SIMULATING SUBMARINE SLOPE INSTABILITY INITIATION USING
CENTRIFUGE MODEL TESTING
S.E. COULTER and R. PHILLIPS
C-CORE, Memorial University of Newfoundland, St. John’s, Newfoundland & Labrador,
Canada, A1B 3X5
Abstract
COSTA is addressing the questions of why seafloor slope failures occur where they do, and
with what frequency they occur. The original program has been complemented by COSTACanada. One of the tasks involves the study of the initiation of slope instability through
numerical and centrifuge modelling.
This paper reviews previous centrifuge studies related to submarine slope failure and
presents the preparations for this task. The initiation of submarine slope instability has been
attributed to triggers such as earthquakes, erosion, oversteepening, wave loading, and
sedimentation. Centrifuge modelling has been used to simulate most of these loading
conditions in similar boundary value problems.
Keywords: Submarine slope instability, triggering mechanisms, centrifuge modelling,
earthquake simulation
1. Introduction
Submarine slope stability has been identified as a major concern in offshore resource
development. Research programs have attempted to characterise and understand submarine
failures worldwide. These programs have included the Arctic Delta Failure Experiment
(ADFEX) 1982-1992, Geological Long Range Inclined Asdic (GLORIA) 1984-1991,
Sediment Transport on Atlantic Margins (STEAM) 1993-1996, ENAM II 1996-1999,
STRATAFORM 1995-2001 and Continental Slope Stability (COSTA) 2000 to present.
The original COSTA program on seafloor stability has been complemented by COSTACanada, Locat and Héroux (2001) and http://www.costa-canada.ggl.ulaval.ca. One of the 6
tasks in this complementary study involves the initiation of slope instability through
numerical and centrifuge modelling.
The initiation of submarine slope instability has been attributed to triggers such as
earthquakes, waves, tides, sedimentation, gas, erosion and diapirism, Hampton et al..
(1996). Centrifuge modelling has been used to simulate many of these loading conditions
and relevant soil conditions in similar seafloor stability studies. Figure 1 shows a submarine
slope failure model of a submerged 8o slope that flowed to an angle of 2o in normally
consolidated silty clay due to an increase of excess pore water pressure in the slope.
Following an overview of centrifuge modelling, other examples reviewed in this paper
29
30
Coulter and Phillips
include submarine failures from: (1) material softening, (2) sedimentation, (3) wave loading,
and (4) earthquakes.
Figure 1. Submarine Slope Failure in Silty Clay.
2. Centrifuge Modelling
Centrifuge model testing is a physical modelling tool for geotechnical engineers. Analogues
to this technique exist in other branches of civil engineering, such as wind tunnel testing in
aeronautical engineering and flume testing in hydraulic engineering. To achieve mechanical
similitude in geotechnical models it is necessary to reproduce the material behaviour both in
terms of strength and stiffness. This behaviour is primarily a function of the effective stress
resulting from self-weight and other external forces.
Centrifuge modelling then is a technique for investigating gravity dependant phenomena,
such as soil slope behaviour, using reduced-scale physical models. As a full-scale soil
structure is in equilibrium under earth’s gravitational field g, a reduced-scale model on a
centrifuge, at radius R, is in equilibrium under an acceleration field of Rω2, where ω is the
rotational speed of the centrifuge. The model will then have its weight increased N times,
where N = Rω2/g.
For a typical value of N = 100, if a model is made at 1/100th scale and is accelerated to
100g, the stresses due to self-weight will be similar to the stresses in the prototype at
homologous points. The model can then reproduce the phenomena of cracking, rupture or,
flow that would be observed in the prototype because the stress dependency of soil
behaviour has been correctly simulated. The principles, scaling laws and some applications
of centrifuge modelling are more fully described by Murff (1996) and Taylor (1995).
