Content uploaded by Juan Carlos Tiznado
Author content
All content in this area was uploaded by Juan Carlos Tiznado on Jun 25, 2019
Content may be subject to copyright.
Earthquake Geotechnical Engineering for Protection and Development of
Environment and Constructions –Silvestri & Moraci (Eds)
© 2019 Associazione Geotecnica Italiana, Rome, Italy, ISBN 978-0-367-14328-2
Centrifuge study of the seismic response of embankments on
liquefiable soils improved with dense granular columns
J.C. Tiznado
University of Colorado Boulder, Boulder, Colorado, USA
Pontificia Universidad Católica de Chile, Santiago, Chile
S. Dashti & B.P. Wham
University of Colorado Boulder, Boulder, Colorado, USA
C. Ledezma
Pontificia Universidad Católica de Chile, Santiago, Chile
ABSTRACT: This paper presents preliminary results from four centrifuge tests, performed
at the University of Colorado Boulder, to evaluate the relative influence of different mitiga-
tion mechanisms provided by dense granular columns (DGC) on the seismic performance of
embankments founded on liquefiable soil deposits. The first test was designed as the baseline,
unmitigated case, and the soil below the embankment in the subsequent three tests was treated
with a grid of DGCs that had an area replacement ratio (A
r
) = 10%. The second test evaluated
the influence of shear reinforcement and enhanced drainage provided by these columns. The
third test isolated the effect of shear reinforcement by preventing drainage. The fourth and
last test combined the effect of shear reinforcement with densification (without drainage) pro-
vided by DGCs. For the conditions investigated, the experimental results showed that densifi-
cation was more effective than drainage in limiting embankment deformations. In addition,
there was no significant difference between settlements measured in tests with and without
drainage, which suggests that shear reinforcement was more important than drainage in
improving the overall embankment’s seismic performance.
1 INTRODUCTION
Over the past few decades, dense granular columns (DGC) have become a common soil
improvement strategy beneath embankments and other geo-structures founded on liquefiable
deposits (Adalier & Elgamal 2004). Case-histories from past earthquakes have consistently
demonstrated the effectiveness of this method in mitigating the effects of soil liquefaction
(Mitchel & Wentz 1991, Mitchell et al. 1995, Hausler 2001, Nikolaou et al. 2016). DGCs are
known to improve site performance through: (i) increasing the density of the surrounding soil
during installation; (ii) enhancing drainage to control net excess pore water pressures; and (iii)
introducing a stiff element that provides shear reinforcement to the surrounding soil (Baez &
Martin 1993). Through these combined mechanisms, depending on the characteristics of the
motion, site, and the geotechnical structure, successful performance has been achieved even
with relatively small area replacement ratios (A
r
) as low as 10% (Hausler 2002). However, the
relative contribution of these mitigation mechanisms to the overall performance of the soil-
embankment system is currently not well-understood, as is necessary for their reliable forward
design. It is not easy to separate these effects with the existing limited, and typically non-
instrumented, case-history observations. Furthermore, there is a lack of physical model stud-
ies evaluating and separating these effects under controlled conditions to, for instance, valid-
ate numerical models.
5282
Previous experimental studies have primarily focused on the influence of prefabricated
drains or granular columns on liquefiable gentle slopes (Adalier et al. 2003, Howell et al. 2012,
Vytiniotis & Whittle 2017, Badanagki et al. 2018), showing that enhancing drainage can help
limit seismic deformations, depending on the characteristics of the motion and columns’A
r
.
Badanagki et al. (2018), for example, showed successful slope performance in centrifuge with
DGC A
r
greater than about 20%, which is different from A
r
in case history observations. Pre-
vious numerical parametric studies representing granular columns as unit cells have also dem-
onstrated their effectiveness in reducing lateral displacements due to a combination of
drainage and shear reinforcement (Elgamal et al. 2009, Asgari et al. 2013, Rayamajhi et al.
