Document Actions

IJEGE-11_BS-Kayley-et-alii

background image
Italian Journal of Engineering Geology and Environment - Book www.ijege.uniroma1.it © 2011 Casa Editrice Università La Sapienza
339
DOI: 10.4408/IJEGE.2011-03.B-039
MODELLING DEBRIS FLOW PROCESSES WITH
A GEOTECHNICAL CENTRIFUGE
P. KAILEY
(*)
, E.T. BOWMAN
(*)
, J. LAUE
(**)
& S.M. SPRINGMAN
(**)
(*)
University of Canterbury, Christchurch, New Zealand
(**)
Institute for Geotechnical Engineering, ETH Zurich, Switzerland (*) Affiliation
and temporally as a debris flow travels down its path,
which complicates understanding and modelling the
mechanics of debris flow motion. Even if one is lucky
(or unlucky) enough to be able to observe a debris
flow event in the field, the boundary conditions and
key parameters influencing flow behaviour may be
difficult or even impossible to measure.
Physical modelling simplifies these processes and
allows boundary conditions to be controlled in the
laboratory, without preconditioning the outcome. This
has made small-scale flume studies of debris flows
an indispensible tool in elucidating some key aspects
of debris flow mechanics. However, there are some
drawbacks to flume studies at the small scale. The
extrapolation of small-scale behaviour to field scale
processes may not always be appropriate, as small
scale flows may not reflect the dominance of Coulomb
stresses and decreasing importance of viscous stresses
in field scale flows (d
enlinGeR
& i
veRson
, 2001). In
previous work, Bowman et alii (2010) showed that the
centrifuge can match particular aspects of large-scale
behaviour well, which would be difficult at the small
scale at 1-g. In particular, the centrifuge tests repro-
duced low Savage and Pore Pressure numbers, ensur-
ing that centrifuge flows are in the frictional regime
and can maintain persistently high pore-pressures, like
large scale flows. The other principle advantage of
centrifuge testing is that they can potentially be more
convenient and flexible than large scale experiments,
as flume geometry can be altered. They also require
ABSTRACT
In this paper, we examine the effect of flow mass
and moisture content on debris flow velocity, dis-
charge, and runout using a series of smallscale flume
tests in a geotechnical centrifuge. We found that an
increase in mass and an increase in moisture content
increased peak velocity during down-slope move-
ment. However, the effect of increased moisture con-
tent is much more pronounced than that of increased
mass. The maximum cross-sectional area observed
did not depend on mass or moisture content, although
may have been affected by the flow rate entering the
centrifuge. Consequently, flow velocity largely deter-
mined the peak discharge of each flow. An increase
in moisture content increased the mobility of the flow
in terms of depositional area and runout. Further, the
runout of the centre of mass of the flows appears to
be linearly related to the momentum of flow material
entering the flume.
K
ey
words
: debris flows, physical modelling, runout, veloci-
ty, centrifuge tests
INTRODUCTION
PHYSICAL MODELLING OF DEBRIS FLOwS
The highly complex, stochastic nature of debris
flows is a direct result of the synergistic interaction
between their fluid and solid phases. Key debris flow
parameters such as particle size distribution, mois-
ture content, velocity, and discharge vary spatially
background image
P. kAILEY , E.T. BOwMAN, J. LAUE & S.M. SPRINGMAN
340
5th International Conference on Debris-Flow Hazards Mitigation: Mechanics, Prediction and Assessment Padua, Italy - 14-17 June 2011
Moisture content within the flow can be just as im-
portant as event volume. For example, in field investi-
gations in the Dolomites and small-scale flume inves-
tigations, moisture content appeared to control runout,
almost regardless of event volume (d’a
Gostino
et
alii, 2010). t
akaHasHi
(2007) also discusses the im-
portance of moisture content in controlling the veloc-
ity distribution of particles with depth.
Preliminary results from five debris flow tests are
presented. Three tests were run using different masses
of solid material (1, 1.75, and 2.5kg) with a uniform
moisture content (by mass) of 33%. Two final tests
were run using 1.75kg of material, with moisture con-
tents of 39% and approximately 41%, respectively.
The purpose of these last tests was to investigate
the influence of moisture content (Table 1). The mois-
ture content in T11 is only approximate. The first at-
tempt at T11 became clogged in the feeder tube. Since
centrifuge time, and time to prepare the material, was
limited, the test was rerun by adding the approximate
volume of fluid lost in the first attempt. While the
range of moisture content tests varies by only 8%, the
effect on flow behaviour, as discussed below, is dra-
matic. This range of moisture content was also con-
venient to test, as at moisture contents below 33%, the
flow consolidates very quickly. Above 41%, it became
more difficult to manage as more material was lost
during the transition from the feeder tube to the head
of the flume, as well as during collection.
