The past decade has provided a wealth of practical experience with longwall pillar design. Design formulas developed specifically for longwalls have proven their value in numerous applications. Analytic methods, including numerical models, have also made important contributions. These advances have been based on basic research into abutment loads, coal pillar mechanics, and the interaction between pillar performance and entry stability.

The goal of this paper is to review the state of the art in longwall pillar design from the standpoint of the practitioner. Actual case histories are presented to illustrate the current capabilities and limitations of pillar design methods. The paper also discusses the advances that can be expected in the future.

Pillar design has long been a central issue for mining research. In the United States, pillar design formulas were first developed for anthracite mining in the first decades of this century (Zern, 1928). Pillar design again became the focus of attention during the mining research “renaissance” of the 1970’s and early 1980’s. Familiar pillar design formulas were updated in the light of new research (Hustrulid, 1976; Bieniawski, 1984), and powerful analytic and numerical approaches began to be brought to bear.

The rapid growth of longwall mining during the past decade brought a new urgency to the subject of pillar design. Longwall pillars perform the essential function of protecting the gate entries which provide the only access to the face. Blockages in the gates disrupt ventilation patterns and close off emergency travelways, posing serious safety hazards. In addition, the impact of any downtime is magnified on a longwall because such a high proportion of the total mine productive capacity is concentrated at the longwall face.

Developing pillar design methods for longwalls required some new thinking. The goal of entry protection differs from the traditional function of pillar design, which is to prevent “squeezes” caused by widespread pillar failure. The long and narrow geometry of typical gate entry layouts makes a classic “squeeze” highly unlikely. Second, the abutment loads applied to longwall pillars are significantly greater and more complex than those assumed by traditional pillar design methods. Effective longwall pillar design requires some knowledge of abutment loads during all phases of longwall mining, from development all the way into the tailgate.

Considerable progress has been achieved in all these areas. Back-analysis of design performance has established the link between pillar design and gate entry stability, and extensive field measurements have provided insight into the magnitude of abutment loads. As a result tested formulas are now available to assist mine planners in sizing longwall pillars. Analytic and numerical methods have also taken great strides. These advances in design methodology have been accompanied by an awareness of the limitations that still remain to be overcome.

The Power of Longwall Pillar Design

Ten years ago, “trial and error” was the only “method” available for sizing longwall pillars. As many mines first introduced longwalls, they first tried pillar dimensions that had been successful in their previous room-and-pillar developments. All too often the results were unfortunate (Artier, 1985; Gauna, 1988; Brass, 1981; Dangerfield, 1981). Then pillar sizes would be increased over the course of several panels until satisfactory conditions were achieved. Other longwalls experienced difficulties only on some panels, and were often at a loss to explain them.

Still others settled early on a successful design, and then continued to use it without regard to whether the pillars might be wastefully oversized.

Pillar design methods have been able to bring some order into the design process. They reduce a multitude of variables, including the depth of cover, pillar widths, seam height, entry width, panel dimensions, and so forth, into a single, meaningful design parameter a “stability factor”. The stability factor, defined as the ratio between the estimated load-bearing capacity of the longwall pillar system and the estimated design loading, can be used in at least four different applications:

– Preliminary design. Few things can cause more distress to a new longwall operator than to have an otherwise highly successful longwall slowed by serious ground control problems on its first or (more often) second panel. Pillar design formulas can give realistic, conservative estimates of pillar sizes for use in feasibility studies or initial longwall panel design. Two recent, successful examples of preliminary design with pillar formulas were reported at Kerr- McGee’s Galatia Mine in Illinois (Doney, 1990) and at Wolf Creek Collieries in Kentucky (Cole et al, 1990).

– Adaptation. No two longwall panels are alike, and as conditions change pillar design formulas help mine planners anticipate rather than react to potential problems. Once a specific stability factor has been shown to be adequate for a given mine, then different designs employing the same stability factor can be employed elsewhere on the property with some confidence. In other words, pillar design formulas can be “calibrated” with site-specific experience.

