases up to about 100 mm at the tailgate corner as the second longwall mines up to the instruments. These deformations are small relative to the yield capacity of standing supports but may approach or exceed the yield limit of cable bolts.

Much larger roof sag measurements have been reported in poor ground or unfavorable stress conditions caused by the orientation of gateroads relative to the major horizontal stress. Lu and Hasenfus [9] measured average roof sag of 150 mm at the headgate corner of a longwall panel that was unfavorably oriented to the major horizontal ground stress. Under these conditions, extreme support measures were required that included installing supplementary 6-m-long cable bolts and injecting polyurethane to stabilize the roof.

In most of the monitored cases, during the periods when the longwall face was remote from the monitoring sites, no roof sag was detected. However, in two cases, mine C and to a lesser extent at mine A in Fig. 3, creep-like roof sag was measured, which appeared to be associated with the presence of clay-rich sediments such as claystone in the roof. These creep-like movements ranged from approximately 1 mm to 10 mm per month.

Since the roof is more than adequately supported during development, gravity-driven bending or sagging of the roof is arrested by the support system. Except for time-dependent, creep-like deformations, the roof deformation observed as the longwall face approaches can be assumed to be driven by stress-induced failure of the roof strata.

Loading and deformation of the fully grouted bolts was not measured because of practical difficulties and regulatory requirements. Cable bolts are usually not fully grouted, and it was possible to measure cable bolt loading using load cells installed between the cable bolt faceplate and the rock face. Initial cable bolt loads during the development stage varied between about 10 kN and 50 kN. As the longwall-induced load increased, cable bolts that were 3.6 m long loaded-up rapidly and achieved their maximum loads after about 75 mm to 100 mm of roof sag. It was found that the cable bolts’ stiffness calculated from field results were lower than those derived from controlled tests. In the field, the measured stiffness of the cable bolts was 2.24 kN/mm compared to 3.50 kN/mm typically measured in controlled tests [11]. The lower stiffness of the cable bolts provides further support for the observation that under small initial roof deformations the contribution of cable bolts will be less than the contribution of the much stiffer fully grouted bolts.

Standing support load and deformation measurements confirmed loading of the standing supports and were used to derive floor heave from the roof-to-floor convergence measurements. This also provided insight into the load transfer from cable bolts to standing supports as the roof deformation increases. Figure 4 shows how longwall-induced roof deformation causes cable bolts to achieve their peak load and start shedding load soon after the longwall face had mined past the monitoring location. As the cable bolts shed load, the weight of detached strata is transferred to the standing supports resulting in a rapid increase in loading of the standing supports. The cable bolts may appear to be redundant under these loading conditions, however, they provide support to the roof from the development stage to the time that standing supports are installed. They also provide interior reinforcement to resist unraveling of the fractured and failed roof strata.

Modeling procedures were developed to investigate both regional stability concerns associated with longwall mining, and local stability issues associated with coal mine entries [2, 13-15]. Models were created at three different scales: the longwall panel scale, the longwall district scale, and the entry scale. Initially the panel-scale models were used to investigate the interactions between the approaching longwall face and gateroad entries. The models were large and took several days to solve for the numerous mining increments needed to capture the longwall induced loading-path of the gateroads. In addition, the panel-scale models could not efficiently model details of the local roof geology and installed supports to conduct meaningful comparative analyses. Consequently, entry-scale models were created that had sufficient detail to model the local roof response and support system interactions. The effect of an approaching longwall on entry stability was simulated by imposing the longwall-induced loading path on the entry model. The loading path was determined either by in-mine monitoring or by the district-scale models.