2016b). They have also provided valuable insight into shear deformations between the col-
umns and their surrounding soil (Rayamajhi et al. 2016a), and have indicated DGCs can sig-
nificantly improve site performance, particularly for A
r
ranging from about 20 to 30%.
However, these studies have neither evaluated the influence of DGCs on the seismic perform-
ance of more complex and realistic geo-structures on top of liquefiable soil profiles, nor
explored the possible effects of installation-induced densification provided by DGCs on the
overall response of the soil-structure system.
In this paper, we present preliminary results of four reduced-scale dynamic centrifuge tests,
performed at the University of Colorado Boulder’s Center for Infrastructure, Energy, and
Space Testing (CIEST) 400 g-ton centrifuge facility, to evaluate the influence of DGCs on the
seismic performance of embankments founded on liquefiable soil deposits. All tests considered
a fully saturated, layered, liquefiable soil profile beneath a dense, dry, coarse sand embank-
ment. The first test (BS) was designed as the baseline, unmitigated case. The soil below the
embankment in the subsequent three tests was treated with a grid of DGCs that had an A
r
=
10%. The second test (RF-DR) evaluated the influence of shear reinforcement and enhanced
drainage provided by DGCs. The third test (RF) isolated the effect of shear reinforcement by
preventing drainage with a latex membrane placed around the columns. The fourth and last
test (RF-DS) combined the effect of shear reinforcement with densification (without drainage)
provided by DGCs. The results of these experiments are aimed to provide insight into the rela-
tive influence of different mitigation mechanisms (i.e., shear reinforcement, drainage, and
densification) on key engineering demand parameters that govern the embankment’s design
(e.g., peak shear stresses and accelerations on the embankment, peak excess pore pressures in
the soil below, and cumulative embankment deformations), and provide data for the calibra-
tion and validation of advanced numerical models in the future.
2 EXPERIMENTAL PROGRAM
A series of centrifuge experiments, with a prototype-to-model scale factor of 70, were per-
formed at the University of Colorado Boulder’s (CU) 400g-ton (5.5 m radius) centrifuge facil-
ity, to systematically evaluate the influence of dense granular columns (DGC) and the relative
importance of the mechanisms of densification, enhanced drainage, and reinforcement, on the
seismic performance of embankments founded on liquefiable soil deposits during 1D horizon-
tal earthquake loading. Figure 1 shows the typical configuration and instrumentation layout
used in these experiments. Dimensions are presented in prototype scale, following accepted
scaling relations (Tan & Scott 1985, Garnier et al. 2007).
For all tests, a dense layer (12 m-thick in prototype scale) of Ottawa sand F65 was dry plu-
viated to attain a relative density (D
r
) of approximately 90% at the bottom of a flexible-shear
beam container. Subsequently, a loose layer of Ottawa sand (4 m-thick) with D
r
≈40% was
modeled as the liquefiable material. This layer was overlaid by a 2 m-thick layer of coarse
Monterrey sand 0/30 with Dr≈90% to create a non-liquefiable crust. The DGCs with diameter
= 1.75 m and a center-to-center spacing of 4.9 m (in prototype scale) were placed vertically (in
a square pattern) at an elevation of 8 m above the bottom of the container, prior to finishing
sand pluvation, to avoid localized densification of sand during their installation, and to keep
the density of the surrounding soil well controlled and uniform within each layer. The engin-
eering soil properties of these materials, as measured at CU, are provided in Table 1.
5283
Test BS was considered a non-mitigated case, absent granular columns. In Test RF-DR,
DGCs were encased within geotextile filters (to avoid clogging in subsequent shakes and allow
water flow) and placed at designated locations shown in Figure 1. In Test RF, DGCs with
geotextile filters were encased within a thin (0.2 mm-thick) latex membrane to prevent drain-
age while providing shear reinforcement. The last test, RF-DS, the DGCs were encased within
latex (similar to Test RF), but the soil surrounding the DGCs was also pluviated at a denser
state of D
r
≈80%. In all tests with mitigation, an A
r
= 10% was used for the DGC’s. A
r
is
defined as the ratio of granular column area to the total treatment area in plan view (Baez &
Martin 1993).