CENTRIFUGE TESTING
APPARATUS AND INSTRUMENTATION
Details regarding the design and instrumentation
used in the experiments have been discussed previously
(b
owman
et alii, 2010). However, a brief description
of the apparatus and some minor changes to the flume,
relevant to the following discussion, is given below.
Experiments were carried out using the ETH Zu-
rich Geotechnical Drum Centrifuge in Switzerland
(s
PRinGman
et alii, 2001). This centrifuge has a maxi-
mum working radius of 1.1m, a maximum design ac-
celeration of 440g and a maximum load carrying ca-
pacity of 2000kg.
The debris flow apparatus was designed to guide
liquefied debris flow material from its head to the in-
ner circumference of the centrifuge drum. The drum
circumference itself was used as the runout zone – i.e.
where the flow comes to rest. Several holes located
less material than large scale flume tests.
The aim of the experiments presented in this pa-
per was to extend the work of b
owman
et alii (2010)
by investigating theinfluence of flow mass and mois
ture content on debris flow behaviour in the centri-
fuge, as these parameters are considered key to the
development of debris flow velocity, discharge, and
runout. It also addresses some experimental difficul-
ties and recommendations for improvement in future
centrifuge studies.
It should be noted that, coming from a geotech-
nical perspective, we use moisture content by mass,
rather than solids concentration by volume as often
used elsewhere in the debris flow literature. The rela-
tionship between them is:
Where C
v
is the solid concentration by volume, w
is the moisture content by mass and Gs is the specific
gravity of solids, taken here to be 2.65. A moisture
content of 33% or 0.33 corresponds to a solids con-
centration of 0.53 by volume.
Debris flow volume is often cited as the most
critical parameter in estimating debris flow hazard,
as larger flows travel faster and farther than smaller
flows, both at the laboratory and field scale. This is
likely due to be due to the prevalence of high pore
pressures which are more likely to be maintained in
a thicker flow due to longer drainage paths (b
owman
et alii, 2010). Previous work has shown that the peak
discharge of the flow can be related to the debris flow
volume (R
iCkenmann
, 1999).
Tab. 1 - Test code and short description of each test. All
tests were conducted over a fixed bed, using a mix-
ture of glycerine and water as a viscous pore fluid
(μ=42cP, or 42 time the viscosity water at 20ºC).
Moisture content was calculated by (mass of pore
fluid/mass of solid)× 100
background image
MODELLING DEBRIS FLOW PROCESSES WITH A GEOTECHNICAL CENTRIFUGE
Italian Journal of Engineering Geology and Environment - Book www.ijege.uniroma1.it © 2011 Casa Editrice Università La Sapienza
341
Unconsolidated debris flow material was intro-
duced “in flight” to the channel by a flexible tube. The
tube extended from the central axis of the centrifuge,
where material was delivered via a funnel, and was
guided by an actuator on the centrifuge tool plate to
the head of the channel, where it exited to flow out-
ward under centrifugal acceleration, down the slope.
This system enabled the material to be prepared and
maintained as a slurry external to the drum (in which
it would otherwise consolidate during spin-up).
After each test, measurements were taken of the
maximum runout and lateral spread of each flow. In
addition, markers running vertically and horizontally
on the drum surface were used as a grid to record spot
depths of the flow deposition (Figure 1). These data
were then used to compare the morphology of deposi-
tion and runout.
CENTRIFUGE SCALING PRINCIPLES
Scaling relationships for geotechnical centrifuge
modelling are shown in Table 2 (Steedman & Zeng,
1995). Note that inertial effects (which scale to 1/N)
and diffusional effects (which scale to 1/N2) scale
differently over the same time period. To resolve this
inconsistency, the prototype pore fluid (assumed to be
water) is replaced with a higher viscosity pore fluid,
which inhibits consolidation, allowing pore pressures
to develop as at prototype scale. A glycerine and water
mixture with viscosity of the pore fluid of approxi-
mately N times higher than water (1cP) is usually cho-
sen. This reduces the time for consolidation by N2 and
inertia by N in the model, resulting in the same over-
along the circumference of the drum allowed fluid to
drain from the consolidating debris flow.
