Академический Документы
Профессиональный Документы
Культура Документы
Abstract: This paper reports the results of two centrifuge tests that were conducted to evaluate the effectiveness of colloidal silica for lique-
faction mitigation. Colloidal silica has been selected as a stabilizer material in soils because of its permanence and ability to increase the strength
of soils over time. The centrifuge model geometry was selected to study the effects of lateral spreading in a 4.8-m-thick liquefiable layer overlain
by a silty clay sloping toward a central channel. The centrifuge test evaluates the response of untreated loose sands versus loose sands treated
Downloaded from ascelibrary.org by Xavier Vera on 11/10/12. Copyright ASCE. For personal use only; all rights reserved.
with 9, 5, and 4% colloidal silica concentrations (by weight). The models were subjected to a series of peak horizontal base accelerations ranging
from 0.007 up to 1:3g (prototype) with a testing centrifugal acceleration of 15g. The results show a reduction in both lateral spreading and
settlement in colloidal silica–treated sands versus untreated sands. The shear modulus at low strains was determined from shear wave velocity
measurements for the untreated and treated loose sands. The hysteretic response during cyclic loading was also determined for various levels of
shaking. The results from the centrifuge tests show an increase in cyclic resistance ratios and a decrease in cyclic shear strains for increasing
colloidal silica concentrations. DOI: 10.1061/(ASCE)GT.1943-5606.0000703. © 2012 American Society of Civil Engineers.
CE Database subject headings: Soil liquefaction; Centrifuge models; Strain.
Author keywords: Colloidal silica; Liquefaction; Centrifuge testing; Shear strain.
The purpose of the centrifuge model tests was (1) to compare the
liquefaction and deformation response of loose sands treated with
various CS concentrations (4, 5, and 9% by weight) with untreated
sands in the liquefiable layer and (2) to characterize the effects of
improvement in terms of accelerations, displacements, and the hys-
teretic stress-strain response. Each centrifuge model consisted of two
Fig. 2. Formation of siloxane bonds as a colloidal silica particle gel slopes that sloped 3 toward a 3-m-wide central channel (Fig. 4). In
(after Moridis et al. 1995) this way, four configurations could be investigated using two models.
The slopes were constructed with the following three distinct layers:
(1) a bottom layer of dense Monterey No. 0/30 sand with a prototype
thickness of 0.75 m overlain by (2) a 4.8-m-thick liquefiable layer of
loose Nevada sand that was overlain by (3) a 1.0-m-thick layer of
compacted Yolo loam (Fig. 4). He data and results are presented in
prototype units hereafter, unless otherwise noted.
In Test 1 (CTC01), one slope was treated with 9% CS (denoted as
CS-9), while the second slope was left untreated (Fig. 4). In Test 2
(CTC02), one slope was treated with 4% CS (denoted as CS-4) and
the second slope was treated with 5% CS (denoted as CS-5) (Fig. 4).
Laboratory tests using CS concentrations of less than 5% have been
shown to gel; however, they do not achieve a rigid gel (i.e., they do
not reach Gel State 11). The purpose of using 4% CS in CTC02
was to compare the effectiveness of liquefaction mitigation for
a weak gel to a strong gel that had achieved the majority of its
strength gain.
Fig. 3. Typical gel time curve of colloidal silica at various normalities
Model Preparation
Use of Colloidal Silica for Liquefaction Mitigation
Soil and Sensor Placement
Laboratory, bench-scale model, and full-scale field tests have
demonstrated the effectiveness of CS treatment against liquefaction Both models were built inside a flexible shear-beam container. The
and liquefaction-induced damage. Gallagher and Mitchell (2002) sand layers were placed in several lifts using a dry pluviation tech-
conducted cyclic triaxial shear tests on loose sand treated with 5– nique with a calibrated pluviator. Each sensor was placed at the
10% by weight CS and the results showed a dramatic increase in desired location during model construction and the model container
deformation resistance. Kodaka et al. (2005) compared the hysteretic was weighed at each lift to confirm the relative density of the soil.
stress-strain behavior of CS-treated sand and untreated sands using The average relative densities in the loose sand layer were estimated
cyclic torsional shear tests and showed the CS-treated sand initially to be 35 and 45% for CTC01 and CTC02, respectively.
