1 Background - The 2D Distinct Element Method

BACKGROUND — THE 2D DISTINCT ELEMENT METHOD 1-1
1 BACKGROUND — THE 2D DISTINCT ELEMENT METHOD
UDEC is described as a “distinct element program.” This numerical method falls within the general
classification of discontinuum analysis techniques. Before describing the distinct element method,
a general overview of discontinuum methods is given.
A discontinuous medium is distinguished from a continuous one by the existence of contacts or
interfaces between the discrete bodies that comprise the system. Discontinuum methods can be
categorized both by the way they represent contacts and by the way they represent the discrete
bodies in the numerical formulation.
1.1 Aspects of Modeling a Discontinuous System
A numerical model must represent two types of mechanical behavior in a discontinuous system:
(1) behavior of the discontinuities; and (2) behavior of the solid material. First, the model must
recognize the existence of contacts or interfaces between the discrete bodies that comprise the
system. Numerical methods are divided into two groups by the way in which they treat behavior
in the normal direction of motion at contacts. In the first group (using a soft-contact approach), a
finite normal stiffness is taken to represent the measurable stiffness that exists at a contact or joint.
In the second group (using a hard-contact approach), interpenetration is regarded as nonphysical,
and algorithms are used to prevent any interpenetration of the two bodies that form a contact.
The choice of contact assumption should be made on the basis of physics rather than on numerical
convenience or mathematical elegance. Depending on the circumstances involved, it is possible for
the same physical system to exhibit different behavior. For example, an assembly of spheres is best
represented with rigid contacts when the friction coefficient is zero and the stress level is very small
(see Papadopoulos 1986). However, if wave propagation is modeled through the same assembly
at higher stress and friction, the contact stiffness must be taken into account in order to obtain the
correct wave speed.
The above comments relate to the magnitude of the contact force. In addition, the contact location
must be identified in the model. For point contacts (or contact almost at a point), the location
of the resultant force vector clearly is at the point of contact. But where contact conditions exist
over a finite surface area on both bodies, the force location is not so obvious. One assumption
might be that the resultant force acts at the centroid of the interpenetration volume. Cundall (1988)
suggests that the location should be regarded as an independent constitutive property, depending
on the relative rotation of the two surfaces in contact. Even if a computer program can relate force
location to geometric variables, there is, at present, very little data from physical tests to substantiate
any physical assumption.
The second type of mechanical behavior that the model must represent is the behavior of the solid
material that constitutes the particles or blocks in the discontinuous system. There are two main
divisions in this representation: the material may be assumed rigid or deformable. The assumption
of material rigidity is a good one when most of the deformation in a physical system is accounted for
by movement on discontinuities. This condition applies, for example, in an unconfined assembly
UDEC Version 4.0

1-2 Theory and Background
of rock blocks at a low stress level, such as a shallow slope in well-jointed rock. The movements
consist mainly of sliding and rotation of blocks and of opening and interlocking of interfaces.
If the deformation of the solid material cannot be neglected, two main methods can be used to
include deformability. In the direct method of introducing deformability, the body is divided into
internal elements or boundary elements in order to increase the number of degrees of freedom.
The possible complexity of deformation depends on the number of elements into which the body
is divided. For example, UDEC automatically discretizes any block into triangular, constant-strain
zones (see Section 1.2.5). In the elastic case, the formulation of these zones is identical to that of
constant-strain finite elements. The zones may also follow an arbitrary, nonlinear constitutive law.
A disadvantage of the method is that a body of complex shape must necessarily be divided into
many zones, even if only a simple deformation pattern is required.
A complex deformation pattern may also be achieved in a body by the superposition of several
mode shapes for the whole body. For example, Williams and Mustoe (1987) rewrite the matrix
equation of motion for an element in terms of a set of orthogonal modes that may or may not be
eigenmodes. Any number of these modes may be added in order to obtain the required complexity
of deformation pattern. The approach is very efficient for bodies of complicated shape that deform
in a simple manner because only a few low modes need to be taken. However, it is not easy to
incorporate material nonlinearity because of the need for superposition.
A somewhat similar scheme was devised by Shi (1989) in his “discontinuous deformation analysis”
(DDA). This method uses series approximations to supply an increasingly complex set of strain
patterns that are superimposed for each block. However, the use of direct strain modes may be
inconsistent (Williams and Mustoe 1987); the comment also applies to the “simply deformable”
element of Cundall et al. (1978).
1.1.1 Computer Programs for Modeling Discontinuous Systems
Many computer programs based upon a continuum mechanics formulation (e.g., finite element and
Lagrangian finite-difference programs) can simulate the variability in material types and nonlinear
constitutive behavior typically associated with a rock mass, but the representation of discontinuities
requires a discontinuum-based formulation. There are several finite element, boundary element
and finite difference programs available which have interface elements or “slide lines” that enable
them to model a discontinuous material to some extent. However, their formulation is usually
restricted in one or more of the following ways. First, the logic may break down when many
intersecting interfaces are used; second, there may not be an automatic scheme for recognizing
new contacts; and, third, the formulation may be limited to small displacements and/or rotation.
For these reasons, continuum codes with interface elements are restrictive in their applicability for
analysis of underground excavations in jointed rock.
A class of computer programs collectively described as discrete element codes provides the capa-
bility to represent the motion of multiple, intersecting discontinuities explicitly. Cundall and Hart
(1989) provide the following definition of a discrete element method: the name “discrete element”
applies to a computer program only if it:
UDEC Version 4.0

(a) allows finite displacements and rotations of discrete bodies, including complete
detachment; and
(b) recognizes new contacts automatically as the calculation progresses.
A discrete element code typically will embody an efficient algorithm for detecting and classifying
contacts. It will maintain a data structure and memory allocation scheme that can handle many
hundreds or thousands of discontinuities.
Cundall and Hart (1989) identify the following four main classes of codes that conform to the
definition of a discrete element method.
1. Distinct element programs use an explicit time-marching scheme to solve
the equations of motion directly. Bodies may be rigid or deformable (by
subdivision into elements); contacts are deformable. “Static relaxation” is
a variation. Representative codes are TRUBAL (Cundall and Strack 1979),
UDEC (Cundall 1980; Cundall and Hart 1985), 3DEC (Cundall 1988; Hart
et al. 1988), DIBS (Walton 1980), 3DSHEAR (Walton et al. 1988) and PFC
(Itasca 1995).
2. Modal methods are similar to the distinct element method in the case of rigid
blocks but, for deformable bodies, modal superposition is used (e.g., Williams
and Mustoe 1987). This method appears to be better-suited for loosely packed
discontinua; in dynamic simulation of dense packings, eigenmodes are appar-
ently not revised to account for additional contact constraints. A representative
code is CICE (Hocking et al. 1985).
3. Discontinuous deformation analysis assumes that contacts are rigid bodies,
and bodies may be rigid or deformable. The condition of no-penetration is
achieved by an iterative scheme; the deformability comes from superposition
of strain modes. The relevant computer program is DDA (Shi 1989).
4. Momentum-exchange methods assume both the contacts and bodies to be
rigid: momentum is exchanged between two contacting bodies during an in-
stantaneous collision. Friction sliding can be represented (for example, see
Hahn 1988).
Another class of codes, defined as limit equilibrium methods, can also model multiple intersecting
discontinuities but does not satisfy the requirements for a discrete element code. These codes use
vector analysis to establish whether it is kinematically possible for any block in a blocky system to
move and become detached from the system. This approach does not examine subsequent behavior
of the system of blocks or redistribution of loads. All blocks are assumed rigid. The “key-block”
theory by Goodman and Shi (1985) and the vector stability analysis approach by Warburton (1981)
are examples of this method.
Cundall and Hart (1989) summarize the attributes of the various discrete element and limit equilib-
rium methods (Figure 1.1). The class of finite element or finite difference methods with slide lines
is not included because of the great variations between programs. There are some programs in this
UDEC Version 4.0

class that exhibit most of the capabilities listed in Figure 1.1, but they do not have both automatic
contact detection and general interaction logic, including finite rotations and interlocking of blocks.
Discontinuous-deformation analysis
Momentum-exchange methods
Limit equilibrium; limit analyses

