International Journal of Rock Mechanics & Mining Sciences 64 (2013) 132–138

Contents lists available at ScienceDirect

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

Fracturing of an Euler–Bernoulli beam in coal mine pillar extraction

C.P. Please a, D.P. Mason b,n, C.M. Khalique c, J.M.T. Ngnotchouye d, A.J. Hutchinson b,
J.N. van der Merwe e, H. Yilmaz e
a
School of Mathematics, Oxford Centre for Collaborative Applied Mathematics, University of Oxford, Oxford 0X1 3LB, United Kingdom
b
School of Computational and Applied Mathematics, University of the Witwatersrand, Johannesburg, Private Bag 3, Wits 2050, South Africa
c
Department of Mathematical Sciences, North-West University, Mafikeng Campus, Private Bag X2046, Mmabatho 2735, South Africa
d
School of Mathematics, Statistics and Computer Science, University of KwaZulu-Natal, Private Bag X01, Scottsville, Pietermaritzburg 3209, South Africa
e
School of Mining Engineering, University of the Witwatersrand, Johannesburg, Private Bag 3, Wits 2050, South Africa

art ic l e i nf o a b s t r a c t

Article history: The paper models the first stage of the process of pillar extraction in a coal mine. The problem of
Received 27 February 2012 understanding how a coal mine roof collapses after secondary cutting of the supporting pillars to create
Received in revised form small supporting snooks is considered. The fracture of the roof is considered when a set of snooks have
14 June 2013
failed and the roof must support itself between two pillars. Models that account for the relative
Accepted 2 August 2013
Available online 20 September 2013
importance of the overburden weight on the roof and the compressive stresses in the roof are examined
using a simple strut and beam theory.
Keywords: & 2013 Elsevier Ltd. All rights reserved.
Coal mine
Pillar extraction
Snook
Bending moment
Euler–Bernoulli beam
Euler strut

1. Introduction
The need to return to old mine workings and extract a signi-
ficant fraction of the remaining coal is becoming of increasingly
economic importance and viability. In the current methods of
secondary mining, the existing large pillars of coal that have been
left to support the roof are cut away and, as the work proceeds back
up the mine, the roof is left to collapse. This collapse needs to occur
in a safe and controlled manner and this is done by carefully cutting
the pillars into smaller structures, called snooks, so that these fail in
a manner that lets the roof fall slowly and the working area remains
free from falling material. Understanding how the snooks fail and
how the roof fractures is central to creating a safe working
environment. Determining how much of the pillars can be cut
and hence how small the snooks can be made determines the
fraction of coal that can be taken by this secondary mining method.
There are two stages in pillar extraction [1]. In the first stage
several pillars are extracted and replaced by snooks. The load on
each snook is increased as more and more pillars are extracted.
When a snook fails, it fails violently. After about seven or eight
pillars have been extracted the snooks fail. The roof layer cracks,
primarily by bending induced tension. Eventually the roof and
n
Corresponding author. Tel.: +27 117176117; fax: +27 117176149.
E-mail address: David.Mason@wits.ac.za (D.P. Mason).
overburden collapse. In this first stage the roof layer can be
modelled as a beam which is clamped at each end at a pillar.
After the first collapse of the roof, the failure will occur at more
frequent intervals. In this second stage of pillar extraction, one end
of the roof layer will be supported by a snook and the other end by
a pillar which is the working face. The end at the snook will no
longer be clamped because of the failure of the surrounding rock
and may be modelled as simply supported or hinged.
In this paper we consider the first stage of pillar extraction.
Each end of the roof layer is clamped at the pillar and a number of
snooks fail. The roof remains in place and supports the local
weight of the rock and the large compressive stresses due to the
overburden. We shall investigate the fracture of the roof and then
summarise our insight into the behaviour.
2. Mathematical model of roof fracture
The problem that we consider is how the roof fractures when
a number of snooks fail. There is considerable previous work on
analyzing such situations which is well reviewed in [2]. For an
elastic beam with no joints and no axial compressive stress
and clamped at pillars at each end, the solutions for the
maximum stress and maximum beam deflection have been
obtained using the beam equation [3,4]. Vertical tensile fractures
form at the pillars and the beam becomes simply supported

