Sunteți pe pagina 1din 11

Soil Dynamics and Earthquake Engineering 19 (2000) 583593

www.elsevier.com/locate/soildyn

Experimental developments for studying static and seismic behavior of


retaining walls with liqueable backlls
M.M. Dewoolkar*, H.-Y. Ko, R.Y.S. Pak
Department of Civil, Environmental and Architectural Engineering, Engineering Center ECOT 441, Campus Box 428,
University of Colorado at Boulder, Boulder, CO 80309-0428, USA
Accepted 26 July 2000

Abstract
The effects of earthquakes on cantilever retaining walls with liqueable backlls were studied. The experimental techniques utilized in this
study are discussed here. A series of centrifuge tests was conducted on aluminum, xed-base, cantilever wall models retaining saturated,
cohesionless backlls. Accelerations on the walls and in the backll, static and excess pore pressures in the soil, and deections and bending
strains in the wall were measured. In addition, direct measurements of static and dynamic lateral earth pressures were made. In some tests,
sand backlls were saturated with the substitute pore uid metolose. Modeling of model type experiments were conducted. The experimental
measurements were found internally consistent and repeatable. Both static and dynamic earth pressure measurements were determined to be
reliable. It was also observed that for the test conguration adopted, a special boundary treatment such as the use of duxseal is optional. Static
and seismic modeling of models were also successful, which indicated that the assumed scaling relations were essentially correct. q 2000
Elsevier Science Ltd. All rights reserved.
Keywords: Retaining wall; Earth pressure; Soilstructure interaction; Liquefaction; Earthquake; Seismic; Centrifuge; Earth pressure transducer

1. Introduction
In the past decades, considerable research has been
conducted to study earthquake related retaining wall
problems. These investigations can be broadly divided in
two categories, namely, theoretical studies and experimental studies. Theoretical studies include analytical and
numerical methods. Experimental studies include shaking
table tests performed under earth's gravity (1g), seismic
centrifuge tests, and any rare occurrences of eld testing.
Most analytical and design procedures such as the ones
suggested by Mononobe and Okabe [13,14], Matsuzawa et
al. [12], and Ishibashi and Madi [7] are generally based on
semi-empirical methods or involve simplistic assumptions.
In most numerical methods, complex theories involving
soil/water interaction incorporating elasto-plastic constitutive models are used. Therefore, such analytical, design, and
numerical methods need to be veried. Validation of such
theoretical studies can be done by comparing their results
with eld measurements. However, real seismic events are
* Corresponding author. Present address: GEI Consultants Inc., 6950 S.
Potomac St., Suite 200, Englewood, CO 80112, USA. Tel.: 11-303-6620100; fax: 11-303-662-8757.
E-mail address: mdewoolkar@geiconsultants.com (M.M. Dewoolkar).

unpredictable and eld conditions are often characterized


with signicant uncertainties. Full scale structures are
generally not adequately instrumented to capture the
complete scenario. An effective alternative is laboratory
methods.
Several laboratory 1g shaking table tests on retaining wall
models have been performed in the past. Some of the early
tests suffered from inadequate instrumentation, unrealistic
frequencies and amplitudes of input vibrations, and lack of
plane strain conditions, which are assumed in most analytical and design methods [15]. These tests generally indicated that the MononobeOkabe (MO) method gives the
magnitude of the total resultant force reasonably well [15]
and the incremental dynamic earth pressure acts at somewhere between 0.45 and 0.55 H from the base depending on
the wall movement [12], where H is the wall height.
However, 1g tests suffer from a lack of similitude of stress
levels between small scale models and any realistic large
scale structures. Because important static and dynamic soil
properties are dependent on effective stress level, the applicability of small scale 1g shaking table tests for quantitative
assessment is limited although their results are useful in
examining qualitative behavior.
This deciency can be overcome with the use of the
centrifuge modeling technique in which scaled models are

0267-7261/00/$ - see front matter q 2000 Elsevier Science Ltd. All rights reserved.
PII: S 0267-726 1(00)00069-5

Downloaded from http://www.elearnica.ir

584

M.M. Dewoolkar et al. / Soil Dynamics and Earthquake Engineering 19 (2000) 583593