KEY
Distinct element method

Does not allow it or
not applicable
Can model it but may be
Modal methods
inefficient or not well-suited
Models it well
Class 4:
Class 2:
Class 3:
For each method... Class 1:
Rigid Deformable
contacts contacts
Rigid Deformable
bodies bodies
Small Large
disp. displacement
Small Large
strain strain
Few Many
bodies bodies
Linear Nonlinear
material material
No Fracture
fracture
Loose Dense
packing packing
Static Dynamic
Forces Forces &

only displacements
Discrete element methods
Figure 1.1 Attributes of the four classes of the discrete element method and
the limit equilibrium method (Cundall and Hart 1989)
UDEC Version 4.0

1.1.2 History of the Distinct Element Method
The formulation and development of the distinct element method has progressed for a period of
over 25 years, beginning with the initial presentation by Cundall (1971). Figure 1.2 shows a
chronological chart of the development of the method and relevant papers by Dr. Cundall and his
associates.
The distinct element method was originally created as a two-dimensional representation of a jointed-
rock mass, but the method has also been extended to applications in particle flow research (see
Walton 1980), studies on microscopic mechanisms in granular material (see Cundall and Strack
1983), and crack development in rocks and concrete (see Plesha and Aifantis (1983) and Lorig
and Cundall (1987)). Distinct element models of jointed-rock problems have been made by many
investigators (e.g., Bardet and Scott (1985), Butkovich et al. (1988), Cundall (1974), Hart et al.
(1990), Heuzé et al. (1990)). The most recent two-dimensional program, UDEC (Cundall (1980),
and Lemos et al. (1985)), was first developed in 1980 to combine, into one code, formulations to
represent both rigid and deformable bodies (blocks) separated by discontinuities. This code can
perform either static or dynamic analyses.
In 1983, work was begun by Dr. Cundall on the development of a three-dimensional version of the
method. This work is embodied in a computer program entitled 3DEC, which has been primarily
used to study rockbursting phenomena in deep underground mines (see Tinucci and Hanson 1990)
and to evaluate three-dimensional stress change due to excavation in jointed rock (see Tinucci and
Israelsson 1991). Recently, 3DEC has been applied to the design of an underground powerhouse in
India (Dasgupta et al. 1995) and to the assessment of the ultimate loads of masonry arch structures
(Lemos 1995).
The latest distinct element developments are the two-dimensional and three-dimensional particle
flow codes, PFC 2D and PFC 3D. These codes can be applied to simulate both granular materials,
such as sand, and bonded materials, such as concrete and rock. Fracturing is simulated in PFC 2D
and PFC 3D via progressive bond breakage under load (see Lorig et al. 1995).
UDEC Version 4.0

04 UDEC Vers. 4.0
03 3DEC Vers. 3.0 PFC3D Vers. 3.0
02 PFC2D Vers. 3.0
01
2000
99 UDEC Vers. 3.1 PFC2D, PFC3D Vers. 2.0

3DEC Vers. 2.0
98
(with FISH)
97
UDEC Vers. 3.0
96
(with FISH)
95 PFC2D, PFC3D Vers. 1.1
94 PFC2D, PFC3D Vers. 1.0
93 UDEC Vers. 2.0 3DEC Vers. 1.5

3DEC Vers. 1.0
88 UDEC Vers. 1.4 (Cundall, 1988, Hart et al., 1988)
87 UDEC Vers. 1.3
86 UDEC Vers. 1.2

3DEC (test bed)
85 UDEC Vers. 1.1 (Cundall and Hart, 1985)
84
83 UDEC Vers. 1.0
82
81
UDEC (test bed)
1980 (Cundall, 1980)
TRUBAL
79 (Cundall and Strack, 1979)
RBM, SDEM, DBLOCK, (FORTRAN)
78 (Cundall et al., 1978, Cundall and Marti, 1979)
77
76
75
General DEM (machine language)
74 (Cundall, 1974)
73
72
DEM (special geometry)
71 (Cundall, 1971)
1970
Figure 1.2 Chronology of the distinct element method
UDEC Version 4.0

1.2 Numerical Formulation
1.2.1 Introduction
In the distinct element method, a rock mass is represented as an assembly of discrete blocks. Joints
are viewed as interfaces between distinct bodies (i.e., the discontinuity is treated as a boundary
condition). The contact forces and displacements at the interfaces of a stressed assembly of blocks
are found through a series of calculations which trace the movements of the blocks. Movements
result from the propagation through the block system of disturbances caused by applied loads or
body forces. This is a dynamic process in which the speed of propagation depends on the physical
properties of the discrete system.
The dynamic behavior is represented numerically by a timestepping algorithm in which the size
of the timestep is limited by the assumption that velocities and accelerations are constant within
the timestep. The distinct element method is based on the concept that the timestep is sufficiently
small that, during a single step, disturbances cannot propagate between one discrete element and its
immediate neighbors. This corresponds to the fact that there is a limited speed at which information
can be transmitted in any physical medium. The solution scheme is identical to that used by the
explicit finite-difference method for continuum analysis. The timestep restriction applies to both
contacts and blocks. For rigid blocks, the block mass and interface stiffness between blocks define
the timestep limitation; for deformable blocks, the zone size is used, and the stiffness of the system
includes contributions from both the intact rock modulus and the stiffness at the contacts.
The calculations performed in the distinct element method alternate between application of a force-
displacement law at all contacts and Newton’s second law at all blocks. The force-displacement law
is used to find contact forces from known (and fixed) displacements. Newton’s second law gives
the motion of the blocks resulting from the known (and fixed) forces acting on them. If the blocks
are deformable, motion is calculated at the gridpoints of the triangular finite-strain elements within
the blocks. Then, the application of the block material constitutive relations gives new stresses
within the elements. Figure 1.3 shows schematically the calculation cycle for the distinct element
method. The equations in this figure are described in the following sections.
UDEC Version 4.0

CONSTITUTIVE kn
Dus
ks Dun
m Fs
Fn
Fn : = Fn - knDun
Fs : = Fs - ksDus
Fs : = min{mFn, |Fs|} sgn(Fs)
RIGID DEFORMABLE
BLOCKS BLOCKS
Fci Fci
CONSTITUTIVE
xi
M
MOTION
.
F i = SFci 1 dui du. j
M = Seij xi Fj
(
DAij = 2 dx + dx Dt
j
(
i
sij = C (sij , DAij , ...)
üi = Fi / m
q=M/I
MOTION
Fei = z sij nj ds
.
.
.
F i = Fei + Fci
üi = Fi / m
.
.
t := t + D t .
back to
Figure 1.3 Calculation cycle for the distinct element method
UDEC Version 4.0

1.2.2 Equations of Motion
The motion of an individual block is determined by the magnitude and direction of resultant out-
of-balance moment and forces acting on it. In this section, the equations of motion that describe
translation and rotation of the block about its centroid are developed. Consider the one-dimensional
motion of a single mass acted on by a varying force, F (t). Newton’s second law of motion can be
written in the form
d u̇ F
= (1.1)
dt m
where u̇ = velocity;
t = time; and
m = mass.
The central difference scheme for the left-hand side of Eq. (1.1) at time t can be written as
d u̇ u̇(t+t/2) − u̇(t−t/2)
= (1.2)
dt t
Substituting Eq. (1.2) in Eq. (1.1) and rearranging yields
F (t)
u̇(t+t/2) = u̇(t−t/2) + t (1.3)
m
With velocities stored at the half-timestep point, it is possible to express displacement as
u(t+t) = u(t) + u̇(t+t/2) t (1.4)
Because the force depends on displacement, the force/displacement calculation is done at one time
instant. Figure 1.4 illustrates the central difference scheme with the order of calculation indicated
by the arrows. The central difference scheme is “second-order accurate” — i.e., first-order error
terms vanish from the solution. This is an important characteristic that prevents long-term drift in
a distinct element simulation.
UDEC Version 4.0

1 - 10 Theory and Background
∆t
2
..
u
.
u
F
∆t ∆t
Figure 1.4 Interlaced nature of the calculation cycle used in distinct element
formulation
For blocks in two dimensions that are acted upon by several forces as well as gravity, the velocity
equations become:
(t)
(t+t/2) (t−t/2) Fi
u̇i = u̇i + + gi t
m
(1.5)

M (t)
θ̇ (t+t/2)
= θ̇ (t−t/2)
+ t
I
where θ̇ = angular velocity of block about centroid;

