International Journal of Rock Mechanics & Mining Sciences 67 (2014) 170–180

Contents lists available at ScienceDirect

International Journal of
Rock Mechanics & Mining Sciences
journal homepage: www.elsevier.com/locate/ijrmms

Calibration of rock fall modeling parameters

Duncan C. Wyllie n
Wyllie & Norrish Rock Engineers Ltd., 850 - 789 West Pender Street, Vancouver, B.C., Canada V6C 1H2

art ic l e i nf o a b s t r a c t

Article history: Rock falls can be a hazard for many facilities in mountainous terrain, particularly in areas with high
Received 14 June 2012 rainfall and freezing temperatures, and where seismic events occur. Design of protection measures
Received in revised form against rock falls, such as fences, barriers and sheds requires information on impact energy to determine
11 September 2013
the strength of the structure, and on trajectories to determine its location and size. This paper presents
Accepted 11 October 2013
Available online 11 November 2013
documentation on rock falls at five locations, in North America and Japan, involving impacts on rock,
talus, colluvium, asphalt and concrete, for which values of the normal and tangential coefficients of
Keywords: restitution have been calculated. The field results show that the normal coefficient of restitution is
Rock fall related to the impact angle with coefficients of about 0.2 for normal impacts and values as great as 2.5 for
Modeling
shallow impacts. These values for the normal coefficient of restitution are essentially independent of the
Coefficient of restitution
slope material. The tangential coefficient of restitution ranges between 0.3 and 0.8 and is related to the
Impact mechanics
Friction friction coefficient at the impact point, and is independent of the velocity and normal force. It is also
Case studies shown that the calculated field values for the coefficients of restitution are consistent with the principles
of impact mechanics.
& 2013 Elsevier Ltd. All rights reserved.

1. Introduction In some high hazard locations it may be appropriate to construct

In mountainous terrain, infrastructure such as highways,
railways and power generation facilities are often subject to rock
fall hazards that can result in economic losses due to service
interruptions and equipment damage, as well as injury to users
and operators of these facilities. As an illustration of these hazards,
Fig. 1 shows a car that struck a falling rock that passed through the
windshield. Rock fall hazards are particularly severe in areas with
heavy precipitation, frequent freeze/thaw cycles and seismic
events [1–3], and the risk can be significant where the population
density is high. These climatic conditions exist, for example, in the
Alps and the west coast of North America. In contrast, climatic
conditions in Hong Kong are not as severe as those in more
northerly latitudes but rock fall risk can be severe because of the
high population density [4].
Since the 1980s, a variety of rock fall protection measures have
been developed that are now widely used and have significantly
improved safety. These protection measures include ditches that
can be designed to reasonably well defined criteria [5,6], and can
be more effective if they incorporate barriers such as gabions or
MSE walls constructed from locally available materials. In addition,
a number of proprietary fence systems have been developed that
use various configurations of high strength steel cables and wires.
reinforced concrete rock sheds that incorporate energy absorbing
features such as flexible hinge elements and a cushioning sand or
Styrofoam layer on the roof [7,8].
Design of protection measures requires data for two basic
parameters of the rock falls – impact energy and trajectory. That
is, information is required on the mass and velocity of the fall to
determine the energy capacity, and on the location of impact
points and trajectory paths to determine the optimum location
and dimensions of the barrier or fence.
Many modeling programs are available to simulate rock fall
behavior, and these are widely used to design protection measures
[9–12]. The reliability of modeling programs depends on a thor-
ough understanding of impact mechanics and trajectory analysis,
and having available precise data on actual rock falls to calibrate
the models. The objective of this paper is to provide carefully
documented data on actual rock events that can be used to
calibrate modeling programs. In addition, the paper discusses the
application of impact mechanics theory to the behavior of rock
falls, and proposes relationships for tangential and normal velocity
changes during the impact process.
2. Documentation of rock fall events

Data on rock fall events are available from both natural events
n
Tel.: þ 1 604 673 9908. where it has been possible to precisely map impact points
E-mail address: dwyllie@wnrockeng.com and trajectories, and from carefully documented, full-scale rock

