NexGen Technologies for Mining and Fuel Industries

ISBN 978-93-85926-40-2

Implicit modelling of excavations in layered rocks

Deepak P. Adhikary and Hua Guo
Coal Mining Program, CSIRO, Brisbane, Australia

ABSTRACT: This paper presents a continuum based formulations for modelling the load-
deformation behaviour of layered rocks, which is often necessary in mining applications (e.g., coal
mining). The modelling of such layered rock masses can be carried out in a discontinuum manner
by explicit introduction of joints using either the finite element or distinct element approach,
however often it is advantageous to devise a continuum-based method when the number of layers
to be modelled is excessively large (i.e., when the layers are thin compared to the dimensions of
the engineering structures). The accuracy of continuum theory based models to accurately describe
the load-deformation behaviour of the layered rocks will be studied in this paper.

1. INTRODUCTION can now be discretised with a coarser mesh

(i.e., subdivided into fewer finite elements)
M odelling the load-deformation behaviour
of rock masses consisting of a large
number of layers is often necessary in mining
than in the discrete models where the size of
the finite elements cannot exceed the layer
thickness. Thus, in smeared joint models, the
applications (e.g., coal mining). Such modelling
size of the elements is solely dictated by
can be carried out in a discontinuum manner by
computational needs rather than by the layer
explicit introduction of joints using either the
thickness.
finite element or distinct element approach
Ubiquitous Joint model proposed in FLAC[3]
(Goodman et al.,[1] Cundall[2]).
When the number of layers to be modelled is is developed for modelling the load
excessively large (i.e., when the layers are thin deformation behaviour of layered materials
compared to the dimensions of the engineering exhibiting strength anisotropy along the layer
structures) it is advantageous to devise a interfaces. Such equivalent continuum models
continuum-based method. A continuum des- may provide reasonably accurate predictions
cription of a layered medium can be formulated when joint slips are minimal i.e., when the
as long as consistency and statistical homo- shearing is in the direction of layering and rock
geneity in joint properties and spacing can be layer bending can be neglected. However,
established. Such a continuum model provides when joint slips are large and loading direction
a large-scale (average) description of the is not aligned with the direction of layering,
material response to loading. The continuum rock layers do bend as they slip against each
model devised in such a manner is often known other. In such cases models based on con-
as smeared (implicit) joint model in a sense that ventional continuum theories may considerably
the joints are implicit in the choice of the overestimate the deformation since the bending
stress-strain relationship adopted for the rigidity of the rock layers are not incorporated
equivalent continuum. A distinctive advantage in such model formulations.
of the smeared joint model is that in a numerical For the case of rock layers with bending
(e.g., finite element) solution the problem region stiffness, such an implicit continuum model can