The depth of cover is the single most important parameter affecting gate entry performance, and in many parts of the coalfields it may change radically from one panel to the next. For example, a Kentucky longwall extracted its first three panels under 1000 feet of cover with excellent conditions. The pillar design employed a 28 and a 100 ft pillar, yielding a stability factor of 1.0. When the same design was used at 1800 ft, the stability factor was reduced to 0.5, and ground conditions were so poor that supplemental support had to be tripled just to maintain a travelway. Pillar design formulas indicated that the width of the large pillar should have been increased to 150 ft to maintain the original stability factor under the greater depth of cover (Barton and Mark, 1989). At a nearby Virginia longwall, pillar widths were increased from 90 ft to 120 ft as the cover has increased from 850 to 1300 ft, and no significant degradation of ground conditions occurred.

Man-made conditions may change as well as natural ones. A common change is a switch from a four- to a three-entry system, or vice-versa. Other mines have kept the same number of entries, but gone from equal to unequal entry centers. Pillar design formulas can show how these changes can be made without beginning all over again “at the bottom of the learning curve”.

– Optimization. Longwall pillars that are too large constitute a needlessly wasted resource. An evaluation of survey data collected from 70 longwalls in the early 1980’s indicated that the stability factor for tailgate loading was greater than 1.5 approximately 50% of the time (Mark, 1990). In some cases it was as high as 4.0! The same data also indicated that little improvement in gate entry performance is achieved once the stability factor exceeds 1.5. Design methods can be used to safely reduce oversized pillars, thereby improving development drivage rates and adding to the life of the mine.

– Failure Analysis. Several factors may be responsible for the failure of a gate entry, and it is not always obvious what to do about it. Pillar design formulas can quickly indicate whether increasing pillar size could be the solution. In the case of the Kentucky longwall described earlier, the increased depth of cover was clearly identified as the problem, and increasing the pillar size could be expected to help. Many similar cases can be cited, from the rolling hills of southwestern Pennsylvania to the dipping coalbeds of Wyoming.

Increasing the size of the pillars will not always solve the problem, however. A relatively large stability factor can be a clear signal that some other solution is required. For example, hazardous roof conditions developed along the length of one tailgate in Alabama despite a stability factor of 1.4. In this case the problem was apparently low-quality roof support. Once “select” hardwood cribs were substituted for second quality “mixed hardwood”, the improvement was sudden and dramatic. In another instance, steel beams had to be used in the tailgate of a Pennsylvania longwall because of roof conditions. The panel was a first panel, however, and its stability factor of 1.7 was much higher than had been safely employed on previous panels. Severe horizontal stress, due to the presence of an overlying stream valley, was identified as the source of the problem.

These are the strengths of longwall pillar design formulas. They have their limitations too. Chief among these are the difficulties with evaluating abutment loading and pillar strength on a site specific basis. Perhaps more importantly, current longwall pillar design methods do not yet explicitly address all aspects of gate entry design. These issues will be discussed below, after the methods are presented.

Longwall Pillar Design Methods

Two basic philosophies are available to guide longwall pillar design. The conventional design approach uses large pillars that are sized to carry all the abutment loads to which they will be subjected. The yielding design philosophy, on the other hand, uses only very small pillars which transfer load. At present the vast majority of U.S. longwalls use some form of conventional pillar design, while only a handful of longwalls (all in Utah) employ yielding pillar designs. Because of its wider application conventional pillar design has been the focus of most research.

Longwall pillar design methods may also be classified according to the form that they take. Design formulas consist of relatively simple equations whose output is a stability factor or a required pillar width. Although they represent highly simplified models of the actual problem, they are extremely useful in practical design. Numerical models are far more complex. When properly designed and calibrated, they can be effectively used on panel design, and they can provide valuable insight into ground control problems not strictly related to pillar stability. Far greater efforts are required to use them, however, so they are often less attractive than the “user-friendly” design formulas.

Longwall Pillar Design Formulas

Four longwall pillar design formulas were recently described in a Bureau of Mines Information Circular (Mark, 1990). These are:

– Analysis of Longwall Pillar Stability (ALPS): ALPS was originally developed at the Pennsylvania State University by Mark and Bieniawski (1986), and was later refined at the Bureau of Mines. ALPS is now based on field measurements conducted at 16 longwall panels, and it has been verified by back-analysis of more than 100 case histories.