1365-1609/$ - see front matter & 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.ijrmms.2013.10.002
 D.C. Wyllie / International Journal of Rock Mechanics & Mining Sciences 67 (2014) 170–180
Fig. 1. Moving vehicle struck rock fall that passed through the windshield.
Fig. 2. Mt. Stephen – barrier constructed with concrete blocks and steel mesh fence
to contain rock falls and snow avalanches.
fall tests. These data have been collected for a variety of slope
geometries and fall heights, and for slope materials comprising
rock, colluvium, talus and asphalt. This section describes five
events covering a range of situations that are intended to encom-
pass many of the commonly occurring rock fall conditions. For
each event, the site topography, geology and rock fall conditions
are described, and values for velocity components are provided for
typical impact conditions [13].
In calculating trajectories at sites where information on precise
impact points and flight paths is not available, it is necessary to
select the two end points for each trajectory and to make an
assumption for the angle at which the rock leaves the slope
surface. These data have been obtained from experience gained
at other, fully documented rock fall sites and from only using
trajectories that are realistic and mathematically feasible.
The insets in Figs. 3–7 show details of the normal and tangential
velocity components at selected impact points, and the calculated
coefficients of restitution that are defined as follows:
Normal coefficient of restitution;
final normal velocity; vf N
eN ¼  ð1aÞ
impact normal velocity; viN
Tangential coefficient of restitution;
final tangential velocity; vf T
eT ¼ ð1bÞ
impact tangential velocity; viT
On the insets, the length and orientation of the arrow is propor-
tional to the velocity vector, where the subscript “i” refers to values at
171
Detail of velocity components
at impact point on rock
A B
Fig. 3. Mt. Stephen – cross section of lower part of slope showing rock fall barrier
and typical rock fall trajectories.
the point of impact (time, t ¼ i), and the subscript “f” refers to values
at the end of the impact process (time, t ¼ f). Further discussion on
the definition of the coefficient of restitution is contained in Section
3.3. In all cases the rocks are also spinning with rotational velocity
(–ω) radian s  1, where clockwise rotation is negative.
2.1. Impacts on rock slopes
Data have been analyzed for falls at locations in Canada, the
United States and Japan, for slopes ranging in height from 2000 m
to 15 m. The following is a discussion on falls at three of these
locations where the falls impacted bare rock slopes.
2.1.1. Rock fall barrier, Mt. Stephen, Canada
Mt. Stephen in the Canadian Rocky Mountains is a source of
rock falls and snow avalanches that originate on a rock face nearly
2000 m high at an overall slope angle of about 501. The geology is
a strong, massive, horizontally bedded limestone containing thin
but widely spaced shale beds; the shale weathers more rapidly
than the limestone resulting in the formation of overhangs and
falls of the stronger rock. As shown in Fig. 2, it has been necessary
to construct a barrier to protect a railway operating at the base
of the slope.
The barrier comprises a double sided MSE wall built with
precast concrete blocks (dimensions 1.5 m long, 0.75 m in section),
supported with Geogrid and gravel fill between the walls, and
a steel cable fence along the top of the wall. The total height of the
structure is 11.6 m. Fig. 3 shows a typical section of the lower
120 m of the slope that was generated from an aerial Lidar survey
of the site. Fig. 3 also shows a range of feasible trajectories of rock
172 D.C. Wyllie / International Journal of Rock Mechanics & Mining Sciences 67 (2014) 170–180

Fig. 4. Oregon test site – typical rock fall trajectory and impact points for 15 m high, 761 rock face from Oregon test site.

Fig. 5. Ehime, Japan – results of rock fall test at showing restitution coefficients for impacts on rock and talus; h is the maximum trajectory height normal to the slope.

falls that impacted the rock slope and were then contained by the
barrier.
It was possible to identify rock fall impact points on both the
steel mesh fence and the concrete blocks, and to define the
coordinates of each point relative to one end of the wall. In total,
466 impacts were documented. Analyses of assumed trajectories
between likely impact points on the slope and the barrier allowed
the impact (vi) and restitution velocities (vf) to be calculated
at each impact point. From these data, the velocity components
and tangential (eT) and normal (eN) coefficients of restitution were
determined. The inset on Fig. 3 shows the calculated velocity
components at impact point A on trajectory S-A-B.
Normal restitution coefficients for five typical impact points on
the rock face are plotted on Fig. 14.
2.1.2. Rock fall test site, Oregon
An extensive rock fall test program was carried out at Kreuger
Quarry in Oregon to determine the required ditch configurations
to contain rock falls in highway ditches [6]. The excavated rock
cuts included cut heights between 8 m and 24 m, and face angles
between vertical and 451; in total 11,250 separate tests were
conducted. For each rock fall, the data collected included the first
impact position, and the rollout distance in the ditch. The rock at
 D.C. Wyllie / International Journal of Rock Mechanics & Mining Sciences 67 (2014) 170–180 173

Source
350

Rock slope

300
Talus
Tree impact

250 Impact points

200
Vertical fall
height (m)

150
Colluvium

100

50
Rock stopped

0
0 100 200 300 400 500 600
Horizontal fall distance (m)

Fig. 6. Tornado Mountain, Boulder A – mapped impact points and broken trees, with detail of velocity components at impact #A26; arrows show locations of trees impacted
by falling rock.

Fig. 7. Impact of a rock fall on a horizontal asphalt surface; last trajectory shown for a fall originating from a height of 138 m.