583
be formulated successfully on the basis of three stress components σ11, σ22, τ = σ21 = σ12.
Cosserat theory (Cosserat and Cosserat[4]). The When the rock layers are aligned in the
Cosserat model provides a large-scale (average) 1-coordinate direction, the moment stress term
description of a layered medium. An important m32 vanishes.
feature of the Cosserat model is that it in-
corporates bending rigidity of individual layers
2.1 Cosserat model
in its formulation and this makes it different
from other conventional implicit models. The four stresses are conjugate to four defor-
Cosserat based equivalent continuum models mation γ 11, γ 22, γ 21, γ 12 measures defined by:
were formulated in (Muhlhaus[5]) and (Adhikary
and Dyskin[6]) where the rock layers were ∂u j
γij = − ε3ijΩ3 … (1)
assumed to be elastic. In,[6] provision was made ∂xi
for plastic deformation along the joints only.
Adhikary and Guo[7] further developed a model and the couple stress m 31 is conjugate to the
incorporating plastic deformation of both joints respective curvature κ 1 defined by:
and rock layers. ∂Ω3
The accuracy of both the Implicit Joint and κ1 = … (2)
∂x1
the Cosserat Models to accurately describe the
load-deformation behaviour of the layered rocks The elastic stress strain relationships are
will be studied in this paper. described by:
σ = ⎡⎣D e ⎤⎦ ee
… (3)
2. THEORETICAL FORMULATIONS
where
A full description of the Implicit (Ubiquitous σ = {σ11, σ 22, σ 21, σ12, m 31} ,
Joint) model for strength anisotropy can be
found in Itasca.[3] A full description of the two e = {γ11, γ 22, γ 21, γ12, κ1} … (4)
dimensional plane strain Cosserat model with
elastic rock layers was previously presented in ⎡ A11 A12 0 0 0⎤
Adhikary and Dyskin[6] (see also references ⎢ A22 0 0 0⎥
⎢ ⎥
cited there) and with elasto-plastic rock layers D= ⎢ G11 G12 0 ⎥ … (5)
was presented in Adhikary and Guo.[7] Hence ⎢ ⎥
we would only concentrate on the major ⎢ symm G22 0 ⎥
differences between the Implicit Joint model ⎢ B1⎥⎦
⎣
and the Cosserat model. For simplicity we would
limit the discussion to two dimensions only. here,
In the Cosserat model using the Cartesian A11 = E … (6)
2 1+ ν 2
coordinates (x1, x2), the material point displace-
1 − ν2 −
ν ( )
ment can be defined by a translational vector
1− ν + E
2
(u1, u2) and by a rotation Ω3, whereas the bkn
material point displacement is defined only by a
translational vector (u1, u2) in the Implicit Joint
1 = 1 + 1 … (7)
G′ G bks
model. Here, axis 3 is aligned to the out of
plane direction and axis 2 is perpendicular to and
the layers.
( )
Df
The two-dimensional Cosserat model has B= 1 − G′ … (8)
b G
4 non-symmetric stress components σ11, σ22, σ21, σ12
where E is the Young’s modulus of the intact
and two couple stresses m31, m32, whereas the layer, ν is the Poisson’s ratio, h is the layer
two-dimensional Implicit Joint model has only thickness, G is the shear modulus of the intact

584
layer, kn and ks are the joint normal and shear where E is the Young’s modulus of the intact
stiffnesses. When the layer thickness h tends to layer, ν is the Poisson’s ratio, h is the layer
zero (i.e., B1 tends to zero) the Cosserat model thickness, G is the shear modulus of the intact
reduces to the standard Implicit Joint model. layer, kn and ks are the joint normal and shear
stiffnesses.
2.2 Implicit joint model The Ubiquitous Joint model described in
FLAC[3] is a strength anisotropy model and thus
The three stresses are conjugate to three
is assumed to have isotropic elastic properties
deformation ε11, ε 22 , γ measures defined by: with reduced strength in the direction of rock
layering.
∂u j
εij = for i = j … (9) Figure 1 presents the stresses and volume
∂xi forces acting on a Cosserat element and Ubi-
quitous Joint element representing a layered
∂u1 ∂u2 rock with layers oriented in the 1-direction.
γ= + … (10)
∂x2 ∂x1
The elastic stress strain relationships are des- 3. DEFICIENCY INHERENT IN THE
cribed by: IMPLICIT JOINT MODEL
σ = ⎡⎣D e ⎤⎦ ee … (11) The Implicit Joint model works well as long as
the rock layers are subjected to translational
where σ = {σ 11 ,σ 22 ,τ } , e = {ε 11 , ε 22 , γ } and deformation without any bending (this may
include slip along the layer interfaces). However
⎡ A11 A12 0 ⎤ when the rock layers undergo bending during
D = ⎢ A21 A22 0 ⎥ … (12) loading the Implicit Joint model (such as
⎢ ⎥
⎢⎣ 0 0 G ⎥⎦ incorporated in FLAC[3] may break down
completely.
here, For simplicity let us assume that the rock
E layer interfaces (joints) have zero shear strength
A11 = … (13) i.e., both the cohesion and the friction angle
2 1+ ν 2
1 − ν2 −
ν ( ) along the layer interfaces are zero and the rock
1− ν + E
2
layer is elastic. Thus if the layered rock is
bkn

σ12≠ σ21 σ12= σ21

m32
fy fy
my
σ22 σ12 σ22
fx fx
σ12
σ21 σ21
m31 σ11
σ11

(a) Cosserat model (b) Ubiquitous joint model

Fig. 1. Stresses and volume forces acting on a Cosserat element and a Ubiquitous joint element.