1365-1609/$ - see front matter & 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.ijrmms.2013.08.001
 C.P. Please et al. / International Journal of Rock Mechanics & Mining Sciences 64 (2013) 132–138 133

at the pillars. The maximum stress is now at the midspan which is

found to be greater than the previous stress at the pillars.
This leads to fracturing at the midspan and other points of the
beam [5,6].
To analyse the problem of fracturing of the roof we consider the
case where the snooks fail completely and the roof has to be
supported by the strength of the rock. The rock can be taken to
have various constitutive relations but here we take the response
to be elastic and we consider the roof to be made of horizontal
layers of rock each acting as a beam. An elastic beam with no joints
or fractures after an excavation is extremely rare [4]. Our model Fig. 1. A combined beam and strut.
will describe the first stage in the fracture of the beam and the
transition from a continuous elastic beam to a voussoir beam
model cut with joints and fractures [2,4,7–9]. We assume that the each end. The derivation of Segel and Handelman can be adapted
roof extends for a long way along the roadway and hence that the to a beam for which each end is clamped instead of simply
behaviour only varies with distance across the roadway from one supported and it can be verified that the same differential
remaining pillar to another. We also assume that the only equation for the displacement is obtained. Here we will give a
displacement is vertical. In practice this beam will be loaded from plausible justification for this differential equation.
three main sources. Firstly the compressive stresses due to the Consider first the problem in which the horizontal axial force P
overburden will create compressive stresses horizontally along the does not act on the beam. Then the bending moment about the
beam, secondly the beam will have to support its own weight and origin, M, which is in the j-direction, has magnitude
thirdly the beam will have stresses transferred to it from adjoining 2
d w
layers of rock. As an initial investigation we ignore the transfer of M ¼ EI 2
; ð1Þ
stresses from adjoining layers. Underground layers of rock tend to dx3
separate on deflection [10,4] and such transfer of stress may be where E is the Young's modulus of the roof rock and I is the second
small in cases where the beam may have displaced sufficiently to moment of area about the x2-axis given by
detach from layers above it. Hence our idealised problem is to Z b Z h 3
investigate buckling of a beam that has a horizontal stress acting 2 2 bh
I¼ dx2 x21 dx1 ¼ : ð2Þ
along its length due the overburden stress and that has to support 2
b
2
h 12
its own weight.
For the static Euler–Bernoulli beam, the transverse angular
This problem is therefore a combination of the conventional
momentum balance equation is [13]
Euler–Bernoulli beam theory and the theory of the Euler strut. The
2
beam is taken to be made of homogeneous material of thickness h. d M
2
¼ q: ð3Þ
We take the mine to be at a depth H. We can expect the horizontal dx3
stress in the rock to be kρgH where k is the lateral stress coefficient, ρ
is the average density of the rocks above the depth H and g is the Consider now a beam in which the axial force P acts. The bending
acceleration due to gravity [1,11,12]. The derivation of this result is moment about the origin, Mn , is
briefly outlined in Section 5. We assume that this stress creates a Mn ¼ Mj þ wðx3 Þi  ðPkÞ
horizontal force exerted at both ends of the beam which we denote ¼ ðM þ wðx3 ÞPÞj ð4Þ
by P ¼ hbðkρgHÞ; where b is the breath of the beam. Moreover, the
where M is given by (1). Hence
roof is subjected to its weight per unit length q ¼ ρghb that we
assume equally distributed along the beam on all the surface. In Mn ¼ M þ wðx3 ÞP: ð5Þ
practice we might wish to make a simple extension to partly allow n
Replacing M in (3) by M we obtain
for other layers above the roof on this beam by adding s to q and still
2 2
denoting it by q where s is the applied surface traction per unit d M d w
þP ¼ q: ð6Þ
length vertically downwards on the top surface of the beam. Both s 2
dx3 dx3
2

and q have the same effect on the displacement and enter the Euler–
Bernoulli beam equation in the same way [13]. Using (1) for M, (6) becomes
We will use the notation and conventions of Segel and Handel- 4
d w d w
2