the test site was a strong, blocky basalt that was excavated with
controlled blasting on the final face to produce a face with few
irregularities.
Fig. 4 shows that test results for a 15 m high cut at a face angle
of 761 (¼H:1V) with a horizontal ditch. The diagram shows
the location of the first impact point in the ditch for the 95th
percentile of the test rocks, and the assumed trajectories for a rock
fall from the crest of the cut initially impacting the face, and then
the ditch. The second impact point in the ditch is an estimated
location based on common rock fall behavior. The insets on Fig. 4
show the calculated velocity components at the impact point on
the cut face, and the values for eT and eN for the first two impact
points. These calculated values show the difference in behavior
of rock falls that impact the rock surface at a shallow angle (on the
rock face) and at a steep angle (in the ditch). That is, for shallow
angle impacts, the normal coefficient is greater than 1, while for
steep impacts the normal coefficient is less than 1. In contrast, the
tangential coefficient is less than 1 for all impact angles. This
behavior is discussed in Sections 3.4 and 3.6 below.
2.1.3. Rock fall test site, Ehime, Japan
In 2003 a rock fall study was carried out on a 42 m high rock and
talus slope at the Ehime test site on Shikoku Island [14,15]. The
slope comprised a 26 m high rock slope in horizontally bedded
sandstone and mudstone with a face angle of 441 and a 16 m high
talus slope at angle of 351. The test involved both natural boulders,
and spherical and cubic concrete blocks containing embedded three
174 D.C. Wyllie / International Journal of Rock Mechanics & Mining Sciences 67 (2014) 170–180
dimensional accelerometers and a high speed data acquisition
system that allowed the precise location, and translational and
rotational velocities to be determined over the full extent of the fall
path. Fig. 5 shows the impact points and trajectories of a typical test.
Also shown in Fig. 5 are the calculated velocity components
and the values for eT and eN at impact point #3 on rock, and impact
points # 6 and #7 on talus. For the rock impact, the value for eN is
greater than 1, and has a similar value to that for the Oregon test
for the shallow impact point on the steep rock face. At both the
Oregon and Ehime test sites, eN values greater than 1 occurred for
shallow impacts where the impact angles (θi) were small (131 and
121 respectively).
2.2. Impacts on talus and colluvium slopes
Information of impacts on talus and colluvium has been
obtained from the Ehime test site in Japan (Fig. 5), and from
two rock falls on Tornado Mountain in southeast British Columbia,
Canada (Fig. 6). The calculated eT and eN values for impact #6 and
#7 at Ehime on talus are shown in Fig. 5. It is of interest that the
eT values of both these impacts are nearly identical, while the
eN values are very different with the trajectory after impact #6
barely leaving the slope surface, while the trajectory after impact
#7 is the longest and highest of the rock fall where eN ¼ 5.48.
The difference in the trajectories is probably due to a combination
of slope roughness and the attitude of the block as it impacted the
surface. Although the eN value of 5.48 is a reliable, measured value,
it has not been used in the analyses presented in this paper
because this is a significantly higher and longer trajectory than
other rock falls at this and other sites studied.
The Tornado Mountain site in south-east British Columbia,
Canada comprises a 50 m high rock face in very strong, massive
limestone, above a colluvium slope at an angle varying from 351
on the upper slope to 221 on the lower slope. The colluvium
is mixture of gravel and soil forming a uniform slope with no
significant irregularities, and no previous rock falls. The slope is
sparsely vegetated with pine trees having diameters ranging from
about 300 mm to 500 mm.
In 2004, two separate rock falls originating on the rock face
traveled a total distance of 740 m down the slope – distances of
340 m vertically and 610 m horizontally. Because no similar rock
falls had occurred in the past and each rock followed a separate
path, it was possible to locate each impact point on the slope and
define its coordinates with a GPS and a laser rangefinder. In total,
45 impact points were identified for Boulder A and 69 impact
points for Boulder B; the final masses for the boulders were about
3750 kg (maximum dimension 1.6 m) and 5600 kg (maximum
dimension 2.5 m) respectively. Both rocks impacted a horizontal
surface on the lower part of the slope that had been excavated in
the colluvium beside a railway and the loss of energy on these
impacts was sufficient to stop the rock within 30 m.
In addition to the impact points on the colluvium, it was also
possible to locate a total of 21 trees that had been impacted and
broken off by the boulders, and measure the height of the impact
and the distance from the adjacent impact point. It is considered
that the trees did not impede the trajectories because of their
small diameter and low strength. Using information on the
coordinates of successive impacts on the slope and the impact
with the tree it was possible for these 21 cases to calculate precise
trajectories and velocities, including angles θf at the completion of
impact and start of the new trajectory. This information on the
average and distribution of θf values was then used to calculate
likely velocities components for all other impacts.
For Boulder A, at impact #26 where the precise trajectory could
be determined from a broken tree, the value of eN is 1.29, a value
that is consistent with the other sites where shallow impacts
occurred (i.e., θi ¼ 221). For all 114 impacts on both fall paths, the
average value of eN was 1.02. The value for eT at impact #26 was
also calculated as shown in Fig. 6, with the average value of eT for
Boulders A and B being 0.65.
The trajectories were also analyzed to determine the maximum
height of the trajectories measured normal to the ground surface.
It was found that the average height was 1.3 m showing that the
rock falls followed shallow, “skipping” paths down the slope.
2.3. Impact on asphalt
A rock fall from a height of 138 m followed a path comprising a
58 m high rock slope in very strong, volcanic rock at an angle of
601, a 70 m high colluvium slope at an angle of 421, and a 10 m
high rock cut. The fall then impacted an asphalt road surface. Fig. 7
shows the final trajectories from the crest of the rock cut to just
after the impact on the road. These trajectories were precisely
defined by a survey of the site and the inset in Fig. 7 shows the
calculated velocity components at the asphalt impact point.
Although this is a single record of an impact with asphalt, the
author has investigated several similar events (confidential,
unpublished) where comparable trajectories were generated for
impacts on asphalt.
The trajectory shown in Fig. 7 is a relatively steep impact (i.e.,
θi ¼ 501) compared to the shallow impacts at the Tornado
Mountain and the Oregon test site, and for this condition the
value of eN is 0.38. As discussed in Section 3.6 below, the value of
eN for steep impacts is low compared to shallow impacts. The
value for eT for this relatively smooth impact surface is 0.24.
3. Impact mechanics
The theory of impact mechanics can be applied to rock falls in
order to understand the impact process and develop algorithms
for modeling rock falls. The particular physical conditions that are
applicable to rock falls are low velocity (i.e., less than about
40 m s  1), and impact of hard, rigid bodies that result in minor
deformations, but high stresses generated over small areas. These
contact conditions, in which no interpenetration of the two bodies
occurs, are referred to as low compliance impacts.
The theory of impact mechanics is well developed [16,17] and
builds on earlier work carried out by Sir Isaac Newton [18] and
others such as Poisson and Hertz in the 19th century. The theories
generally apply to the impact between two bodies, made of an
identical material, that are both translating and rotating, have
unequal masses and are moving in three dimensional space.
For rock falls, the impact conditions are somewhat simplified
because one of the bodies (the slope) is stationary and has infinite
mass. However, the roughness of the slope and irregularity of the
rock fall blocks introduce complexities in the modeling that are
best handled by stochastic analyses [19]. Furthermore, the theory
needs to account for the condition that the two bodies may not be
of the same material.
This section summarizes the application of impact mechanics
theory to rock falls, while Section 4 demonstrates how the field
results discussed in Section 2 above can be related to the theory.
3.1. Forces and impulses generated during impact
The impact process for a non-rotating body moving with
velocity vi and impacting a stationary surface at right angles can
be simulated with an infinitesimal, deformable particle at the
contact point between the two bodies (Fig. 8). The particle acts
as a short, stiff spring that, during the impact, generates equal and
opposite reaction forces –F, F´ at the point of impact that are
 D.C. Wyllie / International Journal of Rock Mechanics & Mining Sciences 67 (2014) 170–180 175