585
subjected to loading such that the layers slip layered rock deforms in such a way that the
along the interfaces and at the same time layer do slip against each other implying zero
undergo bending, though the magnitude of the σ12, σ21 will remain non-zero and will increase
shear stress component along the layer interfaces with layer bending depending upon the bending
will be zero, the shear stress component rigidity of the rock layers. The couple stresses
perpendicular to the layering direction does not arising from layer bending will counter the
vanish and will increase in proportion to layer rotation arising due to the differences in the two
bending. However in the Implicit Joint model components of the shear stresses.
the magnitude of shear stress component σ21
(i.e., the shear stress component acting in the 4. NUMERICAL VERIFICATION
direction perpendicular to the layering direction)
cannot increase as it is restricted to be equal to 4.1 Comparison with analytical result
σ12 (i.e., the shear stress in the layering A simple case as shown in Figure 3 will be
direction which is assumed to be zero, see considered in order to highlight the deficiency in
Figure 2a). This is due to the virtue of the the Implicit Joint model. Here 10 layers are
fundamental assumptions of two equal shear perfectly clamped on the left-hand side and a
components made in the standard continuum traction τs is applied on the right hand side. The
formulation to avoid the elemental rotation. A rock layers are assumed to have Young’s
zero shear stress component in the direction modulus (E) of 5 GPa, Poisson’s ratio of 0.20,
perpendicular to layering implies essentially a thickness of 1m and length (l) of 10 m. The
weak rubber like material with no bending joint normal and shear stiffness is assumed to be
stiffness. Thus such implicit joint models may very big (i.e., 1010 GPa/m) implying no-elastic
yield erroneous and excessively large deforma- anisotropy. The strength anisotropy is intro-
tions. duced by assuming zero joint shear strength.
This could happen easily in the case of under- Since the shear strength in the layer direction is
ground openings excavated in layered rocks zero, the deformation solution should remain
(Figure 2) where the rock layers slip against independent of the x2 direction, which allows
each other and bend into the excavation analytical verification of the results on the basis
inducing tensile breakage of the layers. Since of beam theory, which yields the elastic
the joint shear strength (i.e., the shear strength deflection of the beam as (Timoshenko and
along the layering direction) is generally low, Goodier[8]):
the shear strength in the Implicit Joint model
will be reduced as soon as joint start to slip 4τsl 3
irrespective of intact rock layer strength. u2(l) =
Eh2
(1 − ν2)
… (14)

u1=0
10 m

u2=0 τs
Ω3=0 x
2

x1
10 m
Fig. 2. A schematic showing (a) erroneous shear
stress that may arise in the Ubiquitous joint model Fig. 3. A schematic of the example used in the analytical
(b) Flexural bending failures. verification.

But the Cosserat model does not suffer from the This problem is analysed with a plane strain
same deficiency since it can have two different Cosserat finite element code as well as Ubi-
shear stress components. During loading if the quitous Joint model built in FLAC.[3] The