Model calibration was achieved by developing standardized procedures for estimating the required inputs for the models. Procedures were developed for estimating field-scale rock properties from laboratory tests, for initiating the pre-mining ground stress within the various layers, and for simulating the caved gob behind the longwall face. Through an iterative process, the panel-scale models were calibrated against field monitoring results of abutment stress extent, pillar loading and surface subsidence [2]. For entry-scale models, further calibration was needed because of the change in scale of the models. Model results were calibrated against in-mine extensometer readings and bolt and standing support loading measured at the NIOSH monitoring sites [11, 14, 16]. The main calibration parameters were the rock mass and gob properties that cannot be tested in the laboratory. Support systems are modeled using the strength and deformation parameters as tested in controlled laboratory or in-mine measurements [17]. Grout properties were calibrated against laboratory and in-mine monitoring as described by Tulu et al. [18]. The standardized procedures for creating model inputs were applied to all further numerical modeling studies.

Figure 5 illustrates a panel-scale model demonstrating the model extents and typical excavation details. The model represents two longwall panels with planes of symmetry located along the longitudinal center line of each panel. The model sides were free to move in the vertical direction, while the bottom of the model was fixed. The top of the model is a free surface representing the ground surface. Geologic layers were explicitly modeled as strain-softening ubiquitous joint materials to capture strength anisotropy related to bedding within strata units. Contacts between strata layers were modeled as joint interfaces that follow the Coulomb sliding model, allowing both sliding and separation to occur. Table 2 presents a selection of material types and interfaces ranging from weak to strong, and their mechanical properties as used in the models. The caved gob was modeled as a strain-hardening material extending vertically from the mined floor to three times the mining height and was introduced behind the longwall face as the face was advanced in the models, as described by Tulu et al. [2].

Initial model loading was based on gravity loading in the vertical direction and horizontal stress was initiated using local stress measurement results or according to Mark and Gadde [19] if local measurements were not available. The model was run by first establishing a state of equilibrium prior to mining. The gateroads were excavated next as a single excavation step and solved to equilibrium. The first longwall was then partially excavated as a single step followed by a series of ten mining increments of 5 m each to advance the face past the entries and crosscuts of interest. Gob was introduced in the void created behind the face as mining progressed. For each mining increment the internal unbalanced forces were gradually reduced to avoid shock loading of the system. The process was repeated for the second longwall to investigate gateroad response to tailgate loading conditions. The panel-scale modeling results were evaluated to develop an understanding of the interactions between the gateroad and the longwall, with particular attention being paid to the height of potential roof failure over intersections, crosscuts, and entries. Figure 6 is an example of the results illustrating a close-up view of roof sag in entries and crosscuts near the longwall face.

The district-scale models were created using the same procedures and inputs as described for the panel-scale models except that the district-scale models could simulate the sequential extraction of several longwall panels. Larger element sizes were used to achieve reasonable run times. The stress distributions and overburden deformations were verified against field measurements to ensure that the models were sufficiently accurate at calculating the stress redistribution in the areas of interest around the longwall panels. Figure 7 shows the layout of such a model indicating the model extents and grid meshing on the coal horizon. The longwall-induced stress path in the roof strata of the gateroad entries was determined by extracting the changes in the vertical and horizontal stress components in the plane perpendicular to the entry direction. The calculated stress-path was applied to the detailed entry-scale models.

The entry-scale models simulated a vertical cross section through an entry that has a horizontal thickness equal to a multiple of the support spacing. Typically, the model thickness was 2.4 m. The models used the same constitutive models and modeling approach described earlier. However, in these models, geologic units as thin as 15 cm were modeled, allowing details of the complex geologic layering at many of the case-study sites to be captured adequately. During the calibration studies it was found necessary to account for the weakening effect caused by delamination of bedded rock units in the low confinement zone that exists in the immediate roof of an excavation. This was achieved by considering the “bedding rating” as a parameter to define field-scale strata strength under low confinement [14]. The bedding rating is determined from the Coal Mine Roof Rating (CMRR) [20] and is the sum of the bedding shear strength rating and bedding intensity ratings.