man [13]. The coordinate axes are defined in terms of the EI 4

þP 2
¼ q; ð7Þ
dx3 dx3
undeformed beam. The x1- and x2-axes are along the axes of
principal moment of inertia of the cross-section of the beam with which is the required ordinary differential equation for wðx3 Þ. Eq. (7)
the x1-axis vertically downwards. The x3-axis is horizontal and agrees with the beam equation derived by Segel and Handelman
passes through the centroid of each cross-section. The origin of the [13].
coordinate system is at the centroid of the cross-section of the left We assume that the ends of the beam are clamped by the
end of the beam. Unit vectors i, j and k are directed along pillars at x3 ¼ 0 and x3 ¼ L so that (7) should be considered with
coordinate axis. The problem is illustrated in Fig. 1. the boundary conditions
We use analysis relevant to both an Euler–Bernoulli beam and dw
an Euler strut. The displacement of the beam from the horizontal wð0Þ ¼ 0; ð0Þ ¼ 0;
dx3
position, x3, vertically downwards in the positive x1-direction, is dw
denoted by wðx3 Þ. An outline of the derivation of the differential wðLÞ ¼ 0; ðLÞ ¼ 0: ð8Þ
dx3
equation for wðx3 Þ, has been given by Segel and Handelman [13]
for a beam which is simply supported or hinged at each end. For a Next, we nondimensionalise the problem (7) and (8). We start by
simply supported beam the displacement and its second derivative considering nondimensional variables x3 and w where
vanish at each end. We will consider a beam which is clamped at x3 w
x3 ¼ ; w¼ ð9Þ
each end so that the displacement and its first derivative vanish at L S
134 C.P. Please et al. / International Journal of Rock Mechanics & Mining Sciences 64 (2013) 132–138

and S is a characteristic value of w to be determined. Then (7) with the boundary conditions
becomes 3
dw d w
4 2 2 4 ð0Þ ¼ 0; 3
ð0Þ;
d w PL d w qL dz dz
þ ¼ ; ð10Þ    
dx 43 EI dx 23 EIS 1 dw 1
w ¼ 0; ¼ 0: ð19Þ
2 dz 2
with the boundary conditions
dw We solve Eq. (18) to obtain the general solution
wð0Þ ¼ 0; ð0Þ ¼ 0;
dx 3 z2
dw w¼ þ α cos ðBzÞ þ β sin ðBzÞ þ δz þ ν: ð20Þ
wð1Þ ¼ 0; ð1Þ ¼ 0: ð11Þ 2B2
dx 3
The integration constants are found using the boundary conditions as
We then choose S such that qL4 =EIS ¼ 1 or
1
β ¼ δ ¼ 0; α¼ ;
qL4 3 B
S¼ ð12Þ 2B sin
EI 2
and introduce the nondimensional parameter B, the beam number, B
1 cos
defined by ν¼ 2  2 : ð21Þ
8B B
 1=2 2B3 sin
P 2
B¼L : ð13Þ
EI Substituting in (20), we obtain
Finally, we make the bending moment dimensionless by   !
1 1 2 1
defining wðxÞ ¼ 2 x 
2B 2 4
M     
M¼ : ð14Þ 1 1 B
qL2 þ cos B x  cos ; ð22Þ
B 2 2
2B3 sin
Eq. (1) becomes 2
2 provided
d w
M¼ : ð15Þ
dx 23 B
sin a 0; B a 2nπ; n ¼ 1; 2; 3; … ð23Þ
2
We have assumed that the displacement is sufficiently small that
linear theory can be applied. In the linear approximation, the The displacement will first become singular at B ¼ 2π. We will first
2 2
curvature of the neutral axis of the beam is d w=dx3 [13]. The consider the range 0 o B o 2π and investigate the possible fracture
bending moment is therefore proportional to the curvature of mechanisms for the beam. We will then consider the solution and the
the beam. fracture of the beam for values of B 4 2π.
The problem reduces to solving the ordinary differential It can be verified that the solution of the differential equation
equation (16) subject to the boundary conditions (17) is indeed (22).
4 2
d w d w
þ B2 ¼ 1; ð16Þ
dx 43 dx 23 4. Discussion of the results
subject to the boundary conditions
We consider the analytical solution (22) for the displacement
dw wðxÞ. Graphs of w(x) for a range of values of the beam number, B,
wð0Þ ¼ 0; ð0Þ ¼ 0;
dx 3 are presented in Fig. 2. The deformation of the beam agrees with
dw observations for the values of B considered.
wð1Þ ¼ 0; ð1Þ ¼ 0: ð17Þ
dx 3 For B o2π, the displacement w(x) has three stationary points
The bending moment is then given by (15). which are at x ¼0, 1/2 and 1. It attains its minimum value at x ¼0
We will now consider the behaviour of solutions to this and x¼1 and its maximum value at x¼ 1/2:
problem for various values of B. In the subsequent analysis the    
1 1 4 B
overhead bar on x3 and w will be suppressed, being understood w ¼ 2 tan 1 4 0: ð24Þ
2 8B B 4
that dimensionless variables are being used and to simplify the
notation, x3 will be denoted as x.
x 10−3
0

