International Journal of Rock Mechanics and Mining Sciences 123 (2019) 104102

Contents lists available at ScienceDirect

International Journal of
Rock Mechanics and Mining Sciences
journal homepage: www.elsevier.com/locate/ijrmms

New coal pillar strength formulae considering the effect of interface friction T
Simon Heru Prassetyo , Muhammad Alen Irnawan, Ganda Marihot Simangunsong,
∗

Ridho Kresna Wattimena, Irwandy Arif, Made Astawa Rai

Mining Engineering Program, Institut Teknologi Bandung, Indonesia

ARTICLE INFO ABSTRACT

Keywords: Coal pillars perform the vital function of sustaining the weight of the overburden and protecting the entries and
Coal pillar strength formulae crosscuts during mine development and production. The strength of coal pillars is known to be greatly influenced
Width-to-height ratio by the friction of coal/roof and coal/floor interfaces. Unfortunately, none of the current empirical formulae for
Interface friction pillar strength has considered the effect of this so-called interface friction. This paper develops new coal pillar
Safety factor
strength formulae considering the effect of interface friction. The formulae, which are in linear and power forms,
are derived based on a series of UCS tests on coal specimens at three different interfaces (high: c = 124 kPa and
μ = 0.40; medium: c = 76 kPa and μ = 0.22; and low: c = 55 kPa and μ = 0.13) and at ten width-to-height (w/h)
ratios (w/h = 2–8, 10, 12, and 16). Having compared the formulae against stable and unstable pillar cases from
different areas around the world, this paper finds that: (1) the low-friction formula of the linear form
(
Sp = 2.7 0.12 + 0.88 h
w
) provides the best representation of the pillar cases database; (2) to improve pillar
strength prediction, a safety factor = 1.5 is suggested to be used for all pillar designs using the linear low-friction
formula (except for USA coal pillars with w/h ≤ 6 that need a higher SF up to SF = 2.5); (3) when the friction of
the coal/roof and coal/floor interfaces is not known, it is beneficial to assume that the interface has a low friction
value and thus to use the linear-low friction formula to predict pillar strength.

1. Introduction tributary area. The tributary area is defined as the cross-sectional area
of an overburden block above the pillar and is a summation of the pillar
In underground coal mines, coal pillars play a major role in sus- cross-sectional area and the excavated area. The excavated area is de-
taining the weight of the overburden and protecting the stability of the fined as the rectangular area around the pillar that is formed by the
entries and crosscuts during mine development and production, al- half-width of entries and cross-cuts.
lowing the miners to safely extract the coal.1 Due to these vital func- Because achieving a stable pillar design is essential to creating a safe
tions, the coal pillar has become one of the most fascinating subjects in underground working environment, extensive laboratory work, in situ
the field of rock mechanics, particularly in the field of ground control. tests, and numerical simulations have been conducted in the past few
Ground control engineers seek to design coal pillars in a cost-effective decades to characterize coal pillar strength. From this research effort,
way by minimizing their dimension without sacrificing the stability of most ground control researchers have agreed that the strength of a coal
the entries or gateroads. A few meters' reduction of a typical chain pillar pillar increases with the pillar's width-to-height (w/h) ratio. Therefore,
width in longwall mine may be an attractive incentive in today's high- w/h ratio or pillar dimension has become the major parameter for de-
production longwalls.2 termining coal pillar strength. However, are we satisfied with the cur-
A classical approach to coal pillar design involves the following rent practice of determining pillar strength based merely on pillar di-
three steps: (1) pillar load estimation (usually calculated using the tri- mension?
butary area method), (2) pillar strength estimation (calculated from the On the other hand, coal pillar failure, whether it is violent or not,
available empirical pillar strength formulae), and (3) safety factor cal- occurs mainly because the load applied to the pillar is beyond its
culation (calculated as the ratio of pillar strength to pillar load). To strength. In laboratory scale, the strength of coal specimens is greatly
comply with the tributary area method, the pillar load in this paper is influenced by the following two parameters: (1) the friction between
referred to as the average pillar stress that is calculated as the ratio specimen and machine platens and (2) the specimen size.
between the average overburden stress before excavation and the Unfortunately, among ground control researchers, the contribution of

∗
Corresponding author.
E-mail address: simon@mining.itb.ac.id (S.H. Prassetyo).

https://doi.org/10.1016/j.ijrmms.2019.104102
Received 15 March 2019; Received in revised form 5 August 2019; Accepted 4 September 2019
1365-1609/ © 2019 Elsevier Ltd. All rights reserved.
S.H. Prassetyo, et al. International Journal of Rock Mechanics and Mining Sciences 123 (2019) 104102

