Penalty Function Method For Combined Ÿnite-Discrete Element Systems Comprising Large Number of Separate Bodies

INTERNATIONAL JOURNAL FOR NUMERICAL METHODS IN ENGINEERING
Int. J. Numer. Meth. Engng 2000; 49:13771396
Penalty function method for combined nitediscrete

element systems comprising large number of separate bodies
A. Munjiza; and K. R. F. Andrews

Department of Engineering; Queen Mary and Westeld College; University of London; Mile End Road;
London E1 4NS; U.K.
SUMMARY
Large-scale discrete element simulations, the combined nitediscrete element method, DDA as well as a
whole range of related methods, involve contact of a large number of separate bodies. In the context of the
combined nitediscrete element method, each of these bodies is represented by a single discrete element
which is then discretized into nite elements. The combined nitediscrete element method thus also in-
volves algorithms dealing with fracture and fragmentation of individual discrete elements which result in ever
changing topology and size of the problem. All these require complex algorithmic procedures and signicant
computational resources, especially in terms of CPU time. In this context, it is also necessary to have an
ecient and robust algorithm for handling mechanical contact. In this work, a contact algorithm based on
the penalty function method and incorporating contact kinematics preserving energy balance, is proposed and
implemented into the combined nite element code. Copyright ? 2000 John Wiley & Sons, Ltd.
KEY WORDS: nite elements; discrete elements; contact; interaction
1. INTRODUCTION
The combined nitediscrete element method as a computational tool is aimed at problems involv-
ing transient dynamics of systems comprising a large number of deformable bodies that interact
with each other and that may, in general, fracture and fragment thus increasing the total number of
bodies even further [13]. Each individual body is of general shape and size and is modelled by
a single discrete element. Each discrete element is in turn discretized into nite elements in order
to analyse deformability, fracture and fragmentation. A typical combined nitediscrete element
system comprises a large number (a few thousand to a few million [4]) of separate interacting
solids each associated with separate nite element meshes. In this context, one of the key issues
in the development of the combined nite=discrete element method is the treatment of contact
impact, i.e. the enforcement of the constraint that no point in the space is occupied by more than
one body at the same time. From an algorithmic point of view, there are two aspects to contact
Correspondence to: A. Munjiza, Department of Engineering, Queen Mary and Westeld College, Mile End Road, London
E1 4NS, U.K.
E-mail: a.munjiza@qmw.ac.uk
Received 5 January 1999