– Carr and Wilson’s Method: A. H. Wilson first developed a method for sizing rib and barrier pillars in British mines during the early 1970’s. In 1982, Carr and Wilson modified the original theory to apply to the multientry longwall developments used in the U.S. Carr and Wilson’s method has been used extensively by Jim Walters Resources at their Alabama longwalls,

– Choi and McCain’s Method: Choi and McCain’s method, presented in 1980, was the first design formula developed specifically for the United States. The method was based on field studies, numerical modeling, and observations from Consolidation Coal Co.’s working longwalls in the Pittsburgh seam.

– Hsuing and Peng’s Method: Hsuing and Peng’s method is unique in that it incorporates some properties of the roof and floor into the pillar design. It was developed from three-dimensional finite element modeling performed at West Virginia University, and is described in the second edition of the book “Coal Mine Ground Control” (Peng, 1986a).

All four of these methods are for conventional pillar design. They each consist of the same three basic elements:

– Estimation of the load applied to the pillar system.
– Estimation of the load-bearing capacity of the pillar system,
– A design criterion, usually a recommended “stability factor.”

The methods differ in how they approach each element.

In the first three methods (all but Hsuing and Peng’s), the side abutment load is estimated using an approach first developed by King and Whittaker (1971), as shown in figure 1. The only difference between the methods is their choice of the abutment angle β. ALPS uses β = 21, while the other methods use slightly lower values. For “super critical” panels (figure 1a), the Carr and Wilson’s estimate of the side abutment load is 22% lower than ALPS, and Choi and McCain’s is 15% lower than ALPS. The differences are even less for “subcritical panels” (figure 1b). Both ALPS and Carr and Wilson’s method can also be used to evaluate the distribution of the side abutment load (figure 2).

The critical design loading experienced at a face end (the headgate or tailgate corner) is actually a three-dimensional front abutment, not a two-dimensional side abutment. In ALPS, field measurements were used to develop formulas that estimate the front abutments as a percent of the side abutment. Hsuing and Peng determined the tailgate loading directly from their models.

Two fundamentally different approaches are used for estimating pillar strength and load bearing capacity. The analytic approach, which is derived from the principles of mechanics and which places great emphasis on the triaxial strength of the coal, is used by Carr and Wilson and by Hsuing and Peng. ALPS and Choi and McCain’s method employ empirical pillar strength formulas, which were derived primarily from uniaxial testing of coal specimens.

For design criteria, both ALPS and Choi and McCain’s methods calculate the stability factor as the load-bearing capacity of the longwall pillar system divided by the total design load. The most recent formulation of Carr and Wilson’s method also uses the stability factor approach (Carr and Martin, 1986). These methods can also be used to predict a recommended pillar width once an appropriate stability factor has been set. In the original Carr and Wilson method, the design criterion is the load applied to the pillar closest to the tailgate. Hsuing and Peng’s method is the least flexible with regard to design criteria, and only recommends a pillar size.

Some of the methods also have limitations on entry layouts and pillar geometries. Hsuing and Peng’s method was developed for an eight-ft seam and a three-entry system employing equal-sized pillars. Choi and McCain’s method is also for a three-entry system, but one with a 32-ft “yield” pillar and a large abutment pillar.

The characteristics of the four longwall pillar design formulas are summarized in Table 1.

The predictions of the four formulas are compared in figure 3. The “base case” analyzed in figure 3 is a 6-ft coal seam under 1000 ft of cover, with a 675-ft wide longwall panel. Two three-entry configurations are analyzed, one in which the pillars are equal-sized, and the other in which a 32-ft “yield” pillar is paired with an abutment pillar. The other input parameters and the design criteria used with each formula are described in Mark (1990).

Figure 3 shows that all of the design formulas follow the same general trends. They agree that the depth of cover has a great effect on the required pillar size. In the example, as the cover is increased from 500 to 2000 ft, the formulas indicate that the pillar widths should be increased by factors ranging from 2.5 to 6.0. Most of the formulas also indicate that seam height has a great effect on the required pillar width. Such factors as the panel width, the crosscut spacing, and the entry width are also important, but to a lesser extent.

The Bureau of Mines has developed a large data base of case histories to evaluate the effectiveness of longwall pillar design formulas. In the past year alone, 25 new cases have been added, including nine from western longwalls. The new cases were obtained from Utah, Colorado, Wyoming, Pennsylvania, West Virginia, Alabama, Maryland, and Virginia.