I = moment of inertia of block;

M = total moment acting on the block;
u̇i = velocity components of block centroid; and
gi = components of gravitational acceleration (body forces).
In Eq. (1.5) and those that follow, indices i denote components in a Cartesian coordinate frame,
and summation is implied for repeated indices in an expression.
UDEC Version 4.0

BACKGROUND — THE 2D DISTINCT ELEMENT METHOD 1 - 11
The new velocities in Eq. (1.5) are used to determine the new block location according to:
(t+t) (t) (t+t/2)

xi = xi + u̇i t
(1.6)
θ (t+t) = θ (t) + θ̇ (t+t/2) t
where θ = rotation of block about centroid; and

xi = coordinates of block centroid.
Note that rotations are not stored; incremental rotations are used to update the positions of block
vertices. In summary, each timestep produces new block positions that generate new contact
forces. Resultant forces and moments are used to calculate linear and angular accelerations of each
block. Block velocities and displacements are determined by integration over increments in time.
The procedure is repeated until a satisfactory state of equilibrium or continuing failure results.
Mechanical damping is utilized in the equations of motion (Eq. (1.5)) to provide both static and
dynamic solutions (see Sections 1.2.6 and 4).
1.2.3 Conservation of Momentum and Energy in the Distinct Element Formulation
Many continuum-based computer programs start with a statement of the conservation laws and
then derive the necessary equations from these for the formulation of the numerical schemes. This
approach is used to demonstrate that these codes satisfy conservation of momentum and energy in
their dynamic simulations.
The distinct element method can also be shown to satisfy the conservation laws, but by using
an alternative approach based on the use of Newton’s laws of motion. The equations used in
UDEC are based on the interaction of bodies by means of springs and the response of the bodies
to applied forces (see Figure 1.3). Although the following equations show that the conservation
laws are satisfied exactly by using Newton’s laws of motion, there will be some error introduced
in the computer program by the numerical integration process; however, this error may be made
arbitrarily small by the use of suitable timesteps and high-precision coordinates. Verification of the
formulation in UDEC and demonstration of the accuracy for static and dynamic analysis have been
performed by comparison to closed-form solutions (see the Verification Problems and Example
Applications volume).
Momentum Balance — Consider two bodies (denoted by subscripts a and b) in contact for a period,
T . By Newton’s laws, a common force, F , acts in opposite directions on the two bodies, which
accelerate in proportion to the forces:
ma üa = F (1.7)
UDEC Version 4.0

mb üb = −F (1.8)
Combining the equations and integrating:
T T
ma üa dt = − mb üb dt (1.9)
0 0
(T ) (0)
a − u̇a ) = −mb (u̇b
ma (u̇(T − u̇b )
) (0)
(1.10)
(T ) (0)
a + mb u̇b
ma u̇(T = ma u̇(0)
a + mb u̇b
)
(1.11)
Eq. (1.11) indicates that the total momentum at the end of an arbitrary time period is identical to
that at the beginning.
Energy Balance — Suppose a body with initial velocity v0 is brought to a final velocity of v in a
distance S by a constant force F :
mv̇ = F (1.12)
Using the identity v̇ = v dv/ds,
v S
m v dv = F ds (1.13)
v0 0
assuming m is constant. Hence,
1
m (v 2 − v02 ) = F S (1.14)
2
Eq. (1.14) expresses the fact that the work done by the force is equal to the change in kinetic energy
of the body.
If the force opposing motion is related to the displacement by the equation (F = −ks), where k
denotes the spring stiffness, then Eq. (1.13) is replaced by
v S
m v dv = − ks ds (1.15)
v0 0
UDEC Version 4.0

Hence,
1 1
m (v02 − v 2 ) = kS 2 (1.16)
2 2
In this case, the decrease in kinetic energy equals the energy stored in the spring. The same argument
may be used in reverse to show that the kinetic energy acquired by a body is equal to the decrease
in energy stored in a spring. Hence, the kinetic energy of a body after an elastic collision is equal
to the kinetic energy before the collision.
1.2.4 Rock Joint Representation
1.2.4.1 Contact Detection and Identification
A rock joint is represented numerically as a contact surface (composed of individual point contacts)
formed between two block edges. In general, for each pair of blocks that touch (or are separated
by a small enough gap), data elements are created to represent point contacts. In UDEC, adjacent
blocks can touch along a common edge segment or at discrete points where a corner meets an edge
or another corner. Figure 1.5 illustrates the scheme for representation of contacts. For rigid blocks,
a contact in UDEC is created at each corner interacting with a corner or edge of an opposing block.
If the blocks are deformable (internally discretized), point contacts are created at all gridpoints
located on the block edge in contact. Thus, the number of contact points can be increased as a
function of the internal zoning of the adjacent blocks.
A specific problem with contact schemes is the unrealistic response that can result when block
interaction occurs close to or at two opposing block corners. Numerically, blocks may become
locked or hung-up. This is a result of the modeling assumption that block corners are sharp or have
infinite strength. In reality, crushing of the corners would occur as a result of a stress concentration.
Explicit modeling of this effect is impractical. However, a realistic representation can be achieved
by rounding the corners so that blocks can smoothly slide past one another when two opposing
corners interact. Corner rounding is used in UDEC by specifying a circular arc for each block
corner. The arc is defined by the distance from the true apex to the point of tangency with the
adjoining edges. Examples are shown in Figure 1.6. By specifying this distance rather than a
constant radius, truncation of sharp corners is not severe (compare Figure 1.6(a) to Figure 1.6(b)).
UDEC Version 4.0

BL Initial position
un OC
1 K2 of Block 2
us
1
JO
INT
BL
x OC
K1
us un
2
2
Block Centroid
Figure 1.5 Contacts between two rigid blocks
d=r d >> r
r
d r
d
d = distance to corner
r = radius of rounded corner
(a) rounding of corners using constant rounding length, d
d >> r
d=r
r
r
d
d
(b) rounding of corners using constant radius, r, demonstrating

unacceptable truncation of acute angled corner
Figure 1.6 Definition of rounded corners in UDEC
The point of contact between a corner and an edge is located at the intersection between the edge
and the normal taken from the center of the radius of the circular arc at the corner with the edge (see
UDEC Version 4.0

Figure 1.7(a)). If two corners are in contact, the point of contact is the intersection between the line
joining the two opposing centers of radii and the circular arcs (Figure 1.7(b)). If the edges of two
deformable blocks are in contact (i.e., edge-to-edge contact), the points of contact are still treated
as corner-to-edge contacts, but they are located at the intersections of the normal to the gridpoints
along the edge of one block and the edge of the other block; corner rounding is not used in this
case. If a gridpoint along the edge of one deformable block is created at the same location as a
gridpoint along the edge of another deformable block, two contacts will be created, one contact for
each gridpoint. This provides for better accuracy, particularly if the two gridpoints slide past each
other.
The directions of normal and shear force acting at each corner-to-corner or corner-to-edge contact
are defined with respect to the direction of the contact normal, as illustrated in Figure 1.7.
Corner rounding only applies to the contact mechanics calculation in UDEC. All other calculations
and properties, such as block and zone mass, are based on the entire block. Corner rounding can
introduce inaccuracy in the solution if the rounding is too large. If the rounding length is kept to
approximately one percent (1%) of the representative block edge length in the model, good accuracy
is achieved.
1.2.4.2 Domain Contact Detection
Contact points are updated automatically as block motion occurs. The algorithms to perform
this updating must be computationally efficient, particularly for dynamic analysis, where large
displacements may require deleting and adding hundreds of contacts during the dynamic simulation.
UDEC takes advantage of a network of “domains” created by the two-dimensional block assembly.
Domains are the regions of space between blocks which are defined by the contact points, as are
D1 and D2 in Figure 1.8. During one timestep, new contacts can only be formed between corners
and edges within the same domain, so local updates can be executed efficiently whenever some
prescribed measure of motion is reached within the domain. The main disadvantage of this scheme
is that it cannot be used for very loose systems because the domain structure is ill-defined.
Contact updating is triggered by significant relative motion within a domain. A fictitious displace-
ment is accumulated for each domain, and this displacement is related to the relative motion that has
taken place in the domain since the previous update. The fictitious displacement is the accumulated
maximum relative velocity between any two corners in a domain times the timestep. When this
displacement exceeds a certain tolerance (35% of the rounding length), an update is triggered. This
ensures that contacts are always detected before physical contact is made. During an update, new
contacts are made and old ones deleted depending on the relative motion at each contact — for
example, if the relative shear displacement at a contact exceeds two times the rounding length, a
new contact is formed.
For block motion involving large shear displacements, contact updating must ensure that contact
forces are preserved when contacts are added or deleted such that a smooth transition will exist
between neighboring states. This is particularly important for dynamic analysis with high stress
gradients. The contact update logic in UDEC has been tested for explosion-driven motion involving
large displacement, and is reported by Hart et al. (1987).
UDEC Version 4.0

