Sunteți pe pagina 1din 7

Int. J. Rock Mech. Min. Sci. & Geomech. Abstr. Vol. 28, No. 1, pp.

93--99, 1991

Printed in Great Britain. All rights reserved

0148-9062/91 $3.00 + 0.00


Copyright ~ 1991 Pergamon Press pie

Technical Note
A Coupled Distinct Element-Finite Element Method
for Large Deformation Analysis of Rock Masses
X. D. PANel
M. B. R E E D t

INTRODUCTION

iterating this process, an equilibrium configuration is


obtained. It is thus unnecessary to create a global
stiffness matrix.
Although the problem is in reality 3-D, many practical
excavation situations may be treated as plane strain, and
this is assumed here, so that the numerical analyses are
2-D.

In modelling the deformation of the rock mass surrounding an underground excavation, it is important to
use a numerical method which is capable of producing
the mode of deformation observed in practice. This
mode of deformation depends on the shape of the
excavation, and also on the density of fractures in the
rock (e.g. joints, cracks, bedding planes) relative to the
scale of the excavation. If the rock is intact, it may be
modelled as an elastic continuum; if it is densely fractured it may be treated as a viscoplastic continuum, with
planes of weakness introduced using a multilaminate
model; in both cases, a finite element method (FEM) will
give realistic predictions. If the rock fractures are distributed on a scale comparable to the size of the
excavation, the rock may deform discontinuously, as an
assembly of blocks. This is especially likely where the
excavation has a wide, unsupported roof. The distinct
element method (DEM) has been developed to model
such behaviour.
However, in many practical situations, the two modes
of rock behaviour distinguished above may both be
present. For example, in modelling the rock mass deformation around an advancing longwall coal face, the rock
in the roof of the waste region behind the face will fail
in a "blocky" manner, while around the intact coal seam
at the sides of the gate roadways there will be a mainly
continuous deformation. It is the purpose of this paper
to describe a hybrid numerical method capable of modelling such situations, in which the DEM and FEM are
coupled, so that distinct elements are used in part of the
mesh, while the remainder of the mesh consists of finite
elements.
The conventional DEM uses a process of dynamic
relaxation, by integrating Newton's Law over time; the
size of timestep needed, however, is too small to be used
efficiently with the viscoplastic finite element analysis,
and so a DEM using static relaxation is employed. In the
method, each distinct element or block is given a displacement that would bring it to equilibrium if all other
blocks were to remain fixed. By calculating these displacements for each block independently, and then

THE DISTINCT ELEMENT PROGRAM


BLOCK
The static relaxation technique outlined above is more
fully described by Stewart [1], Stewart and Brown [2] and
Pan [3]. At each iteration step, an equation of the form:
~f7 + A e6ue = 0,

(1)

is solved for each block e, to give the incremental


displacements:
ue = (6x e, @~, t~0")T,
of the block centroid in equilibrium with the sum of the
applied forces f~ at the ith contact point of block e,
including applied load, body force and the forces applied
by neighbouring blocks. These latter forces, as well a s
the stiffness matrices A *, are obtained from a set of
constitutive relations which will now be described.
Block interactions are governed by realistic friction
and stiffness laws which are the relations of force and
displacement at contacts. The deformation properties
of both the intact material and the joint are modelled
collectively at the block interfaces. Normal contact and
shear movements between blocks are simulated by overlapping and sliding, respectively, along the interfaces.
Contacts and associated forces between blocks are
assumed to occur at discrete points where block corners
interact with neigbouring block edges and vice versa.
Each contact between blocks is modelled as a normal
interaction and shear interaction which are usually
assumed to be the linear models shown in Fig. !. Typical
load-deformation characteristics for rock joints are
shown in Fig. 2, and a nonlinear model of the normal
stiffness will be described below.
From Fig. 2a, it appears necessary to use a nonlinear
normal stiffness law which has been chosen as Goodman's hyperbolic relation of stress and deformation for

tDepartment of Mathematics and Statistics, Brunel University,


Uxbridge UB8 3PH, U.K.

93

PAN and REED: TECHNICAL NOTE

94

the incremental form of the constitutive relation


becomes:

..... ....:-:.:.Z'?"