3. Solution of the mathematical model 0.5

1
The problem as posed is on the interval 0 r x r 1 and the
Displacement

resulting algebra of the solution is greatly simplified if the 1.5

symmetry of the solution can easily be exploited. We therefore
2
start by putting the problem on a different interval. To do this we
shift the problem by letting x ¼ z þ 1=2 so that the problem is on 2.5
the interval 1=2 r z r1=2 and the solution will be symmetrical
3
about z¼ 0. We then now solve the differential equation in the half
domain ½0; 1=2 for the variable z: The model now becomes 3.5
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
4 2
d w 2d w
4
þB 2
¼ 1; ð18Þ
dz dz Fig. 2. Graphs of the displacement for values of B in the range 0.3–3.
 C.P. Please et al. / International Journal of Rock Mechanics & Mining Sciences 64 (2013) 132–138 135

Expanding w(x) for small values of B gives 0.7

"   ! #
x2 ð1xÞ2 B2 3 1 2 0.6
wðxÞ ¼  1þ  x þ OðB4 Þ ; ð25Þ
24 30 4 2
0.5
as B-0. Hence,
  " # " # 0.4
1 1 B2 B2
w ¼ 1þ þ OðB4 Þ ¼ 2:604  103 1 þ þ OðB4 Þ ;
2 384 40 40 0.3

ð26Þ 0.2
as B-0: The zero order in B values agree well with Fig. 2. The
0.1
correction to the zeroth order approximation is small when B is
small which explains why the curves for small B are clustered 0
together. In terms of the original dimensional variables and using 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
(2) for I, we have
  " # Fig. 3. Graphs of the magnitude of the curvature jw″ðxÞj plotted against x for B ¼1.0,
L qL4 3 P L2 4 3.0, 4.0, 5.0, 6.0.
w ¼ 3
1þ þ OðB Þ : ð27Þ
2 32Ebh 10bEh3

To lowest order in B the displacement is independent of the

horizontal force P and is due to the weight per unit length q.
To analyse the possible fracture of the beam we examine the
distribution of stresses within the beam. The bending moment of the
beam is proportional to the curvature of the beam (from (1)) and
hence the maximum stress in the beam will correspond to positions
where the magnitude of the curvature is maximum. From (22), we
find that
2 2 0 133
2
d w 16 B 6 B 1C77
¼ 41 cos 4B@x A55: ð28Þ
dx
2
B2 B 2
2 sin
2
In order to investigate the turning points of the curvature, we evaluate
Fig. 4. The ratio of the magnitude of the curvature at the end x¼ 0 (and x¼ 1) to the
" !#
1 magnitude of the curvature at the mid-point x¼ 1/2 plotted against B for 0 o B o 2π.
3
sin B x
d w 2
¼ ð29Þ
dx
3 B To investigate this comparison further consider the expansions
2 sin " #
2 2
d w 1 B2 4
which vanishes for 2
ð0Þ ¼ 1 þ þ OðB Þ ¼ 0:083 þ OðB2 Þ; ð34Þ
dx 12 60
1 nπ
x¼ þ ; n ¼ 0; 7 1; 7 2; … ð30Þ   " #
2 B 2
d w 1 1 7B2 4
For B o2π there is only one turning point, x¼1/2, in the interval 2
¼  1 þ þ OðB Þ ¼ 0:042 þ OðB2 Þ; ð35Þ
dx 2 24 240
½0; 1.
The magnitude of the curvature can now be used to give an as B-0 which is in good agreement with the graph in Fig. 3 for
indication of where the beam is going to fracture. Consider first the B ¼1. Also the ratio of the magnitude of the curvature at x ¼0 (and
range of values 0 o B o 2π: In Fig. 3, the magnitude of the curvature x¼ 1) to the magnitude of the curvature at x¼ 1/2 is
jw″ðxÞj is plotted against x for a number of values of B in the range B B B
0 o B o2π. For the values of B considered the maximum value of the jw″ð0Þj sin  cos
   ¼ 2 2 2; 0 o B o 2π: ð36Þ
magnitude of the curvature occurs at the ends x¼0 and x¼1 of the   B B
w″ 1   sin
beam. The magnitude of the curvature at x¼ 1/2 is always less than  2  2 2
the value at the end points. We now show that this is satisfied for all
values of B in the range 0 o B o 2π. From (28), for 0 o B o2π, For small values of B,
  " #
1 B B B jw″ð0Þj B2
w″ð0Þ ¼ w″ð1Þ ¼ sin  cos 40; ð31Þ    ¼ 2 1 þ OðB4 Þ ; as B-0: ð37Þ
B 2 2 2  
B2 sin w″ 1  80
2  2 
   