BLOCK B
direction
of contact
normal shear
direction
r
edge un
d
BLOCK A
corner Pxi
(a) detail of rounded corner-to-edge contact (rounding length exaggerated)
BLOCK B
direction of
contact normal
r
shear
BLOCK A direction
(b) smooth interaction of corner-to-corner contact
Figure 1.7 Definition of contact normal in UDEC
UDEC Version 4.0

BLOCK 1
1 D1 2 D2 3
L1 L2 L3
BLOCK 2
Finite Difference
Zone
Gridpoints
1 , 2 , 3 Corner-Edge Contacts
L 1 , L 2 , L 3 Length Associated
with Contacts
D1 , D2 Domains
Figure 1.8 Contacts and domains between two deformable blocks
1.2.4.3 Cell Mapping and Searching
There is an alternative contact detection logic built into UDEC. This is called cell mapping. This
logic is used in models where there may be blocks which become detached from other blocks and
may bounce. An example of this type of model is one where blocks are bouncing down a slope.
The cell logic is invoked by use of the CONFIG cell command.
The space containing the system of blocks is divided into rectangular cells. Each block is mapped
into the cell or cells that its “envelope space” occupies. A block’s envelope space is defined as the
smallest box with sides parallel to the coordinate axes that can contain the block. Each cell stores,
in linked-list form, the addresses of all blocks that map into it. Figure 1.9 illustrates the mapping
logic for a two-dimensional space. Once all blocks have been mapped into the cell space, it is an
easy matter to identify the neighbors to a given block: the cells that correspond to its envelope
space contain entries for all blocks that are near. Normally, this “search space” is increased in all
directions by a tolerance, so that all blocks within the given tolerance are found. Note that the
computer time necessary to perform the map and search functions for each block depends on the
size and shape of the block, but not on the number of blocks in the system. The overall computer
UDEC Version 4.0

time for neighbor detection is consequently directly proportional to the number of blocks, provided
that cell volume is proportional to average block volume.
cell space
21 22 23 24 25
A, B, C = rock blocks
16 17 18 19 20
sequential C block
cell numbers 11 12 13 14 15 envelope
A
B
6 7 8 9 10
1 2 3 4 5
cell entries
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 in main array
main array
block entries
O A A B O B A O C O C O C
in main array,
global pointer shown separately
to cell space B O C O C for clarity
O C O A
Figure 1.9 Examples of block mapping to cell space, in two dimensions
It is difficult to provide a formula for optimum cell size because of the variety of block shapes that
may be encountered. In the limit, if only one cell is used, all blocks will map into it, and the search
time will be quadratic. As the density of cells increases, the number of non-neighboring blocks
retrieved for a given block will decrease. At a certain point, there is no advantage in increasing the
density of cells, because all the blocks retrieved will be neighbors. However, by further increasing
the cell density, the time associated with mapping and searching increases. The optimum cell
density must therefore be in the order of one cell per block, in order to reduce both sources of
wasted time.
Scheme for Triggering Neighborhood Searches — As a block moves during the course of the
simulation, it is remapped and tested for contact with new neighbors. This process is triggered by
the accumulated movement of the block: a variable uacc , set to zero after each remap, is updated at
every timestep, as follows:
uacc := uacc + max{abs(du)} (1.17)
UDEC Version 4.0

where du is the incremental displacement of a corner, and the max{ } function is taken over all
corners of the block.
When uacc exceeds ATOL/2.0 (ATOL is half the rounding length), remapping and contact testing are
activated. The contact testing is done for a search volume that is 2∗ATOL larger in all dimensions
than the block envelope. In this way, maximum movement of the block, and any potential neighbor,
is allowed. If any block attempts to move outside the cell space (i.e., the total volume covered by
cells), the cell space is redefined to be 10% larger in the affected dimension. In this case, a complete
remap of all blocks occurs.
The value of ATOL is also used to determine whether a contact is created. BTOL is used to determine
if a contact is deleted. If two blocks are found to be separated by a gap that is equal to or less than
ATOL, a contact is created. Conversely, if an existing contact acquires a separation that is greater
than BTOL, the contact is deleted. The logic described above ensures that the data structure for
all potential contacts is in place before physical contact takes place. It also ensures that contact
searching is only done for moving blocks; there is no time wasted on relatively inactive blocks.
It is important to note that for close-packed block models, the cell logic is not as efficient as the
domain logic. Also, some features such as fluid flow, which depend on the domains, are not available
when using the cell logic.
When using the domain logic, only one block may be defined and then split into additional blocks.
The cell logic does not require this restriction. Therefore, multiple blocks may be defined when
cell logic is selected.
1.2.4.4 Joint Behavior Model
Only two types of contacts are needed by the data structure for representing a system of blocks:
corner-to-corner contacts and edge-to-corner contacts. These are termed “numerical contacts.”
Physically, however, edge-to-edge contact is important, because it corresponds to the case of a rock
joint closed along its entire length. A physical edge-to-edge contact corresponds to a domain with
exactly two numerical contacts in its linked-list (see Section 1.2.11). The joint is assumed to extend
between the two contacts and be divided in half, with each half-length supporting its own contact
stress (see Figure 1.8). Incremental normal and shear displacements are calculated for each point
contact and associated length (i.e., L1 , L2 and L3 in Figure 1.8).
Many types of constitutive models for edge-to-edge contact may be contemplated. The basic joint
model used in UDEC captures several of the features which are representative of the physical
response of joints. In the normal direction, the stress-displacement relation is assumed to be linear
and governed by the stiffness kn such that
σn = −kn un (1.18)
where σn is the effective normal stress increment; and

un is the normal displacement increment.
UDEC Version 4.0

The overlap shown in Figure 1.5, for example, represents a mathematically convenient way of mea-
suring relative normal displacement. This is the soft contact assumption described in Section 1.1.
There is also a limiting tensile strength, T , for the joint. If the tensile strength is exceeded (i.e., if
σn < −T ), then σn = 0. Similarly, in shear, the response is controlled by a constant shear stiffness,
ks . The shear stress, τs , is limited by a combination of cohesive (C) and frictional (φ) strength.
Thus, if
|τs | ≤ C + σn tan φ = τmax (1.19)
then
τs = −ks ues (1.20)
or else, if
|τs | ≥ τmax (1.21)
then
τs = sign(us ) τmax (1.22)
where ues is the elastic component of the incremental shear displacement; and
us is the total incremental shear displacement.
This model is described as the Coulomb slip model and is illustrated in Figure 1.10. In addition, joint
dilation may occur at the onset of slip (nonelastic sliding) of the joint. Dilation is governed in the
Coulomb slip model by a specified dilation angle, ψ. The accumulated dilation is generally limited
by either a high normal stress level or by a large accumulated shear displacement which exceeds
a limiting value, ucs . This limitation on dilation corresponds to the observation that crushing of
asperities at high normal stress or large shearing would eventually prevent the joint from dilating.
In the Coulomb model, the dilation is restricted such that (see Figure 1.10)
if |τs | ≤ τmax , then ψ = 0
and
if |τs | = τmax and |us | ≥ ucs , then ψ = 0.
UDEC Version 4.0