subjected to predetermined acceleration levels to produce


similitude conditions, satisfactorily in most situations.
2. Centrifuge modeling
In a centrifuge rotating at an angular speed v , high centrifugal acceleration levels are created in a scaled model
located at a radius R, producing an acceleration N times
earth's gravity g, where N Rv2 =g: The principle behind
centrifuge modeling is to create a stress eld in a
geometrically similar model, identical to that in a real or
hypothetical prototype so that the stressstrain relationships
at homologous points in the two systems will be the same if
prototype materials are used in the model. A servocontrolled shake table can then be used to simulate any
desired earthquake events in the centrifuge.
Owing to the different time scaling relations for dynamic
(N) and diffusion (N 2) phenomena, a substitute pore uid N
times more viscous than water is useful to reduce the soil
permeability by a factor N, in order that a common time
scale can be used for interpreting the test results. The use
of the more viscous pore uid can slow down the uid ow
rate by a factor N, so that pore pressure generation and
dissipation can both proceed at the same rate, which is N
times greater than that in the prototype.
The ideal way to verify the scaling relations involved is to
compare model test results with the actual prototype behavior. In the absence of a prototype, the behaviors of two or
more models of the same prototype scaled corresponding to
the appropriate g-levels can be compared to each other using
the scaling relations. If the results are found internally
consistent, their model behaviors can be extrapolated to
prototype response. This technique is referred to as modeling of models.
3. Techniques used in previous centrifuge studies on
retaining walls
In the 1980s, the technologies of modeling earthquakes in
centrifuges and data acquisition techniques were fairly rudimentary. Those centrifuge experiments involved relatively
simple model congurations and only a few transducers. For
example, studies conducted by Ortiz et al. [15] and Bolton
and Steedman [2] employed toggle mechanism and bumpy
road methods to generate earthquakes in the centrifuges at
Caltech and Cambridge, respectively. Both investigations
were conducted on cantilever wall models retaining dry,
cohesionless backlls. In the late 1980s, with advances in
data acquisition and earthquake simulation techniques,
experiments on exible [20], anchored [25], and rigid
walls [1,23,24] with a relatively greater number of transducers were conducted. In some of these studies, walls with
water-saturated backlls were examined. Steedman and
Zeng [20] and Whitman and Ting [23] employed substitute
pore uids (silicone oil in the former case and glycerol in the

latter case) in some of their tests. After the 1995 HanshinAwaji earthquake, severe damage was observed at many
port facilities due to failures of numerous quay walls in
the Kobe area. Several investigators [6,8,16,26] from
Japan studied the stability of quay walls using centrifuge
modeling on very similar and more realistic model congurations. Substitute pore uids were used in all four of these
investigations.
Zeng [24], Stadler [19], and Fujiwara et al. [6] used
special materials at the far end container boundary in
contact with the soil to minimize stress-wave reections.
Only Zhang et al. [26] used a laminar container. However,
these investigations did not demonstrate the necessity or
usefulness of such special boundary treatments.
As to the explicit demonstration of the internal consistency,
modeling of models was attempted only in two studies [2,19].
Both investigations were conducted on xed-base cantilever
wall models with dry, cohesionless backlls.
Although determination of the magnitude and distribution
of seismic earth pressures is very important in earthquakeresistant design of retaining walls, most studies did not attempt
to measure lateral earth pressures on model walls. Andersen et
al. [1] and Whitman and Ting [23] tested tilting wall models
retaining dry and saturated sand backlls, respectively. In both
congurations, dynamic thrusts on the walls were deduced
from indirect measurements from load cells, accelerometers,
and pore pressure transducers by using equilibrium equations.
Bolton and Steedman [2] and Steedman and Zeng [20]
deduced bending moments in the walls using strain gage
measurements. The bending moments were compared with
the moments calculated based on the MO theory. Zhang et
al. [26] qualitatively assessed earth pressures on the model
wall using pore pressure measurements near the wall and
some other experimental and theoretical observations.
Ortiz et al. [15] and Stadler [19] used total stress gages to
measure lateral earth pressures from dry sand on cantilever
retaining wall models. Ortiz et al. [15] reported two tests. In
the rst test, lateral earth pressures were not measured
directly. Measured bending moments were differentiated
twice to deduce lateral pressure distribution; however, the
results were considered inaccurate owing to the propagation
of errors inherent in differentiation. In the second test,
Entran Model EPF-200-50 pressure transducers were used.
Their earth pressure measurements seemed reasonable;
however, only four pressure transducers were used. Stadler
[19], using the same transducers as the ones used in the
study presented in this paper, conducted fourteen tests to
measure lateral earth pressures from dry backll. The two
studies were performed simultaneously using the same wall
models, soil, sand pluviation technique, container, centrifuge, shake table, and the type of shaking motion. The
only difference was the presence of the pore uid. Based
on his test results, Stadler [19] concluded that the transducers selected to measure lateral earth pressures did not
perform to expectation and that the lack of good direct
pressure measurements was disappointing and directly

M.M. Dewoolkar et al. / Soil Dynamics and Earthquake Engineering 19 (2000) 583593