..... =========================
...:....,'....**..*...'.':

l
8 f [ + ' = K [ ( v ) " $ v " + ' = K ( l - v " ' V m ' i / ) &'"+*, (5)
RN

where n is the iteration number, N is the normal


direction of contact, J f i s the incremental contact force
and v is the overlap.
The incorporation of (5) into the program requires
that normal stiffness values be associated with corners
Shear interaction
Normal
interact*on
of blocks rather than edges, which means that each
block corner can have a different normal stiffness K~ at
step n.
In the original program, only concentrated external
forces can be applied to the boundaries, which makes
consideration of the in situ stress field somewhat difficult.
Fs
A modification to the program is therefore needed to
make it possible to apply distributed boundary forces in
any direction. This is done by introducing codes to
E
8
Z
transform linearly-distributed edge loads into centroid
force and moment acting on the boundary blocks. The
requirements
of input data for distributed loads can then
Shear displacement
OverLap
be in a similar form to that of the finite element analysis.
Fig. 1. Normal and shear interactions for static relaxation (from
Constrained block displacements at boundaries also
Stewart and Brown [2]).
need to be incorporated into the progam as an application of the boundary displacement condition which is
a joint [4]. The relation can be rewritten:
to occur in coupled problems.
Since the numerical algorithm in static relaxation is of
o.,
= A
(2) an explicit nature whereby incremental displacements
are calculated from the static equilibrium of forces acting
where ~, v and Vm~are indicated on the diagram. For ease on the blocks, a separated block can be longer satisfy the
of computation, ,4 = t = 1 is assumed (,4 > l, t < l, in equilibrium equation as the matrix ,4' is singular. The
fact).
program may stop executing when any block has no
Setting:
contact force (or too few contact points) at its edges or
corners. This limitation might be overcome by introduca~ - ~ = ,~;,
(3)
ing a damping parameter; in this study, however, the
and letting
separated block is either to be temporarily fixed at its
current position or deleted so that calculation can
K = /v~ = "~v
(4)
da~ r-o '
continue.

li!iiiiiiiiiiiiiiii!!iiiiii!iiiii]

[!ii!i!i!i i i i i i i i i i i i l

(a)
O"

I IIII/.

COff

y
/////

"~

.~

"

I ~

,)

/
/
Vm

Au

NormaL dl|ptaoament
Sheor d ispl.ocament

Fig. 2. Typical force-<leformation relation of joints: (a) normal deformation of joints; and (b) shear deformation of joints
(after Goodman [4]).

PAN and REED:

THE FINITE ELEMENT PROGRAM COAL


A finite element program was prepared for 2-D (pla~e
strain or axisymmetric) analyses of displacement and
stresses, using the elasto-viscoplasticity algorithm described in Owen and Hinton [5] (Chap. 8). Eight-noded
quadrilateral elements were employed. The program was
modified to include large displacement theory using the
updated Lagrangian formulation [6]. In addition, loading was achieved by excavation of elements within the
pre-stressed continuum; this was performed by a stepwise reduction in stresses and stiffness for the excavated
elements [3].
Essentially, the algorithm consists of a timestepping
procedure, where at each timestep the global stiffness
equation solved is:
K.6u" = V".

(6)

This gives the nodal displacement vector increment


from time t, to time t,+l, and from this the change
in stress over this time interval at the Gauss integration points may be deduced. The right-hand-side
vector V" includes the residual force vector T" from the
previous timestep, in an attempt to maintain global
equilibrium.

THE COUPLED PROGRAM COUPLE


By representing the rock strata, which constitute the
jointed hard rock around an excavation, with distinct
elements, and the relatively weak and soft rock mass
with finite elements, the conceptual advantages of the
distinct element method are retained, and the full nonlinear problem can be solved. As specific feature of the
hybrid analysis in this research is that both domains of
rock mass may experience relatively large deformation
including slip at the interface.
The interface between the two domains contains
points common to both regions. Determination of
induced contact forces permits direct determination of
the induced total nodal forces on the finite elements.
These induced nodal forces are then added to the
residual force vector ~' and applied as the incremental
load at the current iteration step of the finite element
analysis. The induced nodal displacements of the finite
elements at the interface are thus calculated using
equation (6). These displacements are then fed into
the distinct element domain analysis as displacement