These case histories have been separated into categories of “successful” and “unsuccessful” designs. The unsuccessful designs were those in which intolerable entry conditions occurred, including roof deterioration and falls, severe pillar sloughing and floor heave, and in several cases even pillar bumps. In all cases the problems could be attributed to excessive abutment stresses. The successful designs were those which have been used for many years and in which only minimal ground control problems were reported.

Figure 4 shows the ALPS stability factors calculated for the 80 cases in the data base. Only five unsuccessful designs had a stability factor greater than 1.0, and none exceeded 1.20. On the other hand, nine successful designs had stability factors of less than 1.0. Based on these observations, stability factors of 1.0-1.3 are currently recommended for sizing pillars for tailgate protection with ALPS. The other three longwall pillar design formulas are also capable of segregating the successful cases from the unsuccessful (Mark, 1990).

There is some variability in the required stability factor from mine to mine, as shown in figure 4. For example, longwalls in Virginia and Colorado have successfully employed pillar designs with stability factors of approximately 0.6. At the same time, the stability factor has exceeded 1.0 for unsuccessful designs in Virginia, Wyoming, Pennsylvania, and West Virginia. Identifying the source of this variability, and developing improved design formulas that reduce it, is the central task of current longwall ground control research.

Numerical Models for Longwall Pillar Design

Numerical models constitute the other class of longwall pillar design methods. Peng (1986b) lists the three most important advantages of numerical models over empirical formulas as:

– They consider the stiffness of the individual rock layers in the overburden, providing more accurate estimates of pillar loads;
– They consider the effects of roof and floor strata on the strength of the pillars, and;
– They allow more sophisticated failure criteria to be incorporated.

Nevertheless, the use of numerical models in coal mine ground control remains more of an art than a science. Their utility depends largely on the structural analysis background of the practitioner, the limiting assumptions of the specific numerical technique, and the “engineering judgement” applied in setting up the model. Numerical modeling techniques are seldom used for determining absolute stress magnitudes, but rather to establish general trends and relationships between design variables. They have also been very valuable for exploring theoretical concepts of coal pillar mechanics.

Two classes of numerical models have been used extensively for analysis of longwall pillars. These are the finite element and the boundary element methods. The key difference between the two methods is shown conceptually in figure 5 (Kripakov, 1987). The finite element method requires the region of interest to be divided into a network of elements (figure 5a). In the modeling procedure all of the elements in the network interact, and the stresses and displacements are determined throughout the structure. The most recent finite element codes can also include many different types of inelastic, “non-linear” behavior such as cracking, sliding, or creep. The key disadvantage of the finite element method is that it can be very demanding of computer time. Researchers at West Virginia University have made extensive use of finite element models to evaluate longwall pillar performance (Hsuing and Peng, 1985; Hsuing and Peng, 1987; Su and Peng, 1987). Recently, Park and Gall (1989) used a supercomputer for a longwall model that consisted of 7434 mesh elements and featured progressive failure simulation using the Hoek-Brown criterion.

The boundary element method, shown in figure 5b, requires only that the boundaries of the mine openings be divided into elements (discretized). Loadings may then be applied to the boundaries, while the model treats the intact rock as an elastic continuum. The result is that much less computer time is needed to solve the problem. A variation of the boundary element method is the displacement discontinuity method. In a displacement discontinuity model, the seam is represented as a two dimensional thin slit whose properties can be varied from point to point. The surrounding rock mass is still represented as an elastic continuum, but additional seams, bedding planes, or faults can also be represented. The capabilities of displacement discontinuity models have recently been improved with the addition of non-linear coal and gob elements (Zipf and Heasley, 1990). Displacement discontinuity models have been particularly useful for analyzing the complex load distributions occurring with multiple-seam mining (Kripakov et al., 1986; Malecki et al., 1986).

Limitations of The State-Of-The-Art

It is a relatively easy task to generate an impressive list of limitations to our current longwall pillar design methods. At first

glance it might seem that, taken together, these flaws must invalidate the entire concept. Yet if our current formulas did not already address many of the most important aspects in the design problem they could not have the success rate demonstrated by figure 4. The goal is to “fine tune” the formulas to reduce the variability evident in figure 4. From a research standpoint, it is essential to focus on those issues which promise to make the greatest practical improvements, rather than those of a more purely scientific interest.