affected the way in which the experimental results were


examined. The speculated reasons behind the erratic
measurements were: (1) only about 10% of the total range
of these gages were utilized; (2) owing to the small size of
the transducer, the number of grains of dry sand impinging
against the pressure sensing membrane was most likely not
the same from point to point and test to test; (3) localized
arching of the soil around the gages; and (4) while the global
relative density remained the same from test to test, there
might have been local density variations around the gages.
Kamon et al. [8] and Fujiwara et al. [6] used three transducers to measure seismic lateral earth pressures on the
model walls from saturated backlls. The types and specications of these transducers were not reported. Satoh et al.
[16] used load cells attached to pressure-receiving plates of
size 10 mm 10 mm at four locations on the wall model to
measure seismic lateral pressures from saturated soil backll. In all three investigations, earth pressure measurements
were taken only at three to four locations and, therefore, had
limited potential for accurately determining the magnitude
and location of lateral thrusts.
Several static centrifuge retaining wall tests were also
conducted in the past. Schcherbina [17] conducted static
centrifuge tests on retaining walls to investigate the lateral
earth pressure as a function of soil density and wall base
rotation. In this study, earth pressure cells (pressure strain
gauges within the range of 03 and 06 kg/cm 2) were
used. The type of transducer was not specied. Lyndon
and Pearson [11] studied a method of measuring static
lateral earth pressures by using pressure-beams on
model walls in centrifuge. Their preliminary results showed
good potential in this method. Schurmann and Jessberger
[18] conducted static centrifuge tests on models of sheet pile
walls in dry sand. Static earth pressure distribution was
obtained by differentiating measured bending moments
derived from strain gages twice. Egan and Merrield [5]
developed two similar types of earth pressure cells of
about 32 mm in outer diameter and about 2 mm in thickness.
The performance of these transducers was examined by
evaluating their measurements in at-rest conditions from
static centrifuge experiments and showed promising results.
Egan and Merrield [5] did not report data from seismic
experiments. The size of the transducers was relatively
large to accommodate in small size centrifuge models.
4. Goals of present study
The lack of reliable, direct lateral earth pressure measurements led to general concern about the applicability of pressure cells. Both centrifuge investigators and general
experimenters are usually reluctant to make direct measurements of soil pressures. Weiler and Kulhawy [22] identied
fourteen factors that may affect the accuracy of soil pressure
measurements in any experiment. The factors can be
grouped into four main categories [5]: (i) inclusion effects

585

related to the disturbance of the stress eld arising from the


presence of a cell; (ii) cell/soil reaction, a function of the
relative stiffness of the cell with respect to the soil; (iii)
placement effects; and (iv) environmental inuences and
dynamic response.
For the seismic-resistant design of retaining walls, the
magnitude and line of action of lateral earth pressures, especially from saturated soil, are still in question. To advance the
current understanding of the issue, the feasibility of measuring
lateral earth pressures reliably was studied in this research.
Entran Flatline Pressure Transducers, Model No. EPL-200100S, were used to measure static and dynamic lateral earth
pressures generated from saturated, cohesionless soil backlls.
These transducers are addressed as earth pressure transducers herein. These transducers consisted of semiconductor
strain gages bonded to a circular, stainless steel sensing
membrane. This model had a range of 0100 psi
(6900 kPa). The diameter of the sensing membrane was
4.6 mm. The transducers were housed in epoxy potted seals
and mounted in paddle shaped units. The nominal overall
dimensions were 10 5.1 1 mm 3.
In addition, modeling of models type of experiments were
conducted to check the internal consistency of the measurements and to validate the scaling relations involved. Also,
repeatability of measurements and experimental procedures
was checked by conducting two very similar experiments.
5. Centrifuge model conguration
The tests were conducted using the 400g-ton geotechnical
centrifuge [10] at the University of Colorado at Boulder.
Earthquake-like motions were generated using the servocontrolled, electro-hydraulic shake table [9] on board the
centrifuge. Base motions consisted of twelve roughly sinusoidal cycles at 1 Hz prototype frequency.
A typical model conguration is depicted in Fig. 1a. The
particular experimental conguration shown in Fig. 1a was
adopted for test MMD12. The congurations of other tests
were very similar to that of test MMD12.
To simplify the analysis, model walls were bolted to the
container base to simulate a xed-base condition. Thus, only
the exural mode of wall movement was allowed. Owing to
difculties in replicating reinforced concrete in small
scale models, the wall models were made of T6061-T6
aluminum plates (density 2787.7 kg/m 3, Young's
modulus 69 10 6 kPa and Poisson's ratio 0.3). The use
of a different modeling material for the model walls was not a
concern since structural failure of the wall was not the focus of
this investigation. Flexural stiffnesses of the model walls were
scaled to represent hypothetical reinforced concrete cantilever
retaining walls. The pertinent scaling relations are listed in
Table 1. Young's modulus of aluminum is about twice that
of reinforced concrete. The mass density of aluminum is about
10% higher than that of reinforced concrete introducing a
small error in the inertial behavior of the model. Since this

586

M.M. Dewoolkar et al. / Soil Dynamics and Earthquake Engineering 19 (2000) 583593

Fig. 1. A typical test conguration (test MMD12).

error is relatively insignicant, the scaling relations given in


Table 1 are approximately valid for dynamic situations as well.
The backll was contained in a latex membrane bag. Oilsoaked latex membranes were used between the container
walls and the latex bag containing the backll, in order to
minimize friction along the container boundaries. A layer of

absorbing material called duxseal was used to minimize


unwanted stress-wave reections from the far end container
boundary. A physical gap of 1.6 mm was maintained
between the model wall ends and the side walls of the
container to allow unrestricted movement of the model
under plane strain conditions.