Table 1 the former to coal pillar strength has attracted less attention than the
Empirical formulae for determining coal pillar strength. latter.
No. Developer Pillar strength formula Eq.
1.1. . Empirical coal pillar strength formulae
4
1 Bauschinger Sp = Scube 0.778 + 0.222
w (3)
h
2 Bunting8 w (4) For the past few decades, ground control researchers have devel-
Sp = 6.9 0.7 + 0.3
h oped many coal pillar strength formulae, with a general agreement that
3 Greenwald et al.7 w1/2 (5)
Sp = 19.2 the strength of a coal pillar increases with the pillar's width-to-height
h5/6
4 Holland dan Gaddy6 w (6) (w/h) ratio. There are two general types of expressions for predicting
Sp = K
9
h
the strength of a cubical coal pillar: the linear formula in Eq. (1) and the
5 Salamon dan Munro w0.46 (7)
Sp = 7.2
h0.66
power formula in Eq. (2):
10
6 Bieniawski Sp = Scube 0.64 + 0.36
w (8)
h w
Sp = Scube A+B
7 Holland 11
Sp = Scube
W (9) h (1)
H
12
8 Wagner and Madden (10)
Sp = 7.2
Ro0.5933
V 0.0667
0.5933
{( ) 1} + 1
R
Ro Sp = Scube
w
h (2)
9 Salamon and Wagner13 R2.5 + 181.6 (11)
Sp = 0.00786
V 0.0667 where Sp is the coal pillar strength, Scube is the strength of a cubical coal
10 Sheorey et al.14 Sp = 0.27 c h 0.36 +
z w
1 (12) pillar at w/h = 1, w and h are the width and the height of the coal pillar,
160 h
11 Madden15 w0.63 (13) respectively, and α, A, and B are the constants. There is also an ex-
Sp = 5.24
h0.78 ponential form of the pillar strength formula based on a linear logistic
5
12 Mark and Chase , known (14)
Sp = Scube 0.64 + 0.54
w
0.18
w2 regression model from coal pillar stability data.3However, this ex-
as the Mark-Bieniawski
ponential formula is not considered in this paper.
h lh
formula
13 Galvin et al.16, known as w0.5 (15) Table 1 presents the list of empirical formulae for determining the
Sp = 6.88
the UNSW formula h0.7 strength of coal pillars in the linear and power forms. These formulae
14 van der Merwe17 for Sp = 1.5
w (16) were developed from many coal fields around the world and are listed
h
“weak coal”
in chronological order. It should be noted that, in the derivation of their
15 van der Merwe17 for Sp = 3.5
w (17)
“normal coal”
h pillar strength formulae, the coal source and the machine specification
that were used by the developers for coal testing are not reviewed in
this paper. However, it is believed that such in-depth review would
benefit ground control practitioners to fully comprehend and ap-
preciate the significance of each of the existing empirical pillar strength
formulae.
As expected, all pillar strength formulae in Table 1 are functions of
pillar dimension as represented by either the w/h ratio or the absolute
value of the individual dimension (i.e., w, h, l, and V). In Table 1, K is
the Gaddy factor depending on the characteristic of the coal tested, σc is
the uniaxial compressive strength (UCS) of the coal specimen (MPa), R
is the w/h ratio, Ro is the critical w/h ratio, V is the pillar volume (m3), ε
is the rate of increase of pillar strength, z is the depth cover (m), and l is
the pillar length (m).
Fig. 1 compares several representative pillar strength formulae
among those listed in Table 1. This comparison uses pillar height
h = 2.1 m. The values of Scube are varied among developers and are
shown in Table 2. For the linear formulation, the Bauschinger4 and
Mark-Bieniawski5 formulae predict the lowest and the highest pillar
strength, respectively. Meanwhile, for the power formulation, the
Holland and Gaddy6 and Greenwald7 formulae predict the lowest and
the highest pillar strength, respectively. However, in general, the pre-
Fig. 1. Comparison of several representative coal pillar strength formulae. dicted strengths of coal pillars are bounded by the predicted strengths
of the Holland and Gaddy6 power formula (lower bound) and the Mark-
Bieniawski5 linear formula (upper bound).

Table 2 1.2. . Effect of interface friction on coal pillar strength

Scube and K for several representative coal pillar strength formulae.
Formula Developer K Scube (MPa) Whether it is violent or not, coal pillar failure occurs simply because
the load applied to the pillar is beyond the pillar's strength. In addition
Linear Bauschinger4 – 6.4 to the pillar dimension, the other factor that affects pillar strength is the
Bunting8 – 6.9
friction at the coal/roof and coal/floor interface (Fig. 2), known as the
Mark-Bieniawski5 – 6.4
Power Greenwald et al.7 – 19.2
interface friction. 1In the laboratory setting, the strength of coal spe-
Holland and Gaddy6 5095a 4.2 cimens in the uniaxial compressive strength (UCS) test is greatly in-
Salamon and Munro9 – 7.2 fluenced by the friction between the specimen and the machine platens
Galvin et al.16 – 6.9 and by the specimen w/h ratio.19–26
a To see the effect of interface friction and w/h ratio on the strength of
K is taken from18 and is calculated from various tested coal seams. Scube and Sp
coal specimens, Fig. 3 presents the stress-strain curves of coal specimens
are then first calculated in psi and then converted into MPa.
tested with UCS tests at different interface frictions and w/h ratios.

2
S.H. Prassetyo, et al. International Journal of Rock Mechanics and Mining Sciences 123 (2019) 104102

Fig. 2. A coal pillar in between coal/roof and coal/floor interfaces, modified from .27