1 1 B B For B¼0, the magnitude of the curvature at the ends x¼0 and x¼ 1 is
w″ ¼  sin o 0: ð32Þ
2 B 2 2 twice the value at the center of the beam in agreement with the result
B2 sin
2 of Obert and Duvall [3] for B¼0. The ratio of the magnitudes decreases
The physical reason why w″ð0Þ 4 0 and w″ð1Þ 4 0 while w″ð1=2Þ o 0 as B increases consistent with the graphs in Fig. 3. Further,
is that x¼0 and x¼1 are the minimum turning points of w(x) while jw″ð0Þj
lim    ¼ 1: ð38Þ
B-2π 
x¼ 1/2 is the maximum turning point. Now
   w″ 1 
 1  B B  2 
w″ð0Þ 4 w″ provided tan 4 ; ð33Þ
2  4 4
In Fig. 4, the curvature ratio (36) is plotted against B for 0 oB o 2π.
which is satisfied for 0 oB o2π: Thus for 0 o B o 2π the magnitude of We see that it decreases steadily from 2 to 1 as B increases from 0 to
the curvature is always greatest at the end points of the beam. 2π in agreement with the limits derived analytically. While the
136 C.P. Please et al. / International Journal of Rock Mechanics & Mining Sciences 64 (2013) 132–138

bending moment remains below the threshold of the tensile strength In Fig. 3 the zeros of the magnitude of the curvature for B 4 0
of the beam the roof will not crack. However, when this value exceeds lie close to the zeros for B¼ 0. To investigate this consider the
the tensile strength the beam will fracture at the end points. expansion of w″ðxÞ, given by (28), for small values of B:
"  (    ) #
2
d w 1 1 2 1 B2 1 4 1 1 2 7
2
¼ x   x  x þ þ OðB4 Þ ; ð39Þ
dx 2 2 12 12 2 2 2 240

as B-0. This result may be rewritten as

" !
2
d w 1 B2
2
¼ ðx0:79Þðx0:21 ðx0:76Þ
dx 2 12
#
ðx0:24Þðx1:16Þðx þ 0:16Þ þ OðB4 Þ : ð40Þ

To zero order in B, the zeros of the curvature are at x ¼0.21 and

x¼ 0.79, which is in good agreement with Fig. 3. The zeros of the
second order term in B in the range 0 o x o 1 are x ¼0.24 and
x¼ 0.76 which explains the observation that the zeros of the
curvature for B 4 0 are close to the values for B¼ 0.
In Fig. 5 the behaviour of the magnitude of the curvature for
values of B 4 2π is compared with the behaviour for 0o B o2π. The
solution for B 4 2π demonstrates the behaviour of the beam when it
has buckled into high frequency modes due to the applied axial force,
P, on the end sections. The curvature is singular at Bn ¼ 2nπ,
n ¼ 1; 2; … which divides the values of B into the intervals
Bn1 o B o Bn ; for n Z 1. In Fig. 5 the representative values, B¼ 1,
B¼8, B¼ 15 and B¼35, were chosen which lie in the intervals
0 o B o 2πðn ¼ 1Þ, 2π o B o 4πðn ¼ 2Þ, 4π oB o 6π ðn ¼ 3Þ and
10π o B o12πðn ¼ 6Þ. There are several internal maxima for the
magnitude of the curvature. The number of internal maxima increases
as n increases. For B¼8, 15 and 35 there are n1 maxima and n local
maxima of less magnitude. The magnitude of the curvature will
depend on how close B is to the singular points Bn1 and Bn . The
beam therefore could fracture at interior points for n Z 2 if the tensile
strength of the beam is exceeded. The internal curvature maxima will
be interpreted in terms of the voussoir beam model in Section 6.
We note as an aside that with this analysis, we do not readily
recover the analysis of the first mode of the Euler strut from the
Bernoulli beam model. This is due to the fact that the beam system
is a fourth order differential equation requiring four boundary
conditions while the classical Euler strut is a second order problem
with just two boundary conditions. The main interest of the model
occurs when the additional boundary conditions are crucial to the
behaviour due to the need to include the bending behaviour of
the beam.