M.M. Dewoolkar et al. / Soil Dynamics and Earthquake Engineering 19 (2000) 583593
Table 1
Scaling relations used in the interpretation of test results (N ; g-level and
E ; Young's modulus, subscript m ; model quantity and subscript p ;
prototype quantity), Em =Ep 2
Quantity
Wall quantities
Lateral earth pressure
Lateral thrust
Lateral thrust per unit width
Bending moment
Bending moment per unit width
Flexural stiffness
Flexural stiffness per unit width
Thickness
Width
Height
Acceleration
Time
Frequency
Deection
Bending stress
Bending strain
Soil quantities
Excess pore pressure
Acceleration
Time
Frequency
Backll surface settlement

Prototype

Model

1
N2
N
N3
N2
N4
N3
(Em/Ep) 1/3N
N
N
1
N
1
N
(Ep/Em) 2/3
(Em/Ep) 1/3

1
1
1
1
1
1
1
1
1
1
N
1
N
1
1
1

1
1
N
1
N

1
N
1
N
1

After mounting the instrumented wall in the container,


oil-soaked latex membranes and a membrane bag were
placed. Dry soil was then pluviated through a hopper from
a predetermined height and opening size to achieve 60%
relative density. The backll soil used was Nevada No.
100 sand (uniform, sub-rounded, medium ne to ne sand,
D50 0:1 mm; specic gravity of 2.67, and maximum and
minimum unit weights of 17.33 and 13.87 kN/m 3, respectively). The sand was tested extensively by The Earth Technology Corporation [21]. The backll was saturated with
either water or methylcellulose solution (metolose). This
solution is obtained by mixing a very small portion (less
than 2% by weight) of metolose powder in distilled, deaired
water to obtain solutions of 40 and 60 cSt viscosities (viscosity of water at 208C is 1.007 cSt) to be used in tests at 40
and 60g, respectively.
It was determined that constitutive properties of sand
remain fairly unaltered with the use of metolose instead of
water as the pore uid and the scaling requirements of
permeability are also satised by metolose [4]. If it is
assumed that the hypothetical prototype backll is made
of water-saturated Nevada No. 100 sand at 60% relative
density, then the scaling relations for the soil in Table 1
can be assumed to be valid for the tests in which metolose
was used as a pore uid.
A total of three model walls was used in the tests
discussed herein. Their dimensions are shown in Fig. 1b.
The stems of the walls were welded to the base plates. Tests
were conducted at 40g on walls A and B and at 60g on wall

587

C. Thus, all three models represented hypothetical 9.14-m


(30-ft) tall prototype walls. Walls B and C were 1.5 times
thicker than wall A and in terms of the exural stiffness,
walls B and C were 3.375 times stiffer than wall A. Walls B
and C were used to conduct modeling of models.
Accelerations on the walls were measured using miniature accelerometers (PCB Piezotronics Inc. models 303A11
and 303A12). Wall deections were measured using ACLVDTs (Schaevitz models MHR100, MHR250, and
MHR500). Bending strains were measured using strain
gages (manufactured by the Tokyo Sokki Kenkyujo
Company Ltd of Japan, type FLA-2-23-ILJB foil gages,
with a nominal resistance of 120 ohms). There were about
815 pairs of strain gages in a half bridge conguration
along the height of the wall. Earth pressure transducers
were used to measure static and dynamic lateral earth pressures on the walls along the center line at typically 1012
locations. These earth pressure transducers were carefully
ush mounted on the face of the wall. As an example, locations of earth pressure transducers, strain gages, accelerometers, and LVDTs on wall C are shown in Fig. 1c.
Accelerations, settlements, and pore pressures in the soil
backll were measured by miniature accelerometers (PCB
Piezotronics Inc. models 303A11 and 303A12), DC LVDTs
(Schaevitz Type 500 HR-DC), and pore pressure transducers (Druck Type PDCR81), respectively. Horizontal and
vertical accelerations on the shake table were measured
using PCB Piezotronics Inc. accelerometers model
308B03. The directions of the arrows for accelerometers
and LVDTs in Fig. 1a indicate the assumed positives for
the measured quantities.
The earth pressure transducers and pore pressure transducers were calibrated by two different methods. (1) The transducers were immersed in water contained in a sealed
cylindrical pressure vessel. Pressure was applied in increments to the pressure cell and the corresponding voltage
changes were noted to determine the calibrations. (2) The
model walls instrumented with earth pressure transducers,
strain gages, and LVDTs were tested in the centrifuge with
water as a backll. The pore pressure transducers were
placed at the bottom of the pool of water. Water was used
as a backll material since exact hydrostatic pressures could
be calculated. Also, with water as a backll, the behavior of
the wall and the transducers were believed to be unaffected
by friction, non-plane strain conditions, and localized arching effects. The centrifuge was spun up to a g-level which
would produce lateral pressures comparable to the pressures
that would develop during an earthquake event of interest.
Calibration constants for strain gages were obtained by
calculating bending strains in the wall subjected to the
known hydrostatic pressures. Typically, earth pressure
transducers, pore pressure transducers, and strain gages
showed a linear variation with the increasing water pressure
owing to the increasing g-level. LVDTs were calibrated
using a micrometer. Manufacturer's calibrations were used
for accelerometers.