3. Material Softening
The Canadian Liquefaction Experiment (CANLEX) by Robertson et al. (2000) investigated
the liquefaction potential of loose sand deposits under monotonic shear stress increments.
The program included high quality soils testing, numerical analyses and a full-scale field
event designed to cause a flowslide. Centrifuge model tests were added to the CANLEX
program to provide additional physical data and to simulate the physical response of the
Centrifuge Modelling of Submarine Slope Instabilities
31
field event. The oil sand tailings from the field site used in the model tests were strain
softening under triaxial extension. One centrifuge model simulated the failure of a 16o,
8.8m high loose sand submerged slope, Phillips and Byrne (1994). Surcharging the slope
crest caused the model slope to liquefy and flow with deep-seated lateral movements to an
angle of 7o, Figure 2. This submarine slope failure occurred primarily due to the strain
softening response of the sand at the toe of the slope that caused pore pressure increases and
destabilisation of the slope.
Surcharge
Oil Level
Final
PPT 1
PPT 2
Initial
Figure 2. CANLEX Submerged Slope Flowside Failure.
4. Sedimentation
There are other external loading conditions that can cause pore pressure increases in the
vicinity of slopes. Continuous sedimentation has been identified as one such loading
condition. Hurley (1999) has modelled the sedimentation and primary consolidation of a
highly permeable cohesive kaolin sediment from 580% water content. The equivalent of
10m depth of aqueous slurry sedimented and consolidated into a 2m thick silty clay layer,
Figure 3. Continuous measurements of bulk density (from gamma ray attenuation), pore
pressure, and shear and compression wave velocities were made at 100g. The excess pore
pressures generated by this sedimentation process may be sufficient to destabilise a
submarine slope, Figure 4. The excess pore pressure at PPT6, 5m down in the consolidating
clay layer, was 30kPa at the prototype time of 2 years after commencement of
sedimentation. This is equivalent to an excess pore pressure ratio of about unity. The
associated effective shear strength is low. If sufficient material is subject to such pore
pressures then seafloor instability can ensue.
Initial
Level
PPT1
PPT2
PPT3
PPT4
Excess Pore Pressure (kPa)
80
PPT6
z(mm)
PPT1 37
PPT2 188
PPT3 340
PPT4 493
PPT5 645
PPT6 840
60 PPT5
PPT4
40
PPT3
20
PPT5
PPT2
Final
Level
PPT1
PPT6
Figure 3. Sediment Column Test Setup.
0
0
100
Time (mins)
200
Figure 4. Excess Pore Pressures in Slurry Sediment.
32
Coulter and Phillips
5. Wave Loading
Wave loads can also cause excess pore pressures leading to seabed mobility and
liquefaction. The techniques for modelling wave seabed interaction were developed by
Phillips and Sekiguchi (1992). Sekiguchi et al.. (2000) simulated at 50g the response of a
4.5m thick sand layer under 4m water depth subjected to 0.3m high waves with a 6s period,
Figure 5. Significant seabed mobility and liquefaction was observed to propagate down
from the seafloor to the base of the sand layer. Figure 6 shows the excess pore pressure
response ue at the sand base to pressure cycles at the seafloor. Three additional
configurations were modelled with a gravel cap of various lengths on the seabed to develop
a mitigation strategy to minimise damage to the seafloor.
Figure 5. Wave Loading Model Test Setup.
Figure 6. Excess Pore Pressure Response in Seabed.
6. Earthquake Effects
6.1 VELACS
The VELACS (Verification of Earthquake Liquefaction Analysis by Centrifuge Studies)
program considered the effects of earthquake-like loading on a variety of soil models,
Arulanandan and Scott (1993 &1994) and http://ceor.princeton.edu/~radu/soil/velacs/.
VELACS was aimed at better understanding the mechanisms of soil liquefaction and at
acquiring data for the verification of various analysis procedures. The response of each of
nine boundary value configurations was predicted by between 4 to 16 numerical analysts.