The debris flow flume apparatus consisted of a
channel, a strut and a curved support to spread load to
the drum. The 700mm long flume followed the inner
curvature of the drum, such that, at a slope angle of 0°,
it would lie evenly along the drum circumference. The
flume width in this round of tests was decreased from
160mm, as used in previous tests, to 60mm to provide
increased channelisation, and hence increased flow
depth, and velocity towards values more representa-
tive of field scale flows than those obtained previously.
The confinement ratio (flow height/width of channel)
observed in these tests was approximately 4. Previ-
ous field studies in coarsegrained debris flows have
found the onset of deposition to occur at confinement
ratios of 5 to 7 (H
unGR
et alii, 1984; k
inG
, 1996). The
increased flow depth combined with a coarser particle
distribution enabled individual particles to be tracked
in the high speed camera images, as discussed later.
Six pore pressure transducers (PPTs) were pro-
vided along the base of the flume for the measurement
of pore pressure during the experimental debris flows.
Coarse sand particles glued to the base provided a
rough substrate; the smooth aluminium and Perspex
walls ensured relatively plane strain behaviour.
A small monochrome high-speed digital camera (op-
erating at 330 frames per second) was used to observe
the flow through the clear, Perspex channel wall. Small
markers were painted on the window to provide scale and
reference points to track the flow. The flow was lit by a
close array of 8 LEDs embedded in the Perspex window.
Fig. 1 - Photograph of experimental set-up and deposition
from test T14, which has consolidated on the drum
surface after spin-down. The markers in the top of
the photograph serve as reference points for point
measurements of depth. Note that the photo was
taken obliquely, the deposition is plastered verti-
cally to the inner wall of the drum
Tab. 2 - Scaling laws used in geotechnical centrifuge test-
ing, based on N×g = r×ω2, where ω is angular
velocity (STEEDMAN & ZENG, 1995). N is the
g-level, hence at a gravitation acceleration of 40g,
N=40
background image
P. kAILEY , E.T. BOwMAN, J. LAUE & S.M. SPRINGMAN
342
5th International Conference on Debris-Flow Hazards Mitigation: Mechanics, Prediction and Assessment Padua, Italy - 14-17 June 2011
(approximately 48% by weight) was collected from
the Mt. Thomas debris flow site in Northern Canter-
bury. This locality has been a site of ongoing debris
flow activity since 1977, when a series of debris flows
were triggered on recently harvested cut blocks (P
ieR
-
son
, 1980). The material from Mt. Thomas was sup-
plemented in the range of 0.6mm to 0.075mm with
Swiss fluvial material, since there was relatively little
fine sand and silt available from the Mt. Thomas mate-
rial. 41% of the PSD tested was made from the fluvial
material. The lighter colour of this sand also created
more texture in the high speed camera images, which
was useful for post-processing. The remaining 11% of
the mixture came from nonplastic loess collected from
slips in the central north island of New Zealand. This
provided the remainder of the, silt and minor clay in
the particle size distribution used in the tests (Figure
2). All material was carefully sieved, then mixed to
the desired particle size distribution before each test.
Experimental constraints limit the particle size
distribution tested in the centrifuge. The maximum
size particle used is limited by the internal diameter
in the feeder tube In this case, the internal diameter
of the feeder tube was 32mm and the maximum parti-
cle diameter was approximately 8mm. Particles larger
than this cause chronic mechanical arching and flow
blockage. Even with the PSD used, we did have a test
which clogged (test 11, as discussed previously). The
particle size distribution tested represents a compro-
mise between the largest possible particle size, repre-
sented by d
90
(the particle size at which 90% of par-
ticles by mass are smaller than) of 2mm, while still
allowing a relatively high value of Cu (d
60
/d
10
) of 36.7,
which is shown to be an important parameter in other
physical modelling studies of debris flow behaviour
(b
owman
& s
anvitale
, 2009).
all time for these processes as in the prototype with
water. This approach also means that the particle size
distribution (PSD) used in the experiments is the same
as the PSD at the prototype scale in terms of consoli-
dation behavior. In these experiments, all tests were
run at 40g with a 42cP pore fluid, due to the difficulty
of achieving an exactly 40cP solution. This resulted
in the prototype and model PSD shown in Figure 2.
The mass of solid material used in these tests was
varied from 1 kg to 2.5 kg, which corresponds to 40
kg to 100 kg at the prototype scale. The prototype
channel dimensions scale to 28 m long by 2.4 m wide.
Peak flow heights were recorded between 14 and 17
mm high, corresponding to a prototype flow height of
0.56 to 0.68 m.