586
prroblem domain n is discretizedd into 400 isopaara- 4.2 Coomparison withh explicit joint model
m
metric quadrilatteral elements.. Figure 4 shoows The Coosserat finite element modeel result is
thhe comparison n of the anaalytical and the compareed further witth the results of explicit
nuumerical calcu ulations. The elastic deflecttion joint finnite element model
m result where
w each
obbtained from the
t Cosserat model
m agrees quuite joint is modelled expplicitly. Figuree 5 presents
w with the analytical
well a defllection. Howevver, the finiite element meshm and thee boundary
thhe Ubiquitous Joint
J model prooduces excessivvely conditioons used in thee model. Figurres 6 and 7
laarge deflectionn indicating thhe bending of o a present the displacem ment vectors and
a vertical
ruubber like material with no bending
b stiffneess. stress diistribution plots demonstratinng a remark-
A
Additional simuulations with diifferent rock laayer able maatch between the t explicit jooint and the
Y
Young’s modulu us E or layer thhickness h did not Cosseraat model resultss.
m
make any diffeerence in the Ubiquitous Jooint
m
model FLAC[3] results, wherreas the Cosseerat
4.3 Coomparison witth centrifuge exxperiment
m
model results agreed
a well with
w the analytiical
reesult
soolution (Eq. 144).
The Coosserat finite element moddel is used
further to back anaalyse a centrifuge result
publisheed previously inn Adhikary et al.[9] In that
study a foliated slopee model (manuufactured in
the labooratory from sand and gyppsum) with
330 mm m slope heightt, average layeer thickness
of 10 mm,
m layer incllination angle of 80° and
slope anngle of 61° was w spun in a centrifuge
until it failed
f at a g-levvel of about 833.
In thhat study, addiitional laborattory experi-
ments were
w conductedd to obtain thee basic geo-
mechanical propertiess of the rock laayer and the
joints. Table
T 1 summarrises the laboraatory results.
Fiig. 4. Comparisson of the analyytical and numerrical The vallues shown inn the parentheesis are the
reesults. values adopted
a in thee numerical calculations.
c

Fig. 5. Finiite element mesh and boundary coonditions (left: expplicit joint modell; right: Cosserat model).

Fig. 6. Displacement vectors foor joint friction anngle of 20° (left: explicit joint model; right: Cosserrat model).

587
Fiig. 7. Distributiion of vertical sttress for joint friiction angle of 20°
2 (left: explicitt joint model; riight: Cosserat
m
model).

Taable 1. Rock an
nd joint propertiess obtained in the laboratory
l and ussed in the numericcal analysis
Intact rock layer properties
Young’s modulu
us Cohesion Friction angle
a Tenssile strength
Poisson’’s ratio
(GPa) (MPa) (degreee) (MPa)
2.2–2.6 (2.4) 0.20–0.221 (0.2) 1.4–2.6 (2.0) 34–38 (36)
( 1.1–1.4
(1.55 and 1.7)
J
Joint properties
Cohesion Frictionn angle Shear stiffness
S Normal stiiffness
Tenssile strength
(kPa) (degrree) (GPa/m) (GPa/mm)
5–30 (15) 22–26 (25) (100) (100)) (0)

The joint sheear and normaal stiffnesses were

w The labooratory determ
mination of tenssile strength
noot measured ex xperimentally and
a were assum med of a sinngle 10 mm thhick rock layerr was found
too be 100 GPa per
p meter lengthh in the numerrical to be im
mpractical. Hennce, the tensile strength of
caalculations. Th his implies thhat the joints are the rockk matrix in (A
Adhikary et al.,, 1997) was
prractically rigid up to yielding. An average jooint determinned by five point bending tests on
frriction angle ofo 25° is usedd in the analyysis. a muchh larger rectanggular 50 mm by 50 mm
Jooint dilation anngle is assumeed to be zero. The
T beam of o 300mm lenngth. In Adhikkary et al.
exxperimentally measured joiint cohesion was w (1997), no attempt was w made to study the
foound to vary from
f 5 kPa to 30 kPa, and this
t sample size effect on measured
m tensiile strength.
paarameter is ussed as a fittiing parameter.. A Hence, the numerical calculaations are
nuumber of comp puter simulatioons with differrent conductted here for twwo different values
v (i.e.,
jooint cohesions were carried outo so that a best
b 1.5 MPaa and 1.7 MPaa) of the rock layer
l tensile
fitt to experimeental data couuld be obtainned. strengthh.

Fiig. 8. The finite element mesh annd the boundary conditions

c used inn the Cosserat analysis of the centriffuge model.

588
Fiig. 9. Comparisson of Cosserat anda the centrifugee test results. Num
merical calculatioons were conduccted with rock
tennsile strength of 1.5 MPa and 1.7 MPa.