588

M.M. Dewoolkar et al. / Soil Dynamics and Earthquake Engineering 19 (2000) 583593

6. Performance of earth pressure transducers and


internal consistency in the measurements
One of the key measurements in this study was the lateral
earth pressures provided by the earth pressure transducers. It was felt necessary to demonstrate that the earth
pressure transducer measurements were reliable. Also, to
build greater condence in the overall measurements, the
measured model wall quantities were checked for their
internal consistencies. For this purpose, two approaches
were considered. One could either differentiate the
measured bending strains to calculate earth pressures or
integrate the measured earth pressures to calculate bending
strains. Since numerical differentiation introduces greater
errors than numerical integration, the second option was
selected. The procedure is straightforward for static conditions. For the analysis of the shaking events, certain assumptions have to be made.
For this dynamic-soil-pore uid-structure interaction
problem, the inertia effects of a retaining wall, inertia of
the soil-uid mass, effects such as the damping characteristics of the soil and pore uid, and amplication or
attenuation of the base acceleration in the soil layer are
all inter-related and they can all affect the lateral pressures, accelerations, and deections of the wall. An
evaluation of the effect of each variable independently
is not feasible. Therefore, for simplication, it is assumed
that bending strains and deections in a wall could be
decomposed into two components, one resulting from
the earth pressures on the wall and the other resulting
from the inertia of the wall as a result of the relative
(with respect to the ground motion) wall accelerations.
It is assumed that the lateral pressures on the wall incorporate all other effects from such factors as inertial forces
from the soil resulting from seismic loading, damping
characteristics of the soil and pore uid, amplication or
attenuation of the base acceleration in the soil layer, and
possible effects of wall inertia on the soil. This approximation is acceptable since the primary purpose here is

to build condence in the measurements from the earth


pressure transducers by showing a reasonable agreement
between the computed and measured bending strains
and deections, rather than by obtaining an exact
match.
6.1. Mathematical formulation
A free-body diagram of the simplied problem of a cantilever wall is shown in Fig. 2. It is assumed that shear stresses
along the depth of the wall are negligible. This assumption
is acceptable because latex membrane reduces the friction
between aluminum and sand. Also, shear stress on the wall
is only a fraction of effective lateral pressure, which is less
than 30% of the pressure from the pore uid.
Bending moment Mx; t; shear force Vx; t; loading
 t and deection
px; t per unit width, acceleration ux;
ux; t are functions of space x and time t. The boundary
conditions are:
1: M0; t 0;
3: 2u=2xH; t 0;

2: V0; t 2M=2x0; t 0;
4: uH; t 0:

Considering moment and force equilibrium, it can be


shown that
Vx; t

2Mx; t
2x

22 Mx; t
 t
px; t 2 mux;
2x2

where m is the mass per unit length per unit width of the
wall. The above equation can be solved in space at each time
instance of the base motion time history since the pressure
and relative acceleration proles can be determined from
the measurements.
Integrating Eq. (2) twice with respect to x, the
moment at any depth and a particular time instance t1
is calculated as
Mx; t1

Zx Zj
0

pz; t 1 dz dj 2 m

Zx Zj
0


2M
0; t 1 x 1 M0; t1
2x

 z ; t 1 dz dj
u
3

where j and z are dummy integration variables.


The bending strain e b at a particular depth x1 and time
instance t1 can be obtained from the bending moment determined by Eq. (3) as

e b x1 ; t1

a
Mx1 ; t1
2EI

The curvature at any depth and a particular time instance


t1 is related to bending moment as
Fig. 2. Free-body diagram of the wall.

EI

22 u
x; t1 Mx; t1
2x2

M.M. Dewoolkar et al. / Soil Dynamics and Earthquake Engineering 19 (2000) 583593

589

Thus, at each time step, by using a suitable integration


scheme, proles of earth pressures and relative accelerations
on the wall can be obtained and integrated to compute bending strains and deections. Combining calculated strains
and deections at all time steps, complete time histories
of these quantities can be obtained.
6.2. Measurements and analysis

Fig. 3. Acceleration and lateral earth pressure measurements from test


MMD12 in model scale.