The Class-A predictions were then compared to the measured response from nominally the
same physical model test conducted at between 1 to 3 centrifuge centres.
Model 2 involved the simulation of lateral spreading of a submerged slope due to
earthquake effects, Figure 7. The submerged 2o slope comprised a saturated 10m deep layer
of Nevada sand at 40% relative density. The base of the layer was subjected to about 22
cycles to 0.1g lateral acceleration at 2Hz. The toe of the slope at the seafloor, LVDT3 was
observed to move downslope a distance of 0.5 m due to this earthquake, Figure 8.
Centrifuge Modelling of Submarine Slope Instabilities
33
Lateral Spread, cm
60
40
20
0
0
10
20
Time, seconds
Figure 8. VELACS Model 2 Lateral Spread.
Figure 7. VELACS Model 2 Test Setup.
6.2 COSTA-Canada
6.2.1 Earthquake Simulation
C-CORE will commission a C$0.5m earthquake simulator (EQS) on its 5.5m radius
centrifuge in early 2003. This single axis EQS shown in Figure 9 will excite a 400kg
moving payload of 1m length by 0.5m width by up to 0.6m high with a maximum force of
200kN. The peak lateral acceleration of the full payload will be 40g under a constant
(centrifugal) acceleration of 80g. The lateral acceleration envelope is shown in Figure 10
over the frequency range of 40-200Hz.
The COSTA-Canada tests will involve test configurations similar to Figure 7, using the
selected sand discussed in Section 6.3.4. Lower permeability silt layers will be introduced
into the sand layer to simulate a stratified profile. The effect of these layers on pore pressure
generation and subsequent dissipation will be examined. The silt layers will impede
drainage of the sand as shown by Konrad and Dubeau (2002). The migration of pore
pressure towards potential drainage boundaries is also expected to cause continued
movement of the slope after cessation of the earthquake.
Figure 9. C-CORE Earthquake Simulator Layout.
Figure 10. EQS Performance Envelope.
6.2.2 Model Container
Any model container used in dynamic geotechnical tests should cause stress and strain in the
model similar to the soil layer being tested in infinite lateral area and finite depth. This
34
Coulter and Phillips
requires a model box with flexible boundaries. Among the several options available the one
chosen for these tests is the equivalent shear beam (ESB) container introduced by Zeng and
Schofield (1996). The ESB container has been designed with the following principles: (1)
the end walls function as shear beams with equivalent dynamic shear stiffness to the
adjacent soil, (2) each end wall has similar friction to the adjacent soil so that it can sustain
complimentary shear stresses induced by the shaking, (3) the side walls are frictionless to
avoid induced shear stress between soil and side walls, (4) the walls maintain a Ko
condition, and (5) the frictional end walls have similar vertical settlement of the soil layer to
avoid initial boundary shear stresses. The ESB container is constructed of stacked
aluminium rings alternating with rubber layers to achieve its complimentary flexibility,
Figure 11.
Figure 11. Equivalent Shear Beam Model Container.
6.2.3 Substitute Pore Fluid
Centrifuge modelling of dynamic events in saturated soil environments arouses one major
difficulty in recognized scaling laws. There exists a difference in scale factors (N) with
respect to time between static events and dynamic events. Stewart (1998) shows that by
increasing the pore fluid viscosity by N times a reduction in permeability of the soil of N
times can be achieved. It is shown that this reduction in permeability leads to an agreement
of the two scaling relationships. The substitute pore fluid selected for these tests is
chemically known as hydroxypropyl methylcellulose (HPMC), which has been used
previously by Stewart (1998), Dewoolkar et al. (1999a), and Dewoolkar et al. (1999b).
HPMC is considered a superior product for this purpose because it is biodegradable, can be
easily mixed with deionised water to create fluids with a wide range of viscosities, has many
properties similar to water, is inexpensive, and readily available. A kinematic viscosity of
50cSt has been selected for this research. Mixing trials have been undertaken and fluids
with required kinematic viscosities can be reliably reproduced.