-viN +v

Deformable
particle
-F

Fig. 8. Forces generated at contact point during normal impact [17].

parallel to the velocity direction. During impact, the maximum

deformation of the particle is δ. For hard impacting bodies, the
reaction forces F, F´ are large compared to the body force and act
for a very short time, and consequentially gravity body forces do
not affect the changes in velocity that occur during impact.
The reaction forces also produce a finite impulse p that is
defined as a force acting over time. The impulse generated by the
impact continuously changes the relative velocity v of the bodies Fig. 9. Variation in force F during impact (a) relationship between force and
during the time of impact t. The normal component of impulse pN deformation of the impact point; (b) change in force and impulse with time during
impact for plastic (eN ¼ 0), elastic (eN ¼ 1) and partially elastic materials; pcN is
is generated by the normal component of the contact force F
impulse generated up to time of maximum compression (t ¼ i to t ¼ c), and
according to the relationship [17] (pfN – pcN) is impulse generated during restitution phase of impact (t ¼ c to t ¼ f) [17].
dpN ¼ Fdt ð2Þ
where and deformation (δc) are at a maximum at the end of the
dv compression phase, followed by partial recovery (δf), for inelastic
F ¼m impact, at the completion of the recovery phase. The recovery of
dt
kinetic energy is the process that drives the bodies apart in the
dv final phase of the impact.
dpN ¼ m dt
dt Fig. 9b shows the changes in the normal contact force F as a
and function of time. The area under the [t – F] curve up to time
tc represents the impulse generated pcN and energy lost during the
dpN
dvN ¼ compression phase when the kinetic energy of relative motion is
m
converted into internal energy of deformation. The area between
The relative normal velocity v at any time t during the impact times tc and tf represents the impulse generated and energy
can be obtained by integration, for the condition that the normal recovered during the restitution phase, where the change in
velocity at the moment of impact at time (t ¼ i) is vi , where impulse is (pfN – pcN). The changes in energy and velocity during
Z t impact can be quantified in terms of the normal coefficient of
1
vN ¼ dpN restitution, eN that is the ratio of the final normal velocity vfN to
tðiÞ m
the impact normal velocity, viN.
pN On Fig. 9b, for an elastic impact the two areas before and
vN ¼ vi þ
m after maximum compression are identical (eN ¼ 1) shown by the
where dashed curve, while for a perfectly plastic impact no energy is
recovered (eN ¼ 0).
viN o0 ð3Þ
The principle of separating the compression and restitution
The impacting normal velocity is negative because, as shown in phases of impact can also be demonstrated on a [normal impulse
Fig. 8, the positive normal axis is in the direction away from the (pN) – relative velocity (v)] plot as shown in Fig. 10. On this plot,
point of impact. the tangential velocity vT decreases continuously from viT at the
Eq. (3) shows that the normal component of the relative point of impact, to vfT at the point of separation, as the result
velocity is a linear function of the normal impulse. This relation- of frictional resistance on the contact surface, as discussed in
ship between v and pN is the basis for determining the changes in Section 3.4 below. The normal velocity also changes during
impulse during impact, and finding the final impulse pf at the impact, starting with a negative value (–viN) at the point of impact,
termination of the impact process (t ¼ f). decreasing to zero at the point of maximum compression
pcN , and finally reaching a positive value (vfN) at the point of
3.2. Changes in velocity and impulse during impact separation.

The impact process results in compression δ of the deformable 3.3. Coefficient of restitution
particle during the compression phase, followed by expansion
during the restitution phase. The changes in the normal contact The [pN – v] plot in Fig. 10 shows changes in both the normal
force F during impact are illustrated in Fig. 9a where the force (Fc) and tangential velocities, and the magnitudes of the work of
176 D.C. Wyllie / International Journal of Rock Mechanics & Mining Sciences 67 (2014) 170–180

Fig. 10. Relationship between normal impulse pN and tangential and normal relative velocities vT, vN, and energy changes, during impact; EcN is the kinetic energy absorbed
during the compression phase of impact (t ¼ c); (EfN–EcN) is the strain energy recovered during the restitution phase (t ¼ c to t ¼ f).