These curves are re-constructed from studies22,23 that investigated the decreasing interface friction. When the end-platen exhibits high inter-
effect of interface friction and the w/h ratio of coal pillars on the oc- face friction (Fig. 5a), the percent area that the core zone occupies is
currence of coal mine bumps by means of UCS tests on coal specimens 54% for UT coal and 34% for WV coal. Conversely, when the end-platen
from West Virginia (WV) and Utah (UT). exhibits low interface friction (Fig. 5c), the percent area that the core
The key message from Fig. 3 is that with the increase of w/h ratio zone occupies is only 11% for UT coal and 0% for WV coal.
and interface friction, the stress-strain curves tend to go from brittle to Even though confinement has been the subject of many investiga-
elastic-plastic behavior. In particular, the specimens’ post-failure be- tions on the strength of coal pillars, the findings of this study show
havior changes from a complete failure to failure and then strength physical proof of the confinement experienced by the coal specimens. In
regaining or to strain hardening after yielding, indicating a trend of addition, the relationships between the confinement around the pillar
increasing specimen strength. core and failure types from this study are also in agreement with those
Fig. 4 shows a typical division of the four friction zones identified on observed in other studies.28–30
the top end-surface of coal specimens after UCS tests. 22,23The core zone
is defined as the highly-confined zone that shows no friction marks that
1.3. . Objectives of the paper
indicate movement experienced on its surface during the history of
loading. The intermediate zone is defined as the zone that experiences
Thus, it is clear that the strength of coal specimens is not only af-
only slight friction as indicated by the whitish or light-dark scratches on
fected by the w/h ratio but also by the interface friction. Even though it
its surface. The transition zone experiences intense friction as indicated
is not explicitly stated, the existing coal pillar strength formulae in
by the reddish scratches on its surface. Lastly, the rib zone has no or
Table 1 may have considered the existence of friction between the coal/
very little confinement as it is exposed directly to the free face.
roof and coal/floor interface. However, the influence of this interface
Regarding the four friction zones in Fig. 4, previous studies22,23
friction has not been quantitatively obtained from either in-situ tests or
have also found that, at high interface friction, the coal samples tend to
laboratory tests of coal specimens to derive the corresponding pillar
store more energy in the core zone. At low interface friction, the energy
strength formulae.
is more easily dissipated to create fractures in the specimen ribs or the
To fill this gap, the objectives of this paper are twofold: (1) to de-
rib zone. This behavior is in accordance with Fig. 5 that shows that for
velop linear and power formulae of coal pillar strength considering the
both UT and WV coals, regardless of the interface friction, there are
effect of interface friction and (2) to compare the developed formulae
opposite trends of the percent area between the core and rib zones with
against a database of stable and unstable pillar cases from South Africa,
increasing w/h ratios. The percent area of each friction zone is calcu-
the USA, India, and Australia.
lated as the ratio of the area of each friction zone to the total area of the
coal specimen after failure.
Consequently, the rib zone may fail shortly after the compressive 2. Developing pillar strength formulae considering the effect of
load is applied. Cracks propagate progressively from this zone toward interface friction
the transition zone, exposing the transition zone to new free faces. At
any interface friction, the average percent area of the core zones shows 2.1. . UCS tests of coal specimens
an increasing trend with increasing w/h ratios, while the opposite is
observed for the rib zones. Since the core zone is highly confined, the The formulae are developed based on the results of UCS tests on 178
trends observed in Fig. 5 indicate that confinement developed within a coal specimens conducted at three different interface frictions (high,
coal specimen increases with increasing w/h ratio but decreases with medium, and low) and ten width-to-height (w/h) ratios (w/h = 2–8, 10,
12, and 16) as appears in Refs. 22,23. The three interface frictions were

3
S.H. Prassetyo, et al. International Journal of Rock Mechanics and Mining Sciences 123 (2019) 104102

Fig. 3. Stress-strain curve of UCS tests on coal specimens at different interface frictions and w/h ratios22,23

created using three interface contacts during the UCS tests. The contacts regressions, i.e., the regression equations and the coefficient of de-
were Contact 1: coal/sandstone (regarded as high friction); Contact 2: termination R2, are shown in Fig. 7a. It can be seen from Fig. 7a that
coal/grease-lubricated sandstone (regarded as medium friction); and R2 = 0.75–0.89, meaning that 75–89% of the values fit the regression
Contact 3: coal/grease-lubricated steel (regarded as low friction). Each models. The coefficient of correlation R, defined as the square root of
contact surface was tested with a direct shear test machine to determine R2, is therefore R = 0.86–0.94. The range of R values from 0.86 to 0.94
its friction value (cohesion c and tan φ or μ). The friction values for each tells us that for each interface friction, the linear and power regressions
interface contact are as follows: Contact 1 (high friction) has in Fig. 7a are able to capture the significant correlation between w/h
c = 124 kPa and μ = 0.40; Contact 2 (medium friction) has c = 76 kPa ratio and the compressive strength of coal samples Sp. Once the re-
and μ = 0.22; and Contact 3 (low friction) has c = 55 kPa and μ = 0.13. gression is done, the following steps are carried out for each interface
Fig. 6a shows the arrangement of the UCS test on a coal specimen friction and regression type:
under Contact 1 (high friction), while Fig. 6b presents the scatter plot of
the UCS test results for all coal specimens at different interface frictions 1. Calculate the compressive strength of the cubical coal specimen Su.
and w/h ratios that serve as the main data used in this paper. Only the This is done by entering w/h = 1 into each of the regression equa-
UCS tests from the medium and low frictions up to w/h = 12 are used tions. Each of the equations is then normalized to the calculated Su.
for further analysis in this paper. Preliminary regression analysis on the The results of these normalizations are shown in Fig. 7b, while the
high friction results (not shown in this paper) leads to a very low cor- calculated Su are listed in Table 3.
relation coefficient. Hence, the high friction results are discounted from 2. Extrapolate the compressive strength of the cubical coal specimen Su
the analysis. to the compressive strength of the cubical coal pillar Scube. This is
done by first finding the Gaddy factor K using Eq. (18) and then
using the calculated K to find Scube using Eq. (19):
2.2. . Regression and normalization of the UCS test results
K = Su D (18)
Linear and power regressions are used to determine the regression
equation for each friction level of the UCS test results. The results of the Scube = K / 0.9 (19)

4
S.H. Prassetyo, et al. International Journal of Rock Mechanics and Mining Sciences 123 (2019) 104102

Fig. 4. The four friction zones identified on the top end-surface of a coal specimen after uniaxial loading22,23

where D is the sample diameter or width, that is 0.076 m as used in from w/h = 16, exceeding that from the Greenwald7 formula at this
Refs. 22,23. The square root of 0.9 m is used as this paper uses the SI ratio. On the other hand, the predicted strength from the low-friction
unit. The original equation in Ref. 31 uses the square root of 36 in as Su, formula is between those from the UNSW16 and Greenwald7 formulae.
D, K and Scube are in British units. The calculated K and Scube for each
interface friction and regression type are also listed in Table 3. 2.4. . Justification of the approach