demonstrated behavior similar to that of dense sand with significant Accelerometers, pore pressure transducers, linear potentiometers
damping and cyclic mobility with continued loading. (LPs), and linear variable differential transformers (LVDTs) were
The small-strain dynamic behavior of CS sands from resonant placed at the locations shown in Fig. 4. The instrumentation in the
column tests showed higher G for treated soils compared with un- liquefiable layer consisted of 10 piezoelectric accelerometers (PCB
treated soils with minimal differences between CS concentrations Piezotronics, Inc.) and nine pore pressure transducers (PDCR-81
(Spencer et al. 2008). The G for each CS concentration appeared to Series by Druck, Inc.) placed in a vertical array on each side of the
be unaffected up to strains of about 1 3 1023 % and then gradually model. The LPs (Duncan 600 Series by BEI Technologies, Inc.) and
decreased at higher strain amplitudes from about 65 to 45 MPa. the LVDTs (DCTH Series by RDP Group) were placed on the
surface of the model to measure the surface settlement and lateral Table 1. Colloidal Silica Properties
spreading. Bender elements (PiezoSystems, Inc.) were placed within Colloidal silica
the liquefiable layer to measure the shear wave velocity. Open steel concentration CS-9a CS-5b CS-4c
tubes were also placed vertically on each side of the model, which
were filled with colored sand and subsequently used to observe the Normality (N) 0.24 0.33 0.27
lateral movement of the soil profile, including the relative movements Time at Gel State 3 (h) 13 15 84
at the interface between the crust and the loose sand after testing. Resonating time (h)b 16 26 120
N* gel time at maximum 4 4 NA
strength gainc
Colloidal Silica Preparation and Saturation N* gel time at time of 10 12 NA
The target gel state of the CS was selected to allow ample time to shaking
saturate the treated zones of the model before the viscosity increased Density (g/cc) 1.06 1.05 1.04
too much and to provide sufficient curing time prior to model testing. Initial viscosity (cP) 1.3 1.15 1.1
a
The gel times of the CS solution were controlled by altering the ionic Viscosity of solution measured by the Cannon Ubbelohde viscometer.
b
strength through the addition of sodium chloride (NaCl). Prior to Refers to Gel State 10 (Persoff et al. 1999).
c
model preparation, CS gel time tests were conducted to determine Defined as 4* resonating time or 4* Gel State 10 (Gallagher and Mitchell
the normality that resulted in the desired gel time for each CS con- 2002).
centration. The curing time to reach Gel State 10 was determined via
visual and manual inspection of laboratory specimens. The prop-
deionized water upon completion of CS-9 and placement of the crust
erties of each CS grout solution are summarized in Table 1.
(described subsequently). In CTC02, the preparation and saturation
The CS solution was prepared in a saturation tank. First, deion-
of CS-5 and CS-4 followed a similar approach as CS-9.
ized water was mixed with the appropriate amount of granular NaCl
Once saturation was complete, the vacuum pressure was slowly
(Fisher Scientific) using an air circulation pump. Red food dye was
released from the model and the height of the fluid was carefully
added to enable visual inspection of the treatment coverage during
monitored. While the vacuum pressure was released, the solution
model dissection. The appropriate volume of Ludox SM-30 was
elevation decreased, indicating the solution was continuing to fill
added and the entire batch was mixed.
in dry pockets within the liquefiable layer. The vacuum release was
Prior to saturation, a vacuum pressure of approximately 90 kPa
stopped and more CS solution was added to the model. The process
was applied and the model was flooded with carbon dioxide gas to
was repeated until no further drop in fluid height was observed to
displace any air trapped in the soil voids. Troughs with saturation
ensure complete saturation. Saturation occurred at 1g; however,
tubes were mounted on the north and south ends of the container. The
grouting with CS under increasing levels of confinement resulted in
saturation tubes extended to the bottom of the model container. The
increasing strength in the sands (Kodaka et al. 2005). The strength
fluid was introduced in the troughs under vacuum pressure from
gains provided by CS likely would have been higher had they been
a deaired water chamber. For CTC01, the plan was to drip the CS
allowed to cure under the testing acceleration of 15g.
solution into the north trough, after which it would flow through the
saturation tubes such that saturation would occur from the bottom up.