While the prototype length scales and PSD come
close to replicating some small, field-scale debris flows,
this prototype was not chosen to replicate any particular
event and should still be considered highly idealised.
MATERIAL CONSTRAINTS
The material used in these tests was a mixture of
soil from three separate localities, two in New Zea-
land and one in Switzerland. The largest fraction used
Fig. 2 - Model (actual) and prototype PSD used in all
tests. All tests were run at 40g with a 42cP pore
fluid, shifting the prototype PSD slightly to the left
Fig. 3 - High speed camera images from T14, frames (a) 644, (b) 707, (c) 1501. Flow proceeds from left to right. The dot
spacing is 10 mm. The sequence shows (a) the arrival of the front, (b) thickening of the front, and (c) transition to the
watery tail portion of the flow
background image
MODELLING DEBRIS FLOW PROCESSES WITH A GEOTECHNICAL CENTRIFUGE
Italian Journal of Engineering Geology and Environment - Book www.ijege.uniroma1.it © 2011 Casa Editrice Università La Sapienza
343
Observations from the high speed camera also
enabled the surface velocity of the flow over time to
be determined. By tracking individual particles over
several frames (as many frames as each particle was
distinguishable), an instantaneous velocity of the par-
ticle could be determined. These velocity data were`
used to construct the debris flow hydrographs and ve-
locity profiles discussed later in the paper.
FLOw VELOCITY
Pore pressure transducers mounted in the flume
and high speed camera data were used to reconstruct
the velocity of the flows as they travelled down the
flume (Fig. 5). The locations of the data points shown
in Figure 5 are half way between the PPTs, which re-
corded the responses used to calculate the velocity.
Figure 5 also shows velocity recorded by the high
speed camera near the flume outlet. This camera ve-
locity was calculated by tracking how long it took the
flow to traverse the width of the camera frame (ap-
proximately 100mm). Mixing data-points calculated
from porepressure data with visual data is not ideal,
as the pore pressure data could have a small lag in
response time, especially as the PPTs may not regis-
ter an unsaturated, ‘bouldery’ front. The provision of
an additional, wider angle camera recording the flow
from above, or additional PPTs near the channel bot-
tom would solve this issue.
TEST RESULTS
HIGH SPEED CAMERA IMAGES AND FLOw
HEIGHTS
Images from the high speed camera gave a view
of the debris flow as it passed the Perspex window.
A fast, coarse, unconfined flow front dominated by
larger diameter solids preceded the peak discharge in
every flow (Fig. 3).
The surface of each flow was nearly always
slightly higher in the middle than on the edges because
larger particles were often carried in the center of the
flow and their edges would protrude from the surface.
While particles adjacent to the Perspex window were
in focus, particles near the centre were somewhat blur-
ry and indistin because of the limited depth of field
in the camera. This can be seen quite clearly in Fig-
ure 3. To explore the change in flow depth over time,
the flow heights, both at the free surface at the center
and against the window, were measured by use of the
high speed camera images. Measurements were taken
at least every four frames during front passage, then
every several hundred frames in the watery tail por-
tion of the flow when the rate of change of flow height
dropped significantly. These measurements were pre-
cise to within ±0.5mm. A cross-sectional area for each
frame was calculated from these data.
The depth of flow rapidly attenuated in all the
flows, and then slowly decreased in accordance with a
near-power law (Fig. 4). The transition to the tail por-
tion of the power-law plot roughly coincided with an
apparent sudden increase in moisture content (visible
as a change in texture of the soil against the window
and reduction in flow surface roughness), reflecting the
transition to the “watery tail” portion of the flow.
Fig. 4 - Change in flow depth with time for test T15.
Fig. 5 - Debris flow front velocity versus distance down
slope from light switch. T14, T15 and T20 have
varying masses of material at 33% moisture con-
tent, whereas T23 and T21 have a mass of 1.75kg
and greater moisture contents. The slope of the
flume is shown on the secondary x-axis. The mois-
ture content of T11(41%) is accurate to within
±1%, as discussed in the text
background image
P. kAILEY , E.T. BOwMAN, J. LAUE & S.M. SPRINGMAN
344
5th International Conference on Debris-Flow Hazards Mitigation: Mechanics, Prediction and Assessment Padua, Italy - 14-17 June 2011
Both debris flow volume and moisture content sig-
nificantly affected flow velocity. The tests conducted
with 33% moisture content all show the same general
trend; flow velocity increases to a point nearly half-way
down the flume, then begins to decelerate. As expected,
an increase in mass causes the velocity to increase. An
increase in moisture content, however, has a much more
profound effect on velocity than mass. Comparing T11
(1.75 kg at approximately 41%), and T14 (1.75 kg at
33%), an increase of 8% in moisture content increased
the peak velocity by a factor of 3.
