1

Mechanics of Deformable Materials

The Analysis of Stresses and Strains in Rocks

1. Introduction to stress

In its broadest sense, stress is force per unit area. In rock mechanics applications the units kN/m2
(kilonewtons per square metre) and MN/m2 (Meganewtons per square metre) or kPa (kilopascals)
and MPa (Megapascals) are usually adopted for stress (1 pound per square inch = 6.895 kPa, 1 MPa
= 145.0 psi).

If a uniformly distributed force N acts normal to a surface of area A, the normal stress  is given by

 = N/A (1.1)

The normal stress value  is sometimes referred to as the total stress. This terminology is adopted
because there might be a water pressure u acting across the surface. The effective normal stress '
acting on the surface is given by

' =   u (1.2)

The term ‘effective normal stress’ is adopted because the stress carried by water pressure plays no
part in mobilising frictional shear strength on the surface; it is only the effective stress developed by
grain to grain contacts that serves to mobilise frictional shear strength. Although the effective stress
principle is absolutely crucial in saturated and partially saturated soils, which have an interconnected
pore matrix, the principles of effective stress are not applicable to all rock masses. Crystalline rock
materials are effectively impermeable and do not have an interconnected pore matrix that allows
water pressure to exert its effect. Moreover, the presence of a fracture network, which is usually
responsible for transporting water through the rock mass, will apply a further level of complexity to
the problem. Throughout this document, although the dash convention has been omitted for brevity,
the stress and strength parameters can be taken as effective stress parameters.

If a uniformly distributed force S acts parallel to a surface of area A, the shear stress  is given by

 = S/A (1.3)

There is no need to specify total and effective shear stress values because water cannot transmit shear
stresses. Shear stress levels are, therefore, independent of the hydrostatic pressure.

The normal and shear forces within rock masses vary from place to place, so the assumption of a
uniformly distributed force will only be valid if the area A of the surface under consideration is very
small. Under these conditions the stress is referred to as infinitesimal stress and can be considered
to be uniform across the area under consideration.

2. Geological rock stress

The density of the Earth’s crustal rocks averages about 2.75 tonnes/m3, which is equivalent to a unit
weight of about 0.027 MN/m3. We would, therefore, expect the vertical stress in typical rock masses
to increase by about 2.7 MPa for every 100 metres of depth. Stress measurements show this
approximation generally holds, but that there can be substantial variations. For example, measured
vertical stress levels at depths of 200 m in Australian crustal rocks can range from about 4 MPa up to
18 MPa, while measured stresses at depths of 1000 m range from about 6 MPa up to about 27 MPa.
 2

These variations from the ‘ideal’ model are probably due to the wide range of geological processes
that continuously displace and deform crustal rocks. Such processes include plate tectonic activity,
mountain building processes, displacement along faults, volcanic activity and metamorphism. These
geological processes also probably account for the wide variation in measured horizontal stress
levels. We define the ratio K between horizontal stress h and vertical stress v as follows

K = h/v (2.1)

Within about 500 m of the Earth’s surface, measured values of K generally lie in the range 1.0 to 2.5.
As depth increases, K values gradually decline, such that at 1500 m they lie in the range 0.5 to 1.5.
The relatively high values of horizontal stress are probably associated with the lateral stresses
generated by plate tectonics and mountain building processes.

Rock temperature increases substantially with depth due to the heat given off by the slow decay of
radioactive isotopes in the rocks. At high temperatures rocks become soft and eventually melt.
Fluids do not have the capacity to support shear stresses, so at depths greater than about 2 km the
stress state progresses towards a hydrostatic state, where there are no shear stresses and the normal
stresses in all directions are the same. Under these conditions K = 1 in Equation 2.1.

The radius of the Earth is 6,378 km and it has a mass of 5.976 x 1021 tonnes. So the volume of the
Earth is 1.08678 x 1021 m3 with an average density of 5.5 tonnes/m3, or an average unit weight of
0.054 MN/m3. Hence, in theory, the stress level at the centre of the Earth is approximately 350,000
MPa. This estimate is, however, complicated by the fact that as stress levels increase with depth, the
rock is compressed substantially, which increases its unit weight. Another complication is that
gravity, and therefore unit weight, decrease with depth to a theoretical value of zero at the centre of
the Earth. One cubic metre of typical rock at the Earth’s surface weighs approximately 2.75 tonnes
with a unit weight of 0.027 MN/m3. If this cubic metre of rock were taken to a depth of 100 km it
would be theoretically subjected to a stress of about 2,700 MPa, would be compressed to a volume of
about 0.91 m3 and would have a unit weight of about 0.03 MN/m3. At a depth of 1,000 km it would
be theoretically subjected to a stress of about 37,000 MPa, would be compressed to a volume of
about 0.58 m3 and would have a unit weight of about 0.05 MN/m3. These calculations are, however,
somewhat academic. It is unlikely that humans will ever travel to a depth greater than about 10 km
below the Earth’s surface. Setting aside the difficulties of high temperatures and the provision of
adequate ventilation, at a depth of 10 km the stresses in the walls of a hypothetical 10 m diameter
shaft would be over 500 MPa, which would crush even the strongest rocks.
 3

3. Measuring stress

Unlike other inherent properties of matter, such as mass, length, temperature and colour, it is not
possible to measure stress directly. In order to measure rock stress it is necessary to cause it to
change. There are three general approaches to measuring stress.

1. A borehole is drilled to a depth of between 10 and 500 m. Strain gauges for measuring
surface rock strains, or transducers for measuring changes in borehole diameter, are then
installed. The stresses in the instrumented zone are then relieved by extending the borehole
directly or by over-coring. The measured strains (or displacements) are then processed to
calculate the rock stress.
2. A borehole drilled to a depth of up to 5 km is pressurised with fluid over a specific length of
the hole. The pressure required to fracture the rock gives an indication of the stress
magnitude, while the orientation of the resulting fracture(s) gives an indication of the
orientation of the major and minor stresses.
3. There is growing evidence that rocks have a mechanical ‘memory’ of stresses they have been
exposed to. A core of rock will continue to swell after it has been recovered from the
borehole. Measurement of this anelastic strain recovery can be used to estimate stress levels.
There is also evidence that rock specimens exhibit a change in behaviour when they are re-
loaded to stresses up to, and then beyond their in-situ stress levels.

In three dimensions there are three independent normal stresses and three independent shear stresses.
In order to measure the complete state of stress it is therefore necessary to make six independent non-
parallel measurements. Where a rock mass is subjected to hydrostatic stress all the normal stresses
are the same and the shear stresses are all zero, so only one measurement would be required. Stress
measurement techniques will be considered in more detail in the Rock Mechanics module.

4. Some basic principles of vector algebra

In order to understand the principles of stress analysis in three dimensions it is first necessary to
revise some of the basic principles of vector algebra.
 4

X horizontal north

Vx

Vy
a
b Y horizontal east
g
V

Vz

Z vertical down

Figure 4.1 Cartesian coordinate system

We will adopt a right-handed Cartesian* (orthogonal) coordinate system, in which

 The positive direction of the X axis is horizontal to the north

 The positive direction of the Y axis is horizontal to the east
 The positive direction of the Z axis is vertically down

[*Cartesian is the latinised form of the surname of René Descartes, French philosopher and
mathematician (1596–1650). We recognise this etymology by using the capital ‘C’.]

This XYZ Cartesian coordinate system reflects global geographical directions, so it is usually
referred to as a Global Cartesian coordinate system. The Cartesian coordinate system is also
right-handed. The ‘handedness’ of a Cartesian coordinate system relies on an ordering of the axes
(X,Y,Z, or L,M,N or 1,2,3 etc). If we associate the first axis (X, L or 1 etc) with our index finger,
the second axis (Y, M, 2 etc) with our middle finger) and the third axis (Z, N, 3 etc) with our thumb
we can create a representation of a Cartesian coordinate system with our hand by positioning these
digits at right angles to each other. If the physical alignment of the axes corresponds to our right
hand, then we have a right-handed system. A right-handed system can be converted to a left-handed
system by reversing the direction of any one axis; for example the above system would be left-
handed if the Y axis were changed to an orientation of horizontal to the west. We normally work
exclusively with right-handed Cartesian coordinate systems. Note that if we face north, then look
down along the Z axis, the X axis is ‘vertical’ and the Y axis is ‘horizontal’, which is the opposite
way round from their ‘usual’ orientation on two-dimensional graphs. This minor anomaly is caused
by adopting the convention ‘positive Z down’. Although the global coordinate system defined above
will be used for all formal calculations in these notes, some authors adopt coordinate systems with
the Z axis vertically up to enhance the visibility of three-dimensional sketches. It is sometimes
convenient to orientate our coordinate axes in alignment with our local mine coordinates, some
existing geological structure, our excavation geometry or indeed our measured stress directions. In
this case we describe the coordinate system as a Local Cartesian coordinate system, and if feasible
 5

label the axes LMN, or some nomenclature other than XYZ, to emphasise the local nature of these
axes.

A vector is a quantity defined by both magnitude and orientation, such as force or acceleration.
Consider a vector V of magnitude |V| orientated such that it makes an angle a with the positive end
of the X axis, b with the Y axis and g with the Z axis, where a b and g all lie in the range 0 to 180º.
These angles are usually referred to as direction angles.

The components of V along each of the Cartesian axes are given by

Vx = |V| cos a
Vy = |V| cos b  
Vz = |V| cos g

The cosine terms in Equations 4.1 are referred to as direction cosines, as follows

vx = cos a
vy = cos b  
vz = cos g
hence

Vx = |V| vx
Vy = |V| vy  
Vz = |V| vz

We can consider the same vector V in a different set of Cartesian axes, LMN, inclined to the XYZ
axes. In this inclined set of axes the components of V are Vl Vm Vn. If we know the direction
cosines of the L axis the M axis and the N axis relative to the X,Y,Z system we can calculate Vl Vm
Vn from Vx Vy and Vz.

Direction cosines relative to

the axes
Coordinate axis X Y Z
L lx ly lz
M mx my mz
N nx ny nz

We can regard V as being represented by its three separate orthogonal component vectors Vx Vy and
Vz in the XYZ Cartesian system. So Vl is given by the sum of the components of Vx Vy and Vz in
the direction of the L axis, hence

Vl = Vx lx + Vy ly + Vz lz
Similarly
Vm = Vx mx + Vy my + Vz mz (4.4)
Vn = Vx nx + Vy ny + Vz nz

The above three equations can be written in matrix form

 6

 V   x y z  Vx 
V   m 
m z  Vy 
 m  x my (4.5)
 Vn   n x ny n z   Vz 


or [V*] = [R] [V] (4.6)

Equations 4.4 to 4.6 are referred to as the transformation equation of vectors. The three by three
matrix [R] is called the rotation matrix.

 x y z 
R   m x my mz 

(4.7)
n ny n z 
 x

To perform the reverse rotation we can regard V as being represented by its three separate orthogonal
component vectors Vl Vm and Vn in the LMN Cartesian system. So Vx is given by the sum of the
components of Vl Vm and Vn in the direction of the X axis, hence

Vx = Vl lx + Vm mx + Vn nx
Similarly
Vy = Vl ly + Vm my + Vn ny (4.8)
Vz = Vl lz + Vm mz + Vn nz
or
Vx  x mx n x   V 
 V    my n y  Vm 
 y  y (4.9)
 Vz   z mz n z   Vn 

or [V] = [R]T [V*] (4.10)

where [R]T is the transpose of [R]

but [V] = [R]-1 [V*] (4.11)

where [R]-1 is the inverse of [R]

so [R]-1 = [R]T (4.12)

The fact that the inverse of [R] is given simply by its transpose is a special property of the rotation
matrix.

The analysis presented in this section will form a crucial component of the transformation of stress in
three dimensions.

5. Orientation of lines and planes

Before we embark upon the analysis of stress and strain in three dimensions it is important to specify
a clear and unambiguous three-dimensional reference system. It is not particularly important which
particular reference system we adopt, as long as we specify it clearly and use it consistently.
 7

The orientation of any line can be specified in a spherical coordinate system by reference to two
angles; the plunge and trend, which are here defined as follows

Plunge (-90º < b < 90º) is the acute angle, measured in a vertical plane, between a given inclined
line and the horizontal. A line that is directed downwards, ie below the horizontal, has a positive
angle of plunge; an upward directed line has a negative angle of plunge.

Trend (0 < a < 360º) is the geographical azimuth of the vertical plane that contains a given inclined
line, ranging from north (a = 0 and 360) through east (a = 90) south (a = 180) and west (a =
270). Any vertical plane has two geographical azimuth directions 180 apart; trend is measured in
the same direction as plunge.

Trend and plunge are usually recorded simply as a three digit number followed by a two digit
number separated by a forward slash, eg 145/68, 035/15.

The line of maximum dip of a plane is the specific line in a plane that has the steepest angle of
plunge. The line of maximum dip is the downward directed end of this steepest line. The trend and
plunge of this line of maximum dip are referred to as the dip direction ad and the dip angle bd of
the plane.

The strike as of a plane is the trend of the specific line in a plane that is horizontal, ie has zero
plunge, therefore as = ad  90.

The normal to a plane is the downward or upward directed line that makes an angle of 90 to a
specified plane. The trend and plunge of the downward directed normal to a plane are, therefore,
respectively an = ad  180 and bn = 90 - bd. The trend and plunge of the upward directed
normal are given by a'n = ad and b'n = bd - 90, hence a'n = an  180 and b'n = -b n.

We will adopt the right-handed XYZ Cartesian coordinate system defined earlier

 The positive direction of the X axis is horizontal to the north (trend/plunge 000/00)
 The positive direction of the Y axis is horizontal to the east (trend/plunge 090/00)
 The positive direction of the Z axis is vertically down (trend/plunge 000/90)

A given line of trend/plunge au/bu will have Cartesian components ux uy uz given by

ux = |u| cos au cos bu

uy = |u| sin au cos bu }(5.1)
uz = |u| sin bu

where |u| is the magnitude of the vector. When we are dealing with purely orientation data we set |u|
= 1.0. When we are dealing with full vectorial data, such as force, acceleration and geometrical
properties, |u| will be the magnitude of the vector, with appropriate units, kN, m/sec2, m etc. The
magnitude |u| of the vector can be found from the Cartesian components as follows

|u| = ( ux2 + uy2 + uz2) (5.2)

The components of a general vector can be converted to the components of a unit vector simply by
dividing each component by the vector magnitude. The components of this unit vector, u1x u1y u1z,
which are referred to as the direction cosines, are therefore given by
 8

u1x = ux/|u| u1y = uy/|u| and u1z = uz/|u| (5.3)

These components of the unit vector are called direction cosines because they give the cosine of the
angles between the vector and each of the Cartesian axes. These angles, which are measured to the
positive end of each of these axes, are usually called the direction angles, as shown in Figure 4.1.

A vector whose z component is negative has an upward direction. In order to plot this vector on a
lower hemisphere stereographic projection it would be necessary to determine the components of the
reverse, downward directed, vector. These reverse components u'x u'y u'z are simply given by

u'x = ux u'y = uy and u'z = uz (5.4)

A vector with Cartesian components ux uy uz acts along a line of trend/plunge au/bu given by

au = arctan(uy/ ux) + Q
}(5.5)
bu = arctan(uz/(ux2 + uy2))

where Q is a quadrant parameter that ensures au lies in the correct quadrant and in the range 0 to
360. The value of Q depends on the signs of ux and uy as follows:

ux uy Q
0 0 0
<0 0 180
<0 <0 180
0 <0 360

The above tabulated values of Q can be evaluated in Excel with the following nested IF statement:
=IF(A1<0,180,IF(A2>0,0,360))
where ux is in cell A1 and uy is in cell A2.