2.3. . New coal pillar strength formulae considering the effect of interface It is well known that laboratory testing on coal and rocks always
friction shows that larger specimens give lower strength values than smaller
specimens. This is because more discontinuities are present in larger
Combining Fig. 7b and Table 3, the new coal pillar strength for- specimens that in smaller specimens. Even though it is very expensive
mulae considering the effect of interface friction can now be expressed. to perform a large number of in-situ coal pillar strength tests, the results
For the medium interface friction, the formulae are expressed as of laboratory tests at the cm-scale should be used with caution to pre-
w dict the actual coal pillar strength at the meter-scale. If no extrapolation
Linear formula Sp = 3.2 0.14 + 0.86 or other upscaling approach is performed, laboratory tests on small-size
h (20)
coal samples should then be used for comparative purposes only.10,32
w 0.9
Fortunately, pioneer researchers31,33 found a critical cubical coal
Power formula Sp = 3.7
h (21) pillar above which coal pillar strength does not vary with size. This
critical cubical coal pillar is defined as a coal pillar with side dimension
while for the low interface friction, the formulae are expressed as
of 0.9 m. This finding makes it possible to determine coal pillar strength
w by testing a sample cube with a size of 0.9 m. If the sample's side di-
Linear formula Sp = 2.7 0.12 + 0.88
h (22) mension is less than 0.9 m, the suggested extrapolation approach
by31,33 should be used to determine coal pillar strength. This suggested
0.8
w approach is the same used in this paper (Section 2.2). As can be seen in
Power formula Sp = 3.4
h (23) Fig. 8, the calculated Scube in this paper falls slightly below the Scube
To see the behavior of the new coal pillar strength formulae in light predicted in other pillar strength formulae because the influence of
of existing empirical formulae, the new formulae in Eqs. (20)−(23) are interface friction is now considered in the new formulae. Hence, it can
compared to the several representative empirical formulae of the linear be seen that this paper has followed the correct approach to develop
and power forms. The comparison uses the same parameters used in coal pillar strength formulae from laboratory tests considering the in-
Section 1.1, as shown in Table 2. fluence of interface friction.
In the group of linear formulae (Fig. 8a), the predicted strengths
from the medium- and low-friction formulae are close to the Mark-Bi- 3. Comparison against stable and unstable pillar cases
eniawski5 formula; the medium-friction formula predicts the largest
strength among the other linear formulae in Fig. 8a. In the group of The new coal pillar strength formulae in Eqs. (20)−(23) are now
power formulae (Fig. 8b), the predicted strength from the medium compared against a database of stable and unstable pillar cases from
friction is also the largest among the other power formulae starting around the world. This comparison is made for two reasons:

5
S.H. Prassetyo, et al. International Journal of Rock Mechanics and Mining Sciences 123 (2019) 104102

Fig. 5. The average percent area of each friction zone at each w/h ratio and interface friction22,23

1. To evaluate the suitability of the new pillar strength formulae 3.1. . Comparison result
against a real-world database, and
2. To build confidence in the parametric values used in the new pillar Fig. 9 presents the plots of stable and unstable pillar cases from
strength formulae (Table 3) as a result of regression of laboratory around the world. In both cases, the strength of the pillar Sp is estimated
test data. as the applied pillar load σp. As can be seen from Fig. 9, in both the
groups of linear (Fig. 9a) and power formulae (Fig. 9b), most of the
A total of 636 cases are collected, consisting of 392 cases from South estimated pillar loads in the stable cases are less than the estimated
Africa34, 172 from the USA35, 38 from India14, and 34 from Australia. pillar strength. Conversely, most of the estimated pillar loads in the
16
For the cases from South Africa, the database in this paper excludes unstable cases are greater than the estimated pillar strength. This result
the failure cases from the Vaal Basin, Klip River, and Free State coal- is expected and is particularly true when the estimated pillar strength
fields as the coals from these coalfields are considered “weak coal”.34 uses the low-friction formulae. Another interesting fact illustrated in
Fig. 9 is that most of the unstable cases that fall under the estimated

6
S.H. Prassetyo, et al. International Journal of Rock Mechanics and Mining Sciences 123 (2019) 104102

Fig. 6. (a) Arrangement of the UCS test of a coal specimen under Contact 1: high friction and (b) UCS test results of all coal specimens at different interface frictions
and w/h ratios22,23

pillar strength are at w/h ≤ 6. While the modes of these pillar failures Fig. 10 classifies the satisfactory level of the pillar strength predic-
are unknown, this fact corresponds with the results of the coal bump tion from the new formulae. The pillar strength prediction is called:
research conducted by Prassetyo22 and Prassetyo et al.23, in which it
was found that small pillars of less than w/h = 6 are more prone to fail • Satisfactory when the estimated pillar strength from the new for-
violently than big pillars. mulae is greater than the estimated pillar loads in the stable cases or

7
S.H. Prassetyo, et al. International Journal of Rock Mechanics and Mining Sciences 123 (2019) 104102

Fig. 7. Regression and normalization of the UCS test results.

Table 3 pillar strength predictions are presented in Table 4.

Calculated Su, K, and Scube for each interface friction and regression type. The focus of the discussion is now switched to the unsatisfactory
Interface Regression type Su (MPa) K (MPa.m1/2) Scube (MPa)
prediction in the unstable cases because this type of prediction is critical
to assess the suitability of the new formulae in predicting collapse cases
Medium friction Linear 11.1 3.1 3.2 from a real-world database. The unsatisfactory prediction of the un-
Power 12.6 3.5 3.7 stable cases tells us that the estimated pillar loads in the unstable cases
Low friction Linear 9.3 2.6 2.7
Power 11.6 3.2 3.4
are less than the estimated pillar strength from the new formulae. A
higher percentage of unsatisfactory predictions in the unstable cases
means a lower suitability and confidence level of the new pillar strength
less than the estimated pillar loads in the unstable cases. formulae.
• Unsatisfactory when the estimated pillar strength from the new The preceding paragraph mentioned that the new pillar strength
formulae fail to predict pillar strength in the unstable cases by 13–34%
formulae is less than the estimated pillar loads in the stable cases or
greater than the estimated pillar loads in the unstable cases. (17–45 unstable cases). In terms of the interface friction, the low-fric-
tion formulae have lower unsatisfactory prediction in the unstable cases
Note that the terminology “Satisfactory” or “Unsatisfactory” is used than the medium-friction formulae. In terms of the equation forms, the
to classify the satisfactory level of the pillar strength prediction from formulae of the linear form have lower percentages of unsatisfactory
the new formulae, not the satisfactory level of the pillar designs from prediction in the unstable cases than the formulae of the power form.
the stable and unstable cases. For the linear formulae, the percentages of unsatisfactory prediction in
It can be seen from Fig. 10 that for the stable cases, the satisfactory the unstable cases are 13% (17 unstable cases) and 24% (32 unstable
predictions from the new formulae are relatively similar, that is around cases) for low and medium interface friction, respectively. For the
90–96% (456–483 stable cases), while the unsatisfactory predictions power formulae, the percentages of unsatisfactory prediction in the
are around 4–10% (22–49 stable cases). On the other hand, for the unstable cases are 18% (24 unstable cases) and 34% (45 unstable cases)
unstable cases, the satisfactory predictions from the new formulae are for the low and medium interface friction, respectively. This statistic
only 66–87% (86–114 unstable cases), while the unsatisfactory pre- tells us that the low-friction formula of the linear form provides the best
dictions are around 13–34% (17–45 unstable cases). The detailed per- representation of the pillar cases in the database.
centages and numbers of cases of the satisfactory and unsatisfactory Similar behavior also appears when the database is broken down