It is known that the installation of granular columns in the field typically induces ground
densification, as well as an increase in lateral earth pressures in the surrounding soil. However,
Figure 1. Instrumentation layout of the centrifuge experiments in elevation (top) and plan (bottom)
views.
Table 1. Engineering properties of different soil layers used in centrifuge experiments.
Ottawa sand F65 Monterey sand 0/30
Granular material
(embankment & columns)
Specific gravity, G
s
2.65 2.64 2.66
Minimum void ratio, e
min
0.53 0.54 0.62
Minimum void ratio, e
min
0.81 0.84 0.92
Uniformity coefficient, C
u
1.56 1.30 1.54
Hydraulic conductivity, k (cm/s) 1.41x10
-2
(D
r
=40%)
1.19x10
-2
(D
r
=90%)
5.29 x10
-2
2.90
5284
due to the difficulties of column installation in flight during centrifuge experiments, in Test
RF-DS, densification was considered in an approximate and roughly uniform manner. By
making use of the design chart for granular columns developed by Baez (1995) (based on field
data from 18 case histories) that is shown in Figure 2, the initial relative density of the poten-
tially liquefiable layer was converted to an equivalent pre-improvement (N
1
)
60
value (SPT
blow count normalized to an overburden pressure of 1 atm) through the expression:
Dr¼ffiffiffiffiffiffiffiffiffiffiffiffiffi
N1
ðÞ
60
Cd
sð1Þ
C
d
is a parameter that accounts for the grain-size composition of soil. Typical C
d
values for
fine sand used in laboratory tests are about 35 (Idriss & Boulanger 2008). For A
r
=10%, based
on equation (1), the post-improvement (N
1
)
60
value was estimated and converted to an equiva-
lent relative density (D
r
) for the soil surrounding the columns after installation. This resulted
in D
r
≈80% noted previously for the middle sand layer in Test RF-DS (see Figure 1). Both
Monterey and Ottawa sand layers with a D
r
≈90% were considered not to be significantly
affected by DGC installation. Note, however, that this method of simulating installation-
induced densification does not capture the changes in soil fabric under vibration, but it iso-
lates the effect of soil density or void space.
After dry preparation, the models were saturated with a solution of hydroxypropyl methylcellu-
lose in water (Stewart et al. 1998) with a viscosity 70 times greater than that of water, to satisfy
both diffusive and dynamic scaling laws (Taylor 1995). Initially, the soil model was flushed with
CO
2
from the bottom of the container, after which it was kept under constant vacuum. Then, the
vacuum level in the fluid tank was controlled automatically (using a computer system similar to
that proposed by Stringer & Madabushi 2009) to maintain a safe and constant flow rate below
that required for flow-induced liquefaction until saturation was completed. The model was subse-
quently spun to 70 g of centrifugal acceleration, after which several 1D horizontal earthquake
motions were applied to its base in flight using a servo-controlled hydraulic shake table (Ketchum
1989). This paper focuses on the soil-embankment system performance during the first major
motion: the horizontal component of a scaled version of the 1995 Kobe earthquake recorded at
the Port Island station, with Peak Ground Acceleration (PGA) = 0.37 g, mean period (T
m
) = 0.88
s, significant duration (D
5-95
) = 12.1 s, and Arias Intensity (I
a
)=2.2m/s.
3 PRELIMINARY EXPERIMENTAL RESULTS
3.1 Influence of granular columns on embankment performance
The effectiveness of DGCs and the relative importance of mitigation mechanisms of densifica-
tion, enhanced drainage, and reinforcement, on the seismic performance of the embankment
Figure 2. Pre- and post-improvement (N
1
)
60
during the installation of DGCs (After Baez 1995).