Alternatively, au can be determined by using the Excel ATAN2 function, as demonstrated in the
worksheet ‘Orientation’ in the attached Excel spreadsheet ‘Angles and Components’. This
worksheet also demonstrates the application of Equations 5.1 to 5.5, and the use of an ‘IF’ statement
to ensure that au lies in the range 0 to 360. This last conversion is only necessary for the final
results, and is not required for intermediate calculations, since for example cos a = cos (a + 360).
The Excel functions RADIANS and DEGREES can be used to convert between radians and degrees.
Alternatively it may be more convenient to divide or multiply by 180/PI().

Angle of pitch

The angle of pitch p of a line in a plane is the acute angle, measured in the plane, between the line
and the strike of the plane. A plane will, of course, have two strike directions given by as = ad + 90
and as = ad  90, so it is necessary to specify from which strike direction the angle of pitch has been
measured. A line with an angle of pitch p in a plane of dip direction/dip angle ad/bd will have a
trend ap and plunge bp found as follows

If p is measured from as = ad + 90, then ap = ad + arctan(1/(tan p cos bd)

}(5.6)
If p is measured from as = ad  90, then ap = ad  arctan(1/(tan p cos bd)
 9

bp = arcsin(sin p sin bd) (5.7)

Equations 5.6 and 5.7 are implemented in the worksheet ‘Pitch’ in the attached spreadsheet ‘Angles
and Components’.

Angle between two lines

The angle uv between two lines of trend/plunge au/bu and av/bv can be found from the following
expression:

cos uv = [cos(au  av) cos bu cos bv] + [sin bu sin bv] (5.8)

Alternatively, if the direction cosines of the two lines are respectively u1x u1y u1z and v1x v1y v1z, the
angle uv can be found from the dot product (or scalar product) as follows:

cos uv = (u1x v1x) + (u1y v1y) + (u1z v1z) (5.9)

It is important that direction cosines are used in Equation 5.9. If the angle between a pair of non-unit
vectors is required then Equation 5.3 must be applied to convert the components to direction cosines.
There will, of course, be two angles made by the intersection of a pair of lines; one obtuse (cos uv =
ve) and the other acute (cos uv = +ve), adding to 180. The absolute value of cos uv will,
therefore, always yield an acute angle.

Plane defined by two lines

If two lines of trend/plunge au/bu and av/bv, and direction cosines u1x u1y u1z and v1x v1y v1z are
known to lie in the same plane, then the components wx wy wz of the normal to this plane are given
by the vector product as follows

wx = (u1y v1z) – (u1z v1y)

wy = (u1z v1x) – (u1x v1z) }(5.10)
wz = (u1x v1y) – (u1y v1x)

The trend and plunge of this normal can be found from Equations 5.5. The magnitude of the vector
|w| = ( wx2 + wy2 + wz2) is the sine the angle between the two lines. Equations 5.10 are useful for
identifying the plane defined by the intersection of a pair of lines. An additional application of these
equations is their use in defining the third axis in an inclined Cartesian coordinate system.

Intersecting planes

If a1x a1y a1z are the direction cosines of the normal to plane A, and b1x b1y b1z are the direction
cosines of the normal to plane B, then the components of the line of intersection between these two
planes, ix iy iz can be found from the vector product as follows:

ix = (a1y b1z) – (a1z b1y)

iy = (a1z b1x) – (a1x b1z) }(5.11)
iz = (a1x b1y) – (a1y b1x)

The similarity between Equations 5.10 and 5.11 can be explained as follows: The normals to the two
intersecting planes lie in a common plane. The normal to this common plane is the line of
intersection between the two planes. Equations 5.10 and 5.11 therefore employ the vector product to
perform essentially the same calculation.
 10

The magnitude of the vector given by Equations 5.11 |i| = ( ix2 + iy2 + iz2) is the sine of the angle
between the two planes, or to be more precise the sine of the angle between the normals to the two
planes. Equations 5.10 and 5.11 can yield an upward or downward directed vector, depending on the
orientations of the lines and planes involved. This vector can, of course, be reversed if necessary for
plotting on a lower hemisphere projection. If the components of the normals in Equations 5.11 are
not direction cosines, ie the magnitude of the vector is not equal to 1.0, then ix iy iz will still give the
correct orientation for the line of intersection. In this case the sine of the angle between the planes is
given by

|i| = ( ix2 + iy2 + iz2)/(|a| |b|) (5.12)

where |a| and |b| are respectively the magnitudes of the vectors a and b. Equations 5.8 to 5.12 are
implemented in the worksheet ‘Line Pairs’ in the attached spreadsheet ‘Angles and Components’.

Inclined Cartesian coordinate systems

In mining and civil engineering rock mechanics applications it is often necessary to set up Cartesian
coordinate systems that are in alignment with local geological or excavated structures. The
orientations of these local inclined Cartesian coordinate systems must be specified precisely in
order to avoid errors in vector calculations. Local Cartesian coordinate systems can usually be
specified by reference to a plane of known orientation, such as the plane of a fault, the plane that is
normal to a development drive or the plane defined by the intersection of two boreholes etc.

Example 5.1

An inclined borehole of trend/plunge 105/55 intersects an inclined shaft of trend/plunge

330/60. Determine the trends/plunges of a right-handed L,M,N Cartesian coordinate system in
which the N axis is the downward-directed normal to the plane that contains the axes of the
borehole and the shaft, the M axis has a pitch of 35 from the eastern end of the strike of this
plane, and the L axis completes the right-handed system. (This Example has been constructed
to illustrate the key concepts explained in Section 5.)

Solution

The worksheet ‘Line Pairs’ in the attached spreadsheet ‘Inclined Coordinate Systems Example
5.1’ can be used to identify the orientation of the plane that contains the axes of the borehole
and the shaft by implementing Equations 5.10. The normal to this plane has a trend of
035.219 and a plunge –13.604. The downward-directed end of this normal, which is our N
axis, will, therefore have a trend of 035.219 + 180 = 215.219and a plunge of 13.604. The dip
direction/dip angle of this plane that contains the axes of the borehole and the shaft will
therefore be 035.219/76.396. The worksheet ‘Pitch’ in the attached spreadsheet ‘Inclined
Coordinate Systems Example 5.1’ can now be used to determine the trend/plunge of the line
that has a pitch of 35 from the eastern end of the strike of this plane (125.219). The result
gives the trend/plunge of our M axis as 115.867/33.882. The Cartesian axis system can be
completed by again applying Equations 5.10 as implemented in the lower part of the worksheet
‘Line Pairs’ in the attached spreadsheet ‘Inclined Coordinate Systems Example 5.1’. In this
case we need to identify the normal to the plane that contains our M and N axes; this normal
will define our L axis which has a trend/plunge 323.787/52.766. Finally we need to check
that we have a right-handed system, which is indeed the case. If we had generated a left-
handed system we would simply need to reverse the direction of the L axis to create a right-
handed system.
 11

6. Three-dimensional stress

To analyse three-dimensional stress it is convenient to consider the stresses produced by forces

acting on the faces of a very small cubic element within the material. It is usual to orientate the
corners of the cube so that they are parallel to the specified axes of a Cartesian coordinate system.
Since the faces of the cube are very small it can be assumed that the stress values do not change
significantly from one side of the cube to the other.

X
zz
zx
zy
Y
yz
yy
xy 
yx

xz
xx

Figure 6.1 Stress cube

Figure 6.1 shows a cubic element aligned with its edges parallel to a right handed XYZ Cartesian
coordinate system. The three faces of the cube that each have a corner touching the origin of the
coordinate system are here referred to as ‘origin faces’; the other three faces are referred to as
‘obverse faces’. Each face can be identified by the coordinate axis that runs normal to it. So in
Figure 6.1 we can see the origin-x, origin-z and obverse-y faces.

On each face of the stress cube there are force components acting in the X, Y and Z directions.
Dividing these force components by the area of the face produces three stress components on each
face of the cube, giving a total of 18 stress components. The positive directions of nine of these
stress components, acting on the three visible faces, have been drawn and labelled in Figure 6.1.

It is important to spend some time understanding the nomenclature and sign convention. Taking a
typical stress component xy the Greek symbol ‘sigma’ indicates this is a stress component. The first
subscript x indicates this stress component acts on a plane whose normal is in the x direction. The
second subscript y indicates this stress component acts along the y direction. The stress components
xx, yy and zz act along the normals to their respective faces and are referred to as normal stresses.
For brevity the second subscript is sometimes omitted, so the normal stresses become x, y and z.
The other stress components, such as xy yz etc act parallel to their respective faces and are referred
to as shear stresses. To emphasise the shearing nature of these stresses the Greek symbol ‘sigma’ is
often replaced by the Greek symbol ‘tau’ to give components xy yz etc.
 12

The sign convention for normal stresses is:

Compressive normal stresses are positive.

This sign convention is different from the more logical system adopted by mechanical engineers
where tensile normal stresses are taken to be positive. The logic of this latter sign convention is that
tensile stresses produce elongation, which is a positive increment in length. The logic in taking
compressive stresses as positive in rock mechanics is that almost all stresses in rock are compressive,
adopting the mechanical engineering convention would necessitate a pointless repetition of negative
signs. So, for example, normal stress of 1.45 MPa is taken to be compressive; a normal stress of –
6.75 MPa is tensile.

The sign convention for shear stresses is:

(a) If the positive normal stress on a given face acts in the positive direction of the
coordinate axis, then positive shear stresses on this face act in the positive directions of
the relevant coordinate axes.
(b) If the positive normal stress on a given face acts in the negative direction of the
coordinate axis, then positive shear stresses on this face act in the negative directions of
the relevant coordinate axes.

Case (a) applies to the three origin faces of the stress cube; case (b) applies to the three obverse faces
of the stress cube.

In order to satisfy static equilibrium of the stress cube, ie such that it does not displace or rotate, we
require that
xx,origin = xx,obverse
yy,origin = yy,obverse }(6.1)
zz,origin = zz,obverse
and also
xy,origin = yx,origin = xy,obverse = yx,obverse
yz,origin = zy,origin = yz,obverse = zy,obverse }(6.2)
xz,origin = zx,origin = xz,obverse = zx,obverse

These equalities mean that there are only six independent stress components, which define the
stress matrix, as follows

 xx  xy  xz 
  y x  y y  y z (6.3)
 
 zx  zy  zz 

In view of the equalities outlined above, the stress matrix is symmetrical, specifically xy = yx, yz =
zy and xz = zx. Although we have taken care in defining our nomenclature, these equalities mean
that the ordering of the subscripts for our shear stresses is not really important.

Each of the three groups of two normal stresses in Equations 6.1 can be regarded as a single normal
stress. The origin and obverse components are of equal magnitude but act in opposite directions to
ensure static equilibrium against translation of the stress cube.
 13

Each of the three groups of four shear stresses in Equations 6.2 can be regarded as a single shear
stress. The origin and obverse components are of equal magnitude but act in opposite directions to
ensure static equilibrium against rotation of the stress cube.

If a shear stress xy is computed to have a value of, say, 2.73 MPa then the actual directions of action
of the four components (xy,origin = yx,origin = xy,obverse = yx,obverse) are the same as the positive shear
stress directions drawn on the stress cube. If a shear stress xy is computed to have a value of, say,
3.85 MPa then the actual directions of action of the four components are in the opposite direction
to the positive shear stress directions drawn on the stress cube. This property of shear stresses is one
of the more confusing, and least understood, aspects of stress analysis.

7. Stress transformation

The introduction to three-dimensional stress in the previous section was based on the XYZ Cartesian
coordinate axis system. It is, of course, permissible to adopt any system of coordinate axes to
specify the state of stress. For example, we might wish to specify the stress state [] in Equation 6.3
by reference to the LMN Cartesian coordinate system introduced earlier

    m  n 
 *  m  mm  mn  (7.1)
  n  nm  nn 

It is important to emphasise that in defining the stress state [*] the fundamental stress state has
not been changed; it is simply being expressed relative to a different set of Cartesian coordinate
axes.

In order to understand the relation between [] and [*] we must first appreciate that in transforming
stress from one set of axes to another we are changing two things simultaneously: (i) the components
of the stresses relative to the coordinate axes, and (ii) the areas of the planes on which these stresses
act. We have developed a solution for the transformation of vector components in Section 4; we will
now look at the transformation of areas.
 14

Y
R
Q X
ay
O ax
b a

Area A g
P

az S

Figure 7.1 Tetrahedron

In Figure 7.1 we are looking upwards towards the origin O of a Cartesian coordinate system, with the
X axis coming out of the page to our right and the Y axis to our left (rather like looking up into the
corner of a room). A plane QRS of some general orientation and area A has been positioned such
that the normal to the plane OP is of unit length. The direction angles of this normal are
respectively a, b and g for the X, Y and Z axes. The dimensions OQ, OR and OS of the tetrahedron
OQRS are respectively ax, ay and az.

The direction cosines of the unit vector OP are respectively

x = cos a y = cos b and z = cos g  

In the right-angled triangle OPQ

OP
 cos a but OP = 1 and OQ = ax
OQ
so ax = 1/x and similarly ay = 1/y and az = 1/z (7.3)

Let the areas of the right-angled triangles ORS, OQS and OQR be respectively Ax, Ay and Az

a ya z 1 a xa z 1 a xa y 1
we have A x   Ay   and Az   (7.4)
2 2  y z 2 2 xz 2 2 xy

The volume V of the tetrahedron OQRS is given by the area of the base multiplied by one third of
the vertical height, so using Equations 7.3 and 7.4

Axa x 1
V  (7.5)
3 6 x  y  z
 15

A 1
But V
3

1
so A (7.6)
2  x  y z

Finally, combining Equations 7.4 and 7.6 we obtain

Ax = A x Ay = A  y and Az = A z (7.7)

Many texts take the result in Equation 7.7 to be intuitively obvious. The above proof is provided for
those who don’t find this to be the case.

Y
R
Q X
Az zx
A Sx
O
Ax xx
Ay yx

Figure 7.2 The stress vector

Figure 7.2 shows the same view of our Cartesian coordinate system as that presented in Figure 7.1, ie
we are looking up towards the origin O (in the ‘corner of a room’). The plane QRS cuts off the
origin corner of the stress cube in Figure 6.1 to form the tetrahedron OQRS. We now consider the x
components of the stresses acting on each face of the tetrahedron, ie those stresses whose second
subscript is x. The stress xx acts on face ORS of area Ax, yx acts on face OQS of area Ay, and zx
acts on face ORQ of area Az. For static equilibrium

Ax xx + Ay yx + Az zx = A Sx (7.8)

Where Sx is the x component of the stress vector acting on face QRS, drawn reacting in the negative
x direction.