The deection at any depth and a particular time instance


t1 can be obtained as
"
# 

1 Zx Zj
2u
0; t 1 x
Mz; t 1 dz dj 1
ux; t1
EI
2x
0 0
1 u0; t1

The measurements from test MMD12 are presented in


Fig. 3 in model scale. The test was conducted on wall C
at 60g. The locations of transducers are shown in Fig. 1.
The horizontal and vertical accelerations measured on the
shake table are shown in Fig. 3a. The horizontal accelerations and lateral earth pressures on the wall during the shaking are shown in Fig. 3b and c, respectively. The initial
horizontal portion of the pressure time histories indicates
the static quantities before shaking, the oscillations are
during shaking, and the nal at portion is the post-shaking
residual value.
The horizontal base acceleration consisted of 12 roughly
sinusoidal cycles at 60 Hz. The vertical base acceleration
was about 40% of the horizontal acceleration in terms of the
magnitude; however, it had high frequency content (about
15 times that of horizontal motion) and insignicant energy
content. No signicant effects of such a high-frequency
vertical acceleration could be observed in the lateral earth
pressure, strain, and deection measurements. Therefore,
vertical base acceleration was not considered in the analysis.
Relative wall accelerations at arbitrary time instances
during shaking along the wall depth are plotted in Fig. 4a.
For the same time instances, lateral earth pressures are
plotted in Fig. 4b. As seen, linear regression was appropriate
to obtain lateral earth pressure proles along the depth of the
wall at each time instance. However, for relative horizontal
wall accelerations, linear interpolation being clearly inappropriate, third-order spline interpolation was used.
Before shaking, relative accelerations on the wall are
zero. Only lateral earth pressures contribute to the strains
and deections in the wall. The static earth pressures from
tests MMD12 are plotted in Fig. 5a(i). The best t straight
line was obtained by linear regression analysis of the data
points. Bending strain and deection proles were calculated (from Eqs. (3), (4) and (6), with u terms as zero)
using this best t pressure prole. Measured strains and
deections are superimposed on the calculated proles in
Fig. 5a(ii) and (iii), respectively. Calculated values were
smaller than the measured quantities because of the linear
regression of the measured pressures that calculated unrealistic negative pressures at the top of the wall in this particular test.
Dynamic (over and above the static) bending strains and
deections were calculated using Eqs. (3), (4) and (6). A
third-order spline interpolation for relative acceleration
prole and linear regression for lateral earth pressure prole
were obtained for each time step of data acquisition. The

590

M.M. Dewoolkar et al. / Soil Dynamics and Earthquake Engineering 19 (2000) 583593

Fig. 4. Relative accelerations and lateral earth pressures on wall-C in test


MMD12 in model scale.

Fig. 5. Comparison between measured and calculated, static and dynamic


strains and deections from test MMD12.

time interval was 0.0004 s corresponding to the sampling


rate of 2500 Hz. The time histories of calculated quantities
are compared with selected measured strains and deections
in Fig. 5b and c, respectively. As seen, the agreement was
very good. Fairly good agreement between measured and
calculated static and dynamic strains and deections was
also observed in seven other tests on walls A, B and C in
which a similar set of instrumentation was used.
Therefore, it was concluded that earth pressure transducers (Entran Flatline Pressure Transducers, Model No. EPL200-100S) can provide reliable data if they are used to
measure static and dynamic stresses from saturated cohesionless soils. The good agreement between measured and
computed strains and deections also suggested that the
quantities measured on the wall (lateral earth pressures,

bending strains, deections, and accelerations) were internally consistent.


Characteristics of the earth pressure transducers were
evaluated according to criteria suggested by Weiler and
Kulhawy [22] for diaphragm type pressure cells. Since the
earth pressure transducers were commercial products,
certain recommended correction methods [22] could not
be implemented. The transducers met most of the criteria
suggested by Weiler and Kulhawy satisfactorily. As
discussed earlier, the earth pressure transducers provided
reliable static and dynamic lateral stress measurements
from saturated cohesionless soils. In the experiments
conducted as a part of this study, dynamic pressures at
frequencies up to 60 Hz were measured. The same transducers were found to be unreliable when used to measure

M.M. Dewoolkar et al. / Soil Dynamics and Earthquake Engineering 19 (2000) 583593

pressures from Nevada No. 100 sand in dry condition [19].


The test conguration and measurement techniques were
very similar in the two investigations. The presence of
pore uids in the saturated sand tests improved the transmission of pressure from the soil to the sensing membrane of
the transducer. Also, up to about 40% of the total range of
these gages was utilized because of the signicant increase
in the measured pressures because of the presence of pore
uid. Some scatters in the measurements still exist. For
example, transducer EP3 (Fig. 3c) measured pressure
which seemed to be somewhat inconsistent with the rest
of the pressure measurements. However, when a regression
analysis was performed to calculate strains and deections,
such inconsistencies in one or two measurements did not
have signicant effects on the calculations provided that a
signicant number of pressure measurements were made. If
test conditions permit, their measurements should be veried with some other redundant measurements.
7. Repeatability
Two seismic centrifuge tests (MMD1 and MMD2) on
wall A were performed at 40g. The backlls consisted of
water-saturated Nevada No. 100 sand at 60% relative
density. Test MMD1 was identical to test MMD2, except
it did not have duxseal panel at the end of the backll.
Therefore, the backll in test MMD1 was 3.2 cm longer
than in test MMD2.
Static measurements from the two tests were very similar,
which indicated that the static measurements were repeatable. The presence of duxseal boundary was not critical in
such static conditions.
Shaking motions in the two tests were fairly similar.
Dynamic measurements from the two tests were very
comparable. In addition to conrming repeatability, the
good agreement also indicated that the duxseal boundary,
in the form it was used, was not absolutely necessary for this
particular model conguration, soil type, the type of input
motion, and if the length of the backll is sufciently longer
than the height of the wall. In this case, the ratio of backll
length to wall height was about 2.4. However, it would still
be advisable to use a layer of absorbing material at the far
end container boundary as a precautionary measure since
the installation is easy.