5. Numerical estimates

Consider first an estimate of the axial force, P; on the end

sections of the beam [11]. The vertical compressive stress at the
depth H must support the weight of the overburden rock and
therefore
τzz ¼ ρgH; ð41Þ
where τik is the Cauchy stress tensor and ρ is the average density of
the rock above the depth H: The simplest assumption is that
τxx ¼ τyy ¼ τzz ¼ ρgH; ð42Þ
which is the lithostatic state of stress and (42) is Heim's rule.
Stresses in rock tend to become lithostatic because of creep [12].
There are two extreme models. One model is to use the unphysical
assumption that the gravitational force acts only after the rock has
been put in place and therefore that there is no horizontal strain.
Using the inverse Hooke's law, it can be shown that [11]
Fig. 5. Graphs of the magnitude of the curvature jw″ðxÞj for (a) B¼ 1, (b) B¼8, ν
τzz ¼ ρgH; τxx ¼ τyy ¼ τzz ; ð43Þ
(c) B ¼15, and (d) and B¼35. 1ν
 C.P. Please et al. / International Journal of Rock Mechanics & Mining Sciences 64 (2013) 132–138
where ν is the Poisson ratio of the rock. Since 0 o ν o 1=2, it follows
that 0 oν=ð1νÞ o 1. In rock mechanics, the value of the Poisson
ratio usually used is ν ¼ 1=4. The other extreme model is to assume
that when the gravity starts to act, there are no horizontal
stresses:
τzz ¼ ρgH; τxx ¼ τyy ¼ 0:
The three models can be written as
τzz ¼ ρgH; τxx ¼ τyy ¼ kτzz ;
where k is the lateral stress coefficient. We will use (45) and since
the breadth of the beam is b and the height is h,
P ¼ kρgHbh:
The beam number (13) becomes, using (2) for I and (46) for P,
 1=2
L kρgH
B ¼ 3:46 ; ð47Þ
h E
which is independent of the breadth of the beam, b: Consider first
the range 0 o B r2π. The displacement and the curvature first
become infinite at B ¼ 2π. Now B o2π if
 