5285
was evaluated in terms of accelerations and excess pore pressures in the foundation soil as
well as deformations within the embankment. Figure 3 compares the results among different
tests in terms of the time history of accelerations and Arias Intensities as well as response spec-
tra in the center array below the embankment. Figure 4 compares the results in terms of the
time history of excess pore pressure ratio (i.e., r
u
=Δu/σ′
z
, where Δu is the excess pore water
pressure and σ′
z
is the initial vertical effective stress at a given depth) in the center and edge
arrays as well as the settlement and angular distortion of the embankment.
In general, all models exhibited significant de-amplification of accelerations at shorter
periods (higher frequencies) and amplification at higher periods (shorter frequencies) from the
base toward the ground surface due to soil softening. On the other hand, even though large
excess pore water pressures were measured within the bottom layer of dense Ottawa sand,
liquefaction (defined as r
u
=1.0) was typically not reached in that layer particularly under the
center of the embankment with a greater initial vertical stress. As expected, Test RF-DR was
successful in reducing the extent and duration of large excess pore pressures in all layers when
compared to the unmitigated, baseline case BS. Test RF, which provided similar shear
reinforcement but did not enhance drainage (or even laterally inhibited drainage slightly with
the use of latex around the columns), reduced the rate of dissipation and increased net excess
pore pressures particularly after strong shaking at most depths compared to Tests BS and RF-
DR. Interestingly, a more uniformly dense profile below the treated embankment in Test RF-
DS amplified the excess pore pressures at lower elevations (within dense Ottawa sand) in a
similar manner as the case without mitigation (BS), while reducing the extent of softening in
the top layer of Ottawa sand compared to all other tests. In all tests, excess pore pressures at
locations close to the ground surface and below the embankment experienced low r
u
values,
owing primarily to the high drainage capacity of the top Monterey sand layer.
Tests RF and RF-DR showed similar patterns of amplification in low-period accelerations
at higher elevations within the looser layer of Ottawa sand. This increased acceleration
demand within the treated soil was associated with shear reinforcement of the DGCs under
the confining pressure and boundary conditions of the embankment above (as opposed to the
case of a free-face, where shear reinforcement may reduce accelerations in the surrounding
soil, as observed previously by Badanagki et al. 2018). The added drainage in Test RF-DR
Figure 3. Acceleration time histories, response spectra (5%-damped), and Arias Intensity (I
a
) time his-
tories recorded at the center array of all models during the Kobe motion.
5286
further amplified low-period accelerations by dissipating excess pore pressures and increasing
soil stiffness more rapidly, particularly at higher elevations.
As shown in Figure 3, the combined effects of reinforcement and densification with no
drainage (Test RF-DS) resulted in lower spectral accelerations at all depths along the center
array, in a roughly similar manner as the unmitigated case BS. This response may be explained
by the large excess pore pressure recordings at lower elevations, which in a way, reduced the
seismic demand transferred to higher elevations. Figure 4 showed that creating a more uni-
formly dense deposit in RF-DS led to larger r
u
values within the lower dense layer of Ottawa
sand similar to those measured in Test BS (with no DGCs or densification) and significantly
greater than those measured in Tests RF-DR and RF. Even though increased density of the
top Ottawa sand layer reduced the extent of softening within that layer, greater pore pressures
at lower elevations in Test RF-DS reduced soil’s strength and stiffness, de-amplifying high-
frequency accelerations that propagated upward toward the embankment.
Figure 4 also shows the time histories of vertical displacement (settlement) D
v
recorded at
the top and bottom of the embankment along its centerline. Overall, the embankment itself
did not experience notable volumetric strains, as it was placed at a very dense initial state. As
a result, total embankment settlements were almost entirely due to the settlement of the soil
deposit below (as shown in Figure 4), which was caused by a combination of volumetric and
shear type deformations (Dashti et al. 2010). DGCs used in Tests RF-DR, RF, and RF-DS
were successful in reducing settlements with respect to the unmitigated case, BS, with the
common denominator of shear reinforcement. The comparison between Tests RF-DR and
RF-DS, however, shows that the densification mechanism was more effective than drainage in
limiting the embankment’s settlement for the conditions investigated. The combination of
shear reinforcement, de-amplification of low-period accelerations, and reduction in net excess
pore pressures at higher elevations observed in Test RF-DS reduced the embankment’s
Figure 4. Base acceleration, excess pore pressure ratio, vertical displacement, and angular distortion
time histories measured during the Kobe earthquake motion.