Copyright ? 2000 John Wiley & Sons, Ltd. Revised 6 January 2000
1378 A. MUNJIZA AND K. R. F. ANDREWS
in the combined nitediscrete element methodcontact detection and contact interaction. Contact
detection is aimed at detecting couples of discrete elements close to each otheran ecient and
robust contact detection algorithm should detect all contacts at low CPU and memory costs. Con-
tact detection algorithms proposed in recent years [310] are characterized by contact detection
times ranging from N ln N to cN , where cN is the total number of discrete elements and c is a
constant. Once elements in contact are detected, a contact interaction algorithm is employed to
evaluate contact forces.
Problems of contact interaction in the context of the combined nitediscrete element method
have special importance since in the combined nitediscrete element method, the problem of
contact interaction and handling of contacts also denes the constitutive behaviour of the system
because of the presence of large numbers of separate bodies. Thus, the algorithms employed
traditionally paid special attention to contact kinematics in terms of a realistic distribution of
contact forces, energy balance and robustness. One of the early contact interaction algorithms
taking into account energy balance considerations is the contact interaction algorithm based on the
so-called concept of the contact element [1]. In this, the penalty parameters are given physical
meaning through the concept of a contact layer which is in a way a crude approximation of surface
roughness. This early algorithm was originally developed to handle 2D problems and was dicult
to implement in 3D. In addition where overlap of discrete elements in contact exceeds contact
layer, the energy balance is not preserved which is also the case if new surfaces are created due
to fracture process. The contact forces resulting from the algorithm are concentrated which greatly
in uences stress and strain elds close to the boundary and may considerably aect for instance
results of fracture and fragmentation analysis [2].
In this work, a distributed contact force algorithm which is relatively easy to implement and
which preserves energy balance has been developed and implemented into the combined nite
discrete element method. The algorithm assumes discretization of individual discrete elements into
nite elements, thus imposing no additional database requirements on handling the geometry of
individual discrete elements. It also yields a realistic distribution of contact forces over nite
contact areas resulting from overlap of discrete elements that are in contact. Thus, numerical
distortion of local strain elds close to the boundary due to contact is much reduced (an important
aspect when for instance fracture of brittle materials is analysed). The algorithm is based on a
so-called potential function for contact forces which, by denition, results in contact kinematics
that preserves energy balance regardless of the geometry and number of contacts.
The remainder of the paper deals with the basic assumptions and theoretical aspects of the
proposed contact algorithm together with a detailed description of implementation into the com-
bined nitediscrete element code. Finally, numerical results involving the transient motion of a
large number of discrete elements together with contact sequences and energy balances are given.
These examples are specially selected to illustrate the importance of contact in the combined
nitediscrete element methodthe problem of contact interaction and handling of contacts also
denes the constitutive behaviour of the system because of the presence of large numbers of
separate bodies.
2. POTENTIAL FIELD-BASED PENALTY FUNCTION METHOD
The penalty function method in its classical form assumes that two bodies in contact penetrate
each other and this penetration results in a contact force. The standard contact functional for the
Copyright ? 2000 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2000; 49:13771396
PENALTY FUNCTION METHOD FOR NUMBER OF SEPARATE BODIES 1379
Figure 1. Contact force due to an innitesimal overlap around points Pc and Pt .
penalty function method takes the form