Shear
stress 1
s 2
increasing normal
effective stress
3
n
4
ks
1
Shear displacement u s
=0
Dilational
component 4
of normal 3
increasing normal
displacement effective stress
2
u nd Dilation
n
angle 1
Shear
Critical displacement u cs displacement
us
Figure 1.10 Basic joint behavior model used in UDEC
Dilation is a function of the direction of shearing. Dilation increases if the shear displacement
increment is in the same direction as the total shear displacement, and decreases if the shear
increment is in the opposite direction.
The dilation, by default, does not affect the shear strength in UDEC. As an option, the dilation
can be included in the effective friction angle for the joint. In this case, the dilation is added to the
input friction angle if the shear displacement increment is in the same direction as the total shear
displacement, and subtracted if the increment is in the opposite direction. This option can be used
to approximate the effect of cyclic shearing on changes in shear strength of a joint.
The Coulomb model can also be adapted to approximate a displacement-weakening response, which
is often observed in physical joints. This is accomplished by setting the joint friction, cohesion
and tensile strength to reduced values (usually zero) whenever either the tensile or shear strength
is exceeded.
A more comprehensive displacement-weakening model is also available in UDEC. This model, the
continuously yielding joint model, is intended to simulate the intrinsic mechanism of progressive
damage of the joint under shear. The Barton-Bandis joint model (Barton 1982; Barton et al. 1985)
is also available as an option to UDEC. This model is described in Section 3 in Special Features.
UDEC Version 4.0

1.2.5 Block Deformability
Blocks may be rigid or deformable in the distinct element method. The basic formulation for rigid
blocks is given by Cundall et al. (1978). This formulation represents the medium as a set of
distinct blocks that do not change their geometry as a result of applied loading. Consequently, the
formulation is most applicable to problems in which the behavior of the system is dominated by
discontinuities and for which the material elastic properties may be ignored. Such conditions arise
in low-stress environments and/or where the material possesses high strength and low deformability.
For many applications, the deformation of individual blocks cannot be reasonably ignored — i.e.,
blocks cannot be assumed to be rigid. This requirement is discussed in Section 3.3 in the User’s
Guide. “Fully deformable” blocks were developed in UDEC to permit internal deformation of each
block in the model.
Deformable blocks are internally discretized into finite-difference triangular elements. The com-
plexity of deformation of the blocks depends on the number of elements into which the blocks are
divided. Figure 1.11 illustrates the zoning within blocks for a system of continuous and discontin-
uous joints. The use of triangular elements eliminates the problem of hourglass* deformations that
may occur with constant-strain finite-difference quadrilaterals.
The vertices of the triangular elements are gridpoints (see Figure 1.3), and the equations of motion
for each gridpoint are formulated as follows:

σij nj ds + Fi
üi = s
+ gi (1.23)
m
where s is the surface enclosing the mass m lumped at the gridpoint;

nj is the unit normal to s;
Fi is the resultant of all external forces applied to the gridpoint (from block contacts or other-
wise); and
gi is the gravitational acceleration.
* The term “hourglassing” comes from the shape of the deformation pattern of elements within a
mesh. For polygons with more than three nodes, combinations of nodal displacements exist which
produce no strain and result in no opposing forces. The resulting effect is unopposed deformations
of alternating direction.
UDEC Version 4.0

(a) distinct element blocks (b) zoning within blocks
Figure 1.11 Zoning within a model containing a system of continuous and

discontinuous joints
Gridpoint forces are obtained as a sum of three terms:
Fi = Fiz + Fic + Fil (1.24)
Fil are the external applied loads. Fic result from the contact forces and exist only for gridpoints
along the block boundary. Forces from contacts along the two edges adjacent to the gridpoint
contribute to this term. Because a linear variation of displacements is assumed along any edge, the
effect of contact forces applied along an edge may be represented by statically equivalent forces
applied to the edge endpoints. Finally, the contribution of the internal stresses in the zones adjacent
to the gridpoint is calculated as

Fiz = σij nj ds (1.25)
c
where σij is the zone stress tensor; and
UDEC Version 4.0

nj is the unit outward normal to the contour C, which follows the closed
polygonal line defined by the straight segments which bisect the
zone edges converging on the gridpoint under consideration.

A net nodal force vector, Fi , is calculated at each gridpoint. This vector includes contributions
(g)
from applied loads, as discussed above, and from body forces due to gravity. Gravity forces, Fi ,
are computed from
(g)
Fi = gi mg (1.26)
where mg is the lumped gravitational mass at the gridpoint, defined as the sum of one-third of the
masses of triangles connected
to the gridpoint. If the body is at equilibrium, or in steady-state
flow (e.g., plastic flow), Fi on the node will be zero; otherwise, the node will be accelerated
according to the finite difference form of Newton’s second law of motion:
(t+t/2) (t−t/2)
(t) t
u̇i = u̇i + Fi (1.27)
m
where the superscripts denote the time at which the corresponding variable is evaluated.
During each timestep, strains and rotations are related to nodal displacements in the usual fashion:
1
˙ij = (u̇i,j + u̇j,i )
2
(1.28)
1
θ̇ij = (u̇i,j − u̇j,i )
2
Notice that, due to the incremental treatment, Eq. (1.28) does not imply a restriction to small strains.
The constitutive relations for deformable blocks are used in an incremental form so that imple-
mentation on nonlinear problems can be accomplished easily. The actual form of the equations
is:
σije = λ v δij + 2µ ij (1.29)
where λ, µ are the Lame constants;

σije are the elastic increments of the stress tensor;
ij are the incremental strains;
v = 11 + 22 is the increment of volumetric strain; and
UDEC Version 4.0

δij is the Kronecker delta function.
Nonlinear and post-peak strength models are readily incorporated into the code in a direct way
without recourse to devices such as equivalent stiffnesses or initial strains, which need to be intro-
duced into matrix-oriented programs to preserve linearity dictated by the matrix formulation. In an
explicit program, however, the process is much simpler — after each timestep, the strain state of
each zone is known. The program then needs to know the stress in each zone in order to proceed to
the next timestep. The stress is uniquely defined by the stress-strain model whether it is a linearly
elastic relation or a complex, nonlinear and post-peak strength model.
The basic failure model for blocks in UDEC is the Mohr-Coulomb failure criterion with a non-
associated flow rule. Other nonlinear plasticity models available in UDEC are the Drucker-Prager
failure criterion, the ubiquitous joint model and strain-softening models for both shear and volu-
metric (collapse) yield. The block material models are described in Section 2.
Accurate Modeling of Plastic Collapse — UDEC is primarily intended to simulate mechanisms
related to movement along discrete features (such as joints and faults) within a rock mass. However,
in many problems, the failure and collapse of intact material (for example, roof collapse or sloughing
of sidewalls of excavations) must also be accommodated in the model.
When using the block plasticity models, described in Section 2, it is important to recognize that an
overestimation of the collapse load may be calculated for the constant-strain triangular elements in
UDEC.
A common problem that occurs in modeling of materials undergoing active collapse is the incom-
pressibility condition of plastic flow. The use of plane-strain geometries introduces a kinematic
restraint in the out-of-plane direction, often giving rise to over-prediction of collapse load. This con-
dition is sometimes referred to as “mesh-locking” or “excessively stiff” elements and is discussed in
detail by Nagtegaal et al. (1974). The problem arises as a condition of local mesh incompressibility
which must be satisfied during flow, resulting in over-constrained elements.
One method to overcome this problem is referred to as “mixed discretization” (see Marti and Cundall
1982). In this method, the isotropic stress and strain components are treated separately for each
triangular sub-element. The term “mixed discretization” comes from the different discretizations
for the isotropic and deviatoric parts of the stress and strain tensors.
Mixed discretization is used in the Itasca continuum codes FLAC and FLAC 3D. However, the
procedure cannot be generally used in UDEC because of the difficulty in adapting the procedure
for discretization of arbitrarily shaped blocks.
As an alternative, a special zoning generator has been developed (see the GEN quad command in
Section 1.3 in the Command Reference) which creates diagonally opposed triangular elements in
blocks and works for most block shapes. Diagonally opposed triangles have also been demonstrated
by Marti and Cundall (1982) to improve the accuracy of calculations for plastic collapse.
UDEC Version 4.0