Bolts and cable bolts were modeled using the finite-element structural elements available in FLAC3D. Table 3 presents an example of the mechanical properties of a fully grouted bolt and calibrated resin grout properties used in these models. Standing supports were modeled as passive supports installed after the initial roof sag associated with entry development had occurred. Standing support loads were, therefore, generated by roof deformation induced by the approaching longwall face. The standing supports are simulated by a Coulomb strain-softening/hardening material that follows the load-deformation curve of the full-scale standing supports as tested in the NIOSH mine roof simulator [17].

An example of an entry-scale model layout is shown in Fig. 8. The bottom of the model is fixed, and the sides are free to slide in the vertical direction. Load is applied to the top of the model equal to the overburden load. A plane-strain condition was assumed by inhibiting displacements out of the plane of the section. The modeling sequence is as follows: initialize pre-mining vertical and horizontal stress, run the model to initial equilibrium in the pre-mining state, extract the entry, gradually reduce the unbalanced forces to zero to maintain a quasi-static condition. Bolting is installed when the unbalanced forces have reduced by 70%. The objective is for the bolts to experience deformation that would be associated with development of the entry. Standing supports are installed in the model after the development equilibrium state is achieved. Figure 8 also shows a close-up view of an entry model with bolts, cable bolts, and standing supports installed. After the development stage has been solved, the entry is subject to the desired longwall-induced stress path by applying vertical and horizontal loads to the model in a quasi-static manner. The model is loaded in the vertical direction by applying the longwall-induced loads in small increments along the top of the model while the desired horizontal stress changes are applied by velocity boundaries along the sides of the model.

Figure 9 shows an example of typical entry model results indicating rock mass damage, the resulting rock mass deformation, and support loading. The rock mass damage shown in Fig. 9a illustrates how several rock failure modes are captured by the models and how a dome of failing rock is formed above the entry. Figure 9b shows the same entry at a later stage of loading, where the cable bolts have failed, and the failing roof rock is now resting on the standing supports. The load in the bolts and cable bolts is shown, as well as the stress within the standing supports. The models also captured the formation of roof cutters and subsequent cantilevering of the roof. Figure 9b shows the formation of cutters at the right and left corners of the roof.

This approach for modeling longwall-induced strata deformations and entry stability was used during the model calibration studies and produced realistic ground deformations, extent of strata failure, and support loads when compared to in-mine monitoring results [11, 16]. The standardized modeling approach was used to conduct further investigations and parametric studies of gateroad stability.

The parametric studies considered ten different geologic scenarios. Five of the scenarios were based on actual mine sites where field monitoring was conducted, and five generic scenarios were created to investigate specific geological settings. Table 4 lists the ten geological scenarios and the depth of cover ranges considered.

The numerical models were set to simulate depths of cover ranging from 180 m to 600 m. Not all geologic scenarios were assessed at all depths of cover because the weaker roof cases collapsed under development conditions at the greater depths. The models were subject to a series of stress increments of 2 MPa each to simulate the loading path induced by an approaching longwall panel. The stress increments were selected to provide horizontal-to-vertical stress increment ratios of between 0.33 and 3.0. For each analysis, the stress was increased until the roof sag in the model exceeded 300 mm or the stress increase exceeded 30 MPa, which represents the likely range of loading conditions in the longwall panels monitored.

Support systems consisted of roof rock reinforcement by fully grouted bolts or fully grouted bolts and cable bolts. Supplemental support using various types of standing supports was also modeled. The roof reinforcement consisted of various combinations of 1.8-m and 2.4-m-long fully grouted bolts with peak tensile strength of 180 kN. Cable bolts with lengths of 3.0 m, 3.6 m, and 4.9 m and maximum tensile strength of 270 kN were modeled.

Standing supports were modeled as passive supports installed after the initial roof sag associated with entry development had occurred. Standing support loads were, therefore, generated by roof deformation induced by the approaching longwall face. The load-deformation curve for the standing supports was based on testing results obtained at the NIOSH Mine Roof Simulator and presented in the STOP software. Support types evaluated included: 9-point timber cribs, engineered timber cribs, engineered timber props, and cementitious yieldable cribs of various capacities. The supports modeled represent the typical range of standing support capacities currently used in US longwall gateroads.