Z
1
Uc = p(rt rc )T (rt rc ) d (1)
2
where p is the penalty term while rt and rc are position vectors of the points on the overlapping
boundaries of the target and contactor bodies, respectively. In the limit for innite penalty terms
no penetration would occur. However, in practice, large penalty terms are coupled with integration
problems in the temporal domain and in practical applications the penalty function method is
usually coupled with overlaps between bodies in contact.
When implemented into actual codes dealing with contact, the penalty function method in general
deals with either concentrated or distributed contact force. The concentrated contact force approach
usually assumes nodal contact forces being a function of the penetration of individual contactor
nodes into the target, while the distributed contact force is in general evaluated from the shape
and size of overlap between the contactor and target.
In this work, the distributed contact force is adopted for two discrete elements in contact one
of which is denoted as the contactor and the other as the target. When in contact the contactor
and target overlap each other over area S, bounded by boundary (Figure 1).
It is assumed that penetration of any elemental area dA of contactor into the target results in
an innitesimal contact force given by
df = [grad c (Pc ) grad t (Pt )] dA (2)
where c and t are potential functions for the contactor and target elements, respectively, while
df is the innitesimal contact force due to innitesimal overlap dA dened by overlapping points
Pc belonging to the contactor and Pt belonging to the target.
Equation (2) can also be written as
df = dfc dft (3)
where
dfc = grad t (Pt ) dAc ; dAc = dA
(4)
dft = grad c (Pc ) dAt ; dAt = dA
In other words, the contact as described by (2) can be viewed rstly as the elemental area of the
contactor penetrating the target and then the elemental area of the target penetrating the contactor.
Figure 2. Point Pc penetrating target at point A and contact being released at point B.
Thus, for each of the discrete elements in contact, the contact force is calculated as the gradient
of the corresponding potential function. The eld of contact forces is therefore a conservative eld
for both the target penetrating the contactor and the contactor penetrating the target. Thus, if point
Pc of the contactor discrete element penetrates the target through any path dened by end points A
and B, the total work of the contact force due to the potential function ft does not depend on the
path but on the end points A and B only and is given as t (A) t (B). If both points A and B are
on the boundary of the target discrete element, a contactcontact release situation arises, i.e. point
Pc comes into contact with the target at point A of the target and the contact is released through
the point B of the target (Figure 2). Preservation of energy balance requires that the total energy
of the system before and after the contact are the same, i.e. no work is done by the contact force,
which is equivalent to saying that for any points A and B on the boundary of the target discrete
element
t (A) t (B) = 0 i:e: t (A) = t (B) (5)
The same is true for the contactor discrete element, i.e. for any points A and B from the boundary
of the contactor discrete element
c (A) c (B) = 0 i:e: c (A) = c (B) (6)
Thus, provided that the potentials on the boundary of both contactor and target discrete elements
are constant, the contact force as given by (2) preserves the energy balance regardless of the
geometry or the shape of contactor and target discrete elements, size of the penalty term or size
of penetration (overlap).
The total contact force is obtained by integration of (2) over the overlapping area S
Z
fc = [grad c grad t ] dA (7)
S=t c
which can also be written as an integral over the boundary of the overlapping area
Z
fc = n( fc ft ) d (8)
t c
where n is the outward unit normal to the boundary of the overlapping area.
Proof. For an overlapping areas consisting of convex sub-areas, it is enough to prove that (8)
holds for any convex sub-area. It is therefore assumed that a convex sub-area S (Figure 3), has
Figure 3. Convex overlapping area S.
no holes or bubbles and that its boundary is piecewise smooth curve that intersects any line
parallel to co-ordinate axis x in exactly two points producing lines 1 and 2 such that
1 : y = f1 (x); x [a; b]
(9)
2 : y = f2 (x); x [a; b]
and
f2 (x)f1 (x); x [a; b] (10)
From Figure 3 it follows that
I Z Z
(x; y)ny d = (x; f2 (x))ny d + (x; f1 (x))ny d (11)
2 1
where
n = (nx ; ny ) (12)
is the unit normal to the boundary . Substitution of
(x; f2 (x))ny d = (x; f2 (x)) dx
(13)
(x; f1 (x))ny d = (x; f1 (x)) dx
into (11) yields
I Z b Z b
(x; y)ny d = (x; f2 (x)) dx (x; f1 (x)) dx
a a
Z "Z # Z
b f2 (x)
@ @
= dy dx = dx dy (14)
a f1 (x) @y S @y
In a similar way for a convex area S with no holes or bubbles and for boundary of S being
piecewise smooth curve that intersects any line parallel to co-ordinate axis y in exactly two points
producing lines 1 and 2 such that
1 : x = f1 (y); y [c; d]
(15)
2 : x = f2 (y); y [c; d]
and
f2 (y)f1 (y) (16)
it follows that
I Z
@
(x; y)nx d dx dy (17)
S @x
Equations (14) and (17) when combined yield
I I I
(x; y)nx d ; (x; y)ny d = (x; y)n d
Z Z Z
@ @
= dx dy; dx dy = grad dS (18)
S @x S @y S
3. IMPLEMENTATION INTO THE COMBINED FINITEDISCRETE ELEMENT METHOD
In the combined nitediscrete element method individual discrete elements are discretized into
nite elements and each discrete element is represented by its nite elements
c = c1 c2 ci cn
(19)
t = t1 t2 tj tm
where c and t are contactor and target discrete elements, respectively, while m and n are the total
number of nite elements into which the contactor and target discrete elements are discretized. In
this context, the potentials c and t can be represented by a sum of potentials associated with
individual nite elements
c = c1 + c2 + ci + cn
(20)
t = t1 + t2 + ti + tm
Integration over overlapping area may therefore be represented by summation over nite elements
Z
P
n Pm
fc = [grad ci grad tj ] dA (21)
i=1 j=1 ci cj
By replacing integration over nite elements by equivalent integration over nite element bound-
aries (8), the following equation for the contact force is obtained:
Z
Pn Pm
fc = n ci tj (ci tj ) d (22)
i=1 j=1 ci tj
Figure 4. Subdivision of contactor and target into nite elements and integration over
edges of overlapping nite elements.
where integration over nite element boundaries may be written as the summation of integrals
over the edges of nite elements. In other words the contact force between overlapping discrete
elements is calculated by summation over the edges of corresponding nite elements that overlap
(Figure 4).
4. DISCRETIZATION OF POTENTIAL FIELD AND ASSOCIATED