o
#

95

TECHNICAL NOTE

boundary conditions. Thus the relaxation process


continues with nodal displacements and forces being
updated and used in the subsequent hybrid computation
cycles. The iteration can continue until equilbrium is
approximately reached. In principle, the distinct element
program should be run at each time step of the finite
element analysis. However, in the current stage of
the program, it appears not realistic due to the high
computing cost.
The hybrid distinct element and finite element computation is performed by satisfying the conditions of
displacement compatability (usually not continuous)
and force or traction equilibrium at the interface between the two solution domains. The algorithm used in
this research is different from that of the linkage of
distinct element-boundary element described by Lorig
and Brady [7]. In their coupling procedure, a displacement continuity condition at the interface, which is an
elastic body surface was assumed. In this full nonlinear
analysis, however, a discontinuous displacement
phenomenon including slip and separation at the interface is required to be modelled.
It was described in "The Finite Element Program
COAL" that in the 2-D plane strain analysis, the
eight-node isoparametric finite element is used which is
usually with curved sides and a quadratic variation of
the displacement field within the element.
Difficulties exist in applying this curved displacement
boundary condition to the subsequent distinct element
analysis because the distinct elements are defined as rigid
blocks with linear edges. The best solution for this
problem is to introduce a seven-node isoparametric finite
element at the interface. Figure 3 illustrates the element
and its use.
With this element, the displacements ~us at the interface solved from the finite element analysis can coincide
with the required boundary constraints for the distinct
element analysis. Thus, the distinct element solution can
proceed and the force components qd at the interface can
be solved using equation (1) to give:
qd = a e~ Us +

q0d,

in which A e is the explicit stiffness submatrix related t o


the contact force at interface, and q0d represents a set of
initial contact forces.
To satisfy force equilibrium at the interface, equal
and opposite forces are applied to the abutting finite
elements. Note that distinct element corners do

L
i

(7)

5
Fig. 3. Definition of the seven-node finite element.

96

PAN and REED: TECHNICAL NOTE

rate in the finite elements, and updating when this


becomes small.
In subroutine BLOCK, the criterion for the attainment of an approximate static equilibrium state can
be monitored by a measure of either the maximum
incremental force changes or the maximum incremental displacement changes within the distinct
elements. In the practical test of the hybrid program,
another simple criterion which is based directly on the
static equilibrium was found more effective. It can be
written as:

Input control
[
parameters and choose
analysis method

I D.E.M. analysis by I
catting BLOCK
Yes

Store contact forces of I


interfaces for F'.E.M. I
anal ysis : FORFC
J
I

"1

F.E.M. anotysis by I
cottin9 COAL

I {,~f,}7 - {27f,}1 < co',

Update
co-ordinates of
interface
elements for
feeding back to

D.E.M analysis in
next cycLe,by
cottino BACK

(8)

where {Zfi}~'is the total contact force at the interface in


the nth cycle, {Zf~}0 is the initial total contact force at
the interface and co' is a small value controlling the
approximate equilibrium.

RESULTS
The coupled program was tested on three problems, of
increasing complexity.

Loading and excavation


controL, or convergence
check

= 1 : F.E.M.onty~
:2: D.E.M only,

3 =Hybrid analysis./

Fig. 4. The flowchart of program COUPLE.

not have to coincide with finite element nodes at the


interface; the resulting point loads are approximated by
proportionate forces at the end-nodes of the appropriate
element.
The computation sequence for the coupled program is
shown in Fig. 4. The finite element program COAL and
the distinct element program BLOCK are utilized as
two main subroutines with several modifications of
the related codes. For viscoplastic analysis of the continuum region, the time stepping loop is temporarily
stopped to update the boundary condition by the distinct
element analysis in which only the static equilibrium of
the rock blocks is considered. It usually takes several
hundred iterations for the distinct elements to reach
a new equilibrium after changing the displacement
boundary conditions. Here, two criteria are necessary for
the iterations from calculations of COAL to BLOCK
and vice versa.
In subroutine COAL, the criterion for updating
the interface boundary condition is the attainment
of steady-state conditions which are monitored by
accumulating some measure of the viscoplastic strain

