See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/299389166

3D modelling of rock mass falls using the

Particle Flow Code PFC3D .

Conference Paper January 2007

CITATIONS READS

2 121

2 authors, including:

Alexander Preh
TU Wien
66 PUBLICATIONS 92 CITATIONS

SEE PROFILE

All content following this page was uploaded by Alexander Preh on 27 March 2016.

The user has requested enhancement of the downloaded file. All in-text references underlined in blue are added to the original document
and are linked to publications on ResearchGate, letting you access and read them immediately.
 3D modelling of rock mass falls using the Particle Flow Code PFC3D

A. Preh & R. Poisel

Institute for Engineering Geology, Vienna University of Technology, Vienna, Austria

ABSTRACT: Rock mass falls are modelled as the movements of single rock blocks over a surface or as the movement of a
viscous mass over a surface (DAN). In reality a mass of discrete, interacting rock blocks is moving downslope. Thus, the
program PFC (Particle Flow Code) based on the Distinct Element Method was used to model rock mass falls realistically in
3 dimensions, based on physical relations. PFC models the movement and interaction of circular (2D) or spherical (3D)
particles and wall elements, using the laws of motion and of force - displacement. During the calculation, the contacts
between particles and particles or particles and walls are detected automatically. The particles may be bonded together at
their contact points, and the bondage can break due to an impact. For realistic modelling of a runout a viscous damping
routine in case of a particle wall contact was introduced. Numerical drop tests, comparisons with the results of a rock fall
program and back analyses of several rock mass falls provided appropriate damping factors. Thus, the movement types -
bouncing, sliding, rolling and free falling of single rock blocks - and the interaction between the blocks occurring in a rock
mass fall can be realistically modelled by using the PFC adapted code.

1 METHODS TO MODEL A ROCK MASS FALL absorption, friction, etc.) applied in rock fall programs
USING PFC (Hoek 1987, Spang & Rautenstrauch 1988).

PFC models the movements and interactions of stressed

assemblies of spherical particles either in or getting into
contact with wall elements. The particles may be bonded
together at their contact points to represent a solid that may
fracture due to progressive bond breakage.
Every particle is checked on contact with every other
particle at every time step. Thus, PFC can simulate not only
failure mechanisms of rock slopes but also the runout of a
detached and fractured rock mass (Poisel & Roth 2004).
Rock mass falls can be modelled as an All Ball model
and a Ball Wall model. An All Ball model (Fig. 1)
simulates the slope as an assembly of balls bonded together.
The simulation shows the failure mechanism of the slope
due to gravity (Poisel & Preh 2004). After detachment of
the moving mass, the runout is modelled automatically. As Figure 1. All Ball model (Preh 2004)
in every other numerical model, the material represents only
a hypothetical material. In reality, a large number of blocks
are moving downslope. Due to a restricted number of balls,
due to software and computation time reasons, it is
generally not possible to model every rock block with a
single ball.
In the Ball Wall model (Fig. 2) the underlying bedrock
is simulated by linear (2D) and planar (3D) wall elements
(Roth 2003). Therefore, an estimate or a model of the
failure mechanism of the slope and of the detachment
mechanism is needed as an input parameter. However, in
the Ball Wall model the detached mass can be modelled
using more and smaller balls with the same computational
effort in order to approach reality better. The Ball Wall
model offers the possibility to make use of the know-how
related to runout relevant factors (coefficients of restitution,
Figure 2. Ball Wall model (Frhwirth 2004)
2 RUNOUT RELEVANT PARAMETERS The viscous damping model used in PFC introduces
normal and shear dashpots at each contact (Fig. 3). A
According to observations in nature, several kinds of damping force, Di (i = n: normal, s: shear), is added to the
movements of the rock fall process (Broilli, 1974) have to contact force, of which the normal and shear components
be distinguished during the computation (Bozzolo, 1987): are given by
free falling, (5)
Di = Ci Vi
bouncing,
rolling and where Ci (i = n: normal, s: shear) is the damping constant,
sliding. Vi (i = n: normal, s: shear) is the relative velocity at
In order to achieve an appropriate simulation of these contact, and the damping force acts to oppose motion. The
different kinds of movements by PFC, some modifications damping constant is not specified directly; instead, the
have been necessary using the implemented programming critical damping ratio i (i = n: normal, s: shear) is specified,
language (Fish). and the damping constant satisfies

C i = i C icrit (6)
2.1 Free falling
where Ccriti is the critical damping constant, which is given
In order to model the free falling of blocks, neither the
by
acceleration nor the velocity (ignoring the air resistance) is
to be reduced during fall as a consequence of mechanical (7)
Cicrit = 2mi = 2 mk i
damping.
PFC applies a local, non- viscous damping proportional to where i (i = n: normal, s: shear) is the natural frequency of
acceleration, to the movement of every single particle as a the undamped system, ki (i = n: normal, s: shear) is the
default. The local damping used in PFC is similar to that contact tangent stiffness, and m is the effective system
described in Cundall (1987). A damping-force term is added mass.
to the equations of motion, so that the damped equations of
motion can be written