The model results indicated that the change in horizontal stress caused by longwall mining has a significant impact on roof stability in the gateroads. This may explain why the majority of longwall panels in the eastern USA are oriented to ensure that the horizontal stress impact is minimized. Figure 10 shows numerical-model-predicted roof sag results for a gateroad at a depth of 240 m in the Pittsburgh coal bed subject to variable longwall-induced horizontal-to-vertical stress ratios. The horizontal stress component perpendicular to the entry direction is considered in these models. The charts show results up to the loading stage just prior to roof collapse. The results demonstrate how higher horizontal stress increments can produce significantly greater roof deformation and consequently increased roof instability. These results help to explain why managing horizontal stress is one of the major considerations when laying out new longwall districts [21].

Quantifying the impact of different types of supports on roof sag and roof stability was a major objective of the parametric studies. Figure 11 shows modeling results for a gateroad in the Pittsburgh coal bed at 240-m depth of cover that is subject to increasing longwall-induced stress. Six different support scenarios were modeled, and the resulting roof sag is shown against the longwall-induced vertical stress. Results for the standing supports are shown up to the loading stage where roof sag exceeds 25 cm, while the bolt and cable bolt results are again shown only up to the loading stage just prior to roof collapse.

The support systems consisted first of four 1.8-m fully grouted bolts in rows 1.2 m apart. Then, two 3.6-m-long cable bolts were added and evaluated. The cable bolts had 1.2 m of grout at the upper end with a free length of 2.4 m. The further analyses all had the same bolt and cable bolt configuration supplemented by different standing support types. The results demonstrate the reduction in roof sag achieved by the various support systems. The improvements when adding cable bolts to fully grouted bolts is clearly demonstrated. Adding standing supports further limits roof sag. Early installation of standing supports will, therefore, improve roof conditions by limiting deformation and dilation of the roof. The analyses showed that the difference in roof sag between various types of standing supports is less obvious. Field observations have also shown a minor difference in roof sag with different types of standing supports [22]. However, most standing support systems are able to control the damaged roof that would otherwise have collapsed. This can only be achieved if the standing supports are able to yield to the longwall-induced deformations while maintaining adequate support load.

The conceptual model was developed on the premise that the support system is required to control the detached roof strata that would collapse in the absence of any support. The conceptual model subdivides the roof strata into three categories: intact strata that are undamaged by mining-induced effects, failing strata that have been loaded beyond their peak strength but have sufficient residual strength to be self-supporting, and (c) detached strata that would collapse in the absence of support. The dead weight of the detached strata, therefore, represents the load that must be controlled by the support system.

Mining-induced roof sag is used as the main indicator of roof stability. For the conceptual model, any roof sag that occurs after initial development is assumed to be caused by the approaching longwall. The longwall-induced loads result in failure of the roof strata, which in turn results in dilation of the rock. The roof dilation forces the roof downwards and is observed as roof sag. The support system will influence the roof sag and height of detached roof depending on time of installation and support capacity. The effect of coal rib failure and floor heave on roof sag are indirectly included in the model results but were not explicitly evaluated. Creeprelated deformations were not considered in this study.

The response of a gateroad entry and support system to the effect of longwall-induced loading is illustrated in Fig. 12. The loading stages and support response may be described as follows:

According to this model, the success of a support system can be assessed if the height of detached roof and the amount of roof deformation are known. The height of detached roof determines the dead weight that must be supported. The deformation determines the survival of the bolts and cable bolts against elongation failure and determines the response generated in standing supports.

An equation to estimate the longwall-induced roof sag at the entry roof line was derived by multiple linear regression analysis of the parametric study results. The effects of the independent variables were initially assessed to establish which parameters have the most significant impact on roof sag. Parameters that were strongly correlated were combined to reduce interactions in the analysis, while parameters with limited impact on the roof sag were eliminated. Any variables that had p-values above 0.05 (indicating limited significance) were eliminated.