internal energy of deformation generated during impact. The to the impact normal velocity, and is also the square root of the
triangular area EcN represents the work absorbed in compressing negative ratio of the energy recovered during the restitution phase
the deformable region of the contact area and is given by of the impact to the energy lost during the compression phase. For
Z pcN Z pcN an impulse dependent analysis of rock falls, the normal coefficient
p
EN ðpcN Þ ¼ vðpN Þ dpN ¼ ðviN þ N Þ dpN of restitution is independent of friction and the process of slip.
0 0 m
In this paper, the expressions for the normal and tangential
1 p2cN
¼ viN U pcN þ : coefficients of restitution (eN and eT) as shown in Eqs. (1a) and (1b)
2 m
and Fig. 10 are used to define respectively the changes in normal
1 2
¼  :m UviN ð4Þ and tangential velocities during the impact process, while Eqs.
2
(4)–(8) show the relationship between velocity changes and
where pcN ¼  m  viN. energy losses. The equivalence between these equations demon-
During the restitution phase of the impact, some work strates how velocity changes can be used to quantify rock fall
(EfN – EcN) is recovered that is equal to the elastic strain energy impact behavior, and compare impacts that occur in differing
released, and is given by geological and geometric conditions. The value of the tangential
Z pf N Z pf N coefficient eT is related to the friction at the contact point as
p
EN ðpf N Þ  EN ðpcN Þ ¼ vðpN Þ dpN ¼ ðviN þ N Þ dpN discussed in Sections 3.4 and 3.5, while the normal coefficient eN is
pcN pcN m
 2 shown in Section 3.6 to have values that depend on impact angle
1 p fN
¼  :m:v2iN : 1  ð5Þ and normal force.
2 pcN
The concept of the coefficient of restitution was first developed
where viN o 0. by Isaac Newton [18] who suspended identical spheres on pendu-
These expressions for the partially irreversible changes in lums and measured how high they rebounded after collision; the
energy that occur during impact can be used to define the normal measurements included corrections for loses due to air friction. It
coefficient of restitution, eN, as follows: was assumed at the time of these experiments that the coefficients
EN ðpf N Þ EN ðpcN Þ of restitution were material properties. However, study of rock fall
e2N ¼  ð6Þ behavior shows that reductions in velocity during impact between
EN ðpcN Þ
rough, rotating bodies of different materials depend not only on
This definition of the normal coefficient of restitution in terms the materials forming the bodies, but also on the impact condi-
of energy separates the energy loss due to hysteresis of the contact tions such as impact angle, shape of the bodies and rotational
forces from that due to friction between the colliding bodies. velocity.
As shown in Fig. 9, the value for eN can range from eN ¼ 1 for a The coefficients of restitution eT and eN can be used to quantify
perfectly elastic material with no loss of energy during impact, to velocity changes during impact and to understand how conditions
eN ¼ 0 for a perfectly plastic material where no separation occurs at the impact points influence rock fall behavior. In Section 2, the
during impact with no recovery of the initial kinetic energy. diagrams showing vectors of the velocity components at the
The relationships shown in Eqs. (4),–(6) can be combined to various field sites, also show calculated values of eT and eN at
find the following expression for the final normal impulse, pf selected impact points. In Section 4 below, the values of these
pf ¼  m U viN ð1 þ eN Þ ¼ pcN ð1 þ eN Þ ð7Þ restitution coefficients are analyzed to show how they are related
to site conditions
and for the coefficient of restitution,
vf N 3.4. Friction and rotational velocity changes during impact
eN ¼  ð8aÞ
viN
The normal impulse (pN) – relative velocity (v) plot in Fig. 10
and
shows that the tangential velocity decreases during impact from
ðpf N  pcN Þ viT to vfT. This reduction in velocity can be partially attributed to
eN ¼  ð8bÞ
pcN the friction force acting at the contact between the impacting body
As shown in Fig. 10 and expressed in Eqs. (8a) and (8b), the and the slope, and partially to the irregularities of the slope at
normal coefficient of restitution is the ratio of final normal velocity the impact point. Furthermore, the reduction in velocity can be
 D.C. Wyllie / International Journal of Rock Mechanics & Mining Sciences 67 (2014) 170–180
quantified according to Coulomb’s definition of friction, in which
the coefficient of friction μ is the ratio of the tangential to normal
impulses acting at the contact point. That is, the change in normal
impulse δpN during impact can be related to the tangential impulse
δpT as follows:
δpT ¼  μ U δpN
The ratio of the tangential to normal impulses expressed in Eq.
(9) is a constant such that the friction coefficient is also a constant,
and is independent of the sliding speed and the normal force.
A further effect of friction during impact is its influence on the
rotational velocity of a spinning body. As shown in Fig. 11, when a
rotating body impacts the slope, a frictional force (μpN) acting
upslope is generated as a resisting force at the contact point. The
combination of the frictional force (acting upslope) and the
translational velocity (acting downslope through the center of
gravity) generates a moment within the body that tends to change
the rotational velocity during impact. For a sphere rotating with
initial angular velocity (–ωi) as shown in Fig. 11, the translational
and rotational velocities combine with the friction at the impact
point to increase the angular velocity during impact (–ωs). How-
ever, for irregularly shaped, rotating blocks impacting the slope
at a variety of attitudes, the velocity vectors may combine under
some conditions to reduce the velocity during impact. This
accounts for rotational velocities varying with successive impacts.
For example, at the Ehime test site, the rotational velocities of
about 100 test rocks on the 42 m high slope varied between 6 and
33 rad s  1.
Fig. 11 also shows how changes in the tangential and rotational
velocities during impact can be quantified on the [normal impulse
(pN) – relative velocity (v)] plot. For a spherical body with radius r
and rotating at angular velocity (–ω), the velocity at the periphery