Using Equation 7.7 to substitute for Ax, Ay and Az we obtain the important result

x xx + y yx + z zx = Sx (7.9)

Similarly, in the y and z directions

 16

x xy + y yy + z zy = Sy (7.10)

x xz + y yz + z zz = Sz (7.11)

We have now defined the stress vector S (sometimes referred to as the traction vector) which has
components (Sx, Sy, Sz). Equations 7.9 to 7.11 can be expressed in matrix form as

Sx   xx  xy  xz   x 
S    
 yy  yz   y  (7.12)
 y   yx
Sz  zx  zy  zz    z 

Or [S] = [][] (7.13)

We now consider the same stress state, but expressed relative to the LMN Cartesian coordinate
axes, as in Equation 7.1, reproduced below

  m n 

 * m mm mn 
 n nm nn 

We can envisage the cutting plane QRS with the same global orientation as in Figure 7.2, but now
cutting through the stress cube expressed relative to the LMN axes. Proceeding in the same way as
before, we can derive the equivalent stress vector relation, as follows

 S    m n     

S    mn   m 
 m   m mm (7.14)
 Sn   n nm nn    n 

Or [S*] = [*] [*] (7.15)

The vectors [S] [] [S*] and [*] like any other vectors can be transformed from one set of Cartesian
axes to another, as explained in Section 4. Specifically, we define the rotation matrix [R] as in
Equation 4.7 and adopt the same sets of axes XYZ and LMN, so

[S*] = [R] [S]

Using Equation 4.12 [S] = [R]T [S*]

also [*] = [R] []

or [] = [R]T [*]

recalling that [S*] = [R] [S] and by Equation 7.13 [S] = [][]
then [S*] = [R] [] []
hence [S*] = [R] [] [R]T [*]
but by Equation 7.15 [S*] = [*] [*]

then finally [*] = [R] [] [R]T (7.16)

 17

Equation 7.16 is an example of a tensor transformation equation, confirming that mathematically the
components of stress are a second order tensor. Rotation of vector components is a first order
transformation. Rotation of stress components involves transformation of the equivalent stress
vector components but also involves transformation of the areas that the stresses act upon, hence it is
a second order transformation.

It is a relatively straight forward, but tedious, matter to substitute the full matrices in Equation 7.16
using Equations 4.7, 6.3 and 7.1. Multiplication of these matrices yields the following results

ll = lxlxxx + lylyyy + lzlzzz + (lxly + lylx)xy + (lylz + lzly)yz+ (lxlz + lzlx)xz
mm = mxmxxx + mymyyy + mzmzzz + (mxmy + mymx)xy + (mymz + mzmy)yz+ (mxmz + mzmx)xz
nn = nxnxxx + nynyyy + nznzzz + (nxny + nynx)xy + (nynz + nzny)yz+ (nxnz + nznx)xz
lm = lxmxxx + lymyyy + lzmzzz + (lxmy + lymx)xy + (lymz + lzmy)yz+ (lxmz + lzmx)xz }(7.17)
ln = lxnxxx + lynyyy + lznzzz + (lxny + lynx)xy + (lynz + lzny)yz+ (lxnz + lznx)xz
mn = mxnxxx + mynyyy + mznzzz + (mxny + mynx)xy + (mynz + mzny)yz+ (mxnz + mznx)xz

The six separate equations in Equations 7.17 are clearly all of the same form:

ab = axbxxx + aybyyy + azbzzz + (axby + aybx)xy + (aybz + azby)yz+ (axbz + azbx)xz (7.18)

where a and b each represent one of the l, m or n direction cosines.

It is generally easier to perform the matrix multiplication in Equation 7.16 by using the matrix
functions in Excel. The general form of the required Excel expression is as follows, as illustrated in
the attached spreadsheet ‘Stress Transformation Example 7.1’

=MMULT(MMULT(B24:D26,B9:D11),TRANSPOSE(B24:D26))

The first step in implementing this expression is to highlight a three-by-three matrix of cells where
the transformed stress tensor [*] will be written. The above expression is then typed into the
Formula Bar. In this particular example cells B24:D26 contain the rotation matrix [R] and cells
B9:D11 contain the stress tensor []. When the expression has been inserted press
CTRL+SHIFT+ENTER to enter the formula. Excel will add braces {} to the array formula to
indicate that it is a matrix function. The solution will appear in the form of a symmetrical three-by-
three matrix in the cells that have been highlighted. The above expression has been implemented in
the worksheet ‘Stress Trans’ in the attached spreadsheet ‘Stress Transformation Example 7.1’.

How to use the spreadsheet ‘Stress Transformation …’

An input stress tensor of any specification can be input by writing values into the leading diagonal
and upper part of the three-by-three matrix in cells B9:D11. This input stress tensor is specified by
reference to the XYZ Cartesian coordinate system. The trends and plunges of these XYZ axes can
take any values, specified in degrees, in cells B14:C16. It is, of course, important that the axes are
orthogonal and right-handed. The trends and plunges of the rotated LMN axes are specified in cells
B18:C20. Again it is important that these axes are orthogonal and right-handed. The rotation matrix
[R] is computed in cells B24:D26 by applying Equation 5.8 to the specified values of trend and
plunge for the two pairs of coordinate axes. The rotated stress tensor, specified by reference to the
LMN Cartesian coordinate system, is presented in cells B33:D35. The spreadsheet has been
designed to transform stresses from the XYZ axes to the LMN axes. To perform an ‘inverse’
transformation it is simply necessary to specify the LMN stress values and axis orientations as the
input stress tensor, and specify the XYZ axis orientations for the rotated stress tensor, as
 18

implemented in worksheet ‘Inverse Stress Trans’ in the attached spreadsheet ‘Stress Transformation
Example 7.1’.

Example 7.1

A rock mass has the following stress state in MPa, expressed relative to the Cartesian
coordinate axes X (horizontal north), Y (horizontal east) and Z (vertical down):

 xx  xy  xz  15 0 0
 
 yx  yy  yz    0 10 0
 zx  zy  zz   0 0 5


A fault plane cutting through the rock mass is known to have a dip direction/dip angle
150/65. Calculate the shear stress in the plane of the fault along its strike direction (L axis),
the shear stress in the plane of the fault along its line of maximum dip (M axis), and the
normal stress across the fault plane (N axis).

The first step is to calculate the trends/plunges of a right-handed Cartesian coordinate system LMN
(refer to Section 5). The trend of the L axis, which is the strike direction of the fault, is 150º – 90º =
60º, so the L axis has a trend/plunge 060/00. The trend/plunge of the M axis has been given as
150/65. The trend of the N axis, which is the normal to the fault, is 150º + 180º = 330º. The plunge
of the N axis is 90º  65º = 25º, so the N axis has a trend/plunge 330/25. Note this LMN coordinate
system is right handed. The orientations of the XYZ axes, the associated stress tensor, and the
orientations of the LMN axes have been input into the attached spreadsheet ‘Stress Transformation
Example 7.1’ in the worksheet ‘Stress Trans’. The calculated shear stress in the plane of the fault
along its strike direction is ln = 1.962 MPa. The shear stress in the plane of the fault along its line of
maximum dip is mn = 3.351 MPa. The normal stress across the fault plane is nn =  MPa.

It is instructive to determine the physical directions of these transformed stresses. We need to

visualise the stress state on one side of the fault. For convenience we select the upper (or south east)
side of the fault. On this side of the fault a positive (compressive) normal stress nn acts in the
positive direction of the N axis along a line of trend/plunge 330/25, so the positive shear stress ln
also acts in the positive direction of the L axis, ie along a line of trend/plunge 060/00. The shear
stress along the line of maximum dip of the fault mn is, however, negative, so it acts along the
negative direction of the M axis along a line of trend/plunge 330/65. On the lower (or north west)
side of the fault, the ‘other halves’ of the stress pairs nn, ln and mn respectively act along lines with
trends/plunges 150/25, 240/00 and 150/65. Figure 7.3 shows the physical directions of nn, ln and
mn acting on the upper surface of the fault plane.
 19

Figure 7.3 Stresses acting on upper surface of fault plane

The inverse of the above stress transformation is given in the attached spreadsheet ‘Stress
Transformation Example 7.1’ in the worksheet ‘Inverse Stress Trans’ which yields the stress tensor
that we originally started with.

8. Extreme values of shear stress in a plane

In the previous Section we learned how to transform a given stress state from an XYZ axis reference
system to some rotated LMN axis reference system. Let us imagine that the N axis corresponds to
the normal to a discontinuity plane, and that the L and M axes lie in the plane of the discontinuity, as
in Example 7.1 where the discontinuity is a fault. We specify the L axis to lie along the strike of the
plane and the M axis to lie along the line of maximum dip. By applying stress transformation we can
determine the following stresses:

nn the stress normal to the discontinuity plane

ln or ln the shear stress in the discontinuity plane acting along the L axis
mn or mn the shear stress in the discontinuity plane acting along the M axis

[We can also consider ll mm and lm (or lm) which are the normal stresses and shear stress on
planes running normal to the discontinuity plane and acting along the L axis and M axis.]

If we rotate the L and M axes in the LM plane about the N axis, the values of all stresses except nn
will change. We can implement this rotation by changing the angle of pitch of the L and M axes
from their initial values of 0º and 90º respectively in the LM plane, in increments of say 5º right
through to 360º. We will refer to these rotated axes as L' and M'. The trends and plunges of these
rotated axes can be found by applying Equations 5.6 and 5.7. There will always be two unique
angles of rotation, 90º apart, at which one of the shear stresses (l'n' or m'n') will have maximum
magnitude (positive or negative). At this angle of rotation the other shear stress (l'n' or m'n') will be
zero. This unique angle can be determined graphically, or by direct calculation as follows.

At zero rotation, with the L axis along the strike and the M axis along the line of maximum dip, the
shear stresses are given by ln and mn. The magnitude of the peak shear stress, |qr| is given by

|qr| = √(ln2 + mn2) (8.1)

 20

The orientation of this peak shear stress, expressed in terms of its angle of pitch peak measured from
the positive end of the L axis is given by

peak = arctan(mn/ln) (8.2)

If peak is positive the pitch angle is measured downwards from the positive end of the original L
axis; if peak is negative the pitch angle is measured upwards from the positive end of the original L
axis. The orientation of the line given by peak gives the physical sense of the peak shear stress on
the negative N side of the discontinuity. This peak shear stress is important because it defines the
direction of shear on the discontinuity plane in cases where the shear strength is insufficient to
support the predicted peak shear stress |qr|. The wavelength of the variation of ln and mn is 360º of
rotation (see attached spreadsheet ‘Extreme Shear Stresses’).

During this rotation the values of ll mm and lm (or lm) also change, but the wavelength of their
variation is only 180º. Zero values of lm occur at the rotation angles where ll and mm take their
maximum and minimum values. Peak values of lm occur at the rotation angles where ll and mm are
equal.

Example 8.1

An extensive, dry planar fault of dip direction/dip angle 220/35 adjacent to a mine development
is known to have the following Coulomb shear strength parameters: cohesion = 0.1 MPa and
angle of friction = 28º.

The stresses in the rock mass adjacent to the mine development are known to have the
following values:
xx = 25 MPa compressive, horizontal, trend 150º
yy = 15 MPa compressive, horizontal, trend 240º
zz = 6 MPa compressive, vertical.
xy = yz = xz = 0

a. Calculate the shear stress in the plane of the fault along its strike direction (L axis 130/00),
the shear stress in the plane of the fault along its line of maximum dip (M axis 220/35), and
the normal stress across the fault plane (N axis 040/55).
b. Calculate the magnitude and orientation of the maximum shear stress in the plane of the
fault.
c. Does the fault have sufficient shear strength to sustain this stress state? If your results
predict instability along the fault, calculate the additional normal stress, generated by cable
anchors, that would serve just to stabilise the fault.

The solution to this example is summarised in the attached spreadsheets ‘Stress Transformation
Example 8.1’ and ‘Extreme Shear Stresses’.

(a) Implementation of the stress transformation in the attached spreadsheet ‘Stress Transformation
Example 8.1’ gives the following transformed stresses:

ll = 23.830 MPa

mm = 12.824 MPa
nn = 9.346 MPa (normal stress)
lm = 2.633 MPa
ln = 1.843 MPa (shear stress along strike)
 21

mn = 4.778 MPa (shear stress along line of maximum dip)

(b) Applying Equations 8.1 and 8.2 |qr| = 5.122 MPa at a rotation angle of peak = 68.903º from +L.
This line has a trend of 194.78º and a plunge of 32.35º (see attached spreadsheet ‘Extreme Shear
Stresses’).

(c) Applying the Coulomb criterion (See Section 13), the shear strength f of the fault plane is given
by

f = c + n tan 

where c is cohesion, n is the normal stress and  is the angle of friction. Applying the given values
of c and  at the computed normal stress of 9.346 MPa gives f = 5.069 MPa, which is insufficient to
sustain the computed peak shear stress of 5.122 MPa. The additional normal stress n required just
to stabilise the fault can be found by solving the following expression

|qr| = c + (n + n) tan 

So n = (|qr|  c  n tan )/ tan  = (|qr|  f)/ tan  = 0.098 MPa

The attached spreadsheet ‘Peak Shear Stresses’ shows a graph of the variation in the stress values as
we rotate the L and M axes in the LM plane about the N axis.

9. Principal stress

For any general state of stress it is possible to implement a stress transformation such that each of the
resulting shear stresses is zero. In other words, it is possible to rotate the stress cube to find a single
unique orientation such that the shear stresses on each face are zero. The faces then become
principal planes and the normal stress on each face is called a principal stress. The line of action
of each principal stress is called a principal axis. (For example, when we stand still on slippery ice,
which cannot provide any significant shear reaction, we must ensure that the interface between our
shoes and the ice is a principal plane, otherwise we will slip over. We achieve this goal by standing
vertically.)

[Note: there is much confusion about the meaning and spelling of the words “principal” and
“principle”
“Principal” (adjective) means “most significant or important” as in “principal stress”
“Principal” (noun) means “person in charge” as in “principal of a college” or “sum of money lent”
“Principle” (noun) means “rule or convention” as in “the principle of effective stress”.]

Since there are three normal stresses there will be three principal stresses, as follows:

1 is the major principal stress, which is the largest (or least negative) stress
2 is the intermediate principal stress
3 is the minor principal stress, which is the smallest (or most negative) stress

The analysis of stresses adjacent to an excavation will usually produce a general stress tensor (where
the shear stresses are non-zero) expressed relative to a local set of axes. Strength models for the rock
are usually expressed in terms of the principal stresses, so it is important to be able to calculate the
values and orientations of the principal stresses associated with some general stress tensor in order to
assess rock stability.
 22

If our cutting plane in Figure 7.2 is a principal plane, then the shear stresses on this plane will be zero
and the stress vector must act normal to this plane. Hence the direction cosines of the normal to this
plane x, y and z will also be the direction cosines of the principal stress vector Ap, where A is the
area of the cutting plane and p is the principal stress. So the components of the stress vector will be

(Sx, Sy, Sz) = A(px, py, pz) (9.1)

Recalling Equations 7.9 to 7.11, and noting that in developing these equations we were able to
‘cancel out’ the area A of the cutting plane, we obtain

x xx + y yx + z zx = Sx = px

x xy + y yy + z zy = Sy = py }(9.2)
x xz + y yz + z zz = Sz = pz

Rearranging Equations 9.2 in simultaneous form gives

x (xx  p+ y yx + z zx = 0

x xy + y (yy  p+ z zy = 0 }(9.3)
x xz + y yz + z (zz  p= 0

Since at least one of (x, y, z) must be non-zero, the determinant of the matrix of equation
coefficients must be zero:

 xx  p   yx  zx
 xy  yy  p   zy 0 (9.4)
 xz  yz  zz  p 

Evaluating the determinant gives the following characteristic equation

p3 – I1 p2 + I2 p – I3 = 0 (9.5)

where
I1 = xx + yy + zz (9.6)
I2 = xx yy + yy zz + zz xx – (xy2 + yz2 + zx2) (9.7)
I3 = xx yy zz + 2 xy yz zx – (zz xy2 + xx yz2 + yy zx2) (9.8)

Equation 9.5 is a cubic equation so there are three distinct values of p that provide solutions. These
roots of the cubic expression are the three principal stresses. The characteristic equation is plotted in
the attached spreadsheet ‘Principal Stress Example 10.1’.