The final seven parameters selected for roof sag estimation were as follows, in descending order of importance: (1) longwall-induced horizontal stress, (2) initial horizontal stress in the entry roof, (3) field-scale strength of the roof strata within the fully grouted bolt horizon, (4) field-scale strength of the roof strata in the cable-bolt anchorage horizon, (5) longwall-induced vertical stress, (6) standing support capacity, and (7) reinforcement capacity (bolts and cable bolts). The regression equation to predict roof sag has a coefficient of determination (r²) value of 0.78 with a root mean square error of 12 mm.

The regression analysis showed, not unexpectedly, that the parameters that have the greatest impact on roof sag are the stress state and the roof strength, while the support system has a lesser effect. The limited impact of the support system on the longwall-induced roof sag can be explained by the fact that the equivalent pressure exerted by the support system distributed over the full width of the entry is about two orders of magnitude less than the longwall-induced stress that drives the roof strata failure and associated dilation. In-mine monitoring has also shown that changes in standing support systems appear to have limited impact on roof deformations induced by longwall mining [22]. However, the success of the standing supports depends on their ability to maintain sufficient load bearing capacity in spite of the convergence imposed on them by the deforming strata.

The resulting equation to estimate the roof sag (s) in centimeters at the roof line of a gateroad entry has three components: (1) s=(S1+S2+S3+C)2 where S1 represents the stress state, S2 represents the roof strength, S3 represents the support efficiency, and C is a constant determined from the regression analysis. Each component is calculated as follows: (2) S1=0.1345×Hinc+0.0319×Vinc+0.279×SIGHini (3) S2=−0.117×(RSa+2×RSi3.0)−0.0126×RSc (4) S3=−0.0028×SSc−0.0023×(RRb+RRc) (5) C=1.237 where Hinc is the horizontal stress increase induced by the longwall panel (MPa), Vinc is the vertical stress increase (MPa) induced by the longwall panel, SIGHini is the initial horizontal stress in the roof strata (MPa) in the direction perpendicular to entry development, RSa is the weighted average field-scale strength of strata in the bolt anchorage zone (upper 60 cm of fully grouted bolts) (MPa), RSi is the weighted average field-scale strength of strata in the immediate roof (from roof line to start of bolt anchorage zone) (MPa), and RSc is the weighted average field-scale strength of strata in the cable anchorage zone (upper 1.2 m of cable bolts) (MPa). SSc is a parameter representing standing support efficiency, while RRb and RRc are parameters representing the roof reinforcement provided by bolts and cable bolts, respectively.

The field-scale strength of the different roof strata units is derived from the uniaxial compressive strength (UCS) of the rock matrix and the rating of bedding strength according to the CMRR, as described by Esterhuizen et al. [14, 24].

The effect of the fully grouted bolts (RRb) on roof sag is calculated after Mark et al. [25]: (6) RRb=Nb×Lb×CbWe×Sb where Nb is the number of bolts per row, Lb is the bolt length (m), Cb is the bolt ultimate tensile capacity (kN), We is the width of the entry (m), and Sb is the spacing between rows of bolts (m).

The cable bolt effect (RRc) is also calculated using Eq. 6 but replacing bolt dimensions and capacities with the appropriate values for the cable bolts.

The standing support impact on roof sag is determined from the work capacity of the standing supports. This allows the peak and post-peak yield capacity of the different standing supports to be accounted for. The work capacity (SSw) is calculated as the average work (kN m) per cm of compression when compressing a standing support by 30 cm. Table 5 presents work capacity of some typical standing supports.

The standing support impact (SSc) is calculated as follows: (7) SSc=SSwNs×Ss where SSw is the average standing support work capacity over 30 cm of compression (kN m per cm), Ns is the number of standing supports in each row (typically one or two), and Ss is the center-to-center spacing between rows of standing supports (m).