1.2.6 Mechanical Damping
Mechanical damping is used in the distinct element method to solve two general classes of problems:
static (non-inertial) solutions and dynamic solutions. A different form of damping is used for each
class. For static analysis, the approach is conceptually similar to dynamic relaxation, proposed by
Otter et al. (1966). The equations of motion are damped to reach a force equilibrium state as quickly
as possible under the applied initial and boundary conditions. Damping is velocity-proportional —
i.e., the magnitude of the damping force is proportional to the velocity of the blocks.
The use of velocity-proportional damping in standard dynamic relaxation involves three main dif-
ficulties.
1. The damping introduces body forces, which are erroneous in “flowing” regions
and may influence the mode of failure in some cases.
2. The optimum proportionality constant depends on the eigenvalues of the ma-
trix, which are unknown unless a complete modal analysis is done. In a linear
problem, this analysis needs almost as much computer effort as the dynamic
relaxation calculation itself. In a nonlinear problem, eigenvalues may be un-
defined.
3. In its standard form, velocity-proportional damping is applied equally to all
nodes — i.e., a single damping constant is chosen for the whole model. In
many cases, a variety of behavior may be observed in different parts of the
model; for example, one region may be failing while another is stable. For
these problems, different amounts of damping are appropriate for different
regions.
In an effort to overcome one or more of these difficulties, alternative forms of damping may be
proposed. In soil and rock, natural damping is hysteretic; if the slope of the unloading curve is
higher than that of the loading curve, energy may be lost. This type of damping can be produced
numerically, but there are at least two difficulties. First, the precise nature of the hysteretic curve is
often unknown for complex loading-unloading paths. This is particularly true for soils, which are
typically tested with sinusoidal stress histories. Cundall (1976) reports that very different results are
obtained when the same energy loss is accounted for by different types of hysteretic loops. Second,
“ratcheting” can occur — i.e., each cycle in the oscillation of a body causes irreversible strain to
be accumulated. This type of damping has been avoided, since it increases path dependence and
makes the results more difficult to interpret.
Two alternative forms of velocity-proportional damping are provided in UDEC. The first is a
numerical servo-mechanism termed adaptive global damping and is described by Cundall (1982).
Adaptive global damping is used to adjust the damping constant automatically. Viscous damping
forces are used, but the viscosity constant is continuously adjusted in such a way that the power
absorbed by damping is a constant proportion to the rate of change of kinetic energy in the system.
The adjustment to the viscosity constant is made by a numerical servo-mechanism that seeks to
keep the following ratio, R, equal to a given ratio (e.g., 0.5).
UDEC Version 4.0


P
R= (1.30)
Ėk
where P is the damping power for a node;

Ėk is the rate of change of nodal kinetic energy; and

represents the summation over all nodes.
This form of damping overcomes difficulty (2) above, and partially overcomes (1) since, as a system
approaches steady state (equilibrium or steady flow), the rate of change of kinetic energy approaches
zero and, consequently, the damping power tends to zero (see Cundall 1982).
In order to overcome all three difficulties, another form of damping is provided in UDEC in which
the damping force on a node is proportional to the magnitude of the unbalanced force. For this
scheme, referred to as local damping, the direction of the damping force is such that energy is
always dissipated. For deformable blocks, the equations of motion (Eq. (1.27)) are replaced by the
following equation, which incorporates local damping:

t
(t+t/2) (t−t/2) (t) (t) (t−t/2)
u̇i = u̇i + Fi − α Fi sgn u̇i (1.31)
mn
where α is a constant (set to 0.8 in UDEC), and mn is the nodal mass. A similar equation is used
in place of Eq. (1.5) to apply local damping to translational and angular velocities of rigid blocks.
This type of damping is equivalent to a local form of adaptive damping. In principle, the three diffi-
culties reported above are addressed: body forces vanish for steady-state conditions; the magnitude
of damping constant is dimensionless and is independent of properties or boundary conditions; and
the amount of damping varies from point to point (Cundall 1987, pp. 134-135).
UDEC calculations using local damping and adaptive global damping are compared by Cundall
(1987); the methods are shown to converge to the same solution. Local damping may be preferred
for analyses involving sudden load changes or progressive failure (such as caving of many blocks),
for which a different amount of damping is required in different regions of the model. Analyses
with local damping are observed to be slightly underdamped in general.
For a dynamic analysis, the damping in the numerical simulation should approximately reproduce
the energy losses in the natural system when subjected to a dynamic loading. As mentioned above,
in soil and rock, natural damping is mainly hysteretic (i.e., independent of frequency). It is difficult
to reproduce this type of damping numerically because of the problem with path dependence, as
described previously. Alternatively, Rayleigh damping is used in UDEC. This method of damping
for dynamic analysis is described in Section 4.
UDEC Version 4.0

1.2.7 Mechanical Timestep Determination: Solution Stability
The solution scheme used for the distinct element method is conditionally stable. A limiting timestep
that satisfies both the stability criterion for calculation of internal block deformation as well as that
for inter-block relative displacement is determined. The timestep required for the stability of block
deformation computations is estimated as
1/2
mi
tn = 2 min (1.32)
ki
where mi is the mass associated with block node i; and

ki is the measure of stiffness of the elements surrounding the node.
The ratio of mass to stiffness is related to the highest eigenfrequency, ωmax , of a linear elastic
system.
The stiffness term, ki , must account for both the stiffness of the intact rock and that of the discon-
tinuities. It is calculated as the sum of the two components:

ki = (kzi + kj i ) (1.33)
The first term on the right-hand side represents the sum of the contributions of the stiffness of all
elements connected to node i, which are estimated as
2
8 4 bmax
kzi = K+ G (1.34)
3 3 hmin
where K and G are the bulk and shear elastic moduli of the block material, respectively;
bmax is the largest zone edge; and
hmin is the minimum height of the triangular element.
The joint stiffness term, kj i , exists only for nodes located on the block boundary and is taken as the
product of the normal or shear joint stiffnesses (whichever is larger) and the sum of the lengths of
the two block edge segments adjacent to node i.
For calculations of inter-block relative displacement, the limiting timestep is calculated, by analogy
to a simple degree-of-freedom system, as
1/2
Mmin
tb = (frac) 2 (1.35)
Kmax
UDEC Version 4.0

where Mmin is the mass of the smallest block in the system; and
Kmax is the maximum contact stiffness.
The term frac is a user-supplied value that accounts for the fact that a single block may be in contact
with several blocks simultaneously. A typical value for frac is 0.1.
The controlling timestep for a distinct element analysis is
t = min(tn , tb ) (1.36)
1.2.8 Mass (Density) Scaling
Even though explicit calculations execute very rapidly per timestep, some way of increasing the
timestep is desirable in order to reduce computer time. One way to do this is by scaling the mass (or
density) of the solid material. It may be noted that the value of inertial density is irrelevant to the
modeling of static systems, provided that gravity forces are correctly preserved. For nearly static
systems (i.e., ones which only evolve slowly with time), inertial densities may be increased until
they begin to become appreciable compared to other forces in the system. The system response
will not be significantly modified if inertial forces are low. The reason for wanting to increase the
density is that the critical timestep, t, may also be increased because it is determined by density,
ρ:
√
t ∝ ρ (1.37)
This procedure is called density scaling. Density scaling is only effective at improving convergence
if the model is nonuniform (i.e., if the natural timesteps differ for different parts of the model).
Changing the density in a uniform model does nothing to improve convergence.
Density scaling may be selected by the user via the MSCALE command (see Section 1.3 in the
Command Reference). Mass scaling is turned on automatically when local damping is specified
(via DAMP local) or when adaptive global damping is specified (via DAMP auto). For most problems,
a scale factor based on the average block mass or zone mass in the model provides the most rapid
convergence.
UDEC Version 4.0