E 1=2
L o L1 ¼ 1:82h : ð48Þ
kρgH
The beam will fracture at the pillars when the tensile stress at the
pillars exceeds the tensile strength of the beam which may occur
for lengths L o L1 . If L1 is less than the panel width (distance
between the pillars), then the beam will fracture. We see that L1 is
proportional to the thickness, h; of the beam and inversely
proportional to the square root of the depth of the excavation, H:
Thinner beams at deeper level will therefore fracture for shorter
distances between the pillars, which is consistent with experience.
We now estimate the length L1 for representative values of the
parameters. We assume that the beam is made of sandstone. The
density ρ is the average density of the rock mass above the depth H.
The thickness of the beam, h, could vary between 0.1 m and 2 m. Some
models can predict values of the lateral stress coefficient, k, greater
than unity and that it is greatest near the surface [11,14]. The depth of
the coal mine could vary from 120 m to 1000 m. We will take k¼ 2
which applies to an excavation at shallow depths [1]. Representative
values of the parameters for a beam in a coal mine are
E ¼ 3  1010 Pa; ρ ¼ 2:5  103 kg m3 ;
k ¼ 2; g ¼ 9:8 m s2 ;
h ¼ 0:12 m; H ¼ 1201000 m:
Then
h
L1 ¼ 1:42  103 pffiffiffiffim:
H
During the first stage of pillar extraction in a coal mine the length of
the roof layer which collapses is about 150 m corresponding to the
failure of 7–8 snooks. For a mine of depth 120 m, L1 varies from about
13 m for a thin beam of thickness 0.1 m to about 260 m for a thick
beam with h¼2 m. Also, L1 ¼ 150 m for a beam of thickness
h¼1.16 m. These values are in the correct order of magnitude range.
For a mine of depth 500 m, L1 increases from about 6 m to 127 m as
the thickness of the beam, h, increases from 0.1 m to 2 m while for a
mine of depth 1000 m L1 increases from 4.5 m to 90 m over the same
range of beam thicknesses. In deep mines k may be less than 2 which
will increase L1 .
The bord or distance between pillars in a coal mine ranges from
5 m to 7 m and the width of a pillar ranges from 10 m to 20 m
depending on the depth of the mine. Taking average values, when
the bord is 6 m and the pillar width is 15 m, a roof layer of 150 m
between two pillars would correspond to the extraction of 7 pillars
and the failure of the snook remnants. Roof layers which crack
137
with lengths less than L1 will fracture at the clamped end points
supported by the pillars which have not been extracted.
It is of interest to calculate B for a range of beam lengths. This
will relate the bending moment and curvature and also the curves
in the figures to the beam length.Consider the parameter values in
ð44Þ (49) with H¼ 120 m and h ¼1.57 m which is chosen to give
L¼ 150 m when B ¼ 2π ¼ 6:28. Then when L ¼50, B ¼2.09 and
when L¼100 m, B ¼ 4.61.
ð45Þ Consider now values of the beam number B 42π. From Fig. 5
we see that there are several internal maxima of less magnitude
indicating that the beam may fracture at multiple internal points.
When the beam number lies in the range 2ðn1Þπ oB o 2nπ then
ð46Þ from (47),
 1=2
L kρgH
2ðn1Þπ o3:46 o 2nπ: ð51Þ
h E
We take for L the panel width and therefore L ¼150 m and use the
parameter values in (49). Then for the beam to buckle in the nth
mode its thickness h must lie in the range
pffiffiffiffi pffiffiffiffi
0:11 H 0:11 H
o ho : ð52Þ
n n1
For a shallow mine with H¼ 120 m and the mode n¼3,
0:4 m o h o0:6 m while for the mode n¼ 6, 0:2 m o h o 0:24 m.
The higher modes of buckling require a thinner beam. For a deeper
mine with H¼1000 m and the mode n¼ 3, 1:16 m oh o 1:74 m
while for the mode n¼6, 0:58 m oh o 0:7 m. In deeper mines the
corresponding modes of buckling require a ticker beam.
Van der Merwe [1] has considered the formation of a goaf in
coal mining which occurs in the absence of support for overburden
in mined areas. A plate model was considered since the advance of
the mining face is approximately equal to the panel width.
Although van der Merwe observed that the beam analogy to
simulate plate behaviour is then not valid it is of interest to
compare numerical estimates. It was shown that bending induced
tension is the most likely mode of failure in an unjointed roof
layer. With the aid of analytical results of Obert and Duval [3] for
unjointed layers, van der Merve obtained estimates for the mini-
mum span required for failure of the roof layer. If the minimum
span is less than the panel width, then failure will occur. The
minimum span for failure depended on the thickness of the layer.
At a depth of 100 m and with k ¼2 the minimum span for failure of
a 30 m thick layer was found to be approximately 140 m while for
a 10 m thick layer it was approximately 70 m. Although the
ð49Þ
thickness of the roof layers considered are greater than the beam
thickness the estimates for the minimum span for failure are the
same order of magnitude as the length L1 .
ð50Þ
6. Voussoir beam model
The analysis of an elastic beam without joints and fractures
greatly underestimates the stability of the beam [4]. After cracking
further deformation takes place and the beam again become
stable. Evans [7] introduced the voussoir beam model to under-
stand the stability of the fractured beam. When the beam can no
longer sustain tensile stresses a compressive arch forms within the
beam stretching from the lower beam surface at the pillars to the
upper surface at midspan. The theory has been developed further
by several authors [1,4,8,9] and numerical approaches have been