Figure 8 pressents the Cosseerat finite elem

ment simulatiion of layered rock with thee possibility
mesh and the boundary condiitions used in this
m of rock layer bendingg when shearinng direction
exxample. The problem
p was discretised ussing is not aligned
a with the
t direction of layering
12202 eight nod ded isoparamettric quadrilaterrals. (e.g., fleexural topplingg failures of rock
r slopes,
Fiigure 9 preseents the experrimental and the deformaation of undeerground excaavations in
nuumerical resultts obtained witth a joint cohession layered rocks) will proovide erroneouus results.
vaalue of 15 kPa.. It can be seenn that the Cosseerat The explicit joint model and thhe Cosserat
m
model shows a very good aggreement with the model have been used u to analyyse a load
exxperimental reesults in term ms of the faillure deflectioon problem off layered materrials; a very
m
mechanism, coollapse load, and measuured
good aggreement has been
b obtained between
b the
diisplacement. Th he model with rock layer tennsile
numericcal results of thhe two models.
sttrength of 1.5 MPa
M predicted the t slope failurre at
The Cosserat moddel is also used to back
abbout 66 g wherre as the one with
w 1.7 MPa gave g
a failure g-leveel of about 76 compared to the analyse a centrifuge exxperiment conducted on a
exxperimental g-llevel of about 83. small sccale foliated rock
r slope moodel manu-
facturedd in a laboratoryy. The slope displacement,
failure mechanism annd failure loaad obtained
5.. CONCLUS
SIONS from thhe Cosserat model
m are in very good
Thhe analysis off the constitutivve equations (ii.e., agreemeent with those observed in thee centrifuge
thhe requiremen nt that two shhear stress coom- experimment.
poonents in the Ubiquitous Jooint model be the This study has dem monstrated that the Implicit
saame) and the numerical
n simuulation of bendding Joint model
m based onn non-standardd continuum
off a packagee of layeredd rocks cleaarly (e.g., Coosserat modelss) can accurateely simulate
deemonstrate that the standarrd Implicit Jooint the loadd deformation behaviour
b of layyered rocks.
m
models (e.g., Ubiquitous
U Joinnt model builtt in
FL LAC[3] may co ompletely breaak down when the REFER
RENCES
roock layers undeergo bending during
d loading and
[1] Gooodman, R.E., Taaylor, R.L. andd Brekke, T.,
heence could leaad to erroneouus results. Usee of
“A model for the mechanics
m of joinnted rock”, J.
sttandard impliciit joint models should be limiited Soill Mech. Found. Div., Proc.; ASCE 94, 637–
too small deformmation cases whhere possibilityy of 6599, 1968.
roock layer bendiing is negligiblle. Any attemppt to [2] Cunndall, P.A., “Distinct element moodels of rock
usse such standaard implicit joint model for the and soil structures””, In Anal. and Comp.
C Meth.

589
 in Engng Rock Mech. (Edited by Brown E.T.), associative joint plasticity”, Int. J. Numer. Anal.
Allen and Unwin, London, 129–163, 1987. Methods Geomech, 22(4), 245–261, 1998.
[3] Itasca Consulting Group Inc. FLAC (2008): [7] Adhikary, D.P. and Guo, H., “An orthotropic
Theory and Background, Version 6. Cosserat elasto-plastic model for layered
[4] Cosserat, E. and Cosserat, F., “Theorie des rocks”, Rock Mech. Rock Engg., Vol. 35(3),
corps deformables”, Hermann, Paris, 1909. 161–170, 2002.
[5] Mühlhaus, H.-B., “Continuum models for [8] Timoshenko, S.P. and Goodier, J.N., “Theory of
layered and blocky rock”, Comprehensive Rock Elasticity”, McGraw-Hill, New York, pages
Eng., Invited Chapter for Vol. II: Analysis and 567, 1970.
Design Methods, Pergamon Press, 209–230, [9] Adhikary, D.P., Dyskin, A.V., Jewell, R.J. and
1993. Stewart, D.P., “A study of the mechanism of
[6] Adhikary, D.P. and Dyskin, A.V., “A continuum flexural toppling failure of rock slopes”, Rock
model of layered rock masses with non- Mech. Rock Engng., 30(2), 75–93, 1997.

590