8
S.H. Prassetyo, et al. International Journal of Rock Mechanics and Mining Sciences 123 (2019) 104102

Fig. 8. The new coal pillar strength formulae among the current empirical formulae.

into each country (Fig. 11). In the stable cases, the satisfactory pre- predictions in the unstable cases, India is the most suitable country to
dictions from the new formulae are relatively similar for both the use the low-friction formulae, followed by South Africa, the USA, and
medium- and low-friction formulae. In the unstable cases, the percen- Australia. In average, India has the lowest unsatisfactory prediction in
tage of the unsatisfactory predictions from the medium-friction for- the unstable cases (7%), followed by South Africa (14%), the USA
mulae are generally larger (9–42%) than those from the low-friction (21%), and Australia (22%). Moreover, the formulae of the linear form
formulae (4–28%). Looking at the percentage of the unsatisfactory have lower percentages of unsatisfactory prediction in the unstable

9
S.H. Prassetyo, et al. International Journal of Rock Mechanics and Mining Sciences 123 (2019) 104102

Fig. 9. Comparison of the new strength formulae against stable and unstable pillar cases from around the world.

cases than the formulae of the power form (see Fig. 12 and Table 5). histories collected across the USA, satisfactory pillar designs were found
Even though the percentages of unsatisfactory prediction from the when pillar safety factor SF ≥ 1.5. Starting from SF ≥ 1.5, the number
low-friction formulae are generally lower than those from the medium- of satisfactory pillar designs become significantly higher than the
friction formulae, from a practical standpoint, an unstable case is a very number of unsatisfactory pillar designs (e.g., slow pillar squeezes and
bad outcome that must be avoided, even if the design needs to be sudden pillar collapses). In other words, only a few unsatisfactory pillar
conservative. According to Mark36,37, from 140 pillar retreat case designs appeared when pillar SF is set to ≥1.5. Conversely, Mark36,37

10
S.H. Prassetyo, et al. International Journal of Rock Mechanics and Mining Sciences 123 (2019) 104102

Fig. 10. Bar chart showing the comparison of satisfactory and unsatisfactory pillar strength predictions using the new pillar strength formulae for stable and unstable
pillar cases from around the world.

Table 4 found that when SF < 1.5, the number of unsatisfactory pillar designs
Percentage of the satisfactory and unsatisfactory pillar strength predictions for becomes significantly higher than the number of satisfactory pillar
all pillar cases from around the world. designs.
Cases Interface Linear formula Power formula Adopting this concept, SF ≥ 1.5 is now used to evaluate the sa-
tisfactory level of the predicted pillar strength from the new formulae.
Satisfactory Unsatisfactory Satisfactory Unsatisfactory The pillar strength prediction is called:

• Satisfactory when the ratio of S /σ in the stable cases is ≥ 1.5 or

Stable Medium 95 5 96 4
friction (478) (27) (483) (22) p p
Low 90 10 93 7 when the ratio of S /σ in the unstable cases is < 1.5.
p p

Unstable
friction
Medium
(456)
76
(49)
24
(468)
66
(37)
34
• Unsatisfactory when the ratio of S /σ in the stable cases is < 1.5
p p
or when the ratio of Sp/σp in the unstable cases ≥1.0.
friction (99) (32) (86) (45)
Low 87 13 82 18
friction (114) (17) (107) (24) It should be noted that the percentages of satisfactory and un-
satisfactory predictions that appear in Figs. 10 and 12 and in Tables 4
* number of cases is written in brackets. and 5 are calculated based on SF = 1.0.
Fig. 13 presents bar charts showing the comparison of the

11
S.H. Prassetyo, et al. International Journal of Rock Mechanics and Mining Sciences 123 (2019) 104102

Fig. 11. Comparison of the new strength formulae against the stable and unstable pillar cases from each contry.

12
S.H. Prassetyo, et al. International Journal of Rock Mechanics and Mining Sciences 123 (2019) 104102

Fig. 12. Bar chart showing the comparison of satisfactory and unsatisfactory pillar strength prediction using the new pillar strength formulae for pillar cases from
each country.