The parameters I1 I2 and I3 which have the dimensions of stress, (stress)2 and (stress)3 are called the
first, second and third invariants of stress. They are called stress invariants because they are
characteristic of the state of stress and do not depend upon the choice of axes. For example, for a
given state of stress

I1 = xx + yy + zz = ll + mm + nn = 1 + 2 + 3 and similarly for I2 and I3
 23

Although the roots of the characteristic equation can be zero or negative, they are always real (ie
never imaginary) in the mathematical sense. This special property means that we can adopt a simple
algorithm for solving the equation. We first define five further intermediate parameters:

J1 = I12 – 3I2 (9.9)

J2 = ½ (2I13  9I1I2 + 27I3) (9.10)
J3 = √(J13 – J22) (9.11)
J4 = √J1 (9.12)
 = ⅓ arctan(J3/J2) (9.13)

It is advisable to use the Excel ATAN2 function to evaluate Equation 9.13 as follows

=ATAN2(B41,B42)/3

where in this specific case J2 is in cell B41 and J3 is in cell B42

The three principal stresses 1 2 and 3 are given by the following expressions, noting that 1 ≥ 2
≥ 3

⅓ (I1 + 2 J4 cos 

  ⅓ (I1 + 2 J4 cos (   
  ⅓ (I1 + 2 J4 cos (

The above solution method does not work reliably if there is a hydrostatic element in the stress state,
ie if 1 = 2 = 3 or if 1 ≥ 2 = 3 or if 1 = 2 ≥ 3. Such stress states generally only occur in
theory. If such a stress state does occur in practice, a very small adjustment (0.001 MPa) can be
made to two of the normal stresses to deviate them from hydrostatic. Problems can also be
encountered when xx yy and zz are themselves principal stresses. Equations 9.6 to 9.14 have been
implemented in the attached spreadsheet ‘Principal Stress Example 10.1’.

We have identified the notion of hydrostatic stress, where 1 = 2 = 3. Even rocks that have no
shear strength cannot fail under a state of hydrostatic compression. The only ‘failure mechanism’
available under such a stress state is volumetric compression. Under non-hydrostatic conditions, the
extent to which 1 exceeds 3, expressed by the ratio 1/3, provides a measure of how deviatoric
the stress state has become. A highly deviatoric stress state is one of the primary factors that can
lead to rock failure, as discussed further in Sections 13 and 14.

10. Orientation of the principal stresses

The orientations of the principal axes can be determined by solving Equation 9.3, which can be
expressed in the following form


  xx  p   xy  xz   xp 

 y x  y yp  y z 
  
 y p   0 (10.1)
 
 zx  zy  
 zz  p   
 zp 
where
p = 1 for the major principal stress
p = 2 for the intermediate principal stress
p = 3 for the minor principal stress
and xp yp zp are the direction cosines of the three principal stresses in the XYZ system.
 24

The simplest way to solve Equation 10.1 is to take co-factors as follows. Firstly for the major
principal stress we set p = 1 and define

A1 = (yy – 1)(zz – 1) – yz2

B1 = yz xz – xy(zz – 1) }(10.2)
C1 = xy yz – xz(yy – 1)
D1 = √(A12 + B12 +C12) take +ve root
then
x1 = A1/D1 y1 = B1/D1 z1 = C1/D1 (10.3)

Direction cosines for the other two principal axes are found by substituting p = 2 and p = 3

When p = 2 we obtain A2 B2 C2 D2 and x2 y2 z2 for the intermediate principal stress
When p = 3 we obtain A3 B3 C3 D3 and x3 y3 z3 for the minor principal stress

Again, the above solution method does not work reliably if there is a hydrostatic element in the stress
state, ie if 1 = 2 = 3 or if 1 ≥ 2 = 3 or if 1 = 2 ≥ 3.

Equations 10.2 and 10.3 have been implemented in the attached spreadsheet ‘Principal Stress
Example 10.1’. In this spreadsheet the stress values in Equations 10.2 have been taken from a stress
tensor specified by reference to the XYZ Cartesian coordinate system. The computed orientations
for the principal axes are, therefore, also expressed by reference to this XYZ Cartesian coordinate
system. The calculated values of trend and plunge for each axis will, therefore, only be meaningful
if the XYZ axes have been specified to be horizontal north, horizontal east, and vertical down
respectively. A simple vector calculation demonstrates that the three principal axes are orthogonal.

A simple validation of the spreadsheet can be effected by inputting some arbitrary stress tensor into
the ‘Principal Stress ..’ spreadsheet and then noting the trends/plunges of the three principal axes. If
the original arbitrary stress tensor is then input into the ‘Stress Transformation ..’ spreadsheet with
the XYZ axes specified to have trends/plunges of 000/00 (horizontal north) 090/00 (horizontal east)
and 000/90 (vertical down) respectively. The orientations of the LMN coordinate axes can now be
set to the computed trends and plunges of the principal axes. The stress transformation should now
yield the known values of the principal stresses on the leading diagonal of the transformed (LMN)
stress tensor. This validation is implemented in the spreadsheets ‘Principal Stress Example 10.1’ and
‘Stress Transformation Example 10.1’.

Example 10.1

A rock mass has the following stress state in MPa, expressed relative to the Cartesian
coordinate axes X (horizontal north), Y (horizontal east) and Z (vertical down):

 xx  xy  xz   16.8 3.5  4.3

 
 yx  yy  yz    3.5 18.3  2.9
 zx  zy  zz   4.3  2.9 9.7 


Determine the values and orientations (trends/plunges) of the three principal stresses.

The solution to this example is summarised and validated in the attached spreadsheets ‘Principal
Stress Example 10.1’ and ‘Stress Transformation Example 10.1’. The major, intermediate and minor
 25

principal stresses are respectively 23.013, 14.255 and 7.532 MPa. These three compressive principal
stresses have trends/plunges 47.39/20.76, 133.36/10.50, 18.11/66.52 respectively. The opposing
stresses in each pair have trends/plunges 227.39/20.76, 313.36/10.50, 198.11/66.52 respectively.
For illustration purposes these trends and plunges have been evaluated in the attached spreadsheet
‘Principal Stress Example 10.1’ by Equations 5.5 applying the Excel ATAN2 function and also
ATAN with the quadrant parameter Q.

It is important to remember that the ‘Principal Stress ..’ spreadsheet will only give the correct values
of trend and plunge for the principal stresses if the original stress tensor has been expressed relative
to the Cartesian coordinate axes X (horizontal north), Y (horizontal east) and Z (vertical down). The
actual computed values of the principal stresses are unaffected by the choice of axes. If the stress
tensor has been expressed relative to coordinate axes with some other orientations the tensor would
first need to be transformed to the X (horizontal north), Y (horizontal east) and Z (vertical down)
system. This ‘pre-transformation’ strategy has been implemented in the attached spreadsheet
‘General Principal Stress’ by combining the ‘Stress Transformation …’ and ‘Principal Stress …’
computations in a pair of linked worksheets. In the ‘Stress Trans’ worksheet the following stress
state has been specified relative to X, Y and Z axes with trends/plunges 195/00, 285/65 and 105/24
respectively

 xx  xy  xz   8  6 3 
 
 yx  yy  yz    6 12  4
 zx  zy  zz   3  4 10 


This stress state has then been transformed to Cartesian coordinate axes L (horizontal north), M
(horizontal east) and N (vertical down). The rotated stress tensor has then been read into the input
stress tensor on the linked worksheet ‘Principal Stresses’, with the coordinate axes renamed to XYZ.
The principal stresses and their orientations can then be calculated as before.

11. Displacement and strain

Any body, such as a mass of rock, when exposed to a change in the state of stress will usually
experience a change in shape. The nature and magnitude of the resulting displacements will depend
on the stress increment and the elastic properties of the material. The magnitude of these
displacements will also depend on the size of the body under deformation. As force is scaled by area
to define stress, it is convenient to scale displacements by length to define strain. As with stresses
there are normal (or longitudinal) strains and also shear strains. These quantities can best be
understood by considering displacement and strain in two dimensions. as illustrated in the diagram
below.
 26

Figure 11.1 from http://en.wikipedia.org/wiki/Infinitesimal_strain_theory

In two dimensions the element ACDB of side lengths dx an dy has been moved or translated (by
rigid body displacement) and also deformed to create the rhombus acdb (ie ac = cd = db = ba). In
continuum mechanics we are only concerned with deformations, so we remove the translations by
superimposing point A over point a, as indicated by the element ACDB re-drawn with a dashed line.
Displacements in the x and y directions are respectively ux and uy. These displacements will change
continuously as we move in the x and y directions.

u x
The partial derivative should be interpreted to mean the “rate of change of the x-displacement
x
as we move in the x direction”. So over a distance dx the total change in the x-displacement will be
u x u x
dx . Since points A and a coincide, dx is the x-displacement of point b, as shown in Figure
x x
u y
11.1. Similarly, The partial derivative should be interpreted as the “rate of change of the y-
x
displacement as we move in the x direction”. So over a distance dx the total change in the y-
u y
displacement will be dx . In two dimensions we also have the changes in displacement given
x
u y u x
by dy and dy .
y y

The line ab is the hypoteneuse of a right-angled triangle whose other two sides have lengths
u u y u
2

dx , so by Pythagoras ab   dx  u x dx    y dx  . For very small rates

2
dx  x dx and
x x  x   x 
of change in displacements the second term will be very small so we can ignore it, hence

u x
ab  dx  dx (11.1)
x
 27

If the rates of change in displacements are very small we can assume that strain is homogeneous
across our infinitesimal element. This assumption carries the important implication that any plane
in the undeformed material will also be a plane in the deformed material.

Normal or longitudinal strain, which is a direct result of changes in normal stress, is defined as the
increase in length divided by the original length of an infinitesimal element. Strain is, therefore,
a dimensionless quantity with units of metres per metre. Longitudinal strain in the x direction in
Figure 11.1 is referenced to the fibre AB, and is given by xx (Greek letter ‘epsilon’), as follows
(noting that the length of AB = dx)

u x
dx  dx  dx
ab  AB x u
 xx    x (11.2)
AB dx x

Strain values in rock mechanics are usually very low, of the order of 0.001, so for convenience strain
values are sometimes expressed as millistrain (0.001 m/m = 1 mm/m) or microstrain (0.000001 m/m
= 1 micron/m). Because we have taken compressive stresses to be positive we also take contractile
strains to be positive. Although increments in compressive normal stress are associated with
reductions in length, it is possible to observe increases in length under conditions of compressive
normal stress. For example, if the compressive axial stress on a cylinder of rock is reduced from 10
MPa down to 5 MPa the specimen will become slightly longer as it recovers some of the elastic
strain. For this reason we do not refer to compressive and tensile strains, instead we refer to
contractile (+ve) and extensile (-ve) strains. Positive increments of stress will, however, always be
associated with positive increments of strain.

We adopt the same double-subscript nomenclature to describe strain components as that outlined for
stress components in a previous section. To implement our contractile strain is positive convention
we need to insert a negative sign, so longitudinal strain in the x, y and z directions are given by

u x du
 xx    x
x dx
u y du y
 yy    } (11.3)
y dy
u du
 zz   z   z
z dz

Or generically 
final length   initial length  (11.4)
initial length 
where dx, dy and dz are the initial lengths of the element in the x, y and z directions, and dux, duy and
duz are the increases in length in these directions. For example, if a fibre of initial length 0.0000154
m aligned in the x direction is deformed to a final length of 0.0000152 m, the longitudinal strain is


0.0000152  0.0000154  0.013
0.0000154
The application of shear stress to an element will also cause deformations, which can similarly be
expressed as changes in the length of the element in the x, y and z directions dux, duy and duz. In our
‘contractile strain is positive’ convention, shear strain is defined as the increase in angle, measured
 28

in radians, between a pair of lines that were originally at right angles to each other and in
alignment with a pair of coordinate axes, as illustrated in Figure 11.1. Again, we adopt the same
nomenclature as that outlined for stress components and note that positive shear strains are
associated with positive shear stresses. Shear strain in the xy plane is, therefore, given by gxy (Greek
letter ‘gamma’) as follows

g xy  a  b (11.5)

The element in Figure 11.1 is of infinitesimal size. If we examine the small right-angled triangles,
u x u y
for very small values of shear strain we can ignore the terms dx and dy so
x y

u y u x
dx u y dy
tan a  x  y u (11.6)
and tan b   x
u
dx  x dx x
u y y
dy  dy
x y

Also, when the angle a is very small tan a ≈ a (radians). For example, taking a = 1º = 0.01745
radians, tan a = 0.01746. The same applies for the angle b hence, using Equation 11.5.

 u u 
g xy   y  x  (11.7)
 x y 

 u y u x   du du 
As before g xy       y  x  (11.8)
 x y   dx dy 

where duy and dx are the ‘opposite’ and ‘adjacent’ sides of the small right angled triangle whose
hypoteneuse is ab in Figure 11.1, and similarly dux and dy are the ‘opposite’ and ‘adjacent’ sides of
the small right angled triangle whose hypoteneuse is ac. This physical appreciation of calculating
shear strains by calculating the ratios of displacements and initial fibre lengths at right angles is very
important, and will be examined in more detail in the next Section.

Clearly

g yx  b  a   g xy
Similarly
 du du y 
g yz  g zy   z  
 dy dz 
}(11.9)
 du du 
and g xz  g zx   z  x 
 dx dz 

As with stress components, it is sometimes necessary to calculate from the strains xx, yyzz,gxy, gyz
and gxz the longitudinal and shear strains in a local Cartesian coordinate system (say LMN). The
formulae to do this look very similar to those for transformation of stress, and with a small
adjustment they can be made exactly the same. The adjustment is a change in our measure of shear
strain. Transformation calculations are done in terms of tensorial shear strains g t , g t and g t which
xy yz zx
 29

are respectively equal to half of the engineering definition values  g xy , g yz and g zx . The
tensorial measure of longitudinal strains  tx y ,  t and  tzx are respectively equal to  xy ,  yz and
yz

 zx and so do not need to be changed.

The complete matrix of scaled displacements, or strains [] will therefore be as follows

 xx g txy g txz    xx 1
g xy 1
g xz 
  
  g tyx
2 2
 yy g tyz    12 g yx  yy 1
2
g yz  (11.10)
 g zx

t
g t
zy  zz   12 g zx 1
2
g zy  zz 

The properties of the strain matrix are directly analogous to those of the stress matrix, specifically
there are six independent strain components which can be transformed to any inclined set of axes,
such as LMN. If we specify

  1
g m 1
g n 
 *   12 g m 
2 2
 mm 1
2
g mn  (11.11)
 1 g n 1
g nm  nn 
2 2

then [*] = [R] [] [R]T (11.12)

where, as before, [R] is the rotation matrix and [R]T is its transpose. It is important to note that the
tensorial shear strains produced by the above transformation must be doubled to give the shear
strains that accord with our earlier engineering definition. Equation 11.12 implies that the concepts
of principal planes, principal strains and principal axes are directly applicable to the analysis of
strains. Specifically, planes of principal strain are given by the planes of principal stress. The major,
intermediate and minor principal strains 1 2 and 3 can be found from the strain matrix in the same
way that principal stresses are determined. Finally, it is important to note that

 the major principal strain 1 is parallel to the major principal stress 1

 the intermediate principal strain 2 is parallel to the intermediate principal stress 2
 the minor principal strain 3 is parallel to the minor principal stress 3
 30

Example 11.1

D
B
Y
C
E X

H F

G
Z

The corners A, B, D and E of the cube of side length 800 mm shown above have been instrumented
to monitor their displacement by reference to the Cartesian coordinate system X (horizontal north) Y
(horizontal east) and Z (vertical down) as the cube of material is carefully excavated from solid rock.
The displacements (all positive) in the following table were measured.