Abutment Load Estimation

The formation of abutment loads is an extremely complex process. Whittaker and King’s concept (figure 1) is based on geometry alone, ignoring geology. Subsidence research suggests that the geology-specific behavior of the overburden can have a large effect on surface movements (Karmis et al., 1983), and abutment loads are surely an underground expression of the same phenomena. For example, figure 6 shows stratigraphic columns from two neighboring Virginia longwalls. At the first mine the overburden consists largely of a few massive sandstone units, and we might intuitively expect less-complete caving, more bridging across the gob, and a larger effective “abutment angle”. At the second mine, where the overburden contains few massive units, caving might be more complete, resulting in more gob loading and a smaller “abutment angle.”

The actual range of abutment angles that have been measured in the field is approximately 10 to 25 degrees (Mark, 1990). The influence of the assumed abutment angle is greatest for “supercritical” panels. For very “subcritical” panels, where the depth of cover is more than twice the panel width, variations in β have little effect. An important assumption is that the pillars need only carry the load to the midpoint of the panel, an assumption which may not be valid if the previous pillars have failed.

How significant might this variability be in practice? Current Bureau studies are shedding light on the effects of “extremes” of geology and panel geometry on abutment load estimation. Measurements of pillar stresses and the loads in the gob are being made at a site representative of deep-cover conditions with very strong overburden. Preliminary analysis indicates that even here the assumptions used in ALPS yield reasonably accurate estimates of the abutment load. Recent analytical studies have also addressed the influence of overburden behavior on abutment loading (Salamon, 1991).

Numerical models also have difficulties with abutment load estimation. Finite element models should do the best job, because they can simulate multiple layers and progressive failure of the overburden. But both finite element and displacement discontinuity models are hindered by the need for accurate, field scale, rock properties for the overburden and caved material (gob).

Determining the load carried by the waste, which is in turn determined by the dimensions and stiffness of the gob, is a key concern. Little data on gob characteristics is available. As a result, different modelers have used values for the modulus of the gob that range from 20,000 to 160,000 psi (Maleki, 1990; Kripakov et al., 1988; Hsuing and Peng, 1985). Other researchers have found that changing the gob modulus can change the predicted peak pillar stress by a factor of three (Hackett and Park, 1987). To help resolve this issue, the Bureau is currently testing simulated caved material in the lab (figure 7). Preliminary results indicate that the secant (average) modulus is less than 5000 psi for low-strength gob material subjected to a maximum load of 2000 psi. It also appears that the modulus is an approximately linear function of the applied stress over this load range.

Coal Pillar Mechanics

Current methods for estimating the load bearing capacity of coal pillars are also imperfect. Determination of field-scale material properties remains a problem. Failure mechanics, and constitutive relations in general, raise even more fundamental concerns. Research has barely begun to address the effect of time on pillar strength. In practice it appears that the effects of many of these theoretical shortcomings can be overcome, however.

Modern empirical pillar strength formulas (such as the Bieniawski formula that is used in ALPS) require only a single material property, the “in situ coal strength”. The in situ coal strength is defined as the uniaxial compressive strength of coal cube measuring at least three feet on each side. Because such large-scale testing would be prohibitively expensive, the traditional method for determining the in situ coal strength has been to apply a “size-effect scaling factor” to the strength of small coal specimens tested in the lab (Bieniawski, 1984). In the author’s opinion, two serious drawbacks largely invalidate the laboratory test approach. First, it is very difficult to determine a meaningful specimen-size coal strength. Laboratory tests on coal are notoriously unreliable because of sampling bias, integrity loss during specimen preparation, and platen effects during testing. For example, a Kentucky operator recently had three different consultants carry out

compressive strength tests on coal specimens as part of a longwall feasibility study. The average values reported back were 2700, 5000, and 8500 psi a range approximately as great as that reported for all U.S. coal seams put together (Singh, 1981),

Better sampling and testing procedures can improve consistency, but leave the second problem of the size effect. There is now considerable evidence that the size-effect scaling factor is not a constant but varies considerably from seam to seam. As figure 8 shows, two-inch specimens from the Pittsburgh seam are approximately twice as strong as similar specimens from the Pocahontas No. 3.