13
S.H. Prassetyo, et al. International Journal of Rock Mechanics and Mining Sciences 123 (2019) 104102

Table 5
Percentage of satisfactory and unsatisfactory pillar strength prediction using the new pillar strength formulae for pillar cases from each country.
Country Formula Interface Stable cases Unstable cases

Satisfactory Unsatisfactory Satisfactory Unsatisfactory

South Africa Linear Medium friction 96 4 76 24

(309) (13) (42) (13)
Low friction 90 10 91 9
(291) (31) (50) (5)
Power Medium friction 98 2 58 42
(314) (8) 32 (23)
Low friction 96 4 82 18
(308) (14) (45) (10)
USA Linear Medium friction 91 9 66 34
(124) (13) (23) (12)
Low friction 88 12 77 23
(121) (16) (27) (8)
Power Medium friction 91 9 66 34
(124) (13) (23) (12)
Low friction 84 16 80 20
(115) (22) (28) (7)
India Linear Medium friction 93 7 91 9
(14) (1) (21) (2)
Low friction 87 13 96 4
(13) (2) (22) (1)
Power Medium friction 93 7 87 13
(14) (1) (20) (3)
Low friction 93 7 91 9
(14) (1) (21) (2)
Australia Linear Medium friction 100 0 72 28
(16) (0) (13) (5)
Low friction 100 0 83 17
(16) (0) (15) (3)
Power Medium friction 100 0 61 39
(16) (0) (11) (7)
Low friction 100 0 72 28
(16) (0) (13) (5)

* number of cases is written in brackets.

satisfactory and unsatisfactory pillar strength predictions for the un- the linear form of the low-friction formula because the above findings
stable cases from around the world for SF = 1.0 and 1.5. It can be seen show that the linear form of the low-friction formula is the most re-
that when SF = 1.5 is used, the percentages of the unsatisfactory pre- presentative in predicting coal pillar strength. In general, the low-fric-
dictions are greatly reduced to only 5–8% (7–10 unstable cases) for the tion formula of the linear form has lower percentages of unsatisfactory
medium-friction formulae and 4% (5 unstable cases) for the low-friction prediction in the unstable cases than that of the power form.
formulae (Table 6). These percentages are significantly lower than Considering the unstable cases from all countries (Fig. 15a), it can
when SF = 1.0 is used (i.e., 24–34%, or 32–45 unstable cases) for the be seen that most of the unstable pillar cases are bounded inside the
medium-friction formulae and 13–18% (17–24 unstable cases) for the strength envelope of SF = 1.5. The exception is those from the USA
low-friction formulae. (replotted separately in Fig. 15b). Considering the unstable cases from
When considering the unstable cases from each country, the per- the USA only (Fig. 15b), one can see that higher SF up to SF = 2.5 is
centage of the unsatisfactory predictions can also be kept at a maximum needed to achieve a higher percentage of satisfactory pillar design for
of 14% for the medium-friction formulae and 11% for the low-friction pillars with w/h ≤ 6. For pillars with w/h > 6, SF = 1.5 is deemed
formulae when SF = 1.5 is used. India and Australia are now the most sufficient to result in a stable pillar.
suitable countries to use the low-friction formulae with SF = 1.5, fol- Another interesting fact shown in Fig. 15 is that most of the unstable
lowed by South Africa and the USA. On average, India and Australia pillar cases from all around the world occur to pillars with w/h ≤ 8.
have the lowest unsatisfactory prediction in the unstable cases (0%), About 79% or 103 unstable pillar cases occur to pillars with w/h ≤ 4
followed by South Africa (2%) and the USA (11%) (see Fig. 14 and (slender pillar) and about 21% or 28 unstable pillar cases occur to
Table 7). Further observation of Table 7 reveals that when SF = 1.5 is pillars with 4 < w/h ≤ 8 (intermediate pillar). There are no unstable
used, the low-friction formula of the linear form provides the best re- pillar cases reported in the literature for pillars with w/h > 10 (squat
presentation of the pillar cases database. pillar).
Even when the low-friction formulae with SF = 1.5 is used, com-
pared to that from other countries, the percentage of unsatisfactory 3.2. . Note on the use of the new pillar strength formulae
predictions in the unstable cases from the USA appears to be higher,
that is 11% or 4 out of 35 unstable cases. To further explore this finding, Section 3.1 reveals that the low-friction formula of the linear form
databases from around the world and the database from the USA only provides the best representation of the stable and unstable pillar cases
are compared in a w/h vs. SF plot (Fig. 15). Pillar SF is calculated using database. It should be noted that this new formula is derived based on a

14
S.H. Prassetyo, et al. International Journal of Rock Mechanics and Mining Sciences 123 (2019) 104102

Fig. 13. Comparison of satisfactory and unsatisfactory pillar strength predictions for unstable pillar cases from around the world: (a) SF = 1.0 and (b) (SF = 1.5).

Table 6 when using the new formula, readers should keep in mind the limita-
Percentage of the satisfactory and unsatisfactory pillar strength predictions for tion of the w/h ratio at which the formula is derived and compared.
the unstable pillar cases from around the world: SF = 1.0 vs. SF = 1.5 In doing so, when using the linear low-friction formula for pillars
Formula Interface SF = 1.0 SF = 1.5 with w/h > 20, it is suggested that SF and pillar performance be ver-
ified using numerical simulation that considers the stress-strain relation
Satisfactory Unsatisfactory Satisfactory Unsatisfactory of the coal pillar, the roof, the floor, as well as the coal/roof and coal/
floor interfaces. In particular, the numerical simulation should be used
Linear Medium 76 24 95 5
friction (99) (32) (124) (7) to comprehensively observe:
Low 87 13 96 4

Power
friction
Medium
(114)
66
(17)
34
(126)
92
(5)
8
• Pillar behavior that may enter the strain hardening mode after
yielding as observed in laboratory tests for specimens with high w/h
friction (86) (45) (121) (10)
ratios (see Fig. 3).
•
Low 82 18 96 4
friction (107) (24) (126) (5) Progressive failure of a pillar that starts from the rib zone toward the
transition zone and may continue to propagate toward the core
* number of cases is written in brackets. zone, resulting in complete pillar failure.

series of UCS tests on coal specimens carried out at different interface With regard to the numerical analysis, the parametric values de-
frictions and at w/h ratio up to 16. The comparison of this new formula rived from the laboratory test in this paper (Table 3) could be used as a
against the field database is also only carried out up to w/h = 20. Thus, check on the appropriateness of the coal properties used in numerical

15
S.H. Prassetyo, et al. International Journal of Rock Mechanics and Mining Sciences 123 (2019) 104102

Fig. 14. Comparison of satisfactory and unsatisfactory pillar strength prediction for unstable pillar cases from each country for SF = 1.0 and SF = 1.5.