When saturation was near completion, the injection rate decreased
Crust Placement and Saturation of the Untreated Side
dramatically and the saturation procedure was modified such that the
solution flowed over the lip of the trough and saturated the model from The crust was constructed using Yolo loam sieved to pass a #10 sieve.
the top down. The untreated side of CTC01 was saturated with The crust was placed at the optimum water content of about 12–15%
Fig. 6. Model surface: (a) before the test; (b) after the test Fig. 8. Cumulative surface settlement for Shakes 1–5
CTC02, the surface settlements were minimal overall because of the spreading on the boundary of CS-4 is slightly greater than that of the
fact that liquefaction did not occur on either side of the model and untreated sand as a result of the significant cracking that occurred
slope failure did not occur on the crust. However, the surface set- along the boundary where the sensor was located.
tlements were slightly greater in CS-4 compared with CS-5. At the
Pore Pressure and Acceleration
The pore pressure was measured on the untreated side of the model
within the liquefiable layer to identify the occurrence of liquefaction.
The pore pressure ratio, rU , was computed for the untreated sand as the
ratio of excess pore pressure to the initial vertical effective stress. The
pore pressure data of the CS soils are not reported in this paper. The
mechanism for generating pressure on CS gel in the sand pores is not
yet well understood and requires further study. Although liquefaction
is conventionally defined as the time when rU 5 1:0, the significance
of rU for CS sands does not represent the same physical meaning.
The acceleration and pore pressure responses at the midpoint of the
untreated liquefiable layer are shown in Fig. 10 for Shakes 2–5. In
Shake 2 the model was subjected to a peak base acceleration (PBA) of
0:03g. Fig. 10 shows that rU remains low and the acceleration reveals
a uniform response similar to the base shaking motion. During Shake 3
(PBA 5 0:1g), liquefaction was initiated 3 s into the shaking event. At
this same time, amplification was observed in the acceleration record
followed by deamplification in the subsequent cycles. This same
pattern was observed during Shakes 4 and 5; however, the amplifi-
cation and deamplification responses became more dramatic. Based on
the acceleration and pore pressure responses in Fig. 10, the untreated
loose sand layer continued to liquefy for successive shaking events.
From the acceleration records, the amplification in the acceleration
response occurred when rU 51:0, followed by deamplification.
Fig. 9. Cumulative lateral spreading for Shakes 1–5 Fig. 11 compares the acceleration response at the midpoint of the
untreated and treated liquefiable layers for Shakes 2–5. Fig. 11 shows
Fig. 10. Acceleration and pore pressure response at the midpoint of the untreated liquefiable layer for Shakes 1–5
Fig. 11. Acceleration response at the midpoint of the untreated liquefiable layer, CS-9, CS-5, and CS-4 for Shakes 2–5
a uniform response for each soil type during Shake 2 when the applied In CTC01, the VS measurements were planned to be determined
shaking amplitude was small (PBA 5 0:03g). When the shaking using bender elements. However, the bender elements failed; there-
amplitudes were increased in Shakes 3 and 4, the acceleration response fore, an alternative method utilizing the centrifuge shaker was pro-
in the treated soils showed a relatively uniform response in terms of the posed. The shaker was used to generate a single 0.1 s (model scale) sine
acceleration amplitudes compared with that of the untreated soil. On the wave at the base of the model container, which was recorded in vertical
other hand, the response in Shake 5 showed nonuniformity and high- accelerometer arrays (Fig. 4) with a sampling rate of 20,000 Hz. The
frequency amplification spikes in CS-4 and CS-5 compared with CS-9. pulse wave had an average PBA of 0:025g. As the shear wave traveled
In general, the deamplification of the acceleration response ob- through the liquefiable soil, the vertical array of accelerometers picked
served in the untreated sand is directly related to a change in the ratio up the signal and captured the wave propagation through the profile
between input frequency (which was constant) and the soil profile’s (Fig. 12). The distance between the two signal measurement points, L,
natural frequency (which progressively decreased). The significance is the difference in elevation between accelerometers. The travel time,
of these trends indicates that the stiffness in the untreated sand is t, was taken as the time shift of the pulse wave from successive
continuously degrading for each loading cycle. On the other hand, accelerometers in the array using cross-correlation methods. Prior to
the trends in the acceleration response for CS-9 suggest strength and cross correlation, a seventh-order band-pass Butterworth filter was
stiffness is being maintained for cyclic loading up to 0:56g. The applied to the original records to remove higher-frequency noise.
amplification that occurred in CS-5 and CS-4 in Shake 5 suggests In CTC02, bender elements were used to estimate small-strain VS .