(i) Consolidation of blocks on a deformable body


Five rigid blocks, represented by distinct elements, are
placed on an elasto-plastic continuum modelled by a
layer of seven-noded finite elements--see Fig. 5a.
In the test, only the self weight of the blocks is
considered. The yield strength of the finite elements is
chosen such that these elements remain elastic. Various
methods of iteration were tested, all with the same
material properties.
In accordance with the principles of the relaxation
procedure, the number of iterations in each distinct
element analysis cycle has to be large, especially at
the start of analysis (i.e. the first distinct element calculation cycle). Figure 5b shows the influence of the
number of iterations for each cycle on the solution
convergence.
(ii) A tunnel roof undergoing support deformation
Here, an inclined tunnel roof stratum of distinct
elements overlies an excavation in a soft elastic body
which can be modelled by finite elements as illustrated
in Fig. 6. A hydrostatic stress field of magnitude about
10 times greater than the stress at the interface due to
self weight was applied at the boundaries. The simulation was carried out by applying both boundary
load and gravity after excavation, and then estimating the lower, soft body deformations which gave
rise to a downwards displacement of the interface during
the process of calculation. In order to investigate the
discontinuous behaviour of the distinct elements, only
the tunnel roof response is of primary interest in this
example.
The material properties of the roof strata are assumed
as: p = 2.5 t/m 3, KN = 2000 MN/m, Ks = 100 MN/m,
c = 0.001 MPa and in situ stress tro = 2.5 MPa. A maximum vertical displacement at the interface was evaluated

PAN and REED:

9"/

TECHNICAL NOTE
DISTINCT

ELEMENTS

a)

]
I
I

7-NODE FINITE
ELMENTS

0-

TOTAL

Z
O
m

- 15.3

b)

BLOCK SELFWEIGHT = 1 5 . 3 2 8

J,
J
/

>-

15.1

ONE COMPLETE CYCLE= D E M * F E M

ANALYSES

p-

~- 15.9

I
14.7

I
I
I

,,.5

2'o

3'o

&

,o

50

NUMBER OF DEM

ITERATIONS

PER CYCLE

Fig. 5. Consolidation of blocks on a deformable body: (a) blocks and deformable body geometry (DE and FE); and
(b) calculation solutions.

as about 0.2 m and the deformed tunnel roof geometry,


in this case, is shown in Fig. 7a. It is seen from the figure
that the distinct element method in the hybrid calculation is capable of modelling the large movements
of roof blocks, and for comparison Fig. 7b shows the
block deformations obtained when the supporting body
is kept rigid (i.e. a solution involving distinct elements
only).

~ Interface /

I~

!
!
I
Finite Elements
I

.!

Fig. 6. Test of the program on a tunnel roof problem.

(iii) Excavation of a square tunnel


The main purpose of this test example is to verify the
applicability of the program for a fairly large practical
problem.
The rock mass geometry under consideration
shown in Fig. 8 was chosen as being representative
of a particular mining environment. The upper layer
of the rock mass is characterized as a discontinuum
and lower layer as an equivalent continuum. The
excavation proceeds in the continuum close to the
interface at a depth of 600 m and a hydrostatic/n situ
stress of 15 MPa and a gravitational load were applied.

98

PAN and REED:

(o)

TECHNICAL NOTE
Corner force

BLock p l o t
~te'rotlon

Number o f problem

number : 17'00

plot

i t e r a t i o n number :

u n i t s p e r inch : 0 . 8

Number of problem ~nlts per inch : 4 0 9 3 ? 2 9


1EIO0,O0000

Seal.e: I inci~ : - -

Scol,ll : I Inch :

'-lk'~Itli'~&'l&ll''l~'rSr*

,a%%.L,%%%"

~'~'r

t "~

. . . . . . . . .

J J

,~ , ~ /

s " I~I r I

r' r" ,"

" "

(o)

Stress

support

--

Length of stress plot : 3 0 . 0 MPo

the

*. .''.

(b)
Block

plot

Number o f

problem

~tel'otlon number = 4 0 0

units per

"

distribut=on a f t e r deformation and b~ock movement

0.1%

inside the tunnel

.'. . . . .

:X,,''r,

. . . . . . . . . . .

-- :

inch : 0 . 8

(h)

Fig. 9. Force and stress plots after excavation and iterations:


(a) force distribution showing a wide arch shape in the tunnel roof;
and (b) stress distribution showing a stress release zone near tunnel
roof.

Fig. 7. Calculated roof block movements as the results of the support


medium deformation: (a) hybrid method; and (b) distinct elements
only.