All three flows at 33% moisture content began to
slow at a slope angle of approximately 21º (Figure 5),
while the flows at higher moisture contents show de-
celeration at around 15º.
These angles coincide with the range of slope
angles shown to be transitional between entrainment,
transport, and deposition in the field (f
annin
& w
ise
2001). In the Queen Charlotte Islands, f
annin
& w
ise
(2001) found that for coarse-grained, channelized
flows in unconfined reaches, slope angles between
19º and 24º were found to have both deposition and
entrainment. In confined reaches, both deposition and
entrainment occurred on slope angles between 10º to
22º Deposition was the dominate process for confined
reaches at slope angles of less than 10º.
VELOCITY PROFILES
By tracking particles at various depths as the flow
passed by the Perspex window, we attempted to re-
construct how velocity changed with depth as the flow
front passed. One profile was taken at the flow front,
while another was taken in the receding limb of the
flow hydrograph, before the transition into the much
finer, watery tail portion of the flow. Unfortunately,
the epoxy used to seal the flume, and occasional resid-
ual material from previous tests, obscured the deep-
est 3 to 4 mm of flow, preventing a complete velocity
profile to the base of the flume.
The flow front appears much faster and the veloc-
ity profile also less steep than that of the receding limb
of the flow in all tests.
Examples are shown for T14 and T11 in Figures 6
and 7. The different velocity profile between the flow
front and receding limb of the flow is due to two ef-
fects. The flow decelerates with time as is clear from
the reducing discharge with time (see below). How-
ever, the very slow velocities shown for particles
at depth after passage of the front is also caused by
Fig. 6 - Velocity profile with depth for T14 (1.75kg, 33%
moisture content). Frame 646 records the flow
front, while frame 707 captures the velocity m
profile in the falling limb of the flow hydrograph,
just after the peak flow height
Fig. 7 - Velocity profile for T11.Frame 296 records the
flow front, while 312 represents the falling limb,
just after peak flow height
Fig. 8 - Flow depths versus time showing the passage of
bouldery front and transition to finer, more watery
tail
background image
MODELLING DEBRIS FLOW PROCESSES WITH A GEOTECHNICAL CENTRIFUGE
Italian Journal of Engineering Geology and Environment - Book www.ijege.uniroma1.it © 2011 Casa Editrice Università La Sapienza
345
velocity. A hydrograph was plotted frame by frame for
each of the tests by multiplying this average velocity
by the cross-sectional area. The resulting hydrographs
were then integrated to calculate a total event volume.
The total event volumes calculated from the hy-
drograph do show some differences between input
and output measurements. The discrepancy between
the amount of material entering the feeder tube and
the amount of material recovered after the test was
significant (on average 15%, and up to 25% in T11).
This is mainly because not all of the material poured
into the funnel reached the depositional surface. Some
material remained in the feeder tube or was lost dur-
ing the transition from the feeder tube and the head of
the flume, especially in tests which were more fluid
because material splashed off the flume surface in the
abrupt transition between feeder tube and channel. In
addition, a small amount of material may have been
missed in collecting material from the drum surface.
Additionally, the hydrographs do not fully con-
tain the extent of each flow because they are based on
observations from the high speed camera, which was
only able to record the first 18 seconds per test due
to the memory constraints of the in-flight computer.
Therefore, small amounts of flow continued after the
last frame in each test, so that there was an unknown
(albeit small) amount of deposition, which occurred
after the camera had stopped recording that was not
taken into account in the hydrograph.
Despite these limitations, comparing Figures 8
and 9 leads to some interesting qualitative observa-
tions of flow behaviour. That is, given that volume and
moisture content in these tests had little effect on the
maximum flow height, and hence the cross-sectional
area, this suggests that the maximum cross-sectional
area of flow was largely limited by the maximum
flow rate available from the feeder tube. Given that
the maximum cross-sectional area of each of the tests
was similar, the velocity of the flow almost entirely
controlled peak discharge. In turn, Figure 9 shows that
a very small change in moisture content dramatically
increases the flow discharge via an increase in flow
velocity, independent of flow depth.