Displacements, mm
Point x y z
A 0.200 0.100 0.300
B 0.700 0.500 0.300
C
D 0.200 0.400 0.400
E 0.800 0.600 1.000
F
G
H

(a) Calculate the components xx, yy, zz, gxy, gyz, and gzx according to the engineering definition of
strain with the sign convention contractile strain positive.
(b) Calculate the longitudinal strain along a line connecting A to G.
(c) Calculate the change in volume of the cube.

Solution

We start be removing the rigid body displacement by subtracting the x, y and z displacements for
point A respectively from the x, y and z displacements for all points, giving the following table of
relative displacements:

Relative displacements,
mm
Point x y z
A 0 0 0
B 0.500 0.400 0
C
D 0 0.300 0.100
 31

E 0.600 0.500 0.700

F
G
H

All future calculations in this example will be done using these relative displacements.

(a)
For the longitudinal strain in the x direction, fibre AB has increased in length by 0.5 mm. Applying
Equation 11.3
du x
 xx   = 0.5/800 = 0.000625
dx
For the longitudinal strain in the y direction, fibre AD has increased in length by 0.3 mm. Applying
Equation 11.3
du y
 yy   = 0.3/800 = 0.000375
dy
For the longitudinal strain in the z direction, fibre AE has increased in length by 0.7 mm. Applying
Equation 11.3
du z
 zz   = 0.7/800 = 0.000875
dz
For shear strain in the xy plane, for fibre AB duy = 0.4 mm, and for fibre AD dux = 0 mm. As before,
dx = dy = dz = 800 mm, so applying Equation 11.8
 du y du x 
g xy     = {(0.4/800) + (0/800)} = 0.0005
 dx dy 
For shear strain in the yz plane, for fibre AD duz = 0.1 mm, and for fibre AE duy = 0.5 mm.
Applying Equation 11.9
 du du 
g yz   z  y  = {(0.1/800) + (0.5/800)} = 0.00075
 dy dz 
For shear strain in the xz plane, for fibre AB duz = 0 mm, and for fibre AE dux = 0.6 mm. Applying
Equation 11.9
 du du 
g xz   z  x  = {(0/800) + (0.6/800)} = 0.00075
 dx dz 
(b)
One way (but not the only way) of calculating the longitudinal strain along a line connecting A to G
is to calculate the displacements of points C, F, H and G. We can calculate these displacements by
applying (i) the calculated values of longitudinal strains and (ii) the principle that any plane in the
undeformed material will also be a plane in the deformed material. Working through these
calculations, although relatively arduous, is instructive for those who are new to this topic. The
following calculations are presented as a demonstration of some of the principles of three-
dimensional strain analysis. Such calculations would rarely be conducted in practice.

(i) The changes in the x, y and z dimensions of our cube, calculated earlier are respectively
elongations of 0.5, 0.3 and 0.7 mm.
 So the x displacement of point C is given by the x displacement of the adjacent point D plus
the x elongation of the cube, or 0 + 0.5 = 0.5 mm. The y displacement of point C is given by
the y displacement of the adjacent point B plus the y elongation of the cube, or 0.4 + 0.3 = 0.7
mm.
 Similarly the x displacement of point F is given by the x displacement of the adjacent point E
plus the x elongation of the cube, or 0.6 + 0.5 = 1.1 mm. The z displacement of point F is
 32

given by the z displacement of the adjacent point B plus the z elongation of the cube, or 0 +
0.7 = 0.7 mm.
 Similarly the y displacement of point H is given by the y displacement of the adjacent point E
plus the y elongation of the cube, or 0.5 + 0.3 = 0.8 mm. The z displacement of point H is
given by the z displacement of the adjacent point D plus the z elongation of the cube, or 0.1 +
0.7 = 0.8 mm.

(ii)
 In the xy plane, for surface ABCD to remain planar the z displacement of point C must be the
sum of the z displacements of points B and D, or 0 + 0.1 = 0.1 mm.
 In the yz plane, for surface AEHD to remain planar the x displacement of point H must be the
sum of the x displacements of points E and D, or 0.6 + 0 = 0.6 mm.
 In the xz plane, for surface ABFE to remain planar the y displacement of point F must be the
sum of the y displacements of points B and E, or 0.4 + 0.5 = 0.9 mm.

We are now in a position to calculate the x, y and z displacements of point G.

 The x displacement of point G is given by the x displacement of the adjacent point H plus the
x elongation of the cube, or 0.6 + 0.5 = 1.1 mm.
 The y displacement of point G is given by the y displacement of the adjacent point F plus the
y elongation of the cube, or 0.9 + 0.3 = 1.2 mm.
 The z displacement of point G is given by the z displacement of the adjacent point C plus the
z elongation of the cube, or 0.1 + 0.7 = 0.8 mm.
Applying these displacements to point G gives new x, y and z coordinates of 801.1, 801.2 and 800.8
mm, so the vector AG in the deformed element has a length of √{(801.1)2 + (801.2)2 + (800.8)2)} =
1387.43 mm. The initial length of the vector AG was √{(800)2 + (800)2 + (800)2)} = 1385.64 mm,
so it has suffered a strain of –(1387.43  1385.64)/1385.64 = 0.0012918 (elongation).

The complete set of relative displacements is tabulated below.

Relative displacements, mm
Point x y z
A 0.000 0.000 0.000
B 0.500 0.400 0.000
C 0.500 0.700 0.100
D 0.000 0.300 0.100
E 0.600 0.500 0.700
F 1.100 0.900 0.700
G 1.100 1.200 0.800
H 0.600 0.800 0.800

(c) Volumetric strain  is given by the sum of the three normal strains, so  = (0.000625 +
0.000375 + 0.000875) = 0.001875. Hence the change in volume is given by 0.001875 x 8003 =
960,000 mm3 or an expansion of 960,000 mm3 or 0.00096 m3.

An alternative, and much easier, way of tackling parts (b) and (c) of this example is to apply three-
dimensional strain transformation. This strategy is implemented in the spreadsheet ‘Strain
Transformation Example 11.1’ which is derived from our earlier ‘Stress Transformation ..’
spreadsheet simply by changing the input data and the labels for the tensors. The most challenging
aspect of performing this strain transformation is calculating the orientations of the initial and
transformed sets of axes XYZ and LMN. In this spreadsheet the X axis is horizontal north, the Y
axis horizontal east, and the Z axis vertical down, in order to conform with the diagram in the
example. In the strain transformation the AG vector is taken to be the L axis of trend/plunge
 33

045/35.264. (Note that arcsin(800/√3(800)2) = 35.264º). For simplicity the M axis is taken to be
horizontal with a trend of 045 + 90 = 135º, and the N axis completes the right handed system with a
trend/plunge 225/54.736, as explained in Section 5. The spreadsheet solution confirms, and also
expands upon, the results obtained above.

The data in this example have been input into the spreadsheet ‘Principal Strain Example 11.1’ which
was derived from our earlier ‘Principal Stress ..’ spreadsheet simply by changing the input data and
the labels for the tensors. The spreadsheet ‘Principal Strain Example 11.1’ illustrates the direct
analogy between principal stresses and strains and also shows a graph of the characteristic equation
for three dimensional strain.

12. Constitutive relations

Having examined the concepts of three-dimensional stress and strain it is now appropriate to
consider the relation between the two. All materials are exposed to some form of stress tensor. If
one or more components of this stress tensor are changed there will be a consequent change in the
strain tensor. Conversely, all materials possess some definable shape. If one or more components of
this shape are to be changed it will be necessary to change the applied stress tensor. Changes in
stress are associated with strain increments, and vice versa. This association between stress and
strain is referred to as the constitutive behaviour of the material.

The form of the constitutive equations, which describe the constitutive behaviour, depends upon the
mechanical properties of the material under consideration. We will adopt the generalised form of
Hooke’s law, which states that strain components are a linear function of stress components. If
we introduce a constant of proportionality we obtain

=S (12.1)

where S is called the compliance of the material. Clearly, for a given value of stress, a higher value
of the constant S will produce a larger amount of strain, hence the compliance is a measure of how
soft or compliant the material is. Writing down the inverse form gives

=D (12.2)

where D is called the stiffness of the material. In this case, for a given value of strain, a higher value
of the constant D will necessitate a higher value of stress, hence the stiffness is a measure of how
rigid or stiff the material is.

In general we have six components of stress, each of which can, in theory, generate any one of our
six components of strain. For example, if we applied an axial stress to a cylinder of rock we would
observe contractile axial strains and extensile lateral strains. In theory we would need to specify 36
independent material constants for compliance, as follows.

 xx  S11 S12 S13 S14 S15 S16   xx 

  S S22 S23 S24 S25 S26   
 y y   21  y y
  zz  S31 S32 S33 S34 S35 S36    zz 
     (12.3)
 g xy  S41 S42 S43 S44 S45 S46   xy 
 g y z  S51 S52 S53 S54 S55 S56   y z 
     
 g zx  S61 S62 S63 S64 S65 S66    zx 
 34

or [] = [S] [] (12.4)

where the 6 by 6 matrix [S] is called the compliance matrix. Reciprocity requires that the
compliance matrix is symmetrical so, in theory, there are only 21 independent compliance terms.
Similarly, in the inverse form there are only 21 independent stiffness terms in the stiffness matrix
[D]

[] = [D] [] (12.5)

Although it is easy to visualise how a normal stress increment xx could induce normal strain
increments xx yy and zz, it is difficult to imagine a mechanism whereby an increment in normal
stress could produce an increment in shear strain, nor how an increment in shear stress could produce
an increment in normal strain. The 3 by 3 sub-matrices in the top right and bottom left of the
compliance and stiffness matrices must, therefore, be zero.

If the rock material has the same mechanical properties in all directions, ie it is isotropic, then the
compliance and stiffness terms will be the same in all directions. Under these conditions only two
independent elastic constants are required completely to define the constitutive relations for the
material, as follows:

 xx   1   0 0 0   xx 
    1    
 y y  0 0 0   y y
  zz  1     1 0 0 0    zz 
    
0 2 1   
(12.6)
 g xy  E  0 0 0 0   xy 
 g y z  0 0 0 0 2 1    0   y z 
    
 g zx   0 0 0 0 0 2 1      zx 

where E = Young’s modulus and  (Greek letter ‘nu’) = Poisson’s ratio

If we carry out the matrix multiplication in Equation 12.6 we obtain:

xx = (xx – (yy + zz))/E

yy = (yy – (xx + zz))/E
zz = (zz – (xx + yy))/E }(12.7)
gxy = 2 xy(1 + )/E
gyz = 2 yz(1 + )/E
gzx = 2 zx(1 + )/E

We have encountered the elastic constants Young’s modulus and Poisson’s ratio before. For
example, if we consider a simple uniaxial compression test on a cylinder of rock where the loading
axis is in the Z direction, then zz > 0 and xx = yy = xy = yz = zx = 0. Applying Equations 12.7

xx = yy = – zz/E (12.8)

zz = zz/E (12.9)

and gxy = gyz = gzx = 0 (12.10)

hence E = zz/zz (12.11)

 35

and  = xx E/zz = yy E/zz = xx/zz = yy/zz (12.12)

Equations 12.11 and 12.12 show that one way of interpreting the elastic constant Young’s modulus is
to note that it is given by the ratio of axial stress to axial strain in a uniaxial compression test.
Poisson’s ratio can be interpreted as the ratio between lateral strain and axial strain in a uniaxial
compression test. The negative sign appears in Equation 12.12 because lateral strain will be
extensile and axial strain will be contractile in such a test, so the ratio of these quantities will be
negative. Introducing the negative sign ensures that Poisson’s ratio takes a positive value.

When dealing with shear stress and shear strain it is convenient to derive an additional elastic
constant G called the shear modulus, as follows

G = E/2(1 + ) (12.13)
so that
gxy = xy/G
gyz = yz/G }(12.14)
gzx = zx/G

Under isotropic conditions, the inverse form of Equations 12.6 is

 1    1 1 0 0 0 
  
 xx   1      
   1 1 0 0 0   xx 
 y y
 
 1      y y
  zz   1 1

0 0 0    zz 
   1  2    (12.15)
 xy   0 0 0 0 0   g xy 
 y z 2
 1  2    g y z
   0 0 0 0 0  
  zx   2  g zx 
 0 1  2  
0 0 0 0
 2 

where  is Lamés constant, given by

E

1   1  2  (12.16)

The three terms in the bottom three rows of the stiffness matrix in Equation 12.15 are, when
multiplied by Lamés constant, equal to the shear modulus G. It is interesting to note that when  =
0.25 (and  is often close to this value for rocks) then G =  = E/2.5. Clearly, shear modulus and
Lamés constant are not separate independent properties of the material, since they are each derived
from Young’s modulus and Poisson’s ratio.

Before carrying out the matrix multiplication in Equation 12.15 it is helpful to note that volumetric
strain  is given by

 = xx + yy + zz (12.17)

Hence
xx =   + 2 G xx
yy =   + 2 G yy
zz =   + 2 G zz }(12.18)
xy = G gxy
 36

yz = G gyz
zx = G gzx

Equations 12.6 and 12.15 are implemented in the worksheet ‘Constitutive’ of the attached
spreadsheet ‘Constitutive Relations’ by applying the MMULT function. In this particular
spreadsheet the input data consist of the elastic constants Young’s modulus and Poisson’s ratio, and
the stress tensor increment []. The compliance matrix (Equation 12.6) is then applied to determine
the strain tensor []. The stiffness matrix (Equation 12.15) can then be applied to determine the
stress tensor increment [], which should, of course, agree with the input values.

In many rock mechanics applications it is desirable to conduct a two-dimensional analysis by taking

a cross-section through an elongated structure such as a shaft, tunnel or borehole.

One approach is to regard the two-dimensional structure as a thin plate in the XY plane. The
surface of the plate is assumed to be unstressed, so the out-of-plane stresses zz = yz = zx = 0. This
condition is referred to as plane stress. The worksheet ‘Plane Stress’ in the attached spreadsheet
‘Constitutive Relations’ shows that if we apply a two-dimensional plane stress tensor xx ≠ 0 yy ≠ 0
xy ≠ 0 zz = yz = zx = 0 to our thin plate, the normal strain in the z direction is non-zero. In
practical terms, the application of compressive normal stresses on the boundaries of our plate in the
XY plane has caused the plate to get ‘fatter’ (negative normal strain) in the Z direction.

An alternative approach is to regard the two-dimensional structure as a thin cross-section in the XY

plane through a structure that extends a long distance in the Z direction. This approach is generally
more realistic when dealing with long axi-symmetric structures such as shafts, tunnels and boreholes.
The surface of the cross-section is assumed to be constrained by the presence of adjacent material in
the Z direction, so the out-of-plane strains zz = gyz = gzx = 0. This condition is referred to as plane
strain. The worksheet ‘Plane Strain’ in the attached spreadsheet ‘Constitutive Relations’ shows that
if we require a two-dimensional plane strain tensor xx ≠ 0 yy ≠ 0 xy ≠ 0 zz = yz = zx = 0 for our
thin cross-section, the normal stress in the z direction must be non-zero. In practical terms, the
application of contractile normal strains on the boundaries of our thin cross-section in the XY plane,
which are associated with compressive normal stresses, has necessitated the development of a
compressive normal stress in the Z direction in order maintain the condition of plane strain.