COMPUTATIONAL ASPECTS
Combined nitediscrete element problems involve a large number of separate bodies that are free
to move and interact with each other. Average problems may comprise a few thousand or even
a few millions of separate bodies [4]. In addition, each body is represented by a single discrete
element, which in turn is discretized into one or more nite elements. Thus, evaluation of contact
forces at each time step may involve a large number of contacting couples of discrete elements.
Because discrete elements are discretized into nite elements, each contacting couple of discrete
elements is in fact represented by a whole set of couples of nite elements (Figure 4), making
the total number of contacting couples of nite elements even larger.
In this context, the summation as given by (22) usually involves a large number of contacting
couples of nite elements and the total CPU time and overall eciency of the contact algorithm
critically depend on implementation of nite element to nite element contact. This is illustrated by
some implementation details for overlapping nite elements of triangular shape, which are given
below.
It is assumed that the edges of a triangular nite element are straight and that the current
geometry of a triangular nite element is uniquely dened by co-ordinates of its nodes as shown
in Figure 5. As explained earlier, the potential should be constant on the boundary of a discrete
element. This constraint is unconditionally met if the potential is dened in such a way that
it is constant on the boundaries of each nite element. Thus for a triangular nite element it is
assumed that is given by
(
0 for outside points
(P) = (23)
min{3A1 =A; 3A2 =A; 3A3 =A} for inside points
Figure 5. Potential at any point P of triangle-shaped nite element.
Figure 6. Contact of contactor and target triangles and contact of an edge

of the contactor triangle with the target triangle.
where Ai ; i = 1; 2; 3, is the area of corresponding sub-triangles as shown in Figure 5. The potential