The stage hydrographs and velocity data show
some similarity to field and large scale flume tests,
as well as some differences. In a field-monitored de-
bris flow in the Illgraben catchment in Switzerland,
m
C
a
Rdell
et alii (2007) report velocities of 1.4 m/s
friction against the outer walls of the flume. In the
high speed camera images, flow at the centre, which
is not in contact with the side-walls, appears lighter
than particles pressed against the window, due to the
lighting set up. This enables the difference in surface
velocity between edge and centre to be examined. At
the very front of the flow, the flow margins were not in
contact with the flume walls. Hence, this represents a
velocity profile without the influence of friction from
the walls of the flume. In the receding limb, the par-
ticles tracked are sliding against the window or wall.
FLOw DEPTH AND DISCHARGE
The flow depth of each test was recorded versus
time and is shown in Figure 8. The most notable ob-
servation from these graphs is that maximum flow
depth appeared to be approximately the same from
test to test, irrespective of total flow mass and mois-
ture content. This is likely to be a result of the bound-
ary conditions as further discussed below.
Hydrographs were constructed for each test in order
to explore how the discharge of each flow changed with
time, as shown in Figure 9. Regression of flow depth
against velocity gave a moderately linear relationship
for each test. The resulting function was used to estimate
an instantaneous surface velocity for each frame, based
on the surface flow height observed in that frame.
Assuming the flow velocity decreases linearly
with depth at any point during the flow [a reasonable
first order approximation for a stony debris flow based
on t
akaHasHi
(2007) and Figg. 6 and 7], the average
flow velocity should be half of the observed surface
Fig. 9 - Hydrographs: calculated discharge versus time
for each debris flow test
background image
P. kAILEY , E.T. BOwMAN, J. LAUE & S.M. SPRINGMAN
346
5th International Conference on Debris-Flow Hazards Mitigation: Mechanics, Prediction and Assessment Padua, Italy - 14-17 June 2011
for a gently sloping channel of 5 to 10º. Maximum
flow depths were just over 1m. Video recordings indi-
cated a strong conveyor belt-like circulation of parti-
cles and surface velocities were approximately twice
the average velocity recorded in the camera imagery,
closely resembling the behaviour in the centrifuge.
In large scale flume experiments, d
enlinGeR
&
i
veRson
(2001) show stage hydrographs with peak
flow heights of 0.2 m at 33 m from the head of the
flume, reducing to 0.1 m at 67 m, compared to peak
flow depths in our experiments of approximately
15mm (corresponding to 0.6 m at prototype scale) re-
ducing to 5 mm (0.2 m at prototype scale) in the tail.
While the shape of the peak response of the stage hy-
drograph is similar, the its trailing end differs signifi-
cantly from the centrifuge tests. While the centrifuge
flows display a thick, long running watery tail portion
of the flow, the flow height in the large-scale flume
tests quickly diminishes. This can be explained by
the differing boundary conditions between tests. The
large flume flows were triggered by a sudden release
of a block of material, while material in the centrifuge
tests was released more gradually via the feeder tube
The velocity in the USGS flume studies is mark-
edly higher than in the centrifuge, at up to 10m/s
(d
enlinGeR
& i
veRson
, 2001). This can be attributed
to the much steeper, homogenous slope of the flume
(31º) over its 95 m length. The centrifuge flows have
much less time and length to accelerate and much
less time for waveforms to accelerate, elongate, and
extenuate when compared to the longer flume USGS
flume (d
enlinGeR
& i
veRson
, 2001).
RUNOUT AND DEPOSITION
Point measurements of depth were used to con-
struct contour plots of the deposition (Figg. 10 and
11). The morphology of deposition was strikingly
similar in all tests. In contrast to many debris flows
in the field and in large flume experiments (e.g. d
en
-
linGeR
& i
veRson
2001), the width of lateral spread
exceeded the runout of deposition. This is likely to
be due to the rapid deceleration of the flow within
the channel before opening to a horizontal deposition
zone. In these experiments, unlike many other flume
studies, the slope angle was continuously reducing
from the head, at 36º, to the base, at near 0°. This
means that the flow in all tests was moving relatively
slowly upon reaching the deposition zone. In small
Fig. 10 - D e p o s i t i o n
of T15 (1kg),
T14 (1.75kg)
and
T20
(2.5kg)
All
flows run at
33%
mois-
ture content.
Contour in-
tervals
are
0.2cm. Depth
i n c r e a s e s
from cool to
hot
colors.