of the body (r Uω) is parallel but opposite in direction to the
translational tangential velocity vT. If the magnitudes of (r Uω) and
vT are unequal, then slip will occur between the moving body and
the stationary ground, with the slip velocity vS being given by
vS ¼ vT –ðr U ωÞ
The frictional force (μp) that is generated at the contact between
the body and the slope and acts upslope for the rotational direction
shown in Fig. 11, has the effect of both increasing the rotational
velocity from (–ωi) at impact, and decreasing the tangential velocity.
Depending on the details of the impact process, the values of (r ω)
and vT may equalize during impact, in which case slip will cease at
impulse ps and from this point the body will roll with no change in
angular velocity. In other conditions related to how the irregular
body impacts the slope, the rotational velocity may increase during
impact and slip continues throughout the impact.
The change in rotational behavior during impact from slip to
rolling at impulse ps also changes the rate at which the tangential
Fig. 11. Change in rotational velocity during impact and transition from slip (t ¼ s) to rolling mode (t ¼ s to t ¼ f).
177
velocity decreases. That is, during the slip phase the friction
force results in a relatively rapid decrease in vT, compared with
the rolling phase when friction effects are diminished. This
behavior is illustrated in Fig. 11 by the different in gradients of
the [pN – vT] lines.
The effect of friction on the tangential and rotational velocities
ð9Þ
during contact can be understood by considering that for a
frictionless, smooth contact, no change in either velocity would
occur because no shear resistance is generated at the contact
surface. That is, the [pN – vT] and the [pN – vs] lines would be
horizontal.
Another characteristic of a rotating body is that the rotational
velocity only changes during impact as the result of the frictional
force that acts during contact between the body and the slope.
However, during the trajectory phase of the fall when contact
between the body and the slope ceases, no forces act to change the
rotational velocity and this remains constant during the trajectory.
This also means that the rotational energy remains constant
during the trajectory.
3.5. Tangential coefficient of restitution
Fig. 11 illustrates the relationships between vT and vs that can
develop during impact. Because it is difficult to determine these
relationships for actual impacts, it is useful to simplify the impact
process by just comparing the initial and final tangential velocities,
viT and vfT . As shown in Eq. (1b), the ratio (vfT / viT) defines the
tangential coefficient of restitution, eT. Since the gradient of the
[pN – vT] line is negative, the value of eT will be less than 1.0, and
will also be an approximate value for the coefficient of friction
since this parameter has a significant role in reducing the
tangential velocity during impact.
Section 2 shows the calculated values for eT at selected impact
points for the different slope materials examined in the field
studies, while Section 4.1 discusses the calculated eT values and
how they relate to the field conditions.
ð10Þ Most of the programs used to model rock fall behavior use the
tangential coefficient of restitution as one of the input parameters.
The results of the field studies provide guidelines on appropriate
values for eT to use in modeling rock falls.
3.6. Relationship between impact angle and normal
coefficient of restitution
The [pN – v] diagram can be used to examine the effect of the
impact angle θi on the impact process for a rotating body as shown
in Fig. 12. The two dashed lines in Fig. 12 show how the normal
component of the velocity vN changes from a negative value at the
point of impact (–viN), to zero at the point of maximum compression
178 D.C. Wyllie / International Journal of Rock Mechanics & Mining Sciences 67 (2014) 170–180
(pcN), to a positive value vfN at the completion of the impact process.
The slope of the line representing this change in velocity is a
function of both the size and shape of the particle, and the contact
friction.
For shallow impacts for which θi´ is small (upper dashed line)
and a “light” impact occurs, the normal component of the impact
velocity is smaller than the normal component of the final velocity
(viN´ o vfN´) and the value of the normal coefficient of restitution
eN is greater than 1. In contrast, for a steep impact for a larger value
of θi (lower dashed line) where a “heavy” impact occurs, the value
of the normal impact velocity is greater than the final normal
velocity (viN 4 vfN) and the normal coefficient of restitution is less
than 1. The intersection point of the two lines with the impulse
axis shows that the value of pcN′ for the shallow, light impact is less
than the value of pcN for the steeper, heavy impact, consistent with
the observation that the less indentation of the slope occurs with
the shallow impact.
A value for eN that is greater than 1 means that, for the
particular impact geometry, the final normal velocity (vfN) is
greater than the impact normal velocity (viN). A value of eN greater
than 1 does not mean that energy is created during the impact
process because the ratio of the final velocity to the impact
velocity (vf / vi) is less than 1 demonstrating the overall loss of
kinetic energy that occurs [20,21]. In Section 4 below, the actual
values of eN from the field data are plotted to show the relation-
ship between eN and the impact angle θi (see Fig. 14).
In Fig. 12, the two dashed lines are parallel because the impacts
are for identical conditions – same shaped body, materials and
rotational velocity – except for the impact angle, demonstrates
that the coefficient of restitution varies with impact conditions
and is not a material property.
4. Rock fall modeling parameters
The discussion of impact mechanics in Section 3 above illus-
trates how modeling of rock falls requires values for tangential
and normal coefficients of restitution that are applicable to the
Fig. 12. Effect of impact angle θi on the normal coefficient of restitution, eN.
particular site conditions. Section 3.4 showed how the reduction in
tangential velocity defined by the tangential coefficient of restitu-
tion, eT is related to the friction force generated between the slope