591

was 2.4. In test MMD12, this ratio was 3.5. In test MMD6,
the sand was saturated with 40-cSt metolose and the test
was conducted at 40g. In test MMD12, the viscosity of
the metolose solution was 60 cSt and the test was conducted
at 60g. Transducer locations were selected such that they
would correspond to the same prototype locations. To facilitate a direct comparison, selected results from these tests
are presented at prototype scale according to the scaling
relations listed in Table 1. Detailed comparisons are made
in Ref. [4].
Static lateral earth pressures, bending strains, and deections of the walls are compared in Fig. 6a. As seen, the
agreement was very good. Therefore, it was concluded
that static modeling of models was achieved.
Fig. 6b shows selected dynamic measurements from tests
MMD6 and MMD12. The plots include horizontal base
motion (accelerometer AC1), dynamic tip deection (transducer LV1), dynamic bending strain towards the bottom of
the walls, and dynamic lateral thrust and its line of action.
As was determined by Fig. 4b, linear regression through
earth pressure data points was appropriate to calculate the
lateral thrust and its line of action on the wall.
The base input motions were intended to be the same;
however, there were some differences in terms of magnitude
and shape. In terms of frequency content, both motions had
predominantly 1 Hz prototype frequency. The shape of the
input base accelerations dictated the dynamic behavior of
the walls. In test MMD6, the rst three to four cycles of the
input motion were larger than the rest. On the other hand,
the cycles in the base motion of test MMD12 were progressively larger. These differences in the base motions were
also reected in the time histories of deections, bending
strains and thrusts. The excess pore pressures and acceleration measurements in the backlls indicated that the soil
liqueed in both tests.
At the end of shaking, an increase in the static thrust of
about 50% occurred in both tests. Before shaking, the static
thrust of 569 kN/m was calculated based on Coulomb's
active earth pressure theory. At the end of shaking, after
complete liquefaction, the lateral thrust was estimated to
be 860 kN/m. Thus, the dynamic thrust can be expected to
be about 50% of the static thrust if the soil liquees completely. The experimental results agreed with this analysis.
Thus, although some differences existed due to the variations in the input base motions, there was sufcient evidence
to conclude that seismic modeling of models was achieved.

8. Verication of scaling relations


In the absence of prototype measurements for comparison
with model test results, scaling relations were validated by
conducting modeling of models. Two tests were conducted
on wall B (test MMD6) and wall C (test MMD12) as a part
of the modeling of models study. The model congurations
of these two tests were very similar. The backlls contained
uniform Nevada No. 100 sand at 60% relative density. The
ratio of the backll length to the wall height in test MMD6

9. Conclusions and discussion


The centrifuge modeling technique was adopted to study
the seismic behavior of xed-base, cantilever retaining walls
with liqueable backlls. The base motions consisted of
twelve roughly sinusoidal cycles at 1 Hz prototype
frequency. In order to achieve a common time scale for
dynamic and diffusive pore pressure phenomena,

592

M.M. Dewoolkar et al. / Soil Dynamics and Earthquake Engineering 19 (2000) 583593

metolosewater solutions were used as a substitute pore


uid. Direct measurements of static and dynamic lateral
earth pressures were made.

It was determined that static as well as dynamic measurements of lateral earth pressures, accelerations, bending
strains and deections made on the model walls were internally consistent. The earth pressure transducers (Entran Flatline Pressure Transducers, Model No. EPL-200-100S)
provided reliable measurements of lateral earth pressures
when used with saturated Nevada sand.
Static as well as dynamic measurements were repeatable.
The measurements on the wall and in the soil near the wall
also indicated that special boundary treatment, such as the
use of duxseal absorbing material, is not critical for this
particular model conguration, soil type, the type of input
motion, and especially when the length of the backll is
sufciently longer than the height of the wall. A suitable
absorbing boundary may be necessary if the behavior of the
soil near the container end is of interest. Static and seismic
modeling of models were achieved validating the experimental procedures and underlying similitude assumptions.
Thus, the selected centrifuge model conguration and
instrumentation generated internally consistent and repeatable measurements. Also, modeling of models indicated that
the involved scaling relations were essentially correct.
Therefore, further detailed interpretations [3] of the experimental measurements could be performed to study the
complex dynamic-soil-pore uid-structure interaction
problem of seismically loaded retaining walls with liqueable backlls.
Acknowledgements
Partial nancial support provided by the United States
Bureau of Reclamation, Denver, Colorado for this research
is acknowledged.
References

Fig. 6. Measurements from tests MMD6 and MMD12 at prototype scale.

[1] Andersen GR, Whitman RV, Germaine JT. Seismic response of rigid
tilting walls. In: Ko HY, McLean F, editors. Centrifuge 91, Rotterdam: Balkema, 1991. p. 41723.
[2] Bolton MD, Steedman RS. The behaviour of xed cantilever walls
subject to lateral shaking. In: Craig, editor. Application of centrifuge
modeling of geotechnical design, Rotterdam: Balkema, 1985.
[3] Dewoolkar MM, Ko HY, Pak RYS. Seismic behavior of cantilever
retaining walls with liqueable backlls. ASCE Journal of Geotechnical and Geoenvironmental Engineering 2000 (accepted).
[4] Dewoolkar MM, Ko HY, Stadler AT, Astaneh SMF. A substitute pore
uid for seismic centrifuge modeling. ASTM Geotechnical Testing
Journal 1999;22(3):196210.
[5] Egan D, Merrield CM. The use of miniature earth pressure cells in a
multigravity environment. In: Kimura, Kusakabe, Takemura, editors.
Centrifuge 98, Rotterdam: Balkema, 1998. p. 5560.
[6] Fujiwara T, Horikoshi K, Sueoka T. Dynamic behavior of gravity type
quay wall and surrounding soil during earthquake. In: Kimura, Kusakabe, Takemura, editors. Centrifuge 98, Rotterdam: Balkema, 1998.
p. 35964.
[7] Ishibashi I, Madi L. Case studies on quaywalls stability with liqueed
backll. In: Proceedings of Fourth US National Conference on Earthquake Engineering, vol. 3, 1990. p. 72534.