considered in [15,16].
The roof collapse in secondary mining is mainly due to the
failure of a voussoir beam. There are four primary modes of failure
of the voussoir beam [2,4,17]. They are buckling (snap-through)
failure, lateral compression or crushing of the central voussoirs or
the voussoirs at the pillars, shear slip failure at the pillars and
138 C.P. Please et al. / International Journal of Rock Mechanics & Mining Sciences 64 (2013) 132–138
diagonal fracturing. Shear slip occurs in thick beams with small
length to thickness ratios ðL=h o 10Þ while buckling and crushing
failures occur in thin beams with larger length to thickness ratios.
When the beam number B o2π we have seen that the elastic
beam will first crack at the pillars. For values of B 42π the
magnitude of the curvature has internal maxima which could lead
to fracturing at multiple internal points if the tensile strength of
the beam is exceeded. When the beam first fractures at the pillars,
there will be a change in the boundary conditions from clamped to
hinged which will determine the progression of the fractures [3,4].
7. Discussion and conclusions
A simple model of the behaviour of the roof layer during the
first stage of pillar extraction has been given that accounts for the
compressive stresses created by the overburden and due to the
weight of the roof layer. Several pillars, as many as seven or eight,
are extracted and replaced by snooks before these pillar remnants
fail violently due to the increased load acting on them. The roof
layer forms a beam clamped by the pillars at each end. The
analysis has indicated that such a beam could fracture adjacent
to the supports at the pillar at each end of the beam and that this
would occur when the stresses exceed the tensile strength of
the beam.
The model could be developed further by investigating the
progression of fractures in the beam and the transition from a
continuous elastic beam, with an axial compressive force acting at
each end, to a voussoir beam intersected by cracks and joints.
A roof made of a series of thin layered beams could also be
considered where some stresses are transferred between the
beams. This should allow the fracturing of the roof to be con-
sidered with some account of the structure of the rock layers.
Acknowledgments
We thank the reviewer for valuable and constructive comments
and for bringing to our attention the significance of the voussoir
beam model in the simulation of the roof layer over a mine span.
The problem of pillar extraction in a coal mine was submitted to
the Mathematics in Industry Study Group which was held at the
University of the Witwatersrand, Johannesburg, from 10 to 14
January 2011. The authors thank R. Kgatle, B. Matebese, K.R. Adem,
E. Yohana, A.R. Adem, B. Bekezulu, M. Hamese and B. Khumalo for
their participation in the research group.
References
[1] van der Merve JN. Fundamental analysis of the interaction between over-
burden behavior and snook stability in coalmines. Journal of South African
Institute of Mining and Metallurgy 2005;105:63–73.
[2] Brady BHG, Brown ET. Rock mechanics for underground mining. 2nd ed.
Chapman & Hall: London; 1993 [chapter 8].
[3] Obert L, Duvall WI. Rock mechanics and the design of structures in rock. John
Wiley and Sons; 1966 [chapter 5].
[4] Diedericks MS, Kaiser PK. Stability of large excavations in laminated hard rock
masses: the voussoir analogue revisited. International Journal of Rock
Mechanics and Mining Sciences 1999;36:97–117.
[5] Simpson B, Ahmed M. Failure of a linear Voussoir arch: a laboratory and
numerical study. Canadian Geotechnical Journal 1992;29:188–94.
[6] Milne D. Underground design and deformation based on surface geometry.
Ph.D. Thesis, Mining Department, University of British Columbia, Canada;
1996.
[7] Evans WH. The strength of undermined strata. Transactions of The Institution
of Mining and Metallurgy 1941;50:475–500.
[8] Beer G, Meek JL. Design curves for roofs and hanging walls in bedded rock
based on ‘voussoir’ beam and plate solutions. Transactions of The Institution of
Mining and Metallurgy 1982:A18–22.
[9] Sofianos AI. Analysis and design of an underground hard rock voussoir beam
roof. International Journal of Rock Mechanics and Mining Sciences and
Geomechanics Abstracts 1996;33:153–66.
[10] Fayol M. Sur les movements de terrain provoques par l'exploitation des mines.
Bulletin Society of Industries and Mining 1885;14:818.
[11] Jaeger JC, Cook NGW, Zimmerman RW. Fundamentals of rock mechanics. 4th
ed.Oxford: Blackwell; 399–403.
[12] Jaeger JC, Cook NGW. Fundamentals of rock mechanics.London: Methuen;
356–357.
[13] Segel LA, Handelman GH. Mathematics applied to continuum mechanics.
New York and London: Macmillan; 193–219.
[14] McCutchen WR. Some elements of a theory of in situ stress. International
Journal of Rock Mechanics 1982;19:201–3.
[15] Alejano LR, Taboada J, Bastante FG, Rodriguez P. Multi-approach back-analysis
of a roof collapse in a mining room excavated in stratified rock. International
Journal of Rock Mechanics and Mining Sciences 2008;45:899–913.
[16] Lorig L, Brady BHG. An improved procedure for excavation design in stratified
rock. In: Proceedings of the 24th US symposium on rock mechanics—theory –
experiment – practice; 1983. p. 577–85.
[17] Sterling RL. The ultimate load behaviour of laterally constrained rock beams.
The state of the art in rock mechanics. In: Proceedings of the 21st US
symposium on rock mechanics; 1980. p. 533–42.