and the body, while Section 3.6 showed how the reduction in
normal velocity defined by the normal coefficient of restitution eN
is related to the impact angle.
This section of the paper shows the actual values of eT and eN
for the rock fall sites documented in Section 2 above are related to
the theory of impact mechanics discussed in Section 3. In sum-
mary, the field data have been analyzed for five rock fall sites
encompassing four different slope materials and 57 impacts.
The sixth location was a laboratory test where blocks of rock
were dropped on a concrete floor and the rebounce heights were
measured.
4.1. Tangential coefficient of restitution and friction
The reduction in tangential velocity during the impact process,
as shown on the [pN – v] plot in Fig. 10, can be quantified in terms
of the tangential coefficient of restitution, eT. The values of eT
measured at the five rock fall locations described in Section
2 above have been plotted in Fig. 13 showing a total of 57 impact
points for rock (12 points), talus (one point), colluvium (43 points)
and asphalt (one point). The plot also lists the average eT values for
the rock and talus slopes.
Fig. 13 shows considerable scatter in the eT values, as would be
expected for the site conditions where the slope surfaces are
rough and the rock falls are irregular blocks. In fact, no significant
difference is evident in the values of eT for rock, talus and
colluvium. The lowest value for eT of 0.24 is that for asphalt, and
it is reasonable that this relatively smooth surface will have a
lower friction angle than the rougher, natural slope surfaces.
The values plotted in Fig. 13 appear to be independent of the
impact velocity and angle, which is consistent with the Coulomb’s
law of friction in which the friction coefficient is independent of
velocity and normal force.
As a comparison with the field values of eT, direct measure-
ments have been made of friction for blocks of rock sliding on
 D.C. Wyllie / International Journal of Rock Mechanics & Mining Sciences 67 (2014) 170–180
1.2
1
Tangential coefficient
0.8
of restitution, eT
0.6
0.4
0.2
0
0 20 40 60 80
Calculation #
Fig. 13. Values for tangential coefficient of restitution eT for the rock fall sites
described in Section 2.
various geological materials [22]. These tests involved pulling a
natural block of rock with a mass of 433 kg on surfaces comprising
concrete, a gravel road and soil. The friction coefficient was
calculated from the shear force that was recorded by a load cell
on the pulling cable, and the shear displacement that was
measured with a laser displacement meter. It was found that the
average friction coefficients μ for these materials were
μconcrete ¼ 0:59; μgravel ¼ 0:68; μsoil ¼ 0:90:
These measured values of the friction coefficient are compar-
able to the field values plotted in Fig. 13 considering the wide
scatter in the results. The lowest value for the friction coefficient is
that of the smoothest surface, concrete.
4.2. Normal coefficient of restitution
The [pN – v] diagram in Fig. 12 shows how impact mechanics
defines the change in the normal velocity during impact, and how
the value of the normal coefficient of restitution eN varies with the
impact angle, yi. To find the actual relationship between eN and yi,
an analysis of the field data discussed in Section 2.0 has been
carried out, the results of which are summarized in Fig. 14. The
data plotted on Fig. 14 are the average values for the rock, talus
and colluvium, as well as values for concrete and asphalt.
The field measurements show that the lowest value of eN is 0.18
for the laboratory test where the block of rock was dropped in a
concrete surface, i. e., yi ¼ 90° [22]. It can be assumed that similar
tests for less dense materials such as gravel and colluvium, values
for eN would be less than that for concrete, while eN values for
impacts on strong rock may be close to 0.18.
The data plotted on Fig. 14 shows how the value of eN decreases
as the impact angle yi increases from very shallow impacts (low
values of yi) to a minimum value of about 0.1 for normal impacts.
For shallow impacts, i.e., yi less than about 20°, the value of eN is
greater than 1. As discussed in Section 3.6 above and demon-
strated in Fig. 12, values of eN greater than 1 does not mean that
energy is being created during impact, but only that the impact
geometry causes the final normal velocity to be greater than the
impact normal velocity.
The symbols in Fig. 14 show [eN - yi] values for the five different
slope materials listed in the legend. It is noted that values for each
slope material are not grouped together, but tend to be follow the
best fit line according to the value of the impact angle. This would
indicate that eN is more related to the impact geometry than the
slope material.
In Fig. 14, the best fit line for the field data has been plotted
showing that the relationship between the normal coefficient of
179
Fig. 14. Relationship between the impact angle θi and the normal coefficient of
restitution eN for the rock fall sites described in Section 2.0.
restitution and the impact angle is given by
eN ¼ 19:5 ðyi Þ1:03 ð11Þ
with a correlation coefficient for this power equation of r ¼ -0.48.
That is, Fig. 14 indicates that values of eN depend more on the
impact geometry than on the properties of the slope material.
The theory of impact mechanics may be used to examine the
relationship between eN and the impact angle for eccentric impact
conditions. That is, for a non-spherical, rotating body impacting
the slope at an oblique angle where the equations of motion each
involve both normal and tangential forces, the effects of friction
and normal forces are not separable (Stronge 2000). Goldsmith
(1960) presents a graphical solution for eccentric impact that
relates the normal and tangential impulses throughout the dura-
tion of the impact to the friction coefficient, the normal coefficient
of restitution and the impact angle.
The field data showing the conditions for which eN is greater
than 1 are supported by findings of others [23,24,25].
5. Conclusions
Careful documentation of rock falls in a variety of geologic and
topographic conditions has provided data on velocity and energy
changes during the impact process. Data has been collected for
sites where rock falls have occurred on rock, talus and colluvium