The disparity is rapidly reduced as the specimen size is increased, because of the different size effects in the two seams. Fortunately, experience has shown that site-specific coal strength data is seldom necessary for longwall pillar design. For several years this author has consistently assumed the in situ coal strength used in ALPS is 900 psi. There has been no indication that the variability in required stability factor (figure 4) can be explained by variations in coal strength. It may be that while the laboratory strength varies considerably from seam to seam, the in situ strength is more consistent, as suggested by figure 8. Barton and Bandis (1982) observed a similar scale effect on the strength of rock joints.

It is also possible that the uniaxial compressive strength is less important when considering the strength of the “squat” (large width-to-height ratio) pillars typical of longwalls. It is widely accepted that the strength of squat pillars is largely derived from the confinement that is generated within them. Analytic and numerical models therefore focus on the triaxial strength of coal. Although it has been subject to less study, it would seem that the triaxial strength would be subject to the same reliability and scaling problems as the uniaxial strength. More fundamentally, the behavior of coal under very high confining pressures, like those occurring in the core of a “squat” pillar (a pillar with a large width-to-height ratio), is not well understood.

Figure 9 illustrates how current theories predict the strength of coal pillars. It compares five empirical formulas, three analytic formulas, and two numerical model studies using a pillar with a width-to-height ratio of five as the baseline. The figure shows that while the most of the theories predict similar strengths for slender pillars, they diverge widely in their predictions of the strength of very wide pillars.

The theories compared in figure 9 can be divided into three categories. The first group, which includes the numerical models of Hsuing-Peng and Kripakov, Wilson’s formula, and Salamon’s “squat” formula, indicate that the pillar strength increase exponentially as the pillar width is increased. These theories all assume that confinement grows rapidly within squat pillars, allowing the core stress to build indefinitely. A second group of theories, comprised of the Salamon-Munro and Holland-Gaddy empirical formulas and Barron’s analytic formula, indicate the core stress, and the pillar strength, tend towards some maximum, limiting value. A final group of empirical formulas, those of Bieniawski, Sheory, and Obert-Duvall, take a middle road, implying a linear increase in pillar strength.

Several recent developments may prove useful in resolving these issues. New theoretical concepts proposed by Barron (1984) have raised the issue of the brittle-ductile transition and its applicability to coal pillars. Mark et al. (1988) developed techniques for integrating empirical and analytical coal strength formulas, and for indirectly determining in situ triaxial strength parameters from field tests. Numerical modeling and field observations reported by Iannacchione (1990) indicate that the strength of many pillars may be determined by the low-strength frictional contact between the pillar and the roof or floor, rather than any characteristic of the coal itself. Finally, measurements have recently been conducted in the cores of large, highly stressed longwall pillars (Campoli and Barton, 1990) and these could provide the information needed to develop and verify constitutive relations for highly-stressed coal.

One final aspect of pillar mechanics that has immediate practical implications for pillar design is the effect of time. It is well known that the strength of rock decreases the longer the load is applied. In longwalls, the maximum design loading at the tailgate is applied for a relatively short time. As stated earlier, the design criterion for this application is ALPS stability factor in the 1.0-1.3 range. It should be expected that long-term applications, such as bleeders or mains, should have other criteria. At one gassy Pittsburgh seam longwall four separate bleeder systems had been blocked by roof falls. Analysis indicated that the problems occurred when the stability factor dipped below 1.5. It appears that such long-term applications may require the stability factors originally suggested by Bieniawski, or 1.5-2.0. The effect of time also becomes more important when designing for “super” longwall panels, because of the longer period of time that the gates will be in use.

Gate Entry Stability

Abutment loads and pillar mechanics relate directly to longwall pillar stability. But for longwalls the bottom line is protecting the gate entries. There is an undeniable empirical correlation between gate entry stability and pillar design, as established by the case histories represented in figure 4. But there is no reason to believe that pillar size is the only factor determining gate entry performance. This author’s experience is that two other factors must be considered in a successful gate entry design. These are roof rock qualitv and artificial support/entry width.