5287
permanent settlement by about 50% when compared to Test BS. In addition, there was no
significant difference between settlements measured in Tests RF-DR and RF, suggesting that
shear reinforcement was more important than drainage in limiting net settlements below the
embankment.
To quantify changes in the angle of embankment’s slope over time, its angular distortion
was approximated as the difference in vertical displacements between the edge (D
v edge
) and
the toe (D
v toe
) of the embankment slope, divided by the horizontal distance (d) between those
two locations after construction or before shaking. Figure 4 shows the angular distortion time
histories experienced by the north and south slopes of the embankment during the Kobe
motion. In general, Tests BS and RF-DS showed a non-symmetrical deformation pattern,
whereas Tests RF-DR and RF exhibited similar combinations of volumetric and shear
deformations at both slopes. Nevertheless, on the north side, the angular distortion patterns
among tests were similar to the settlements, showing that combined reinforcement and densifi-
cation was most successful in reducing distortions, and that drainage did not provide add-
itional benefits to reinforcement in limiting those deformations. It must be noted, however,
that these conclusions are only applicable to the conditions (geometry, soil properties, and
ground motion) investigated here and cannot be generalized before additional numerical
investigation.
4 SUMMARY AND CONCLUSIONS
Preliminary results from four dynamic centrifuge tests studying the influence of DGC on the
seismic performance of embankments on top of liquefiable soil deposits were presented. Over-
all, it was found that DGCs used in Tests RF-DR, RF, and RF-DS were effective in improv-
ing performance when compared to the unmitigated case. However, the results from Tests
RF-DR and RF-DS showed that densification was more effective than drainage in limiting
the embankment’s deformations. Additionally, no significant difference was found between
settlements measured in Tests RF-DR and RF, suggesting that shear reinforcement was more
important than drainage in limiting net settlements beneath the embankment.
ACKNOWLEDGMENTS
The authors would like to acknowledge the Chilean National Commission for Science and
Technology (CONICYT) for the financial support given to the first author through the grant
CONICYT-PCHA/Doctorado Nacional/2015-21150231. The authors also extend their appre-
ciation to the Department of Civil, Environmental, and Architectural Engineering as well as
all staff engineers and undergraduate student researchers at the University of Colorado Boul-
der’s Center for Infrastructure, Energy, and Space Testing (CIEST) for their assistance during
the execution of the centrifuge tests.
REFERENCES
Adalier, K., Elgamal, A., Meneses, J., & Baez, I.J. 2003. Stone columns as liquefaction counter-measure
in non-plastic silty soils. J. of Soil Dynamics and Earthquake Eng. 23 (7), 571–584.
Adalier, K., & Elgamal, A. 2004. Mitigation of liquefaction and associated ground deformations by
stone columns. Engineering Geology, 72(3-4), 275-291.
Asgari, A., Oliaei, M., & Bagheri, M. 2013. Numerical simulation of improvement of a liquefiable soil
layer using stone column and pile pinning techniques. J. of Soil Dynamics and Earthquake Eng.51
(2013)77–96.
Baez, J. I., & Martin, G. R. 1993. Advances in the design of vibro systems for the improvement of lique-
faction resistance. In Symposium Ground Improvement (pp. 1-16).
Baez, J. I. 1995. A design model for the reduction of soil liquefaction by vibro-stone columns Ph.D.
thesis, Dept. of Civil and Environmental Engineering, Univ. of Southern California.
5288
Badanagki, M., Dashti, S., & Kirkwood, P. 2018. Influence of Dense Granular Columns on the Perform-
ance of Level and Gently Sloping Liquefiable Sites. Journal of Geotechnical and Geoenvironmental
Engineering, 144(9), 04018065.