16
S.H. Prassetyo, et al. International Journal of Rock Mechanics and Mining Sciences 123 (2019) 104102

Table 7
Percentage of satisfactory and unsatisfactory pillar strength predictions for unstable pillar cases from each country: SF = 1.0 vs. SF = 1.5
Country Formula Interface SF = 1.0 SF = 1.5

Satisfactory Unsatisfactory Satisfactory Unsatisfactory

South Africa Linear Medium friction 76 24 98 2

(42) (13) (54) (1)
Low friction 91 9 98 2
(50) (5) (54) (1)
Power Medium friction 58 42 96 4
(32) (23) (53) (2)
Low friction 82 18 98 2
(45) (10) (54) (1)
USA Linear Medium friction 66 34 86 14
(23) (12) (30) (5)
Low friction 77 23 89 11
(27) (8) (31) (4)
Power Medium friction 66 34 86 14
(23) (12) (30) (5)
Low friction 80 20 89 11
(28) (7) (31) (4)
India Linear Medium friction 91 9 100 0
(21) (2) (23) (0)
Low friction 96 4 100 0
(22) (1) (23) (0)
Power Medium friction 87 13 96 4
(20) (3) (22) (1)
Low friction 91 9 100 0
(21) (2) (23) (0)
Australia Linear Medium friction 72 28 94 6
(13) (5) (17) (1)
Low friction 83 17 100 0
(15) (3) (18) (0)
Power Medium friction 61 39 89 11
(11) (7) (16) (2)
Low friction 72 28 100 0
(13) (5) (18) (0)

* number of cases is written in brackets.

models. While this numerical analysis is beyond the scope of this paper, (1) The low-friction formula of the linear form provides the best re-
this paper provides a combination of experimental (UCS tests on coal presentation of the pillar cases database. In terms of the interface
specimens at various interface frictions and w/h ratios), empirical friction, the low-friction formulae have lower percentages of un-
(derivation of new pillar strength formulae), and observational methods satisfactory prediction in the unstable cases than the medium-fric-
(comparison against field cases) for predicting pillar strength. Thus, tion formulae. In terms of the equation forms, the formulae of the
when the result of this paper is combined with numerical analysis, one linear form have lower percentages of unsatisfactory prediction in
may achieve an optimum pillar design approach. the unstable cases than the formulae of the power form.
(2) To improve the pillar strength prediction, SF = 1.5 is suggested to
be used for all pillar designs using the linear low-friction formula.
4. Conclusions Except for USA coal pillars, higher SF up to SF = 2.5 is suggested to
achieve a higher percentage of satisfactory pillar designs for pillars
This paper develops new coal pillar strength formulae considering with w/h ≤ 6 (slender pillar).
the effect of interface friction. The formulae, which are in linear and (3) When the friction of the coal/roof and coal/floor interfaces is not
power forms, are derived based on a series of UCS tests on coal speci- known, it is beneficial to assume that the interface has low friction
mens carried out at three different interface frictions and at ten width- value and thus to use the linear low-friction formula to predict
to-height (w/h) ratios. The three interface frictions are regarded as high pillar strength.
friction (c = 124 kPa and μ = 0.40), medium friction (c = 76 kPa and
μ = 0.22), and low friction (c = 55 kPa and μ = 0.13), while the w/h Having analyzed the results, it appears that the linear low-friction
ratios are at w/h = 2–8, 10, 12, and 16. The results for high friction are formula has the potential to be used concurrently with other pillar
not included in the analysis because they lead to a very low correlation strength formulae or pillar design programs such as that developed by
coefficient. Mark and Agioutantis38 to achieve a satisfactory pillar design. How-
Combining the regression and normalization of the UCS test results, ever, when using the new formula, readers should keep in mind the
new coal pillar strength formulae considering the effect of interface limitation of the w/h ratio at which the formula is derived and com-
friction are then developed and compared against a real-world database pared against the field database.
of stable and unstable pillar cases.
The results of this paper highlight three important messages:

17
S.H. Prassetyo, et al. International Journal of Rock Mechanics and Mining Sciences 123 (2019) 104102

Fig. 15. Comparison of w/h ratio vs. SF for stable and unstable pillar cases between (a) all countries and (b) the USA only using low-friction formula of the linear
form.

Acknowledgements on coal mine ground control. Int J Min Sci Technol. 2015;25:1–6.
3. Wattimena RK, Kramadibrata S, Sidi ID, Azizi MA. Developing coal pillar stability
chart using logistic regression. Int J Rock Mech Min Sci. 2013;58:55–60.
The authors gratefully acknowledge a research grant from the 4. Bauschinger J. Mitteilungen aus dem Mechanisch Technishen Laboratorium der K. 1876;
Program of Research, Community Service, and Innovation of the 1876.
Institut Teknologi Bandung (P3MI-ITB), fiscal year 2018. The authors 5. Mark C, Chase F. Analysis of Retreat Mining Pillar Stability. Proc. New Technol. Gr.
Control Retreat Mining-NIOSH IC 9446. Pittsburgh, PA: National Institute of
also wish to thank Dr. Christopher Mark (the Principal Roof Control Occupational Safety and Health; 1997:17–34.
Specialist at MSHA-USA) and Dr. Rudrajit Mitra (the Chair of Rock 6. Holland CT, Gaddy FL. Some aspects of permanent support of overburden on coal
Engineering at the School of Mining Engineering at the University of beds. Proc. West Virginia Coal Min. Inst. 1957; 1957:43–66.
7. Greenwald HP, Howarth HC, Hartmann I. Experiments on Strength of Small Pillars of
Witwatersrand, South Africa) for critically reviewing earlier drafts of Coal in the Pittsburgh Bed. Washington, DC: U.S. Dept. of the Interior, Bureau of Mines;
the manuscript. Their thoughts and suggestions have substantially im- 1939.
proved the technical and academic merits of this manuscript. In parti- 8. Bunting D. Chamber pillars in deep anthracite mines. Trans AIME. 1911;42:236–245.
9. Salamon MDG, Munro AH. A study of the strength of coal pillars. J South African Inst
cular, Dr. Mark suggested the concept of coal pillar safety factor for the
Min Metall. 1967;68:55–67.
analysis in this paper. 10. Bieniawski ZT. In situ strength and deformation characteristics of coal. Eng Geol.
1968;2:325–340.
References 11. Holland CT. Pillar design for support of the over burden in coal mines. Proc. 9th Can.
Symp. Rock Mech. Ottawa: Mines Branch, Dept. of Energy, Mines and Resources;
1973:114–139.
1. Peng SS. Coal Mine Ground Control. third ed. Morgantown, WV: Syd Peng Publisher; 12. Wagner H, Madden BJ. 47 Fifteen years' experience with the design of coal pillars in
2008. shallow South African collieries: an evaluation of the performance of the design
2. Peng SS. Topical areas of research needs in ground control - a state of the art review procedures and recent improvements. Des. Perform. Undergr. Excav. ISRM Symp.