F(i) + F(i)d = M (i) A (i) ; i = 1K 6 (1)

mx
&&(i) for i = 1K3; (2)
M (i) A (i) =
I
& (i 3) for i = 4K 6
where F(i), M(i), and A(i) are the generalized force, mass, and
acceleration components, I is the principal moment of
inertia, & is the angular acceleration and &x& is the
translational acceleration; F(i) includes the contribution from
the gravity force; and Fd(i) is the damping force
d
F(i) = F(i) sign( (i) ) i = 1K 6 (3) Figure 3. Viscous damping activated at a contact with the linear
contact model (Itasca 1999)
expressed in terms of the generalized velocity

x& (i) for i = 1K3; In rock fall programs, the rebound height of blocks
(i) = (4) touching the bedrock is calculated using restitution
(i 3) for i = 4K 6. coefficients. The restitution coefficient Ri (i = n: normal,
The damping force is controlled by the damping constant , s: shear) is defined as the ratio of the contact velocity before
whose default value is 0.7 and which can be separately and after the impact and can be defined as
specified for each particle. vif
This damping model is the best suited for a quick Ri = (8)
vii
calculation of equilibrium. There arises, however, the
disadvantage of the movements of the particles being where vif (i = n: normal, s: shear) is the velocity of the block
damped as well. Therefore, the local damping has been after impact and vii (i = n: normal, s: shear) is the velocity
deactivated for all kinds of particle movements. of the object before impact. The relation between the
restitution coefficient Ri and the critical damping ratio i
2.2 Bouncing can be estimated by simulating drop tests (Fig. 4).

Elastic and plastic deformations occur in the contact zone

during the impact of a block. Both the kinetic energy of the
bouncing block and the rebound height are reduced by the
deformation work. The reduction of the velocity caused by
the impact is modelled with the help of a viscous damping
model integrated in PFC.
 rebounds of three differently damped particles have been
investigated: the first one using the default value of the non
viscous damping ( = 0.7), the second one using the viscous
damping with a critical damping ratio n of 0.27 and the
third one using the viscous damping with a critical damping
ratio n of 0.12.
The critical damping ratios selected for this comparison
equal the restitution coefficient RN = 0.5 (n = 0.27) and
0.71 (n = 0.12) according to the drop tests back calculated
by Itasca (Fig. 4) This is verified by the course of the
particle velocities (Fig. 5).
According to the conservation of energy, a restitution
coefficient RN = 0.5 results in a rebound height of a quarter
of the drop height and a restitution coefficient RN = 0.71
results in a rebound height of half the drop height. This is
verified by the course of the rebound height (Fig. 6).
Figure 4. Relation between restitution coefficient and critical Thus, it is possible to control the number of impacts as
damping ratio (Itasca 1999) well as the magnitude of the bouncing height, using the
viscous contact damping model.
Spin has an impact on both the direction and the velocity
of the rebounding block. Therefore, it is essential to
consider the spinning when modelling the runouts of rock
falls. PFC determines the motion of a each single particle by
the resultant force and moment vectors acting upon it, and
describes it in terms of the translational motion of a point in
the particle and the rotational motion of the particle
(Equations 1 and 2). Figure 7 depicts the flight trajectories
of three particles bouncing at different spins.

Figure 5. Drop tests, course of velocity: dashed line local

damping = 0.7, full line RN = 0.71, dotted line RN = 0.5

Figure 7. Rebound angle influenced by the particle spin

(black line rebound course; green, yellow and red balls particle
position after rebound)

Furthermore, with PFC the interaction of friction and spin

is considered, since the influence of the spin increases with
the increase of frictional resistance.
By modelling rock mass falls, it was shown to be
necessary to distinguish between ball-ball contacts and ball-
wall contacts. This was done by using the programming
language Fish.
Figure 6. Drop tests, rebound height: dashed line local damping
= 0.7, full line RN = 0.8, dotted line RN = 0.5 2.3 Rolling
The most important runout relevant effect is rolling
Figures 5 and 6 show the results of a drop test for resistance, because it is known that pure rolling of blocks in
verification of the viscous damping model, where the
the model leads to more extensive runouts than observed in The angular acceleration is defined by a finite difference
nature. relation in order to express the increment of the angular
The rolling resistance is caused by the deformation of the velocity per time increment (Equation 12). Thus, the friction
rolling body and/or the deformation of the ground (Fig. 8) force Frr iis defined by the conservation of momentum
and depends strongly on the ground and the block material.
rr . (12)
Frr = m r
t

Equation 10 and equation 12 yield

2 (13)
m r 2 rr = Frr rrr N c rr
5 t

2
m r 2 rr = m rr r rrr m g c rr .
5 t t
Figure 8. Deformation of the surface and distribution of contact
stresses Therefore, the angular deceleration is

g c rr (14)
rr = t
Due to these deformations, the distribution of contact 2
stresses between the ground and the block is asymmetric r (rrr r)
5
(Fig. 9). Replacing the contact stresses by equivalent static
contact forces results in a normal force N, which is shifted
rrr = r 2 c 2rr .
forward by the distance of crr , and a friction force Frr ,
opposing the direction of the movement.
The rolling resistance is implemented by adding the
r calculated increment of the angular velocity, to the angular
ex
velocity calculated automatically by PFC at every time step
r (Equation 14).
ey
i(t ) = i(t ) + rr,i (15)