Changes in the width of entries have a direct impact on roof deformation and stability. Intersections of two or more entries are generally less stable than single entries because larger roof spans are exposed, resulting in increased roof deformation and height of potential roof strata failure [26]. The results of initial modeling and analysis of intersections [27], combined with a review of field-measured data, were used to formulate a procedure to adjust for the effect of entry and intersection width on roof stability. The calculated roof sag (s) is modified to account for changes in the excavation width as follows: (8) sw=0.182×W×s where sw is the width-adjusted roof sag, W is the entry width in meters (not the intersection dimension), and s is the roof sag in centimeters calculated by Eq. 1. The constant value of 0.182 in Eq. 8 will result in an adjustment factor of 1.0 when the entry width is 5.5 m, which was the base entry width used in the parametric studies.

The increased roof sag at intersections is estimated by setting W in Eq. 8 equal to 1.5 times the entry width for four-way intersections and 1.3 for three-way intersections. These increased excavation spans approximate the average diagonal spans at the intersections. These factors may be adjusted to assess the impact of increased intersection spans caused by coal sloughing at the pillar corners or poor mining practices.

To obtain realistic estimates of the height of detached roof, statistics of actual roof falls were reviewed and compared to numerical model results in this study. Figure 13 shows the cumulative distribution of the height of reportable noninjury roof falls in US coal mines [28]. The maximum height of the reported roof falls is approximately equal to the typical 5 to 6-m width of entries found in US coal mines. The figure also shows the height of detached roof predicted by the numerical models conducted as part of this research. In the models, the height of detached roof was defined as the vertical distance from the centerline of the roof to a point where the sag exceeds 2.5 cm. The results show that there is a satisfactory agreement between the model outcomes and reported roof-fall heights.

Further evaluation of the model results showed that the height of detached roof was related to both the strength of the roof strata and the magnitude of roof sag measured at the entry roof line. The roof sag estimate adequately accounts for the effect of longwall-induced stress changes. The results also showed that the maximum height of roof strata failure is approximately the same as the entry width, similar to field observations. Observations and model results seem to indicate that the vertical growth of roof failure becomes limited as the failed zone achieves an arched shape.

An equation to estimate the height of detached roof (Y) in meters was derived from the results of the parametric study models. The equation is based on least squares error fitting an exponential curve to the modeling results for roof sag exceeding 2.5 cm: (9) Y=1.34×SFR×ln(sw)−4.75×ln(SFR×1.34) where sw is the width-adjusted roof sag (cm) using Eq. 8 and SFR is the thickness-weighted average safety factor of the roof strata units within 5 m of the entry roof. The safety factor of each strata unit is calculated as the ratio of the field scale strength of the strata unit to the horizontal component of the stress acting in the direction perpendicular to the long axis of the entry.

The roof sag and height of detached roof can be used to determine the expected loading and deformation of bolts and cable bolts. The approach presented here expresses support performance as a safety factor against failure of the supports. According to the conceptual model, failure by elongation, loss of anchorage, or overloading by the dead weight of the detached roof should be considered. The method of calculation and simplifying assumptions to facilitate these calculations are described below.

Failure by either loss of anchorage or overloading is both affected by the load sharing between bolts and cable bolts. A simplifying assumption was made that bolts and cable bolts share the dead weight of the detached roof in proportion to their axial stiffnesses.

Failure by elongation can occur if stress-driven fracturing and dilation of the roof elongates the bolt and cable bolt support units beyond their tensile strain limit. The safety factor against elongation failure is calculated as the ratio of the elongation limit of the support to the expected amount of roof sag. The study results showed that the typical 1.8-m-long fully grouted bolts are unlikely to fail by elongation because they become encapsulated by the detached roof before they reach critical elongation. Once encapsulated by the detached roof, they will sag within the roof strata without significant increase in elongation. The cable bolts, with their greater length, are more likely to fail by elongation.