A total o f 234 rectangular rock blocks and 126 finite


elements were included in the model. The material
parameters for the blocky rock are: K m = 3000 MNm,
Bl.ock p l o t
Iterotion

I
I

I
I

I
I

I
I

IiI

I
I

I
I

of

problem

583.0000

I I
IIi
11
I
I I I I
I I I
I I I I I I I I
I I I I I I I 1
I I I I I I I IiI
I

Number

number

I
I

I
I

units

p e r inch = 2 . 2 ~ ) 0 0 0

Scol.e : 1 inch =

I
I

i
I

I
I

= = = I z
DistlntBemenlSl
I I I I I
I I I l I
I I I I I
I I I I I
iII
I i

I
I

I
=

I
I

I
I

I l
I I
I I
I I
I I

I
I

I
I
I I
I I
I I
I I
I I
I

I
I

I
I
I I
I I
ii II
I

I
I

I
I
I I
I I
I ii
I
I I I
I

~nlte~ements

Fig. 8. Excavation of the tunnel by reducing the stress and stiffness of


the elements.

Ks0 -- 1333 MN/m, c = 0.1, q~ = 20 , p = 0.025 MN/m ~,


and the input rock mass constants for finite elements are:
E = 8000 MPa, v = 0.25, p = 0.25 MN/m 3, c = 20 MPa,
~0=40 .
A pre-mining stress field of 15 MPa was applied before
the excavation. The gravitational load, compared to this
stress level, has little influence. This "hydrostatic" stress
field will be disturbed by the presence of a discontinuous
rock mass. The force distribution at the interface of
the distinct element domain was calculated from the
consolidation o f blocks under the applied loads and the
converged solution was achieved after 582 iterations o f
relaxation.
The excavation of the tunnel was modelled by
reducing the stress and stiffness of the elements representing the tunnel profile, in ! 0 steps. In each excavation
step, the iterations were carried out until a temporary
equilibrium was reached and the interaction between
the block movements and continuum deformation was
thus simulated.
After 99% of the stress and stiffness o f the tunnel
has been removed, small arch effects were seen in
both roof and floor. At this stage, the rock mass still
behaved elastically and no slip occurred between
rock blocks. With further iteration and excavation,
a relatively large plastic deformation occurred in the
continuum so that slip between the interfaces continued
and stress (or force) in the blocks was further released.
As a result, as shown in Fig. 9, a large influence zone
and a wide arch-shaped distribution o f stress and

PAN and REED: TECHNICAL NOTE


Toter dlspt, plot

HMotlon

---

REFERENCES

Number of problem units per inch : 0 . 0 0 2 8 5

number = 1 8 0 0 . 0 0 0 0 0

~.x

SCoLe:I inch =

Ii I/ //I-"

Fig. 10. Plot of displacement of the roof blocks.

force f o r m e d . C o r r e s p o n d i n g with the a b o v e stress a n d


force plots, d i s p l a c e m e n t s o f r o o f b l o c k s are p l o t t e d in
Fig. 10.

Accepted for publication 24 May 1990.

99

1. Stewart I. J. Numerical and physical modelling of underground

excavations in discontinuous rock. Ph.D. Thesis, Imperial College,


University of London (1981).
2. Stewart I. J. and Brown E. T. A static relaxation method for the
analysis of excavations in discontinuous rock. In Design and
Performance of Under~roand Excavations (E. T. Brown and J. A.
Hudson, Eds), pp. 149-155. British Geotechnical Society, London
0984).
3. Pan X. D. Numerical modelling of rock movements around mine
openings. Ph.D. Thesis, Imperial College, University of London

0988).
4. Goodman R. E. Methods of Geological Engineering in Discontinuous Rock. West, St Paul, Minn. (1976).
5. Owen D. R. J. and Hinton E. Finite in Plasticity. Pineridge Press,
Swansea 0980).
6. Bathe K. J. Finite Element Procedures in Engineering Analysis.
Prentice-Hall, Englewood Cliffs, New Jersey 0982).
7. Lorig L. J. and Brady B. H. G, A hybrid computation scheme for
excavation and support design in jointed rock media. In Design
and Performance of Underground Excavations (E. T. Brown and
J. A. Hudson, Eds), pp. 105-112. British Geotechnical Society,
London (1984).

S-ar putea să vă placă și