equals 1 at the centre of triangle, zero at edges of the triangle and zero for all points outside
the triangle.
According to (23) the problem of interaction between two triangles can be reduced to interactions
between contactor triangle and the edges of target triangle coupled with interactions between target
triangle and the edges of the contactor triangle as shown in Figure 6.
Figure 7. Distribution of contact force between the target triangle and an edge of the contactor triangle.
In Figure 7, contact between the edge AB of the contactor triangle and the target triangle is
shown. In order to minimize the number of necessary operations and therefore CPU time to process
this contact, a local co-ordinate system given by local axes u and v is introduced and transformation
of the nodal co-ordinates of the target triangle into local co-ordinate system is performed
pi = ([ri rA ] u; [ri rA ] v) (24)
Using the local nodal co-ordinates for the target triangle, characteristic intersection points between
the edge AB and the target triangle are obtained together with the corresponding values of the
potential function. Potential for each intersection point is calculated by interpolation between the
central node (node 3, where the potential is equal to 1) and the corresponding edge node (nodes
0; 1 or 2 where potential is equal to zero) for each intersection point.
The total contact force exerted by the target triangle onto the edge AB is given by the area of
the potential as shown in Figure 7, i.e.
Z L
1
fc; AB = u p(v) dv (25)
u2 0
where p is the penalty term and the term u2 comes from the fact that vectors u and v are not unit
vectors. This is computationally convenient since the evaluation of integral (25) does not involve
a computationally more expensive square root. In addition, in between the intersection points the
potential is given by straight lines, which reduces the integration to an area calculation as shown
Figure 8. Equivalent nodal forces.
in Figure 7. The contact force obtained is represented by equivalent nodal forces at points A and
B together with corresponding nodal forces at nodes of the target triangle as shown in Figure 8.
The whole process is repeated for the remaining edges of the contactor triangle and in this way
contact forces due to penetration of the contactor triangle into the target triangle are obtained from
the potential t as dened for the target triangle. In order to take into account the contact force
due to the potential c associated with the contactor triangle, contact of each edge of the target
triangle to the contactor triangle is analysed and corresponding nodal forces for both contactor
and target triangle are updated. The potential associated with the contactor triangle is dened in
exactly the same way as the potential associated with the target triangle and evaluation of nodal
forces proceeds using the same procedures described above.
5. NUMERICAL EXAMPLES
The contact interaction algorithm described in the previous sections has been implemented into
the combined nitediscrete element code. The code includes algorithmic procedures for transient
dynamics of fracturing solids consisting of a large number of fragments of general shape. Each
fragment is represented by a single discrete element discretized into nite elements. The code
therefore includes algorithmic procedures for deformability of individual discrete elements, fracture
and fragmentation algorithms, contact detection between moving fragments, contact interaction
between fragments and a whole set of application specic algorithms.
Figure 9. Beam-shaped heap of rigid fragments.
The set of numerical examples presented in this section has been solved using the code with the
aim of testing the proposed contact interaction algorithm. In this context, all the problems presented
involve transient motion of a large number of rigid fragments so that no energy dissipation occurs.
At any time the total energy consists of the kinetic energy of moving fragments and potential
energy due to the overlapping fragments that are in contact.
Four heaps of the same shape and size were considered (Figure 10). The heap A is made of a
smaller number of relatively large fragments, while heaps BD comprise larger number of smaller
fragments.
The transient motion of the heap A is given in Figure 11. Initially all 80 of the heap fragments
are at rest. As the projectile approaches the heap, contact between the projectile and fragments
produces acceleration of the heap fragments and deceleration of the projectile. Also interaction
between the heap fragments results in fragments further from the point of impact acquiring velocity
and moving apart from each otherthus, the packing of the heap becomes looser.
The actual analysis did not make use of the initial symmetry of the problem. The problem as
a whole was analysed and motion of all 80 fragments together with interaction among fragments
was traced. Contact forces were evaluated at each time step while a central dierence explicit
time-marching scheme was employed to calculate new velocity elds and new nodal co-ordinates.
It is worth noting that after a considerable motion of individual fragments the symmetry of the
system is still preserved.
The transient motion of the heap B due to the impact of a rigid projectile is shown in Figure 12.
Again, the initial impact induces signicant disturbance close to the middle section of the heap.
Through the interaction of individual rigid fragments this disturbance is spread toward the edges
of the heap. Additional constraint of small support-like triangles being xed to the ground restricts
the fragments closest to these supports.
The problem consisting of 320 rigid fragments is characterised by initial symmetry and after the
impact and considerable motion of the fragments, the symmetry is preserved. This is without any
symmetry being taken into account during analysis and the problem being analysed as a whole,
i.e. by considering the simultaneous transient motion of each of the 320 interacting fragments.
The motion sequence for heap C is shown in Figure 13. Again, as the projectile hits the
centre of the heap, fragments close to the point of impact are accelerated. However, this time the
total number of fragments is much larger (1280) and each fragment is therefore much smaller
in comparison to the size of projectile. The projectile therefore becomes submerged among the
fragments of the heap.
Figure 10. Beam-shaped heaps of fragments: heap A80 fragments; heap B320 fragments;
heap C1280 fragments; and heap D5120 fragments.
The later stages of transient motion clearly indicate separation of individual fragments and
transition toward the stage where no two fragments are in contact (Figure 14). The reason for this
is that the heap is not conned in space and the motion of the fragments is almost free except for
the interaction with each other.
The initial symmetry of the system is again preserved despite the fact that symmetry was not
taken into account during the analysis, i.e. the problem was analysed as a whole and the motion
and interaction of all 1280 fragments was traced at each time step of the explicit time integration
scheme.
Heap D consists of the smallest fragments. The size and the shape of the heap is the same as
heaps AC. Thus the total number of closely packed fragments is much largerthe heap consists
Figure 11. Predicted transient motion sequence for heap A.
of 5120 fragments initially packed in such a way that they touch but no two fragments overlap
each other and no contact force is generated.
First contact occurs after the projectile makes contact with fragments close to the centre of the
heap. The projectile penetrates the heap and fragments of the heap are initially extracted by the
Figure 12. Predicted transient motion sequence for heap B.
Figure 13. Predicted transient motion sequence for heap C.
projectile. This results in fragments close to the point of impact moving upwards (Figure 15).
This initial contact is followed by contacts between the heap fragments. Thus, the fragments move
away from each other reducing the average packing density of the heap.
This motion results in the shape of the heap being gradually changed until the projectile nally
passes through the heap (Figure 16).At this stage the shape of the heap has changed signi-
cantly and individual fragments have acquired signicant velocity, thus, the remaining contacts are
released and eventually the projectile and fragments move at constant velocity.
This time the initial symmetry of the system is in general preserved although localized loss
of symmetry is observed. The reasons for this are outside the scope of this paper and can be
Figure 14. Second part of the predicted transient motion sequence for heap C.
found elsewhere [1]. However, it is worth noting that the total number of individual fragments
is relatively large and local bifurcation type corner to corner contact that is sensitive to rounding
errors may occur.
The motion sequences gradually changes from the motion sequence for heap A to the motion se-
quence for heap D and, in some sense, the motion sequence of heap A represents an approximation
of the motion sequence of the heap D, while the motion sequence of heap D by similar reasoning
Figure 15. First part of the predicted motion sequence for heap D.
approximates the motion sequence of a heap consisting of an innite number of innitesimally