Dashti, S., Bray, J. D., Pestana, J. M., Riemer, M., & Wilson, D. 2010. Mechanisms of seismically
induced settlement of buildings with shallow foundations on liquefiable soil. Journal of Geotechnical
and Geoenvironmental Engineering, 136(1),151-164.
Elgamal, A., Lu, J. & Forcellini, D. 2009. Mitigation of liquefaction-induced lateral deformation in a
sloping stratum: three-dimensional numerical simulation. J. of Soil Dynamics and Earthquake Eng.
ASCE, 135(11),1672-1682.
Garnier, J., Gaudin, C., Springman, S.M., Culligan, P.J., Goodings, D., Konig, D., Kutter, B., Phillips,
R., Randolph, M.F., & Thorel, L. 2007. Catalogue of scaling laws and similitude questions in centri-
fuge modeling. Intern. J. of Phy. Mod. In Geotech., 7(3),1-24.
Hausler, E. A. 2002. Influence of ground improvement on settlement and liquefaction: A study based on
field case history evidence and dynamic geotechnical centrifuge tests. Ph.D. thesis, Dept. of Civil and
Environmental Engineering, Univ. of California.
Howell, R., Rathje, E. M., Kamai, R., & Boulanger, R. 2012. Centrifuge modeling of prefabricated verti-
cal drains for liquefaction remediation. Journal of Geotechnical and Geoenvironmental Engineering, 138
(3),262-271.
Idriss, I. M., & Boulanger, R. W. 2008. Soil liquefaction during earthquakes. Earthquake Engineering
Research Institute.
Ketchum, S.A., 1989. Development of an Earthquake Motion Simulator for Centrifuge Testing and the
Dynamic Response of a Model Sand Embankment. Ph.D. Thesis, Department of Civil, Environmen-
tal, and Architectural Engineering, University of Colorado, Boulder, CO.
Mitchell, J. K., & Wentz, F. J. 1991. Performance of improved ground during the Loma Prieta Earthquake.
Berkeley: Earthquake Engineering Research Center, University of California (Vol. 91, No. 12).
Mitchell, J. K., Baxter, C. D., & Munson, T. C. 1995. Performance of improved ground during earth-
quakes. In Soil improvement for earthquake hazard mitigation (pp. 1-36). ASCE.
Stringer, M., & Madabhushi, S. 2009. Novel computer-controlled saturation of dynamic centrifuge
models using high viscosity fluids. Geotechnical Testing Journal, 32(6),559-564.
Rayamajhi, D., Ashford, S. A., Boulanger, R. W., & Elgamal, A. 2016a. Dense granular columns in
liquefiable ground. I: Shear reinforcement and cyclic stress ratio reduction. J. of Soil Dynamics and
Earthquake Eng., vol. 142, no. 7, pp. 4016023.
Rayamajhi, D., Boulanger, R. W., Ashford, S. A., & Elgamal, A. 2016b. Dense granular columns in
liquefiable ground. II: effects on deformations. Journal of Geotechnical and Geoenvironmental Engin-
eering, 142(7), 04016024.
Stewart, D. P., Chen, Y. R., & Kutter, B. L. 1998. Experience with the Use of Methylcellulose as a Vis-
cous Pore Fluid in Centrifuge Models. Geotechnical Testing Journal, 21(4): 365-369.
Tan, T. S. & Scott, R. F. 1985. Centrifuge scaling considerations for fluid-particle systems. Geotechnique
35(4),461–470.
Taylor, R. N. 1995. Geotechnical centrifuge technology, 1st Ed., Blackie Academic and Professional.
New York; London.
Vytiniotis, A., & Whittle, A. J. 2017. Analysis of PV Drains for Mitigation of Seismically Induced
Ground Deformations in Sand Slopes. Journal of Geotechnical and Geoenvironmental Engineering, 143
(9), 04017049.
5289