Centrifuge Modelling of Submarine Slope Instabilities
35
6.2.4 Selected Sand
The sand that has been selected for use in the COSTA-Canada centrifuge tests is Fraser
River sand from the west coast of Canada. A considerable amount of this sand has been
processed and delivered to C-CORE. Fraser River sand has been used extensively by
project collaborators at the University of British Columbia in numerous laboratory tests
relating to liquefaction studies, including those noted in Uthayakumar and Vaid (1998) and
Vaid et al. (2001). This sand is uniform, grey coloured, and medium grained with
subangular to subrounded particles. Fraser River sand features an average mineral
composition of 40% quartz, 11% feldspar, 45% unaltered rock fragments, and 4% other
minerals. The engineering properties presented in the literature have been confirmed within
reasonable tolerances for the batch of Fraser River sand that is to be used at C-CORE for
the centrifuge tests.
6.2.5 Model Preparation
The sand models will be placed into the ESB by means of air pluviation, similar to methods
discussed by Ueno (1998). A hopper containing the test sand will be moved over the model
container at a specified drop height that will ensure a relative density of approximately 3032%. Some densification in the range of 8-10% is expected during movement of the test
slope in the laboratory and saturation with the substitute pore fluid. The saturation of the
model ground will be done under vacuum in a method similar to that explained by Ueno
(1998). A high degree of saturation is required to avoid any capillary action amongst the
sand particles that may produce cohesion. Vacuum will be applied to the model container to
remove the air and then carbon dioxide will be introduced to displace less soluble air
particles that may be present in the voids. Following this step, de-aired pore fluid will be
introduced slowly under vacuum conditions by means of a small differential driving head so
as not to cause any disturbance in the sample.
6.2.6 Instrumentation
Sufficient saturation will be verified in flight using a miniature air hammer to generate Pwaves, similar to the device used by Malvick et al. (2002). The arrival of these P-waves will
be measured by two in-line accelerometers that will be placed a known distance apart in the
sand model during pluviation. The arrival times of the generated P-wave signals will be
compared. As shown by Ishihara et al. (2001) a P-wave velocity of greater than 750m/s
indicates a degree of saturation in excess of 99%, which can be considered adequate to
ensure that liquefaction would be insensitive to the degree of saturation.
Additional instrumentation will be installed during pluviation and used to examine the
reaction of the model slope before, during, and after earthquake actuation. Pore pressure
transducers equipped with sintered bronze stones will be used to monitor the changes in
pore pressure. Miniature accelerometers will be placed at key locations to measure the
shaking response of the slope. Linear variable differential transformers and laser transducers
will be employed to measure surface deformations. All instrumentation will be connected to
the new high-speed 24 channel data acquisition system that will be integrated with the new
EQS. With this new system data acquisition can be performed automatically while the
shaker control is in operation.
36
Coulter and Phillips
7. Summary
Centrifuge modelling has been shown to be a valuable tool for investigations into various
triggering mechanisms of submarine slope instabilities. It is upon this foundation that
COSTA-Canada hopes to build. The preparations for the COSTA-Canada centrifuge tests
are advancing and model preparation methods are currently being perfected. Testing will
commence shortly after commissioning of the new earthquake simulator in early 2003.
Initial results can be expected by mid 2003 and the results of this work will be offered in
future publications.
8. Acknowledgements
This work is supported by C-CORE and COSTA-Canada, a NSERC Collaborative Research
Opportunities Grant.
9. References
Arulanandan, K., and Scott, R.F. (eds.), 1993 & 1994. Verification of Numerical Procedures for the Analysis of Soil
Liquefaction Problems. International Conference on the Verification of Numerical Procedures for the
Analysis of Soil Liquefaction Problems, Vols. 1 and 2. Balkema, Rotterdam, Netherlands.