1.2.9 Boundary Conditions
Either stress (load) or displacement (velocity) may be applied at the boundary of a UDEC model.
The condition is applied to the centroid of blocks along the boundary for a rigid block model. For
deformable blocks, displacements are specified in terms of prescribed velocities at given gridpoints;
Eq. (1.27) is not invoked at those gridpoints. At a stress boundary, forces are derived as follows:
Fi = σijb nj s (1.38)
where nj is the outward normal vector of the boundary segment; and

s is the length of the boundary segment over which the stress, σijb , acts.
The force, Fi , is added into the force sum, in Eq. (1.23), for the appropriate gridpoint.
1.2.10 Boundary-Element Representation of the Far Field
When performing static analyses, the problem of defining boundary conditions for a finite numerical
model of an unbounded medium can be adequately handled by coupling the block assembly to a
boundary-element representation of the far field. Because nonlinear behavior is usually confined
to the vicinity of the structure or excavation under study, the assumption of a linear elastic far
field is justified. A hybrid rigid block-boundary element model was developed by Lorig (1984)
for the analysis of underground excavations in rock. A half-plane formulation for the boundary
element region was used by Lemos (1983) in a coupled distinct element-boundary element model
appropriate for the study of foundations or shallow excavations. A similar scheme is implemented
in UDEC. The boundary element formulation follows the work of Brady and Wassyng (1981).
The boundary-element region is represented by a stiffness matrix, K, which relates the forces and
displacements at the interface of the two domains. Either an infinite plane or a half-plane solution
can be used. The elastic moduli of the far-field domain should reflect the deformability of the
jointed rock system. At every timestep, the motion of the blocks defines the displacements at the
interface. The boundary-element domain provides elastic reaction forces given by
F = −K u (1.39)
Dynamic analysis, discussed in Section 4, usually starts from some in-situ condition. Normally, a
simple uniform stress field is assumed. A more realistic stress distribution can be simulated with
a hybrid distinct element-boundary element model. Then, before the dynamic input is applied, the
boundary-element boundaries can be replaced by non-reflecting boundaries, provided the boundary-
element reaction forces are maintained throughout the dynamic loading phase.
UDEC Version 4.0

1.2.11 Data Structure
In UDEC, the discrete-element information is all stored within a linked-list data structure that
corresponds to the topological structure of the physical system. Each physical entity (such as a
block, corner or contact) is represented by a data element that is linked by pointers (or indices)
to the data structure from a main storage array (see Section 4 in the FISH volume). A guiding
principle in the design of the data structure is to reduce computer time at the expense of using more
memory to store data. The linked-list data structure is not well-suited to supercomputers (such as a
CRAY) that derive their speed from vector processing, because data are not organized sequentially
in memory.
The topological nature of the data structure permits a direct translation of rock joint structural data
into UDEC. An automatic joint generation model exists in UDEC to generate joint patterns that
resemble those in natural rock (see Section 3.2.2 in the User’s Guide). The joint geometry model
is described in statistical terms and can be used to create blocky structures consisting of arbitrarily
shaped (including concave) polygons. The physical characteristics describing block size and shape
and the contact locations between blocks are translated directly into data elements and linked to the
data structure.
For deformable blocks, each block is independently discretized into a mesh of triangular elements.
The automatic mesh generator performs the internal discretization for arbitrarily shaped blocks and
stores the zone and gridpoint information in the data structure.
Other data for constitutive models, material properties and initial and boundary conditions are also
linked to the data structure during model generation. The structure facilitates the assignment of
different constitutive models, properties and conditions to user-selected locations in the model. For
example, different joint sets can be assigned unique material properties. See Section 4 in the FISH
volume for more details on UDEC ’s data structure. The data structure can easily be accessed via
FISH. See Section 1 in the FISH volume for more information and examples.
UDEC Version 4.0

1.3 References
Bardet, J.-P., and R. F. Scott. “Seismic Stability of Fractured Rock Masses with the Distinct Element
Method,” in Research and Engineering Applications in Rock Masses (Proceedings of the 26th
U.S. Symposium on Rock Mechanics), Vol. 2, pp. 139-150. Boston: A. A. Balkema, 1985.
Barton, N. “Modelling Rock Joint Behavior from In-Situ Block Tests: Implications for Nu-
clear Waste Repository Design (Technical Report), Office of Nuclear Waste Isolation, ONWI-308,
September, 1982.
Barton, N., S. Bandis and K. Bakhtar. “Strength, Deformation and Conductivity Coupling of Rock
Joints,” Int. J. Rock Mech. Min. Sci., 22, 121-140 (1985).
Brady, B. H. G., and A. Wassyng. “A Coupled Finite Element-Boundary Element Method of Stress
Analysis,” Int. J. Rock Mech. Min. Sci., 18, 475-485 (1981).
Butkovich, T. R., O. R. Walton and F. E. Heuze. “Insights in Cratering Phenomenology Provided
by Discrete Element Modeling,” in Key Questions in Rock Mechanics: Proceedings of the 29th
U.S. Symposium (University of Minnesota, June 1988), pp. 359-368. Rotterdam: A. A. Balkema,
1988.
Cundall, P. A. “A Computer Model for Simulating Progressive Large Scale Movements in Blocky
Rock Systems,” in Proceedings of the Symposium of the International Society of Rock Mechanics
(Nancy, France, 1971), Vol. 1, Paper No. II-8, 1971.
Cundall, P. A. “Rational Design of Tunnel Supports: A Computer Model for Rock Mass Behaviour
Using Interactive Graphics for the Input and Output of Geometrical Data,” U. S. Army Corps of
Engineers, Missouri River Division, Technical Report MRD-2-74; NTIS Report No. AD/A-001
602, 1974.
Cundall, P. A. “Explicit Finite Difference Methods in Geomechanics,” in Numerical Methods
in Engineering (Proceedings of the EF Conference on Numerical Methods in Geomechanics
(Blacksburg, Virginia, June, 1976)), Vol. 1, pp. 132-150, 1976.
Cundall, P. A. “UDEC — A Generalized Distinct Element Program for Modelling Jointed Rock,”
Peter Cundall Associates, Report PCAR-1-80; European Research Office, U.S. Army, Contract
DAJA37-79-C-0548, March, 1980.
Cundall, P. A. “Adaptive Density-Scaling for Time-Explicit Calculations,” in Proceedings of the
4th International Conference on Numerical Methods in Geomechanics (Edmonton, 1982), pp.
23-26, 1982.
Cundall, P. A. “Formulation of a Three-Dimensional Distinct Element Model — Part I: A Scheme
to Detect and Represent Contacts in a System Composed of Many Polyhedral Blocks,” Int. J. Rock
Mech., Min. Sci. & Geomech. Abstr., 25, 107-116 (1988).
Cundall, P. A., and R. D. Hart. “Development of Generalized 2-D and 3-D Distinct Element
Programs for Modeling Jointed Rock,” Itasca Consulting Group; U.S. Army Corps of Engineers,
Misc. Paper SL-85-1, 1985.
UDEC Version 4.0

Cundall, P. A., and R. D. Hart. “Numerical Modeling of Discontinua,” Engr. Comp., 9(2), 101-113
(1992).
Cundall, P. A., and O. D. L. Strack. “A Discrete Numerical Model for Granular Assemblies,”
Géotechnique, 29, 47-65 (1979).
Cundall, P. A., and O. D. L. Strack. “Modeling of Microscopic Mechanisms in Granular Material,”
in Mechanics of Granular Materials: New Models and Constitutive Relations, pp. 137-149.
Amsterdam: Elsevier Scientific Publications, B.V., 1983.
Cundall, P. A., J. Marti, P. J. Beresford, N. C. Last and M. I. Asgian. “Computer Modeling of Jointed
Rock Masses,” U.S. Army Engineer Waterways Experiment Station, Vicksburg, Mississippi, Tech.
Report N-78-4, August, 1978.
Cundall, P. A., and J. Marti. “Some New Developments in Discrete Numerical Methods for Dynamic
Modelling of Jointed Rock Masses,” in Proceedings of the Rapid Excavation and Tunnelling
Conference, 1979 (Atlanta, June 1979), Vol. 2, pp. 1466-1477. Rotterdam: A. A. Balkema, 1979.
Dasgupta, B., R. Dham and L. J. Lorig. “Three-Dimensional Discontinuum Analysis of the Under-
ground Power House for Sardar Sarovar Project, India,” in Proceedings of the Eighth International
Congress on Rock Mechanics (Tokyo, September 1995), Vol. 1, pp. 551-554. T. Fujii, Ed. Rot-
terdam: A. A. Balkema, 1995.
Goodman, R. E., and G.-H. Shi. Block Theory and Its Application to Rock Engineering. New
Jersey: Prentice Hall, 1985.
Hahn, J. K. “Realistic Animation of Rigid Bodies,” Computer Graphics, 24(4), 299-308 (1988).
Hart, R. D., J. Lemos and P. Cundall. “Block Motion Research: Analysis with the Distinct Element
Method,” Itasca Consulting Group/Agbabian Associates, DNA-TR-88-34-V2, December, 1987.
Hart, R., P. Cundall and J. Lemos. “Formulation of a Three-Dimensional Distinct Element Model
— Part II: Mechanical Calculations for Motion and Interaction of a System Composed of Many
Polyhedral Blocks,” Int. J. Rock Mech., Min. Sci. & Geomech. Abstr., 25, 117-126 (1988).
Hart, R. D., J. Lemos, L. J. Lorig and P. A. Cundall. “Tunnel Prediction Using Distinct Elements:
Volume II — Computer Code Modification and Verification,” DNA, Technical Report DNA-TR-
90-56-V2, December, 1990.
Heuzé, F. E., O. R. Walton, D. M. Maddix, R. J. Shaffer and T. R. Butkovich. “Analysis of Explosions
in Hard Rock: The Power of Discrete Element Modeling,” Lawrence Livermore Laboratory, Preprint
UCRL-JC-103498, March, 1990.
Hocking, G., G. G. W. Mustoe and J. R. Williams. CICE Discrete Element Code — Theoretical
Manual. Lakewood, Colorado: Applied Mechanics Inc., 1985.
Itasca Consulting Group, Inc. Particle Flow Code in 2 Dimensions and Particle Flow Code in 3
Dimensions, Version 1.1. Minneapolis: ICG, 1995.
UDEC Version 4.0