18
S.H. Prassetyo, et al. International Journal of Rock Mechanics and Mining Sciences 123 (2019) 104102

Cambridge, U.K: British Geotechnical Society; 1984:391–399. Virginia University Bulletin, Series 37, No. 8. IV. Morgantown, WV: West Virginia
13. Salamon MDG, Wagner H. Practical experiences in the design of coal pillars. Proc. University; 1937.
21st Int. Conf. Saf. Mines Res. Rotterdam: A.A. Balkema; 1985. 26. Pariseau WG, Hustrulid WA, Swanson SR, Van Sambeek LL. Coal Pillar Strength Study
14. Sheorey PR, Das MN, Barat D, Prasad RK, Singh B. Coal pillar strength estimation (The Design of Production Pillars in Coal Mines). Bruceton, PA: U.S. Dept. of the
from failed and stable cases. Int J Rock Mech Min Sci Geomech Abstr. Interior, Bureau of Mines; 1977.
1987;24:347–355. 27. Peng SS. Ground Control Failures−A Picturial View of Case Studies. Morgantown. WV:
15. Madden BJ. A re-assessment of coal-pillar design. J South African Inst Min Metall. Syd Peng Publisher; 2007.
1991;91 27–23. 28. Barron K. Analytical approach to the design of coal pillars. Can Inst Min Metall Bull.
16. Galvin JM, Hebblewhite BK, Salamon MDG. UNSW pillar strength determinations for 1984;77:37–44.
Australian and South African conditions. Rock Mech. Ind. Proc. 37th U.S. Rock Mech. 29. Gale WJ. Geological issues relating to coal pillar design. Proc. Symp. Geol. Longwall
Symp. Rotterdam: A.A. Balkema; 1999:63–71. Min. Springwood, NSW: Conference Publications; 1996:185–191.
17. van der Merwe JN. New pillar strength formula for South African coal. J South 30. Mark C. The state-of-the-art in coal pillar design. SME Trans. 2000;308:123–128
African Inst Min Metall. 2003;103:281–292. (originally preprint 99–86).
18. Du X, Lu J, Morsy K, Peng SS. Coal pillar design formulae review and analysis. Proc. 31. Hustrulid WA. A review of coal pillar strength formula. Rock Mech. 1976;8:115–145.
27th Int. Conf. Gr. Control Min. Morgantown, WV: West Virginia University; 32. Bieniawski ZT. The effect of specimen size on compressive strength of coal. Int J Rock
2008:254–261. Mech Min Sci. 1968;5:325–335.
19. Babcock CO. Constraint is the prime variable in pillar strength. Proc. 4th Int. Conf. Gr. 33. Bieniawski ZT. New design approach for room-and-pillar coal mines in the USA. Proc.
Control Min. Morgantown, WV: West Virginia University; 1985:105–116. 5th ISRM Congr. Rock Mech. Rotterdam: Balkema; 1983:E27–E36.
20. Khair AW. Effect of Coefficient of Friction on the Compressive Strength of Model Coal 34. van der Merwe JN, Mathey M. Update of coal pillar strength formulae for South
Pillars Master’s thesis West Virginia University; 1968. African coal using two methods of analysis. J South African Inst Min Metall.
21. Meikle PG, Holland CT. The Effect of Friction on the Strength of Model Coal Pillars. 2013;113:841–847.
1965; 1965:322–326 SME-AIME Prepr. No. 65FM80, Chicago, IL. 35. NIOSH. Analysis of Retreat Mining Pillar Stability v.6.2.02. August 201. Pittsburgh, PA:
22. Prassetyo SH. The Influence of Interface Friction and W/H Ratio in the Violence of Coal National Institute of Occupational Safety and Health; 2013.
Specimen Failure Master’s thesis Morgantown, WV: West Virginia University; 2011. 36. Mark C. Empirical methods for coal pillar design. Proc. Second Int. Work. Coal Pillar
23. Prassetyo SH, Rashed G, Li Y, Luo Y, Peng SS. The influence of interface friction and Mech. Des. Pittsburgh, PA: National Institute of Occupational Safety and Health;
W/H ratio on the violence of coal specimen failure–A comparison between a bump 1999:145–154.
and non–bump prone mines. Proc. 30th Int. Conf. Gr. Control Min. Morgantown, WV: 37. Mark C. The evolution of intelligent coal pillar design: 1981-2006. Proc. 25th Int.
West Virginia University; 2011:1–13. Conf. Gr. Control Min. Morgantown, WV: West Virginia University; 2006:325–334.
24. Daniels J, Moore LD. The crushing strength of coal. Eng Min J. 1907;84:263–268. 38. Mark C, Agioutantis Z. Analysis of coal pillar stability (ACPS): a new generation of
25. Lawall CE, Holland CT. Some Physical Characteristic of West Virginia Coals. West pillar design software. Int J Min Sci Technol. 2019;29:87–91.

19