Dewoolkar, M.M., Ko, H.Y., and Pak, R.Y.S., 1999a. Centrifuge Modelling of Models of Seismic Effects on
Saturated Earth Structures. Geotechnique, 49: 247-266.
Dewoolkar, M.M., Ko, K.Y., Stadler, A.T., and Astaneh, S.M.F., 1999b. A Substitute Pore Fluid for Seismic
Centrifuge Modeling. ASTM Geotechnical Testing Journal, 22: 196-210.
Hurley, S.J. 1999. Development of a Settling Column and Associated Primary Consolidation Monitoring Systems
for use in a Geotechnical Centrifuge. Master of Engineering Thesis, Memorial University of
Newfoundland.
Hampton, M.A., Lee, H.J., and Locat, J., 1996. Submarine Landslides. Reviews of Geophysics, 34:33-59.
Ishihara, K., Tsuchiya, H., Huang, Y., and Kamada, K., 2001. Recent Studies on Liquefaction Resistance of
Sand – Effects on Saturation. 4th International Conference on Recent Advances in Geotechnical
Earthquake Engineering and Soil Dynamics. San Diego, USA.
Konrad, J.M. and Dubeau, S., 2002. Cyclic Strength of Stratified Soil Samples. 55th Canadian Geotechnical
Conference, Niagara Falls, Canada.
Locat, J., and Héroux, M.C., 2001. COSTA-Canada: A Canadian Contribution To The Study Of Continental Slope
Stability – An Overview. 54th Canadian Geotechnical Conference, Sept. 2001, Calgary, Canada.
Malvick, E.J., Kulasingam, R., Kutter, B.L., and Boulanger, R.W., (2002). Void Redistribution and Localized Shear
Strains in Slopes During Liquefaction. Physical Modelling in Geotechnics: ICPMG ’02. Balkema.
Rotterdam, Netherlands.
Murff, J.D., 1996. The Geotechnical Centrifuge in Offshore Engineering. OTRC Honors Lecture, Offshore
Technology Conference, Houston, USA.
Phillips, R., and Sekiguchi, H., 1992. Generation of Water Waves in a Drum Centrifuge. Techno-Ocean '92,
Yokohama, Japan.
Phillips, R., and Byrne, P.M., 1994. Modelling Slope Liquefaction Due to Static Loading. 47th Canadian
Geotechnical Conference, Halifax, Canada.
Robertson, PK, …Phillips, R...et al. (34 authors), 2000. The Canadian Liquefaction Experiment: An Overview.
Canadian Geotechnical Journal, 37: 499-504.
Sekiguchi, H., Sassa, S., Sugioka, K., and Miyamoto, J., 2000. Wave Induced Liquefaction Flow Deformation and
Particle Transport in Sand Beds. GeoEng 2000, Melbourne, Australia.
Stewart, D.P., Chen, Y.R., and Kutter, B.L., 1998. Experience with the Use of Methylcellulose as a Viscous Pore
Fluid in Centrifuge Models. ASTM Geotechnical Testing Journal, 21: 365-369.
Taylor, R.N., 1995. Geotechnical Centrifuge Technology, Blackie Academic & Professional. London, UK.
Ueno, K., 1998. Methods for Preparation of Sand Samples. Centrifuge 98. Balkema. Rotterdam, Netherlands.
Uthayakumar, M., and Vaid, Y.P., 1998. Static Liquefaction of Sands Under Multiaxial Loading. Canadian
Geotechnical Journal, 35: 273-283.
Vaid, Y.P., Stedman, J.D., and Sivathayalan, S., 2001. Confining Stress and Static Shear Effects in Cyclic
Liquefaction. Canadian Geotechnical Journal, 38: 580-591.
Zeng, X., and Schofield, A.N., 1996. Design and Performance of an Equivalent-shear-beam Container for
Earthquake Centrifuge Modelling. Geotechnique, 46: 83-102.