The following example illustrates the relationship between these three elements. Geotechnical surveys were conducted underground at three longwalls operated by a major coal company. The data was summarized as follows:

Mine X

Depth of cover: 650 ft
Pillar design: Three entries on 55 and 70 ft centers
Roof rock: Strong, massive siltstone and sandstone
Primary tailgate roof support: 6-ft resin bolts on 5-ft centers
Secondary tailgate roof support: Three rows of wood cribs

Mine Y

Depth of cover: 1000 ft
Pillar design: Three entries on 90 and 60 ft centers
Roof rock: 4-ft of laminated shale overlain by siltstone
Primary tailgate roof support: 5-ft point- anchor resin bolts on 5-ft centers
Secondary tailgate roof support: Two rows of wood cribs

Mine Z

Depth of cover: 900 ft
Pillar design: Two entries on 100 ft centers
Roof rock: Weak, slickensided mudstones and laminated shales
Primary tailgate roof support: 6-ft point- anchor resin bolts and trusses
Secondary tailgate roof support: One row of yielding steel posts

From this information the ALPS stability factors were calculated, and ratings were applied to the roof quality and tailgate roof support as follows:

Where: “A” = Much greater than average
“B” = Greater than average
“C” = Average
“D” = Less than average
“E” = Much less than average

Based on pillar design alone, Mine X and Mine Y both seemed like candidates for tailgate failure. When all factors were considered, however, Mine Z was actually at greater risk than Mine X. Serious tailgate roof control problems later did occur at both Mines Y and Z, while Mine X was untroubled.

This author believes that it is these factors, roof quality and artificial support, that explain the lion’s share of the variability in figure 4. Current research efforts are therefore focussed on developing procedures for including them directly in the design process. The first step is to establish quantitative relationships between the three variables. Geotechnical surveys have now been conducted at nearly 50 longwalls, and the data that has been collected is being evaluated using multivariate statistical techniques. The goal is to develop guidelines for selecting the appropriate ALPS stability factor and the artificial support based on site specific geologic evaluation.

At present, these factors can be indirectly incorporated into the design process by the technique of “calibration” that was described earlier. The site-specific stability factors determined through back-calculation largely reflect the influence of roof quality and support. So long as neither of these factors changes significantly, the stability factors can be used as the design criteria for future panels.

Yielding Pillar Design

Although 95% of U.S. longwalls employ some form of conventional pillar design, interest in yield pillar design remains strong. The reason is that conventional pillar designs can have some important disadvantages in deep-cover, multiple seam, and/or bump prone conditions. These conditions are rare today, but they will be increasingly prevalent in the future.

The greatest disadvantage of conventional pillar designs in deep cover is that the required pillar size can become extremely large. In the Warrior basin, where the cover approaches 2200 ft, pillars measuring 220 by 300 ft are required just to maintain a stability factor of 1.0. Such large pillars are quite expensive, due to the value of the coal that they contain and the additional time that is necessary to develop them. Yield pillars under similar depths might be one-tenth as wide. Conventional pillars have further disadvantages in multiple-seam environments because they can cause destructive stress concentrations in overlying and underlying seams. Yield pillars are incapable of storing stress, offering significant advantages under these extreme conditions.

The development of design guidelines for yielding pillar systems still presents considerable challenges. Yielding pillar systems require the development of a destressed zone within a pressure arch. Therefore the load transfer mechanics of the roof above the pillars are of immediate concern. These will be largely determined by the geology of the roof. The post-failure load-deformation relations for the pillars themselves are also essential. The importance of the time of pillar yielding, whether during development or during longwall mining, also needs to be determined. It will also be necessary to determine whether certain roof types are more compatible with yielding pillar designs than others, and what the requirements are for artificial support. The span of the yielding pillar system, which is largely determined by the number of entries, will be the critical design parameter.

Conclusions

The longwall method is currently the safest and most productive for mining coal underground. For a longwall face to reach its full potential, stable gate entries must be maintained. Effective pillar design is often the single most important step a longwall operator can take to protect the gate entries.

During the last ten years powerful new pillar design methods have been developed specifically for longwalls, and these methods have removed much of the uncertainty which used to surround gate entry performance.

Current research continues to advance the art and science of longwall pillar design. The next big step for empirical formulas will be to quantify the effects of roof quality and artificial support. The goal is to develop a design methodology for longwall gate entries, of which pillar design will be a critical part.

Fundamental research into abutment load formation and pillar mechanics will help improve the reliability of numerical models. Key factors that are currently being investigated include gob characteristics, the effects of geology, depth, and panel geometry on abutment loading, the strength of squat coal pillars, and the influence of time on pillar strength. Finally, research into the mechanics of yield pillar systems will prepare for the difficult deep-cover, multiple seam, and bump-prone conditions that will become increasingly common in tomorrow’s mines.