Failure by loss of anchorage can occur if the detached roof encroaches on the anchorage zone of the supports and reduces the anchorage capacity. The anchorage zone is assumed to be the upper 60 cm of the solid bar bolts and 1.2 m for cable bolts. The load-carrying capacity of the bolts was assumed to linearly decrease from full capacity to zero as the height of detached roof reaches from 60 cm below to 60 cm above the top of the bolts. This assumption is partially justified by the study of reported roof falls [28], which showed that 69% of reported roof falls extend between 0 and 60 cm above the top of the bolted horizon. This seems to indicate that bolts become ineffective as the detached roof extends more than 60 cm above the bolted horizon. The safety factor against loss of anchorage is calculated as the ratio of anchorage capacity to the share of dead weight load carried by the supports.

If the supports do not fail by elongation or loss of anchorage, they may fail by overloading. The safety factor against overloading is calculated as the ratio of the ultimate tensile strength of the supports to their share of the dead weight load.

In these calculations, it was assumed that standing supports are initially passive and only respond to roof-to-floor convergence that occurs after installation. The longwall-induced roof sag is used to calculate the contribution of the roof to the convergence. The floor heave component of convergence is difficult to estimate because standing supports may punch the floor and would not be subject to the same amount of compression as the observed roof-to-floor convergence. A practical solution may be to assume that floor heave is a fixed percentage of the roof sag, based on knowledge of the floor geology and in-mine observations.

Standing support loads are determined from the estimated convergence and the load-deformation curve of the support as tested at the NIOSH Mine Roof Simulator and presented in the Support Technology Optimization Program (STOP) [6]. The safety factor of the standing supports is calculated as the ratio of the current capacity of the supports to the dead weight load of the detached roof rocks minus the portion of the dead weight borne by the roof reinforcement system (bolts and cable bolts). For supports with continuously increasing load capacity, such as 9-point timber cribs, the capacity is based on 15 cm of convergence.

In accordance with the conceptual model, the stability of the roof depends on the ability of the support system to prevent the collapse of the detached roof strata. Therefore, the roof stability is expressed as a safety factor, which is the ratio of the support capacity to the dead weight of the potential detached roof.

Gateroads are usually supported using different support units, each with its own capacity. The maximum safety factor of the support units is used to represent the overall stability of the roof. The safety factor can be converted to an equivalent “roof stability rating” or similar parameter for ease of communication as presented in Esterhuizen et al. [23].

The ability of the presented calculation procedures to estimate roof sag is compared against field measurements at three NIOSH monitoring sites and one published case history that represent a variety of roof strengths and depths of cover. For the calculation of roof sag, the longwall-induced horizontal and vertical stress changes were estimated from numerical model results following the procedures outlined in this paper. The calculated results and field-measured roof sag are presented in Fig. 14, showing the effect of first panel and second panel mining. The results demonstrate that the calculation procedures adequately predict the large displacements associated with unfavorable horizontal stress and weak roof geology at a depth of 240 m measured by Lu and Hasenfus [9] at Mine J. The same procedures are able to predict relatively minor displacements in a gateroad entry at mine B at a depth of 600 m in strong roof strata. The correlation between the calculated and measured roof sag for initial development, first panel loading, and second panel loading is shown in Fig. 15.

At mine A, the monitoring instruments survived inby the face, and it was possible to measure load shedding of the cable bolts and transfer of load to the standing supports as the roof sag approached 100 mm [11]. Figure 4 shows the in-mine monitoring results for this case. The geotechnical and support data were used to calculate the expected roof sag in the tailgate as the second panel mined up to the entry. The roof sag for points inby the face is not predicted by the calculation procedures. For the inby points, the field-measured roof sag was used to demonstrate how the equations predict the support response. Results are presented in Fig. 16, which compares well to the field-measured support response presented in Fig. 4.

The ability of the calculation procedures to provide realistic estimates of overall roof stability is demonstrated by assessing a headgate roof fall at Mine F in the Pittsburgh coal bed. The gateroad consisted of three entries that were each 4.9-m wide. The overall width of the gateroad was 65 m consisting of 30-m-wide and 35-m-wide pillars, with the larger pillar against the tailgate entry.