13. Failure criteria for rock: the Coulomb criterion

A failure criterion for a rock is a mathematical expression that defines the stress state at which the
rock will fail, and is usually expressed in terms of the stress tensor and the material properties of the
rock. The term ‘failure’ implies that the rock has completely disintegrated. It is, however, possible
for the rock to become unserviceable, in an engineering sense, if substantial plastic strains have
developed. In this context the term ‘yield criterion’ is often adopted. Failure criteria can be
developed fundamentally from the mechanical analysis of an assumed failure mechanism, or they
can be developed empirically by modelling the observed behaviour of rock during laboratory and/or
site tests.

One of the most widely applied fundamental failure criteria for rocks is the Coulomb criterion.
This criterion assumes that rock failure occurs as a result of shear displacement along a shear surface
cutting through the rock. This is quite a sensible assumption since we can observe shear failure in
laboratory specimens and also at a large scale in rock masses. Shear failure is predicted to occur
when the shear stress  on a particular surface exceeds the shear strength f, where

f = c + n tan    (13.1)
 37

where c is the cohesion,  is the angle of friction and n is the normal stress on the shear failure
surface. Cohesion, which is the shear strength at zero normal stress, is related to the cementing
between the rock grains and crystals on the shear surface. Friction is, of course, the frictional
strength mobilised as a result of the normal stress acting across the shear surface. Although in rock
mechanics we recognise the importance of water pressure and the principle of effective stress, in
most rocks there is not an interconnected pore network so for simplicity failure criteria are usually
expressed in terms of total stresses . It is, however, a relatively simple matter to re-work the
following analysis using the effective normal stress ′n = n – u, where u is the pore water pressure.
Development of the Coulomb criterion from the relatively simple Equation 13.1 through the
application of two-dimensional stress transformation provides an instructive example of how failure
criteria can be developed.

1

X
L  M

Y

n 
3  3
n

1

Figure 13.1 Normal and shear stresses on an inclined plane

Figure 13.1 shows a two-dimensional view of the normal stress n and the shear stress  acting on an
inclined plane cutting through a cylindrical or prismatic specimen of rock subjected to an axial
principal stress 1 and a lateral or confining stress 3. The normal to the inclined plane makes an
angle  to the 1 axis.

Recalling our three-dimensional stress transformation Equations 7.17, we let xx = 1 and yy = 3,
while the remaining stresses zz = xy = yz = zx = 0. We define our rotated L axis to be aligned
with n, and our rotated M axis to be aligned with the inclined plane, so lx = cos(), ly =
cos(sin(), mx = cos(90 – ) = sin(), and my = cos()

Substituting the above values into expressions for ll and lm in our original Equations 7.17

ll = lxlxxx + lylyyy + lzlzzz + (lxly + lylx)xy + (lylz + lzly)yz+ (lxlz + lzlx)xz
lm = lxmxxx + lymyyy + lzmzzz + (lxmy + lymx)xy + (lymz + lzmy)yz+ (lxmz + lzmx)xz

gives

ll = n = lxlxxx + lylyyy = 1 cos2 + 3 sin2     (13.2)

 38

and
lm = = lxmxxx + lymyyy = 1 cos sin  3 sin cos = (1  3) sin cos (13.3)

Adopting the following well-known trigonometrical identities

sin cos = ½ sin2
cos2 = ½(1 + cos2) and
sin2 = ½(1  cos2)
we have
  n = ½(1+ 3) + ½ cos2 (1  3) (13.4)
and
  = ½ sin2 (1  3) (13.5)

At the point of failure the shear strength is balanced by the shear stress, ie f in Equation 13.1 is
equal to  in Equation 13.5. Using Equations 13.4 and 13.5 to substitute for n and f in Equation
13.1 gives

½ sin2 (1  3) = c + tan  [½(1+ 3) + ½ cos2 (1  3)]

Which on rearrangement gives

2c  3 sin 2  tan 1  cos 2

1  (13.6)
sin 2  tan 1  cos 2

Application of the following trigonometrical identities to Equation 13.6

cos 2 = 2 cos2 – 1 = 1 – 2 sin2, tan  = sin /cos  and sin2cos2

and further rearrangement gives the following alternative form of Equation 13.6

2c  3 tan 
1  3  ` (13.7)
 tan  
sin 21  
 tan  

If the inclined plane is a discontinuity with cohesion cd angle of friction d and orientation d, then
the theoretical strength 1d of the specimen by the failure mechanism of shear along the discontinuity
surface is given by Equations 13.6 and 13.7 by replacing c,  and  by cd, d and d respectively, so

2c d  3 sin 2d  tan d 1  cos 2d  2c d  3 tan d 

1d   3  (13.8)
sin 2d  tan d 1  cos 2d   tan d 
sin 2d 1  
 tan d 

If there is no discontinuity cutting through the specimen, or if the strength and/or orientation of the
discontinuity are such that shear failure does not occur along the discontinuity plane, then when 1
reaches a sufficiently high level the specimen will theoretically fail by creating a shear surface
through intact rock. The rock will theoretically ‘choose’ the failure surface that gives the minimum
strength according to Equations 13.6 and 13.7. In Equation 13.7 we can identify the minimum value
for 1 by maximising the denominator term
 39

 tan  
sin 21   (13.9)
 tan  

Differentiating Equation 13.9 with respect to , then equating the result to zero shows that the
minimum value for 1 occurs when


  crit  45o  (13.10
2

If we substitute this value for  into Equation 13.6 then

the term sin 2 becomes sin(90º + ) = cos  and

the term cos 2 becomes cos(90º + ) = sin 

Making these substitutions in Equation 13.6 gives

2c  3 cos   tan 1  sin 

1  (13.11)
cos   tan 1  sin 

Multiplying the ‘top and bottom’ of Equation 13.11 by cos , and noting that cos2sin2 = 1, gives
the important result

2c cos  3 1  sin 
1   (13.12)
1  sin  1  sin 

Equation 13.12 is the equation of a straight line when 1 is plotted against 3, as shown in Figure
13.2, noting that the region below the line represents a stable stress state. The vertical-axis intercept
of this straight line gives the theoretical value of 1 at which the rock specimen will fail when 3 = 0,
which is of course the uniaxial compressive strength of the rock c. So, according to the Coulomb
criterion

2c cos 
c  (13.13)
1  sin 

1

Unstable

c
Stable

t 3

Figure 13.2 Coulomb failure criterion

 40

The slope of the line in Figure 13.2 is given by

1  sin   
tan    tan 2  45o   (13.14)
1  sin   2

Students of soil mechanics will recognise tan  as the passive earth pressure coefficient for a soil
whose failure is governed by the Coulomb criterion.

Simple geometrical analysis of Figure 13.2 shows that the uniaxial tensile strength t of the rock is,
theoretically, given by

2c cos 
t  (13.15)
1  sin 

Finally, we can write Equation 13.11 in the following alternative compact form

1  2c tan   3 tan  (13.16)

The worksheets ‘Coulomb data’ and ‘Coulomb graph’ in the Excel spreadsheet ‘Coulomb’ show the
calculations for, and also the graphs of, the shear strength f given by Equations 13.1 and 13.4, and
shear stress  given by Equation 13.5 plotted against the angle  in the range 0 to 90º. For the
purposes of this illustration the value of 1 has been taken as the strength value predicted by
Equation 13.16. The curves for f and  just touch at the critical value of crit given by Equation
13.10. It is interesting to note that the curves for f and  are very close to each other over a range of
 values at least 5º either side of the critical value. This theoretical result implies that in practice we
might expect the shear failure surface through intact rock to lie within the range crit ± 5 º. The
worksheet ‘Denominator’ in the Excel spreadsheet ‘Coulomb’ shows the graph of Equation 13.9
plotted against the angle  in the range 0 to 90º, illustrating the maximum at the value of  given by
Equation 13.10.

Example 13.1

A cylindrical specimen of rock contains a pre-existing through-going plane of weakness

whose normal makes an angle of d = 47° with the axis of the cylinder. This plane of
weakness has Coulomb shear strength parameters: cohesion cd = 1.2 MPa, angle of friction d
= 38°. The intact rock material has Coulomb shear strength parameters: cohesion c = 2.5
MPa, angle of friction  = 40°.

Assuming that the plane of weakness and the intact rock material both obey the Coulomb
failure criterion, calculate the separate theoretical axial stresses at failure of (a) the plane of
weakness and (b) the intact rock material failing by shear along an inclined failure plane, in a
triaxial test conducted at a confining pressure of 1.5 MPa. Comment briefly on the
implications of your result.

The solution to this example presented in the spreadsheet ‘Coulomb Example 13.1’ shows that the
value of 1 required to cause failure along the plane of weakness is predicted to be 19.02 MPa,
whereas the value of 1 required to cause failure through intact rock is predicted to be 17.62 MPa.
Hence this rock specimen will, in theory, fail through intact rock at an axial stress of 17.62 MPa.
The spreadsheet also demonstrates the application of Equations 13.6 to 13.16.
 41

14. Failure criteria for rock: the Hoek-Brown criterion

The Hoek-Brown criterion has been widely adopted in rock mechanics applications for the following
reasons:

 The Hoek-Brown criterion was developed specifically for rock materials and rock masses.
 Input parameters for the Hoek-Brown criterion can be derived from uniaxial testing of the rock
material, mineralogical examination, and measurement of the rock mass fracture characteristics.
 The Hoek-Brown criterion has been applied for some 30 years by practitioners in rock
engineering, and has been applied successfully to a wide range of intact and fractured rock types
(Hoek and Brown, 1997).

The Hoek-Brown failure criterion for intact and fractured rock masses was first introduced in 1980.
Over the last 30 years or so, the criterion has been widely adopted for modelling the strengths of a
wide range of rock types. It is acknowledged that the Hoek-Brown criterion has its limitations. In
particular, the controlling parameters for this empirical criterion, and the empirical expressions that
link these parameters to the observed mineralogy and rock structure, have been subject to change
over the years. Despite these limitations, the Hoek-Brown criterion has been widely accepted by the
international rock mechanics community. A recent version of the failure criterion, described by
Hoek and Brown (1997) is as follows:
a
 m   
1  3  ci  b 3   s (14.1)
 ci  
where
1 is the maximum (effective) stress at failure
3 is the minimum (effective) stress at failure
ci is the uniaxial compressive strength of the intact rock material
mb is the value of the Hoek-Brown constant m for the rock mass, which might be fractured
s and a are parameters whose value depends on the nature and the degree of fracturing of the rock
mass.

The parameters mb, s and a can be estimated from the Geological Strength Index (GSI in the range 0
to 100) for the rock mass. Broadly speaking the GSI depends on the degree of interlocking of the
blocks in the rock mass, and on the surface quality of the discontinuities in the rock mass.
Determination of the GSI value for a rock mass will be covered later in the course, meanwhile, it is
sufficient to understand that a GSI close to 0 corresponds to a heavily fractured rock mass, and a GSI
close to 100 corresponds to a relatively unfractured rock mass

The Hoek-Brown parameter mb can be estimated from the GSI and mi as follows.

 GSI 100 
 
 
m b  mi e 28
(14.2)

Where mi, which is the Hoek-Brown constant m for the intact rock material, depends on the rock
type and mineralogy. Some typical values of mi are tabulated below.
 42

Table 14.1 Typical values of the Hoek-Brown parameter mi

Typical values of mi
Grain size
Rock type Class/group Coarse Medium Fine Very fine
Clastic Conglomerate Sandstone 19 Siltstone 9 Claystone 4
22* Greywacke 18* Chalk 7
Organic Coal 8 to
21*
Sedimentary
Carbonate Breccia 20* Limestone 10* Limestone
8
Chemical Gypstone 16 Anhydrite
13
Non-foliated Marble 9 Hornfels 19* Quartzite
24
Slightly Migmatite Amphibolite 25 Mylonites
Metamorphic
foliated 30* to 31 6*
Foliated+ Gneiss 33 Schists 4 to 8 Phyllites Slate 9
10*
Granite 33 Rhyolite Obsidian
16* 19*
Light Granodiorite Dacite 17*
30*
Igneous Diorite 28* Andesite 19
Gabbro 27 Dolerite 19* Basalt 17*
Dark
Norite 22
Extrusive Agglomerate Breccia 18* Tuff 15*
pyroclastic 20*
* Estimated values
+
Tested normal to foliation with failure through intact rock

When GSI > 25, which covers about 80% of rock masses, then the Hoek-Brown parameters s and a
are estimated by

 GSI 100 
 
 
se 9
and a = 0.5 (14.3)

Equations 14.2 and 14.3 show that for intact unfractured rock (where GSI = 100) mb = mi, s = 1.0
and a = 0.5 in Equation 14.1. These conditions are assumed to apply for laboratory tests on intact,
unfractured specimens of rock.

When GSI < 25, which relates to rock masses of very poor quality

GSI
s=0 and a  0.65  (14.4)
200

Hoek et al (2002) subsequently introduced a disturbance factor D for the computation of mb and s,
which depends upon the amount of disturbance caused by blast damage and stress relaxation. Values
of D range from 0 for rock adjacent to machine excavated underground openings, to between 0.7 and
1.0 for civil engineering and open pit mine slopes, depending on the quality of blasting (refer to the
 43

Rock Mechanics notes for more information about the parameter D). Hoek et al (2002) proposed the
following revised empirical expressions for mb, s and a

 GSI 100 
 
 2814 D 
m b  mi e (14.5)

 GSI 100 
 
 9  3D 
se (14.6)

  GSI   20 
   
 15   3 
e e
a  0.5  (14.7)
6

Although Equations 14.5 to 14.7 give similar numerical results to Equations 14.2 to 14.4, Equations
14.5 to 14.7 have the merit of covering the entire range of GSI values in a single group of
expressions. In rock mechanics applications it is important to state which set of empirical
expressions have been used.

Example 14.1

(a) Show that for the Hoek-Brown criterion for fractured rock, a graph of ((1 – 3)/ci)2 on
the vertical axis versus (3/ci) on the horizontal axis will, in theory, plot as a straight line for
the case when a = 0.5 (ie GSI >25). What parameters define the slope and intercept of this
straight line? Note that 1 is the major principal stress at failure, 3 is the minor principal
stress at failure, and ci is the uniaxial compressive strength of the intact (unfractured) rock.

(b) Triaxial tests on moderately fractured specimens of sandstone gave the results tabulated
below. Intact specimens of this rock were found to have a uniaxial compressive strength ci
= 35 MPa. By fitting a best straight line to a graph of ((1 – 3)/ci)2 on the vertical axis
versus (3/ci) on the horizontal axis, estimate the Hoek-Brown parameters mb and s for this
fractured rock.

3 MPa 1 measured, MPa

0.5 13.5
1 16
1.5 18
2 20
2.5 22.5
3 24
3.5 26
4 27.5
4.5 29
5 30.5

(c) Numerical modelling shows that at a critical location adjacent to an underground mine,
the moderately fractured rock mass tested in part (b) will be subjected to a stress state 1 = 35
MPa and 3 = 10 MPa. Is the rock mass capable of sustaining this stress state?

The solution to Example 14.1 is presented in the spreadsheet ‘Hoek-Brown Example 14.1’. Here we
see that fitting a best straight line to a graph of test data with ((1 – 3)/ci)2 plotted on the vertical
 44

axis versus (3/ci) plotted on the horizontal axis provides a useful way of estimating mb and s for
fractured rock, or mi for intact specimens of unfractured rock. It must be acknowledged, however,
that it is very difficult (and expensive) to conduct laboratory tests on undisturbed specimens of
fractured rock.