The maximum
contour is
3.0cm in T20
Fig. 11 - D e p o s i -
tion of T23
( 1 . 7 5 k g ,
39% MC),
and T11
( 1 . 7 5 k g ,
approxi-
m a t e l y
41% MC).
background image
MODELLING DEBRIS FLOW PROCESSES WITH A GEOTECHNICAL CENTRIFUGE
Italian Journal of Engineering Geology and Environment - Book www.ijege.uniroma1.it © 2011 Casa Editrice Università La Sapienza
347
squared and the runout for the flows at a moisture con-
tent of 33% (Figure 12), this relationship breaks down
when flows at higher moisture contents are included.
The flows with higher moisture contents run out far
less than predicted, according to a linear squared ve-
locity-runout relationship.
A better predictor of runout for the test flows
conducted at 33% was the mass entering the flume
(Figure 13). However, this was not entirely consistent.
Flows undertaken at different moisture contents plot
with different relationships.
Runout to the centre of mass of the flows in-
creased with both moisture content and flow volume
and showed a linear relationship (Figure 14) with in-
flume experiments, b
owman
& s
anvitale
(2009)
found that deposition morphology (in terms of lateral
spread and length) was largely a function of velocity
at the exit point of the channel.
As expected, the depositional area increased with
increased mass and moisture content. However, the
effect of increased moisture content was much more
important than an increase in mass. In fact, the overall
deposition area between tests T20 and T23 was nearly
identical, despite the fact that test T20 had more mass
than T23. The higher mobility provided by the higher
moisture content of T23 allowed it to spread thinner
and farther than a flow of the same mass and lower
moisture content (T14). This shows that pore pres-
sures are key to reducing friction via a reduction in
effective stress within a debris flow.
The contour plots, while useful for visualizing
the morphology of deposition, were not well suited to
analysing debris flow runout. Consequently, the centre
of mass was calculated for each flow using the point
depth measurements; this was used in the subsequent
analysis of results.
Figures 12 through 14 show relationships between
runout and the square of velocity, mass entering the
flume and peak momentum of the flow, respectively.
Previous studies have found the flow runout to
scale with velocity squared (t
akaHasHi
, 2007). While
there is a clear linear relationship between the velocity
Fig. 12 - Location of centre of mass of deposition (meas-
ured from the flume exit) against velocity squared
(measured at the high speed camera position).
Trendlines for 33% moisture content flows (solid)
and all flows (dashed) are shown. The test run at
41% moisture content is accurate to within ±1%,
as discussed in the text
Fig. 14 - Location of center of mass of deposition (measured
from the flume exit) against momentum of each flow
(based on the velocity at the high speed camera po-
sition). Trendlines are 33% moisture content (solid)
and all data (dashed)
Fig 13 - Location of centre of mass of deposition (meas-
ured from the flume exit) against total mass of ma-
terial entering the flume for each test. Trendlines
are 33% moisture content (dashed) and all data
(solid)
background image
P. kAILEY , E.T. BOwMAN, J. LAUE & S.M. SPRINGMAN
348
5th International Conference on Debris-Flow Hazards Mitigation: Mechanics, Prediction and Assessment Padua, Italy - 14-17 June 2011
creasing momentum at the high speed camera posi-
tion, calculated as mass (in kg) multiplied by velocity
(in m/s) at the high speed camera location.
Hence, it is hypothesised that because these
flows displayed a low and homogenous velocity upon
reaching the flume outlet, and because the peak cross-
sectional area of the flow was limited by the feeder
tube, lateral spread (which is determined by the total
amount of material), largely governed the runout of
these flows. Momentum, as it takes into account both
the total mass and the velocity of the flow, shows a
strong relationship with runout for all tests.
CONCLUSIONS
The five geotechnical centrifuge tests summarised
in this paper demonstrate several important aspects of
debris flow behaviour They also highlight some of the
advantages and challenges of modelling debris flows
in a geotechnical centrifuge.
Five debris flow tests were conducted in a geo-
technical centrifuge at 40 g with variable volumes and
moisture contents. Pore pressure and data derived from
photographs taken with a high speed camera were used
to construct plots of flow velocity with distance, flow
height over time, velocity profiles with depth, and dis-
charge over time, at one point in the flume.
Both an increase in volume and an increase in
moisture content increased peak velocity during down-
slope movement. However, the effect of increased
moisture content is much more pronounced than that
of increased mass. The maximum cross-sectional area
observed was limited by the flow diameter of the feeder
tube. Consequently, flow velocity largely determined
the peak discharge of each flow.
The large difference between the measured veloc-
ity profiles at the flow front, and during the recessional
phase of the flow (but still in the coarse front), is ex-
plained by a general trend of decreasing velocity with
time and friction against the flume walls.