M S According to these considerations, the rolling resistance is

x&s
m g
an eccentricity crr or sag function urr. The deeper the block
sags, the greater is the rolling resistance rr. In classical
mechanics, the rolling resistance is a function of the ratio of
the eccentricity crr to the radius r.

c rr
r = [ ] . (16)
r

This means that spherical blocks of different sizes have

the same runout for the same rolling resistance coefficient.
Figure 9. Calculation of the rolling resistance In nature, however, it can be observed that large blocks
generally have a longer runout than smaller ones. Therefore,
according to the damping model described, the runout is
The deceleration of the angular velocity caused by the calibrated by the sag urr.
rolling resistance is calculated using conservation of Figures 10a-10d show model calculations carried out by
translational momentum (Equation 9) and angular PFC, using the model of rolling resistance just described.
momentum (Equation 10). The detached rock mass was modelled as an irregular
assembly of particles (Fig. 10a) of two different sizes
m &&
x s = Frr (9) (r1 = 0.8 m, r2 = 1.6 m). For both particle sizes, the same
sag of urr = 25 cm was employed. Figure 10b shows the
& rr = M rr , Isphere = 2 m r 2
I (10) position of the runout and figures 10c and 10d show a
5 detailed view of the runout. It therefore became apparent
that the larger particles have a longer runout than smaller
where Mrr is the resulting moment caused by the rolling ones. Fig. 10d shows that within the deposit mass the
resistance, I is the principal moment of inertia and rr is the smaller particles (r = 0.8 m) rest at the bottom and the larger
angular deceleration. particles (r = 1.6 m) at the top. This model behaviour
The kinematic link is established by the condition of pure corresponds closely to observations in nature.
rolling (Equation 11).

x s =
&& & r (11)
Figure 10a. Detached rock mass

Figure 11. Plan view of final state of runout simulation of a rock

mass fall using PFC3D.

REFERENCES

Bozzolo, D. 1987. Ein mathematisches Modell zur Beschreibung

der Dynamik von Steinschlag. Dissertation Nr. 8490 an der
Figure 10b. Final state of the PFC2D calculations ETH Zrich.
Broilli, L. 1974. Ein Felssturz im Groversuch. Rock Mechanics,
Suppl. 3, S. 69-78.
Cundall, P.A. 1987. Distinct Element Models of Rock and Soil
Structure. In Analytical and Computational Methods in
Engineering Rock Mechanics, Ch. 4, pp. 129-163, E. T. Brown,
Ed. London: Allen & Unwin.
Frhwirt, Th. 2004. Numerische Untersuchungen von
Felsmassenstrzen im Bereich der Grohangbewegung
Galgenwald bei Murau mittels PFC3D. Diplomarbeit am Institut
Figure 10c. Deposit
fr Ingenieurgeologie der Technischen Universitt Wien.
Hoek, E. 1987. Rockfall A program in basic for the analysis of
rockfalls from slopes. Dept. Civil Eng., University of Toronto,
Toronto.
Itasca 1999. PFC2D (Particle Flow Code in 2 Dimensions) User's
Guide. Itasca Consulting Group, Inc., Minneapolis.
Poisel, R., Angerer, H., Pllinger, M., Kalcher, T. & Kittl, H.
2007. Risk Management of the Landslide Lrchberg
Galgenwald (Austria). Proceedings of the 11th Int. Congr.
ISRM, Lisbon.
Poisel, R. & Preh, A. 2004. Rock slope initial failure mechanisms
Figure 10d. Deposit, detail view and their mechanical models. Felsbau 22: 40-45.
Poisel, R &, Roth, W. 2004. Run Out Models of Rock Slope
Failures. Felsbau 22: 46-50.
2.4 Sliding Preh, A. 2004. Modellierung des Verhaltens von Massen-
Sliding is calculated by the slip model implemented in PFC bewegungen bei groen Verschiebungen mit Hilfe des Particle
without any further adaptation. Flow Codes. PhD Dissertation, Inst. for Eng. Geology, Vienna
University of Technology.
Roth, W. 2003. Dreidimensionale numerische Simulation von
Felsmassenstrzen mittels der Methode der Distinkten Elemente
3 CONCLUSION (PFC). PhD Dissertation, Inst. for Eng. Geology, Vienna
University of Technology.
The model described above was used to calculate the runout Spang, R.M. & Rautenstrauch, R.W. 1988. Empirical and
distances and the risks of the mass movement Lrchberg mathematical approaches to rockfall protection and their
Galgenwald (Austria; Fig. 11; Poisel et al. 2007). practical applications. Proceedings of the 5th International
With the help of the adapted PFC code it is possible to Symposium on Landslides, Lausanne, Vol. II, pp. 1237-1243.
create a mechanically correct model of rock mass falls. A
realistic prognosis of runouts necessitates a calibration of
relevant runout model parameters by back calculation of
silent witnesses (blocks having fallen down already).

View publication stats