Most of the currently used techniques for analysing the stability of near surface structures, such as
rock slopes, are based on the application of the Coulomb shear strength parameters cohesion c, and
the angle of friction  on some known or anticipated shear surface subjected to a normal stress n.
The most widely used of these techniques are the variants of the method of slices and related upper
bound techniques. If the Hoek-Brown criterion is to be used to model the strength of near surface
fractured rocks, it is necessary to determine equivalent Coulomb shear strength parameters for the
specified level of effective normal stress. Since the Hoek-Brown criterion is non-linear, the values of
cohesion and friction will vary with the value of the effective normal stress at which they are
computed. It is desirable, therefore, to refer to these equivalent Coulomb parameters as
‘instantaneous’ values ci and i adding the subscript “i” to emphasise that they are applicable at one
specified normal stress level.

Instantaneous ci and i parameters for the Hoek-Brown criterion for a shear surface subjected to a
normal stressn can be determined from the following expressions for the case where the GSI > 25,
where the parameter a = 0.5. The first step is to determine intermediate parameters h and  as
follows

16 m b  n  sci 
h  1 (14.8)
3m 2b ci
where
ci is the uniaxial compressive strength of the intact rock material
mb is the value of the Hoek-Brown constant m for the fractured rock mass given by Equation 14.2
s and a are the Hoek-Brown parameters given by Equation 14.3.

 1 
0.5  arc tan 

 h 3 1 
Then  (14.9)
3

 1 
and i  arc tan 
 (14.10)
 4 h cos 2
  1 

The shear strength f of the shear surface is given by

m b ci cot i  cos i 

f  (14.11)
8

Finally ci = f  n tan i (14.12)

The above method for calculating the equivalent angle of friction and cohesion for a Hoek-Brown
material preserves the ‘correct’ value for the Coulomb angle of friction at the specified normal stress,
so it can be referred to as the ‘angle of friction’ strategy. It is important to note that at relatively low
values of normal stress n, the angle of friction i is predicted to be relatively large, while cohesion ci
is relatively small. As the normal stress increases the angle of friction reduces and the cohesion
increases. This effect, which reflects the curved nature of the Hoek-Brown criterion, is illustrated in
 45

Figure 14.1 which is based on calculations in the spreadsheet ‘Hoek-Brown c phi Fig 14.1’. The
rock in this spreadsheet is assumed to have a ci = 80 MPa, mi = 10, and a GSI = 45. Equations 14.8
to 14.11 have been applied to plot the Hoek-Brown failure envelope in normal stress-shear strength
space in Figure 14.1 for normal stress in the range 0 to 25 MPa. The instantaneous values of
Coulomb cohesion and friction have been calculated at a specific selected normal stress n = 8 MPa.
The associated values of ci = 2.801 MPa, and i = 35.806 have been used to plot the Coulomb failure
envelope by applying Equation 13.1. The resulting line is, of course, tangential to the Hoek-Brown
envelope at the selected normal stress of 8 MPa. The prediction of relatively high values of friction
at low normal stresses is consistent with experimental observations in fractured rock masses. Correct
modelling of this behaviour is crucially important for slope stability analysis where stress levels are
generally low.

25
Hoek-Brown shear strength, t, MPa, Equation 14.11
Coulomb shear strength, tf, MPa, Equation 13.1
20 Selected normal stress
Shear strength, MPa

15

10

0
0 5 10 15 20 25
Normal stress, MPa

Figure 14.1 Hoek-Brown criterion plotted in normal stress-shear strength space

A relatively simple link between the Coulomb and Hoek-Brown criteria can be derived for the
special case of intact unfractured rock where, for the Hoek-Brown criterion in Equation 14.1 mb =
mi, s = 1 and a = 0.5. Letting 3 = r ci gives, for the Hoek-Brown criterion

  1hb = r ci + ci √((r mi) + 1)

and for the Coulomb criterion

1mc = ci + r ci tan

where tan is given by Equation 13.14. Equating 1hb and 1mc we have

r ci + ci √((r mi) + 1) = ci + r ci tan

Dividing through by ci gives

 46

r + √((r mi) + 1) = 1 + r tan

Hence

tan = ((r  1) + √(r mi + 1))/r (14.13)

Noting that tan = tan2((/4) + (/2)) we obtain (14.14)

  r  1  r m i  1    
  2arctan  
  4 
(14.15)
  r
   

The above method for calculating the equivalent angle of friction and cohesion for a Hoek-Brown
material preserves the ‘correct’ value for the Coulomb uniaxial compressive strength, so it can be
referred to as the ‘uniaxial compressive strength’ strategy. Table 14.2 shows angles of friction in
degrees predicted by Equation 14.15 for stress ratios r in the range 0.1 to 1.0, and Hoek-Brown
parameters mi in the range 2 to 30. These values are also tabulated in the worksheet ‘Table’ in the
attached spreadsheet ‘Coulomb – Hoek-Brown Example 14.2’.

Table 14.2 Angles of friction (degrees) predicted by Equation 14.15 for a range of stress ratios r, and
Hoek-Brown parameters mi

Hoek-Brown parameter for intact rock, mi

Stress
ratio r 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30
0.1 18.8 28.6 34.7 39.1 42.4 45.0 47.1 48.9 50.4 51.8 52.9 53.9 54.9 55.7 56.4
0.2 18.3 27.4 33.2 37.2 40.3 42.7 44.7 46.4 47.8 49.1 50.2 51.2 52.0 52.8 53.6
0.3 17.8 26.5 31.9 35.8 38.7 41.0 42.9 44.5 45.9 47.1 48.2 49.2 50.0 50.8 51.5
0.4 17.4 25.7 30.9 34.6 37.4 39.6 41.5 43.0 44.4 45.6 46.6 47.6 48.4 49.2 49.9
0.5 17.0 25.0 30.0 33.6 36.3 38.5 40.3 41.8 43.1 44.3 45.3 46.3 47.1 47.9 48.6
0.6 16.7 24.4 29.2 32.7 35.3 37.5 39.2 40.7 42.1 43.2 44.2 45.2 46.0 46.8 47.5
0.7 16.4 23.8 28.5 31.9 34.5 36.6 38.3 39.8 41.1 42.2 43.3 44.2 45.0 45.8 46.5
0.8 16.1 23.3 27.9 31.2 33.7 35.8 37.5 39.0 40.3 41.4 42.4 43.3 44.1 44.9 45.6
0.9 15.8 22.9 27.4 30.6 33.1 35.1 36.8 38.2 39.5 40.6 41.6 42.5 43.3 44.1 44.8
1.0 15.5 22.5 26.8 30.0 32.5 34.5 36.1 37.6 38.8 39.9 40.9 41.8 42.6 43.4 44.1

Example 14.2

An intact, unfractured specimen of limestone is known to have a uniaxial compressive

strength of 45 MPa and a Hoek-Brown mi value of 10. The specimen will be tested under
triaxial conditions at a confining pressure of 18 MPa.
(a) Apply the ‘uniaxial compressive strength’ strategy to calculate the equivalent angle of
friction and cohesion, assuming the material conforms to the Hoek-Brown failure criterion.
(b) By examining the normal stress and shear strength of the critical shear surface cutting
through the specimen, apply the ‘angle of friction’ strategy to calculate the equivalent
instantaneous angle of friction and cohesion, assuming the material conforms to the Hoek-
Brown failure criterion.

The solution to Example 14.2 is presented in the worksheet ‘Example 14.2’ of the spreadsheet
‘Coulomb  Hoek-Brown Example 14.2’. In this solution, more significant figures have been
adopted than would be warranted for geotechnical calculations, to facilitate checking. For part (a),
 47

application of Equations 14.13 and 14.14 (or Equation 14.15) gives an angle of friction of 37.379º.
Equation 13.12 gives a theoretical Coulomb yield strength of 118.623 MPa, which is the same as the
Hoek-Brown yield strength predicted by Equation 14.1. Application of Equation 13.13 gives a
theoretical Coulomb cohesion of 11.125 MPa based on our specified uniaxial compressive strength
of 45 MPa.

For part (b) we do not know the instantaneous angle of friction, so we cannot calculate the
orientation of the critical shear surface. As an initial estimate we can use the angle of friction
calculated in part (a). Adopting this angle of friction, Equation 13.10 gives the inclination angle  of
the critical shear surface as 63.69º, adopting the Hoek-Brown yield strength from part (a), Equation
13.4 gives the normal stress n on the critical shear surface as 37.77 MPa. Equation 14.8 gives the
parameter h as 1.501 and the parameter  as 0.715. Equation 14.10 gives the instantaneous angle of
friction i as 32.725º, Equation 14.11 gives the shear strength f of the shear surface as 40.213 MPa,
and Equation 14.12 gives the instantaneous cohesion ci as 15.94 MPa. Equation 13.13 gives the
Coulomb uniaxial compressive strength of the intact rock c as 58.39 MPa, and Equation 13.12 gives
the Coulomb yield strength of the specimen 1 as 118.755 MPa. This value of Coulomb yield
strength is slightly different from the value found in part (a) because of the discrepancy between our
initial estimate of the angle of friction (37.379º) and our calculated value of 32.725º. In ‘(b) Final
cycle’ of this worksheet, the estimated angle of friction on the critical shear surface i has been
adjusted so that it is the same as the computed instantaneous angle of friction i given by Equation
14.10 at 31.861º, which gives the Coulomb yield strength of the specimen 1 as 118.623 MPa, which
is the same as part (a)

In the Hoek-Brown (1997) criterion, applying Equations 14.2 to 14.4, the Hoek-Brown parameter a
will be greater than 0.5 when GSI < 25, rising to a theoretical value of 0.65 when GSI is zero. Since
the derivation of the equivalent Coulomb parameters in Equations 14.8 to 14.12 is based on the
assumption that a = 0.5, these equations will not be applicable for Hoek-Brown (1997) when GSI <
25. This limitation is relatively significant in the analysis of rock slope stability, since rocks that are
close to the ground surface tend to be weathered and more heavily broken. If the modified empirical
expressions in Equations 14.5 to 14.7 proposed by Hoek et al (2002) for the Hoek-Brown criterion
are adopted for calculating mb, s and a, the parameter a will only be equal to 0.5 in the rare case
when GSI = 100 and will vary continuously with GSI, rising to approximately 0.6665 when GSI is
zero. Clearly, alternative strategies for calculating the equivalent Coulomb parameters need to be
adopted for the case when a ≠ 0.5 for current and future versions of the Hoek-Brown criterion.
Consideration of these alternative strategies, which involve numerical calculations, is beyond the
scope of this course.

15. Analysis of stresses around underground excavations

The design of an underground excavation generally follows the following broad strategy:

 An initial design is developed that satisfies the need to gain access to, excavate or ventilate
the ore body
 The stresses in the rock immediately adjacent to the excavation wall are estimated by means
of analytical or numerical methods
 If the predicted stresses are found to be greater than the rock mass strength then further
calculations may be required to identify the extent of the zone of failed rock. If the failure
zone is extensive then it will be necessary to modify the excavation design or to design a
suitable rock support system.
 48

 If the predicted stresses are found to be less than the rock mass strength, then it is necessary
to assess the likely impact of mechanical discontinuities in the rock mass. In particular, the
likelihood of slip along a major discontinuity, or the failure of discrete rock blocks, need to
be assessed.

Estimation of the stresses adjacent to an underground excavation is usually based on numerical

modelling techniques using boundary element or finite element packages. These numerical
modelling methods are discussed in the Appendix to these notes. A number of valuable insights into
such matters as localised rock yield, stress concentration and zones of stress influence can, however,
be obtained from analytical, closed-form solutions such as those outlined below. In addition, these
closed form solutions can be used to check and validate the results produced by numerical models.

In two-dimensional analysis of axially symmetrical structures it is generally convenient to represent

the state of stress and strain in terms of polar coordinates r . The angle  (0 to 2) gives the anti-
clockwise angle of rotation from a reference axis. The distance r gives the radial distance from the
centre of the axially symmetrical structure. In two dimensions the normal stress components are rr
(radial stress) and  (tangential or circumferential stress) with the associated shear stress r.
Two-dimensional polar coordinates can be extended to three-dimensional cylindrical coordinates
by defining the Z axis to correspond to the axis of the axially symmetrical structure. In this case the
additional normal stress component is zz (axial stress) and the associated shear stresses are rz and
z. Two-dimensional and three-dimensional strains can be specified in the same way. It is
important to remember that the actual orientation of the radial and tangential stress or strain
components will vary with location around the axially-symmetrical structure. Polar and cylindrical
coordinate systems are adopted because the states of stress and strain around axially symmetrical
structures generally vary directly as functions of the parameters r and . The parameters r and 
therefore not only serve to specify the orientation and location of the stress/strain components, but
also appear in functions that define the values of the stress/strain components themselves at the
specified location.

The Kirsch equations can be utilised to model the effects of in situ stress in an idealised rock mass
(continuous, homogeneous, isotropic , linear elastic). (Kirsch E G, 1898, Die Theorie der Elastizität
und die Bedürfnisse der Festigkeitslehre. Zeitschrift des Vereines deutscher Ingenieure, 42, 797–
807). Figure 15.1 shows a cross-section through a cylindrical hole, such as a borehole or tunnel, of
radius a, subject to remote in situ stresses xx and yy, where xx = K y.
 49

u ur
rr
yy
r


r

a


xx = K yy X

Figure 15.1 Stresses around a circular opening

The radial, tangential and shear stresses at an anti-clockwise angle  measured from the X axis, and a
radial distance r from the centre of the hole are given by:

rr = 0.5yy{(1 + K)[1 – (a2/r2)] – (1 – K)[1  (4a2/r2) + (3a4/r4)] cos 2

 = 0.5yy{(1 + K)[1 + (a2/r2)] + (1 – K)[1 + (3a4/r4)] cos 2 }(15.1)

r = 0.5yy{(1 – K) [1 + (2a2/r2) – (3a4/r4)] sin 2

The stresses in Equations 15.1 give the final state of stress after excavation, not the stress
increments resulting from excavation of the circular opening. In later analysis the subscript ‘k’ will
be added to the above stress components (krr, k and kr) to emphasise that they have been
derived from the Kirsch equations. Equations 15.1 show that rr is at its smallest and  is at its
largest (ie rr, and  are at their most deviatoric) when r = a. This observation is important because
rock failure will always be initiated at the point where the stresses are at their most deviatoric, ie
adjacent to the wall of the excavation.

The radial and tangential displacement increments induced by excavation are given by

ur = (yy a2)/(4G r){(1 + K) – (1 – K)[(4 – 4) – (a2/r2)] cos 2

}(15.2)
u = (yy a )/(4G r){(1 – K)[(2 – 4) + (a /r )] sin 2
2 2 2

where G is the shear modulus (Equation 12.13) and  is the Poisson’s ratio of the rock.

Equations 15.1 and 15.2 are not applicable if the remote in situ stresses are not principal stresses.
The general case where the remote in situ stresses are not principal stresses is given below.
 50

(Although these general equations lie outside the scope of this course, they have been included for
reference.)

Consider a circular hole of radius a, subject to a remote in situ stress tensor given by Equation 15.3
and an internal pressure pi. The X and Y axes are as shown in Figure 15.1; the positive direction of
the Z axis points into the diagram.

 xx  xy  xz 
 
xyz 
  yx  yy  yz 

(15.3)
  zx  zy  zz 


The radial (rr), tangential (), axial (aa) and associated shear stresses at an anti-clockwise angle 
measured from the X axis, and a radial distance r from the centre of the hole are given by

  rr  r  ra 
 
ra
   r    a  (15.4)
  ar  a  aa 

Before writing down the equations it is convenient to define the following intermediate parameters:

S = 0.5 (xx + yy) T = 0.5 (xx  yy)

B = a2/r2 C = a4/r4

Assuming conditions of plane strain and a Poisson’s ratio , the stress components are given by

rr = S (1 – B) + [T (1 + 3C – 4B) cos 2 + [xy (1 + 3C – 4B) sin 2 + B pi

 = S (1 + B)  [T (1 + 3C) cos 2  [xy (1 + 3C) sin 2  B pi

aa = zz 4 (TB cos 2 + Bxy sin 2 }(15.5)


 r = [(T sin 2 + (xy cos 2] (1  3C + 2B)

a = (yz cos   xz sin )(1 + B)

ra = (yz sin   xz cos )(1  B)

Equations 15.5 are required when the remote stresses are not principal stresses, for example when an
inclined cylindrical hole is excavated through a stress field where the principal stresses are vertical
and horizontal.