small fragments.
In all four heaps, the system starts at a state where no two fragments are in contact and converges
toward the state where all fragments move apart from each other. Thus the energy due to overlap
of individual fragments in contact is zero before the impact and is also zero after all fragments in
contact have separated from each other.
This is shown in Figure 17, where the total kinetic energy of all fragments for the four heaps
is shown. All four heaps start with the same initial kinetic energy, which is equal to the initial
kinetic energy of the projectile. As impact occurs, the projectile slows down and contact between
Figure 16. Second part of the predicted motion sequence for heap D.
individual fragments results in some of the kinetic energy of the system being taken as potential
energy due to overlapping fragments in contact. This temporary reduction in kinetic energy is
largest for heap A and smallest for heap D. In fact, in terms of kinetic energy, there is a gradual
change from heap A to heap D.
Figure 17. Kinetic energy of beam-shaped heaps AD as function of time.
For all four heaps, as the transient contacts are released due to the fragments moving away from
each other, the initial kinetic energy of the system is restored. Contacts have therefore resulted in
some of the kinetic energy being transformed into the potential energy due to contacts between
fragments and nally the transfer of potential energy into kinetic energy takes place as contacts
are released. Thus, the algorithm, as proposed and implemented into the combined nitediscrete
element code, has preserved energy for all four heaps regardless of the number of contacts, duration
of the contacts or the size of overlap between fragments in contact, which is in agreement with
theoretical analysis as given earlier.
6. CONCLUDING REMARKS
The contact interaction algorithm proposed in this paper is a natural contact solution for the com-
bined nitediscrete element method where each discrete element is discretized into nite elements.
The algorithm proposed enables relatively simple contact procedures between nite elements of
triangular shape to be employed for discrete elements of general shape and size, provided that
these discrete elements are discretized into nite elements (which is the essence of the combined
nitediscrete element method). It uses only the data supplied in the in-core database for contact
free nite element analysis and is easily linked to contact detection algorithms such as NBS contact
detection algorithm. The algorithm does not require detection of boundary surfaces.
The proposed algorithm results in preservation of energy balance regardless of the number of
discrete or nite elements involved, the shape and size of discrete elements in contact and the
shape and size of overlap between discrete elements. It also results in realistic distribution of
contact forces.
The numerical examples presented conrm the analytical results and clearly demonstrate the
ability of the algorithm to handle contact situations where the transient motion of thousands of
interacting bodies is involved.
REFERENCES
1. Munjiza A, Owen DRJ, Bicanic N. A combined nite-discrete element method in transient dynamics of fracturing
solids. International Journal of Engineering Computations 1995; 12:145 174.
2. Munjiza A, Andrews KRF, White JK. Combined single and smeared crack model in combined nite-discrete element
method. International Journal for Numerical Methods in Engineering 1999; 44:4157.
3. Gen-Hua Shi. Discontinuous deformation analysisa new numerical method for the statics and dynamics of block
system. PhD Thesis, Department of Civil Engineering, University of California, Berkeley, 1988.
4. Munjiza A, Andrews KRF. NBS contact detection algorithm for bodies of similar size. International Journal for
Numerical Methods in Engineering 1998; 43:131149.
5. Oldenburg M, Nilsson L. The position code algorithm for contact searching. International Journal for Numerical
Methods in Engineering 1994; 37:359 386.
6. Bonet J, Peraire J. An alternating digital tree (ADT) algorithm for 3D geometric searching and intersection problems.
International Journal for Numerical Methods in Engineering 1991; 31:117.
7. Malone JG, Johnson NL. A parallel nite element contact=impact algorithm for non-linear explicit transient analysis:
part Ithe search algorithm and contact mechanics. International Journal for Numerical Methods in Engineering
1994; 37:559 590.
8. Mustoe G, Williams JR. Proceedings of the 1st U.S. Conference on Discrete Element Methods, Golden, CO, 1989.
9. Williams JR, Mustoe G. Proceedings of the 2nd U.S. Conference on Discrete Element Methods, MIT, MA, 1993.
10. Zhong ZH, Nilsson L. A contact search algorithm for general 3-D contactimpact problems. Computers and Structures
1990; 34:327 335.