changes in the behavior of the grouted medium. The bender element tests used the signal stacking procedure by
Brandenberg et al. (2008). Ten pulses were sent out in quick suc-
Shear Wave Velocity cession and the signals were stacked to improve the signal-to-noise
ratio (SNR). Once the SNR was satisfactory, the multiple waves were
The shear modulus degradation curve is characterized using both the stacked into a single signal. The data were transmitted and received
maximum shear modulus, Gmax , and the variation in modulus ratio, through a data acquisition system and processed from a built-in
G/Gmax , with cyclic strain amplitude; Gmax can be obtained by software program (i.e., LabVIEW). A square source wave (9 V am-
measuring the shear wave velocity (VS ) at low strain amplitudes by plitude) was sent and sampled at a frequency of 90,000 Hz. Here, L is
the tip-to-tip distance from the source to the receiving bender and
Gmax ¼ r × Vs2 ð1Þ t is the time of first arrival of the receiving signal (Fig. 13). The source
and receiver benders were placed approximately 100 mm apart at four
where VS is measured as different elevations in the liquefiable soil (Fig. 4). The duration of the
bender element test was approximately 2 ms.
The values of VS from CTC01 and CTC02 cannot be directly
VS ¼ L ð2Þ
t compared because of the differences in the measurement methods.
The bender element tests typically induced strains lower than 3 3
where L 5 effective distance between two signal measurement points 1024 % while the shear strains induced from the pulse wave were
and t 5 travel time for the shear wave to travel through the medium. determined to range about 0.01%. This difference in strains can
Stress-Strain Analysis
Fig. 13. Travel time determination from the Bender element tests Calculation of Shear Stress
The shear stresses were determined at the midpoints between the
accelerometers in the vertical arrays (Fig. 14). The calculations fol-
cause a 30% difference in the shear modulus as demonstrated by lowed 1D shear-beam behavior where the shear stress at any depth, z,
Spencer et al. (2008). The step-wave technique used in CTC01 also can be obtained by integration of the soil density, r, times acceler-
had lower resolution, in part because of limits on the sampling ation, a
frequency and input wave frequency. In general, both methods can
ðz
be used to assess trends in G; however, the bender elements offer
a more accurate means to obtain Gmax . tðzÞ ¼ r × a × dz ð3Þ
The VS and G measurements are summarized in Table 4 for CTC01 0
and CTC02, respectively, after a small shaking event (PBA 5 0:007g)
was applied to the model. For CTC01, VS of the untreated sand at 30 Unlike field data, surface accelerations (when z 5 0) cannot be
kPa confinement was measured as 174 m/s. The results are generally directly measured in centrifuge models because the instruments re-
consistent with the studies by Kasantikul (2009), which reported VS quire sufficient contact with the soil and must be buried within the
160–180 m/s for loose-to-medium sand at 30 kPa confinement. The model. As a result, rigid body motion in the centrifuge was assumed
G for CS-9 showed an increase of about 12% compared with the for the first soil element in the array. This is a reasonable assumption
untreated sand. These results are also consistent with the initial small- because the distance from the surface to the first accelerometer was
strain G obtained through the resonant column tests for loose Nevada only about 0.8 m. Based on rigid body motion, the acceleration at the
sand treated with CS (Spencer et al. 2008). The values of VS from surface is considered equal to the acceleration measured at Node 1
CTC02 indicated an increase of about 14% in G between CS-4 and from Fig. 14. The equations used for calculation of shear stress are
CS-5. Although strength gains in soils are reported to increase with presented in Kamai and Boulanger (2010). The shear stresses were
increasing CS concentration, such large differences between the two normalized by the initial effective overburden pressure at the be-
treatment concentrations are not expected to occur. The discrepancy is ginning of each shake to establish the CSR. The CSR was determined
likely a result of the fact that CS-4 represents treatment with a weak as the ratio of shear stress at the midpoints of each element by the
grout characteristic of a highly deformable, nonflowing gel. corresponding initial effective vertical stress at a similar depth.
Fig. 15. CSR versus shear strain for Shakes 3–5 at the midpoint of the liquefiable layer (D 5 3:1 m) for untreated, 9% CS, and 5% CS
stress-strain path observed from Cycles 6 and 19 may be a result of the 0.7, 0.4, and 0.3% for the untreated sand, CS-5, and CS-9, respectively.
Fig. 16. CSR versus shear strain for Shake 3 (PBA 5 0.1–0:15g)
Fig. 17. CSR versus shear strain for Shake 5 (PBA 5 0.56–0:69g)