Returning to the two-dimensional case in Equations 15.1, if xx = yy then K = 1 and Equations 15.1
simplify to:

rr = yy [1 – (a2/r2)]

 = yy [1 + (a2/r2)] }(15.6)

 51

r = 0

For the two-dimensional case, an internal radial pressure pi inside a circular hole of radius a, will
theoretically generate at a radial distance r from the centre of the hole radial, tangential and shear
stresses given by the following expressions:

irr = pi a2/r2

i = – pi a2/r2 }(15.7)

ir = 0

In the current context, this internal pressure pi can be regarded as a support pressure generated by
rock bolts, shotcrete or some other support system. Support pressures in underground excavations
would normally be less than about 0.2 MPa.

If we let the radial, tangential and shear stresses predicted by the Kirsch Equations 15.1 be
respectively krr, k and kr, we can add these stresses to those generated by internal pressure
(Equations 15.7) by simple superposition as follows:

rr = krr + irr

 = k + i }(15.8)

r = kr + ir

The stability of the rock mass subjected to the stresses in Equations 15.8 will depend on its strength,
as discussed in Sections 13 and 14.

Equations 15.1 can be used to demonstrate how an excavation concentrates stress in the rock
immediately adjacent to the excavation boundary. On the excavation boundary r = a, so adopting the
k subscript, Equations 15.1 become

krr = 0

k = yy{(1 + K) + 2(1 – K) cos 2 }(15.9)

kr = 0

Extreme values of tangential stress occur when  = 0 and when  = 90. At these points cos 2 is +1
and –1 respectively, so tangential stress at the excavation boundary expressed as a ratio of the remote
in situ stress is given by

kyy = (3  K) when  = 0 and

}(15.10)
kyy = (3K  1) when  = 90

The terms (3 – K) and (3K – 1) can be regarded as stress concentration factors. When K = 1 the
stress is concentrated by a factor of 2 all round the perimeter of the excavation.
 52

Assuming there is no internal pressure, Equations 15.9 show that the radial stress is zero at the
excavation boundary, so the rock is in a state of uniaxial compression and its yield strength is given
by the uniaxial compressive strength of the rock. Equation 13.12 for the Coulomb criterion shows
that the yield strength of rock is substantially increased by the effect of confining pressure. For
example, a rock with a cohesion of 1.4 MPa and an angle of friction of 35 has, by Equation 13.12, a
theoretical uniaxial compressive strength of only 5.379 MPa. A confining pressure of 0.12 MPa,
generated by a support pressure on the boundary of the excavation, can be represented by setting 3
to 0.12 MPa in Equation 13.12, giving a yield strength of 5.822 MPa, which is an increase of some
0.443 MPa or nearly four times the confining stress. This simple analysis demonstrates that the
action of rock support is, in reality one of generating a small confining pressure which dramatically
increases the strength of the rock, allowing it better to withstand the stress concentrations adjacent to
the excavation boundary. If we accept this model then it is more appropriate to call rock bolts and
shotcrete ‘rock reinforcement’ rather than ‘rock support’ because they strengthen the rock rather than
supporting it like columns in a building. This matter is explored further in Example 15.1 below.

When K = 0 the stress concentration factors are 3 and –1 respectively, implying that tensile stresses
will develop at the  = 90 location. A stress field with K = 0 is unlikely to occur deep within a rock
mass, but in regions adjacent to a large free surface or underground excavation, the stress in one
direction may become relieved almost down to zero and thereby create a K = 0 stress field. In such a
stress field, the creation of tensile stresses at the  = 90 location in a circular excavation is a serious
matter. Rock masses are not good at sustaining tensile stresses, so yield and cracking is likely to
occur under these conditions. If it is the horizontal stress that has been relieved to zero, the  = 90
location will be in the roof of the excavation, so rock yield will be compounded by the effects of
gravity; indeed engineers may, perhaps erroneously, ascribe the cause of the rock failure to the
gravitational dislocation of rock blocks in the tunnel roof.

Equations 15.1, 15.7 and 15.8 provide a two-dimensional analysis of stress. If we do not wish to
apply the relatively complex Equations 15.5 nor adopt three-dimensional numerical modelling, we
can achieve a pseudo-three-dimensional analysis by considering the stress zz parallel to the axis of
the circular opening. The minor principal stress 3 is taken to be the smallest, or most negative of
{rr  zz}. The major principal stress 1 is taken to be the largest, or most positive of {rr 
zz}. In assigning these values we are ignoring the fact that r is not necessarily zero, and that rr
and  are not, therefore, strictly principal stresses. The value of 3 can be input into Equation
13.12 (Coulomb) or Equation 14.1 (Hoek-Brown) to determine the value of 1 at failure, here termed
1f. This value of 1f is then compared to the computed value of the major principal stress 1 at the
specified point within the rock mass.

If 1f > 1 then the rock is predicted to be stable

If 1f < 1 then the rock is predicted to yield

Although localised zones of yielding can be tolerated, substantial zones of rock mass yield would
necessitate a re-design of the excavation geometry and/or the support system.

There are a number of yield criteria that take each of the three principal stresses into account when
calculating rock strength. Consideration of these three-dimensional yield criteria is beyond the scope
of this course.

Example 15.1

An 8 metre diameter horizontal tunnel is to be constructed to provide access to a new area of

an underground mine. Stress measurements show that in this region of the mine the pre-
 53

existing vertical stress is 2 MPa; the horizontal stress normal to the tunnel axis is 5 MPa, and
the horizontal stress along the tunnel axis is 3.5 MPa. The excavation will be supported by
rock bolts and shotcrete which will supply a radial support pressure of 0.12 MPa. The rock
mass strength can be modelled by the Coulomb criterion with an angle of friction of 35 and
cohesion of 1.4 MPa.

Taking the above stresses to be principal stresses, apply the Kirsch equations to delineate the
extent of the zone of rock adjacent to the tunnel wall that will be likely to fail at the 3 o’clock
( = 0) and the 12 o’clock ( = 90º) locations. What support pressure would you recommend
in order to limit the radial extent of the failure zone to no more than 0.4 m beyond the tunnel
wall at these locations?

The attached spreadsheet ‘Kirsch-Coulomb Example 15.1’ utilises Equations 15.1, 15.7 and 15.8 in
the solution of the above problem. The points of interest lie along radial lines at rotation angles  = 0
and 90. At each point of interest 1 and 3 are respectively the largest and smallest of {rr  zz}.
The value of 3 is input into Equation 13.12 with c = 1.4 MPa, and  = 35 to determine the value of
the major principal stress at failure, here termed 1f. The ratio 1f/1 is taken to be a measure of the
factor of safety F of the rock against failure. By applying the ‘Goal Seek’ function in Excel it is
possible to determine the values of radial distance r where F = 1, which is taken to be the limit of the
zone of failure adjacent to the tunnel. For  = 0 there is no predicted failure zone since F = 1.66 at
the most critical location adjacent to the tunnel wall where r = 4.00 m. For  = 90 the failure zone is
predicted to extend to 4.549 m, or 0.549 m into the rock mass. Noting that the 12 o’clock location is
the more critical, we can specify r = 4.4 m then again use ‘Goal Seek’ to identify the value of support
pressure pi that gives F = 1.0 at this location 0.4 m from the tunnel wall. In this case the required
support pressure is 0.497 MPa. When tunnelling in weak yielding rock it is rarely feasible to apply a
support pressure that will completely eliminate all rock failure. For example, setting r = 4.0 m, a
support pressure of 1.624 MPa would be required to eliminate all rock failure adjacent to the tunnel.
Such support pressures would be difficult to achieve both from engineering and financial
perspectives.
 54

16. Non-circular excavations and zones of stress influence

The analysis presented above is only applicable to circular excavations. There are no standard
analytical solutions for stresses and displacements around the more complex excavation geometries
encountered in mining operations. Some useful results have, however, been obtained for elliptical
excavations.

yy

q = ax/ay



ay
xx = K yy X

ax
Y

Figure 16.1 Cross-section through an elliptical excavation

Figure 16.1 shows an elliptical excavation with an axis length ax in the x direction and an axis length
ay in the y direction. As before, the remote in situ stresses are xx and yy, where xx = K yy. Note
that we stipulate the ax axis is parallel to the xx stress axis, so the ellipse is aligned with the stress
field. If we define the axis ratio q = ax/ ay, the tangential stresses at the surface of the excavation at
an anti-clockwise angle  measured from the xx axis, are given by:

At  = 0 and 180 yy (1 – K + 2q)

}(16.1)
At  = 90 and 270 yy (K – 1 + [2K/q])

If q = 1, the opening is circular and Equations 16.1 become the same as Equations 15.9. If q is very
large the excavation is long and thin in cross-section, rather like a long-wall coal mine. In this case,
for large q, Equations 16.1 become:

At  = 0 and 180 very high

}(16.2)
At  = 90 and 270 yy (K – 1)
 55

At the  = 90 and 270 locations, the Equations 16.2 show that the tangential stresses will be tensile
if K < 1.0. We can re-examine Equations 16.1 to identify the conditions under which the tangential
stresses will be compressive. We require

yy (1 – K + 2q) > 0 and yy (K – 1 + [2K/q]) > 0

Dividing by yy and simplifying the inequalities gives the following conditions under which the
tangential stresses will be compressive:

(K – 1)/2 < q < (2K/(1K)) (16.3)

The excavation geometry is optimised when the compressive tangential boundary stresses have their
smallest value. This optimum geometry occurs when q = K, and so from Equations 16.1

yy (1 + K) (16.4)

This result tells us that the optimum design for an opening in a biaxial stress field yy, Kyy is given
by an ellipse orientated with its ax dimension in the x direction and with an axis ratio q = ax/ ay = K.

For example, if q = K = 1.5 then ax = 1.5 ay and xx = 1.5 yy. The ax axis will be the major axis of
the ellipse aligned parallel to the major in situ stress xx. Conversely, if q = K = 0.7 then ax = 0.7 ay
and xx = 0.7 yy. The ay axis will be the major axis of the ellipse aligned parallel to the major in situ
stress yy. Hence, in all cases, for optimal excavation design the major axis of the ellipse should be
aligned parallel to the major in situ stress with the axis ratio q = K. This result is useful where there
is some flexibility in the design of a non-circular underground excavation.

Returning to a circular excavation, Equations 15.6 show that the stress perturbation caused by
excavation decreases with distance from the wall of the excavation. Rearranging Equations 15.6
shows that the radial and tangential stresses, expressed as a ratio of the remote in situ stresses are as
follows

krr/yy = 1 – (a2/r2)
}(16.5)
kyy = 1 + (a /r )
2 2

The radial stress perturbation ratio is –(a2/r2) and the tangential stress perturbation ratio is +(a2/r2).
For example, when a/r is 0.5 then r = 2a, so we are looking at a point one excavation radius beyond
the excavation wall. At this point the stress perturbation will be (0.5)2 or 25%. When a/r is 0.2 we
are looking at a point four excavation radii beyond the excavation wall; at this point the stress
perturbation is only 4%. This point of 4% stress perturbation, two excavation diameters from the
excavation wall, is generally regarded as the limit of influence of the stress concentration induced by
a circular, or approximately circular, excavation.

Clearly, the actual size of the zone of stress influence will depend on the size of the excavation. The
zone of stress influence of a 30 m ‘diameter’ stope will extend to some 60 m beyond the wall of the
stope. A 5 m diameter development drive located 20 m from the stope will therefore be driven
through rock subjected to stress concentration effects of the stope, rather than the ‘pristine’ in situ
stress field. The stope will, however, be outside the zone of stress influence of the development
drive. This development drive will create its own additional stress concentration within this distorted
stress field, which could lead to substantial rock yield. This situation will be exacerbated if the stope
is extended and backfilled. The zone of stress influence may well expand and retreat, exposing the
development drive to the damaging effects of cyclic loading and blast vibrations.
 56

A pair of 30 m diameter excavations separated by 80 m will be outside each other’s zones of stress
influence. There will, however, be a zone of rock some 40 m thick that lies within the zones of stress
influence of both excavations. Equations 15.1 cannot be used to calculate the stress state in this zone
of overlap, because they give the final stress state adjacent to a single circular excavation. In this
case we need to consider the stress increments produced by each excavation. The following example
illustrates this procedure.

yy = 30 MPa

30 m 80 m 30 m

xx = 40 MPa

zz = 35 MPa
Stope A Stope B

Figure 16.2 Cross-section through two adjacent stopes

Example 16.1

Two stopes 30 m across in a deep mine are to be excavated with their walls separated by 80
m of rock, as shown in Figure 16.2. For the purposes of this analysis these stopes can be
regarded as circular excavations, with their horizontal axes parallel to the Z Cartesian
coordinate axis. The remote in situ stresses prior to mining are xx (horizontal) 40 MPa, yy
(vertical) 30 MPa, and zz 35 MPa (normal to the plane of the diagram). Estimate the vertical
and horizontal stresses along a line drawn between the centres of the two stopes after their
excavation.

The attached spreadsheet ‘Superposition Example 16.1’ summarises the calculations for this
problem. The important point to note is that, since we are considering the stress state along a line
drawn between the centres of the two ‘circular’ stopes, their radial stresses will be horizontal and
parallel to the X Cartesian coordinate axis, and their tangential stresses will be vertical and parallel to
the Y Cartesian coordinate axis. This property would not apply along any other direction. The
coincidence of the radial and horizontal stresses, and the tangential and vertical stresses allows us to
add and subtract stress increments directly. This process, which is known as superposition, can only
be done when the component stresses are parallel. In order to consider stress increments along some
other axis we would need to undertake a two-dimensional stress transformation to calculate the xx
and yy stress components from {rr  r} for each stope.

The attached spreadsheet ‘Superposition Example 16.1’ calculates stresses at 9 points at distances
ranging from 35 to 75 m from the centre of Stope A, which correspond to distances of 75 to 35 m
 57

from the centre of Stope B respectively. At each point, the stress increment in the x direction xx
produced by the excavation of stope A is given by krr  xx; similarly the stress increment in the y
direction yy is given by k  yy. Similar calculations for each point give the stress increments
produced by the excavation of Stope B, xx and yy. The final stress state at each point is
given by the algebraic sum xx + xx + xx and yy + yy + yy. These results are
tabulated in the spreadsheet and also plotted graphically. The radial and tangential stresses at
distances of 15 to 35 m from the perimeter of Stope A, ignoring Stope B, are also plotted on this
graph. The step in these graphs at 35 m from Stope A demonstrates the influence of the second
excavation, which in this case tends to increase the stress concentration effects by about 7.7% for
radial stress and 3.8% for tangential stress. These effects would have been more substantial if the
stopes had been closer together.

It is not feasible, and probably not desirable, to pursue further the analysis of stresses adjacent to
complex underground excavation geometries. It is easier, and more reliable, to analyse these stresses
by applying numerical methods outlined in the Appendix to this document.

© Copyright Mining Education Australia (MEA)

These lecture notes were developed as common teaching material for Mining Geomechanics across
member universities of Mining Education Australia. Main contribution for this material was made by
J. Watson and S. Priest.