Penalty Function Method For Combined Ÿnite-Discrete Element Systems Comprising Large Number of Separate Bodies

Uploaded by

Document Information

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

Penalty Function Method For Combined Ÿnite-Discrete Element Systems Comprising Large Number of Separate Bodies

Uploaded by

Copyright:

Available Formats

INTERNATIONAL JOURNAL FOR NUMERICAL METHODS IN ENGINEERING

Int. J. Numer. Meth. Engng 2000; 49:13771396

Penalty function method for combined nitediscrete

A. Munjiza; and K. R. F. Andrews

KEY WORDS: nite elements; discrete elements; contact; interaction

Received 5 January 1999

2. POTENTIAL FIELD-BASED PENALTY FUNCTION METHOD

Figure 1. Contact force due to an in nitesimal overlap around points Pc and Pt .

penalty function method takes the form

Figure 3. Convex overlapping area S.

producing lines 1 and 2 such that

3. IMPLEMENTATION INTO THE COMBINED FINITEDISCRETE ELEMENT METHOD

4. DISCRETIZATION OF POTENTIAL FIELD AND ASSOCIATED

Figure 5. Potential at any point P of triangle-shaped nite element.

Figure 6. Contact of contactor and target triangles and contact of an edge

where Ai ; i = 1; 2; 3, is the area of corresponding sub-triangles as shown in Figure 5. The potential

pi = ([ri rA ] u; [ri rA ] v) (24)

Figure 8. Equivalent nodal forces.

Figure 9. Beam-shaped heap of rigid fragments.

Figure 11. Predicted transient motion sequence for heap A.

Figure 12. Predicted transient motion sequence for heap B.

Figure 13. Predicted transient motion sequence for heap C.

approximates the motion sequence of a heap consisting of an in nite number of in nitesimally

Figure 17. Kinetic energy of beam-shaped heaps AD as function of time.

You might also like

Figure 1. Contact force due to an innitesimal overlap around points Pc and Pt .

approximates the motion sequence of a heap consisting of an innite number of innitesimally