M.M. Dewoolkar et al. / Soil Dynamics and Earthquake Engineering 19 (2000) 583593
[8] Kamon K, Mimura M, Matsuda S, Nagayama S, Misaki S. Experimental studies on the seismic response of gravity caisson quay wall.
In: Kimura, Kusakabe, Takemura, editors. Centrifuge 98, Rotterdam:
Balkema, 1998. p. 3338.
[9] Ketcham SA, Ko HY, Sture S. Performance of an earthquake motion
simulator for a small geotechnical centrifuge. In: Ko HY, McLean F,
editors. Centrifuge 91, Rotterdam: Balkema, 1991. p. 3618
(Boulder, Colorado).
[10] Ko HY. The Colorado centrifuge facility. In: Corte JF, editor. Centrifuge 88, Rotterdam: Balkema, 1988. p. 7375 (Paris, France).
[11] Lyndon A, Pearson RA. Pressure distribution on a rigid retaining wall
in cohesionless material. In: Craig, editor. Application of centrifuge
modeling of geotechnical design, Rotterdam: Balkema, 1985. p. 271
80.
[12] Matsuzawa H, Ishibashi I, Kawamuara M. Dynamic soil and water
pressures of submerged soils. Journal of Geotechnical Engineering
1985;111(10):116176.
[13] Mononobe N, Matsuo M. On the determination of earth pressures
during earthquakes. In: Proceedings, World Engineering Congress,
vol. 9, 1929. p. 17785.
[14] Okabe S. General theory of earth pressure and seismic stability of
retaining wall and dam. Journal of the Japanese Society of Civil
Engineers 1924;10(5):1277323.
[15] Ortiz LA, Scott RF, Lee J. Dynamic centrifuge testing of a cantilever
retaining wall, Earthquake Engineering and Structural Dynamics, vol.
11. New York: Wiley, 1983 (p. 25168).
[16] Satoh T, Tsurugasaki K, Nagata T, Miyake M. Deformation of caisson
type quay wall during earthquake. In: Kimura, Kusakabe, Takemura,
editors. Centrifuge 98, Rotterdam: Balkema, 1998. p. 33944.
[17] Schcherbina VI. Earth pressure studies on retaining walls by
centrifugal modelling. In: Corte JF, editor. Centrifuge 88, Proceed-

[18]

[19]
[20]

[21]

[22]

[23]

[24]

[25]
[26]

593

ings of the International Conference on Geotechnical Centrifuge


Modeling, Rotterdam: Balkema, 1988. p. 4216.
Schurmann A, Jessberger HL. Earth pressure distribution on sheet pile
walls. In: Leung CF, Lee FH, Tan TS, editors. Centrifuge 94, Proceedings of the International Conference, Rotterdam: Balkema, 1994.
p. 95100.
Stadler AT. Dynamic centrifuge testing of cantilever retaining walls.
PhD thesis, University of Colorado at Boulder, August 1996.
Steedman RS, Zeng X. Centrifuge modeling of the effects of earthquakes on free cantilever walls. In: Ko HY, McLean F, editors.
Centrifuge 91, Rotterdam: Balkema, 1991. p. 42530.
The Earth Technology Corporation, Irvine, CA. Verication of Liquefaction Analysis by Centrifuge Studies Laboratory Testing Program
Soil Data Report, Project No. 90-0562. Submitted to the National
Science Foundation, USA, 1992.
Weiler WA, Kulhawy FH. Factors affecting stress cell measurement
in soil. ASCE Journal of Geotechnical Engineering 1982;108(12):
15291548.
Whitman RV, Ting NH. Experimental results of Experiment No. 10.
In: Arulanandan K, Scott RF, editors. Verication of numerical
procedures for the analysis of soil liquefaction problems, vol. 1.
1993. p. 88191.
Zeng X. Seismic response of gravity type quay wall. In: Lee FH,
Leung CF, Tan TS, editors. Centrifuge 94, Rotterdam: Balkema,
1994. p. 1916.
Zeng X, Steedman RS. On the behavior of quay walls in earthquakes.
Geotechnique 1993;43(3):41731.
Zhang JM, Shamoto Y, Sato M. Dynamic soil and water pressures on
waterfront rigid walls. In: Kimura, Kusakabe, Takemura, editors.
Centrifuge 98, Rotterdam: Balkema, 1998. p. 34550.

S-ar putea să vă placă și