slopes, and on asphalt, and for laboratory tests on soil, gravel and
concrete. The calculated normal and tangential velocity compo-
nents at each impact point have been used to determine the
normal and tangential coefficients of restitution for the different
slope materials. The results show that the tangential coefficient of
restitution (eT) is related to the friction coefficient of the surface
and values of eT vary between 0.2 and 0.9, with little correlation
between the type of surface; a smooth asphalt surface had the
lowest eT of 0.24.
With respect to the normal coefficient of restitution (eN), it has
been found that values of this parameter are correlated with the
impact angle. For near normal impacts (y0 ¼ 50–90°), the value of
eN is about 0.1–0.2, corresponding to a drop test on a horizontal
surface. For shallow impacts (y0 ¼ 10–25°), the value of eN often
exceeds 1.0 for oblique impacts of rough, non-spherical bodies.
Values of eN greater than 1 do not mean that energy is being
created, but that the combination of the impact angle and rotation
are producing normal velocities that are greater after impact than
before impact; overall energy loss occurs at every impact. The
180 D.C. Wyllie / International Journal of Rock Mechanics & Mining Sciences 67 (2014) 170–180
results also show that values of eN are generally independent of
the slope materials.
The theory of impact mechanics for impact between a translat-
ing and rotating, rough body and a stationary slope relates the
impact and final velocity components to the coefficients of
restitution and friction.
Acknowledgments
The author acknowledges the support of the Canadian Pacific
Railway for providing access to the Tornado Mountain and Mount
Stephen sites, and for enabling the study of rock fall behavior and
design of protection measures at both these two sites, and many
other locations. This work over the past forty years has been vital
in developing a detailed understanding of rock falls in natural
conditions. In addition, the assistance of Mr. Thierry Lavoie and Mr.
Phillip Lesueur, while students at the University of British Columbia,
in collecting and analyzing field data, and Cheng-wen Tina Chen in
editing the manuscript, is greatly appreciated.
References
[1] Transportation Research Board (TRB). Landslides, Investigation and Mitigation.
National Research Council, Special Report 247: Washington DC; 1996.
[2] Kobayashi Y, Harp EL, Kagawa T. Simulation of rock falls triggered by earth-
quakes. J Rock Mech Rock Eng 1990;23(1):1–20.
[3] Jibson R.W., Harp E.L. Inventory of Landslides Triggered by the 1984 North-
ridge, California Earthquake. U. S. Dept. of the Interior, Open - file 95-213, 17
pages: USGS; 1995.
[4] Chau KT, Wong RHC, Liu J, Lee CF. Rockfall hazard analysis for Hong Kong
based on rockfall inventory. J. Rock Mech Rock Eng 2003;36(5):383–408.
[5] Ritchie A.M. An evaluation of rock fall and its control. Highway Research
Record 17, Highway Research Board, NRC, Washington DC NRC; 1963:13–28.
[6] Pierson L.A., Gullixson C.F., Chassie R.G. Rockfall Catchment Area Design Guide.
Research Report SPR-3(032): Oregon Department of Transportation – Research
Group: Federal Highway Administration; 2002.
[7] Japan Road Association Rock Fall Control Manual (in Japanese); 2000.
[8] Yoshida H., Nomura T., Wyllie D.C., Morris A.J. Rock fall sheds – application of
Japanese designs in North America. In: Proceedings of the first North
American Landslide Conference. AEG Special Publication No. 22., ed. Turner
AK, Schuster RL. Vail, CO., 3–8 June 2007. p. 179–96.
[9] Jones C, Higgins J, Andrew R. Colorado Rockfall Simulation Program, Version
4.0 (for Windows). Colorado Geological Survey; 2000.
[10] Komura T., Muranishi T., Nisizawa K., Masuya H. Study on parameters
concerning impact of falling rock on field slopes and rock fall simulation
method. In: Proceedings of the fourth Asia-Pacific conference on shock and
impact loads on structures. Singapore; 21–23 November 2011. p. 345–52.
[11] Nishikawa Y, Masuya H, Moriguti Y. Three dimensional simulation of rock fall
motion with consideration of roughness of the slope surface. Trans Japan Soc
Comput Eng Sci 2012;20120003.
[12] RocScience. Computer program RocFall. Toronto, Canada: RocScience Inc;
2011.
[13] Wyllie DC, Rock fall engineering, 2014. Reproduced with permission from Boca
Raton: CRC Press, 2014 (in publication).
[14] Ushiro T., Tsutsui H. Movement of rock fall and a study of its prediction. In:
Proceedings international symposium on geotechnical and environmental
challenges in mountainous terrain. Kathmandu, Nepal; 6–7 November
2001:366–75.
[15] Ushiro T, Kusumoto M, Shinohara S, Kinoshita K. An experimental study
related to rock fall movement mechanism (in Japanese). J Jpn Soc Civil Eng F
2006;62(2):377–86.
[16] Goldsmith W. Impact – the theory and physical behaviour of colliding solids.
2001 edition. New York: Dover Publications; 1960.
[17] Stronge WJ. Impact mechanics. U.K. Cambridge: Cambridge University Press;
2000.
[18] Newton I. Philosophiae naturalis principia mathematica. Cambridge, U.K.:
Cambridge University Press; 1687.
[19] Transportation Research Board (TRB). Rockfall – characterization and control.
National Research Council, Washington DC, editors, Turner K.A, Schuster RL,
658 pages.
[20] Chau KT, Wong RHC, Wu JJ. Coefficient of restitution and rotational motions of
rock fall impacts. Int J Rock Mech Mining Sci 2002;39:69–77.
[21] Buzzi O, Giacomini A, Spadari M. Laboratory investigation of high values of
restitution coefficients. J Rock Mech Rock Eng 2012;43:35–43.
[22] Masuya H., Ihara T., Onda S., Kamijom A. Experimental study on some
parameters for simulation of rock fall on slope. In: Proceedings of the fourth
Asia-Pacific Conference on shock and impact loads on structures. Singapore;
21–23 November 2011. p. 63–9.
[23] Giacomini A, Spadari M, Buzzi O, Fityus SG, Giani GP. Rockfall motion
characteristics on natural slopes in eastern Australia. Eurock 2010 (Lausanne,
Switerland, 15 to 17, June).
[24] Asteriou H, Saroglou G, Tsiambaos T. Geotechnical and kinematic parameters
affecting the coefficient of restitution for rock fall analysis. Int J Rock Mech
Mining Sci 2012;54:103–13 (September).
[25] Spadari M, Giacomini A, Buzzi O, Fityus SG, Giani GP. In situ rock fall testing in
New South Wales, Australia. Int J Rock Mech Mining Sci 2012;49:84–93.