Lemos, J. V. “A Hybrid Distinct Element-Boundary Element Computational Model for the Half-
Plane.” M.S. Thesis, University of Minnesota, 1983.
Lemos, J. V., R. D. Hart and P. A. Cundall. “A Generalized Distinct Element Program for Modeling
Jointed Rock Mass (A Keynote Lecture),” in Proceedings of the International Symposium on
Fundamentals of Rock Joints (Bjorkliden, September, 1985), pp. 335-343. Luleå, Sweden:
Centek Publishers, 1985.
Lemos, J. V. “Assessment of the Ultimate Load of a Masonry Arch Using Discrete Elements,”
in 3rd International Symposium on Computer Methods in Structural Masonry (Lisbon, April,
1995), 1995.
Lorig, L. J. “A Hybrid Computational Model for Excavation and Support Design in Jointed Media.”
Ph.D. Thesis, University of Minnesota, 1984.
Lorig, L. J., and P. A. Cundall. “Modeling of Reinforced Concrete Using the Distinct Element
Method,” in Fracture of Concrete and Rock, pp. 459-471. Bethel, Conn.: SEM, 1987.
Lorig, L. J., W. Gibson, J. Alvial and J. Cuevas. “Gravity Flow Simulations with the Particle Flow
Code (PFC),” ISRM News J., 3, 18-24 (1995).
Marti, J., and P. A. Cundall. “Mixed Discretization Procedure for Accurate Solution of Plasticity
Problems,” Int. J. Num. Methods & Analy. Methods in Geomech., 6, 129-139 (1982).
Nagtegaal, J. C., D. M. Parks and J. R. Rice. “On Numerically Accurate Finite Element Solutions
in the Fully Plastic Range,” Comp. Meth. Appl. Mech., 4, 153-177 (1974).
Otter, J. R. H., A. C. Cassell and R. E. Hobbs. “Dynamic Relaxation (Paper No. 6986),” Proc.
Instn. Civ. Engrs., 35, 633-656 (1966).
Papadopoulos, J. M. “Incremental Deformation of an Irregular Assembly of Particles in Compres-
sive Contact.” Ph.D. Thesis, M.I.T, Department of Mechanical Engineering, 1986.
Plesha, M. E., and E. C. Aifantis. “On the Modeling of Rocks with Microstructure,” in Rock
Mechanics — Theory-Experiment-Practice (Proceedings of the 24th U.S. Symposium on Rock
Mechanics, Texas A&M University, 1983), pp. 27-35. New York: Association of Engineering
Geologists, 1983.
Shi, G.-H. “Discontinuous Deformation Analysis — A New Numerical Model for the Statics and
Dynamics of Block Systems,” Lawrence Berkeley Laboratory, Report to DOE OWTD, Contract
AC03-76SF0098, September 1988; also Ph.D. Thesis, University of California, Berkeley, August,
1989.
Tinucci, J. P., and D. S. G. Hanson. “Assessment of Seismic Fault-Slip Potential at the Strathcona
Mine,” in Rock Mechanics Contributions and Challenges, pp. 753-760. Rotterdam: A. A.
Balkema, 1990.
Tinucci, J. P., and J. Israelsson. “Site Characterization and Validation — Excavation Stress Effects
Around the Validation Drift,” SKB, Stripa Project Technical Report 91-20, August, 1991.
UDEC Version 4.0

Walton, O. R. “Particle Dynamic Modeling of Geological Materials,” Lawrence Livermore National

Laboratory, Report UCRL-52915, 1980.
Walton, O. R., R. L. Braun, R. G. Mallon and D. M. Cervelli. “Particle-Dynamics Calculations
of Gravity Flows of Inelastic, Frictional Spheres,” in Micromechanics of Granular Material, pp.
153-161. Amsterdam: Elsevier Science Publishers, 1988.
Warburton, P. M. “Vector Stability Analysis of an Arbitrary Polyhedral Rock Block with Any
Number of Free Faces,” Int. J. Rock Mech. Min. Sci. & Geomech. Abstr., 18, 415-427 (1981).
Williams, J. R., and G. G. W. Mustoe. “Modal Methods for the Analysis of Discrete Systems,”
Computers & Geotechnics, 4, 1-19 (1987).
UDEC Version 4.0

UDEC Version 4.0

1 Background - The 2D Distinct Element Method

Uploaded by

Copyright:

Available Formats

1 Background - The 2D Distinct Element Method

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

1 Background - The 2D Distinct Element Method

Uploaded by

Copyright:

Available Formats

BACKGROUND — THE 2D DISTINCT ELEMENT METHOD 1-1

1 BACKGROUND — THE 2D DISTINCT ELEMENT METHOD

1.1 Aspects of Modeling a Discontinuous System

UDEC Version 4.0

1.1.1 Computer Programs for Modeling Discontinuous Systems

UDEC Version 4.0

UDEC Version 4.0

Limit equilibrium; limit analyses

Distinct element method

Forces Forces &

Discrete element methods

UDEC Version 4.0

1.1.2 History of the Distinct Element Method

UDEC Version 4.0

04 UDEC Vers. 4.0

03 3DEC Vers. 3.0 PFC3D Vers. 3.0

02 PFC2D Vers. 3.0

99 UDEC Vers. 3.1 PFC2D, PFC3D Vers. 2.0

94 PFC2D, PFC3D Vers. 1.0

93 UDEC Vers. 2.0 3DEC Vers. 1.5

92 UDEC Vers. 1.8 3DEC Vers. 1.4

91 UDEC Vers. 1.7 3DEC Vers. 1.3

1990 UDEC Vers. 1.6 3DEC Vers. 1.2

89 UDEC Vers. 1.5 3DEC Vers. 1.1

86 UDEC Vers. 1.2

83 UDEC Vers. 1.0

Figure 1.2 Chronology of the distinct element method

UDEC Version 4.0

1.2 Numerical Formulation

UDEC Version 4.0

UDEC Version 4.0

1.2.2 Equations of Motion

Substituting Eq. (1.2) in Eq. (1.1) and rearranging yields

With velocities stored at the half-timestep point, it is possible to express displacement as

u(t+t) = u(t) + u̇(t+t/2) t (1.4)

UDEC Version 4.0

where θ̇ = angular velocity of block about centroid;

UDEC Version 4.0

(t+t) (t) (t+t/2)

where θ = rotation of block about centroid; and

1.2.3 Conservation of Momentum and Energy in the Distinct Element Formulation

UDEC Version 4.0

Combining the equations and integrating:

Using the identity v̇ = v dv/ds,

assuming m is constant. Hence,

UDEC Version 4.0

1.2.4 Rock Joint Representation

1.2.4.1 Contact Detection and Identification

UDEC Version 4.0

Figure 1.5 Contacts between two rigid blocks

(a) rounding of corners using constant rounding length, d

(b) rounding of corners using constant radius, r, demonstrating

UDEC Version 4.0

UDEC Version 4.0

(a) detail of rounded corner-to-edge contact (rounding length exaggerated)

(b) smooth interaction of corner-to-corner contact

Figure 1.7 Definition of contact normal in UDEC

UDEC Version 4.0

σije = λ v δij + 2µ ij (1.29)