An increase in moisture content increased the mo-
bility of the flow in terms of inundated depositional area
and runout. The runout of all flows can be related to the
velocity, mass, and momentum. However, the momen-
tum of the flow at the high speed camera position was
the best predictor of runout in these experiments.
While aspects of centrifuge flow behaviour com-
pare well with some field observations, the limited
number of tests and experimental boundary conditions
limit comparison with others. This paper provides an
example of the kinds of data which can be generated
from centrifuges tests, as well as some of the challeng-
es and opportunities of using the technique. Based on
these set of tests, a steeper flume configuration and larg-
er diameter feeder tube, which would increase velocity,
peak discharge, and the range of PSD tested, would be
useful in the future.
Future work will compare the results from these
flume tests with those predicted by analytical equations
presented in the literature, as well as compare the effect
of using a Newtonian vs. non-Newtonian pore fluid.
ACKNOWLEDGEMENTS
This research was made possible by a New Zea-
land Earthquake Commission Bi-annual grant, a Uni-
versity of Canterbury Doctoral Scholarship, and a New
Zealand Postgraduate Study Abroad Award. Thanks
go to the geotechnical group at ETH for their patience
and generosity. A debt of gratitude is owed especially to
Markus Iten, who was indispensable and irreplaceable
before, during and after the experiments. The last two au-
thors are members of the Competence Centre of Environ-
mental Sustainability at ETHZ and contribute to a range
of projects under this thematic partial funding for research
into natural hazards (TRAMM, COGEAR, APUNCH).
REFERENCES
b
owman
e.t., l
aue
, J., & s
PRinGman
s. (2010) - Experimental modelling of debris flow behaviour using a geotechnical centri-
fuge. Canadian Geotechnical Journal, 47 (7): 742-762.
b
owman
e.t. & s
anvitale
n. (2009) - The role of particle size in the flow behaviour of saturated granular materials. Paper read
at 17th International Conference on Soil Mechanics and Geotechnical Engineering, at Alexandria.
d’a
Gostino
v., C
esCa
m. & m
aRCHi
l. (2010) - Field and laboratory investigations of runout distances of debris flows in the
Dolomites (Eastern Italian Alps). Geomorphology, 115 (3-4): 294-304.
d
enlinGeR
R.P. & i
veRson
R.m. (2001) - Flow of variably fluidized granular masses across three-dimensional terrain: 2. Numeri-
cal predictions and experimental tests. J. Geophys. Res, 106: 553-566.
f
annin
R.J. & w
ise
m.P. (2001) - An empirical-statistical model for debris flow travel distance. Canadian Geotechnical Journal,
background image
MODELLING DEBRIS FLOW PROCESSES WITH A GEOTECHNICAL CENTRIFUGE
Italian Journal of Engineering Geology and Environment - Book www.ijege.uniroma1.it © 2011 Casa Editrice Università La Sapienza
349
38 (5): 982-994.
H
unGR
o., m
oRGan
G., & k
elleRHals
P. (1984) - Quantitative Analysis of Debris Torrent Hazards for Design of Remedial Meas-
ures. Canadian Geotechnical Journal, 21 (4): 663–677.
k
inG
J. (1996) - Tsing Shan debris flow. In Special Project Report SPR 6/96. Geotechnical Engineering. Office, Hong Kong
Government, Hong Kong. p. 133.
m
C
a
Rdell
b.w., b
aRtelt
P. & k
owalski
J. (2007) - Field observations of basal forces and fluid pore pressure in a debris flow.
Geophysical Research Letters, 34 (7): 74-76.
P
ieRson
t. (1980) - Erosion and deposition by debris flows at Mt. Thomas, North Canterbury, New Zealand. Earth Surface Proc-
esses, 5: 227-247.
R
iCkenmann
d. (1999) - Empirical Relationships for Debris Flows. Natural Hazards, 19 (1): 47-77.
s
PRinGman
s., l
aue
J., b
oyle
R., w
Hite
J, & z
weidleR
a. (2001) - The ETH Zurich Geotechnical Drum Centrifuge. International
Journal of Physical Modelling in Geotechnics, 1 (1): 59-70.
s
teedman
R.s. & z
enG
x. (1995) - Dynamics. In: t
ayloR
R.n. (1995, ed.) - Geotechnical Centrifuge Technology. Blackie Aca-
demic & Professional, Glasgow.
t
akaHasHi
t. (2007) - Debris flow: mechanics, prediction and countermeasures. Routledge. Taylor & Francis Group, London.
Statistics