SEVENTH FRAMEWORK PROGRAMME
Capacities Specific Programme
Research Infrastructures
Project No.: 227887
SERIES
SEISMIC ENGINEERING RESEARCH INFRASTRUCTURES FOR
EUROPEAN SYNERGIES
Work package [WP8 – TA4 EQUALS]
DYNCREW
Experimental Investigation of the Dynamic Behaviour of Cantilever
Retaining Walls
- Final Report -
User Group Leader: Prof. A. Evangelista
Revision: Final
July, 2013
SERIES 227887 DYNCREW Project
i
ABSTRACT
The document reports the outcomes of an experimental test campaign, under 1-g shaking table
testing, for the validation of the performance of cantilever retaining walls to dynamic loading.
Cantilever retaining walls represent a popular type of retaining system, widely considered
advantageous over conventional gravity walls as it combines economy and ease in construction
and installation. Apart from few available limit analysis solutions, the seismic behaviour of this
type of walls is little explored. In the frame of the research presented in this report, a systematic
investigation on scaled wall models was carried out at the Earthquake and Large Structures
Laboratory (EQUALS) - University of Bristol. The experimental program encompasses different
combinations of retaining wall geometries, soil configurations and input ground motions (white
noise, sine dwells and actual recorded motions from the Italian and American database). The
response analysis of the systems at hand aimed at shedding light onto the salient features of the
problem, such as: (1) the magnitude of the soil thrust and its point of application; (2) the relative
sliding as opposed to rocking of the wall base and the corresponding failure mode; (3) the
importance/interplay between soil stiffness, wall dimensions, and excitation characteristics, in
affecting the above. The data obtained by the experimental investigations were in good
agreement with the results by the theoretical models used for the analysis and are expected to be
useful for the better understanding and the optimization of earthquake design of this particular
type of retaining structure.
Keywords: Cantilever Retaining Walls, Earthquake Design, Dynamic Earth Pressures, Soil-
Structure-Interaction, Experimental Analysis, Shaking Table Testing
SERIES 227887 DYNCREW Project
ii
SERIES 227887 DYNCREW Project
iii
ACKNOWLEDGMENTS
The research leading to these results has received funding from the European Union Seventh
Framework Programme [FP7/2007-2013] for access to EQUALS, University of Bristol under
grant agreement n° 227887 [SERIES].
SERIES 227887 DYNCREW Project
iv
SERIES 227887 DYNCREW Project
v
REPORT CONTRIBUTORS
University of Naples Federico II Aldo Evangelista
Anna Scotto di Santolo
University of Sannio Armando Lucio Simonelli
Augusto Penna
Pamela Imbriale
Carmine Lucadamo
University of Patras George Mylonakis
Panos Kloukinas
University of Bristol Matt Dietz
Colin Taylor
SERIES 227887 DYNCREW Project
vi
SERIES 227887 DYNCREW Project
vii
CONTENTS
CONTENTS .....................................................................................................................................v
List of Figures .............................................................................................................................. xiii
List of Tables ............................................................................................................................ xxxi
1 The DYNCREW Project ...............................................................................................................1
1.1 INTRODUCTION .................................................................................................................1
2 The experimental program ............................................................................................................1
2.1 THEORETICAL BACKGROUND ......................................................................................2
2.2 EXPERIMENTAL DESIGN AND SETUP ..........................................................................4
2.2.1 The Earthquake Simulator .............................................................................................4
2.2.2 The Equivalent Shear Beam container (shear “stack”) .................................................5
2.2.3 Instrumentation and recording hardware .......................................................................7
2.2.4 Shaking table model geometry and instrumentation ...................................................10
2.2.5 Soil material and soil layers properties .......................................................................15
2.2 EXPERIMENTAL PROCEDURE .....................................................................................19
2.3.1 Model Configurations ..................................................................................................19
2.3.2 Iterative testing procedure ...........................................................................................20
2.3.3 White noise testing ......................................................................................................21
2.3.4 Dynamic testing with harmonic, sinusoidal excitation ................................................21
2.3.5 Dynamic testing with real earthquake records ............................................................22
3 Experimental results....................................................................................................................24
3.1 DYNAMIC PROPERTIES OF THE MODEL ...................................................................24
SERIES 227887 DYNCREW Project
viii
3.2 EXPERIMENTAL RESULTS FOR DISPLACEMENTS, SEISMIC LOADS
AND FAILURE MECHANISMS EXPERIMENTAL DESIGN AND SETUP ................37
4 Interpretation of displacements in the light of sliding block theory predictions ........................47
5 Interpretation of dynamic bending moments ..............................................................................57
6 Conclusions .................................................................................................................................64
References ......................................................................................................................................66
Appendix .......................................................................................................................................69
SERIES 227887 DYNCREW Project
ix
List of Figures
Figure 2.1 Pseudo-dynamic analysis of cantilever retaining walls for seismic loading: (a)
Seismic Rankine stress field for the problem under consideration, (b) Seismic Rankine
stress tensor and stress, (c) Modified virtual back approach by stress limit analysis, (d)
Stability analysis according to Meyerhof (EC 7) ......................................................................... 3
Figure 2.2 The shaking table of EQUALS – BLADE (University of Bristol, (UK) .................. 5
Figure 2.3 General and inside aspect of the Equivalent Shear Beam container (“shear
stack”) (a) side walls lubricated with silicone grease and covered with latex membrane, (b)
wooden floor and (c) transverse walls roughened by sand-grain adhesion ................................. 6
Figure 2.4. Schematic diagram of the EBS container (Bhattacharya et al., 2012) ..................... 7
Figure 2.5 Illustration of the model and experimental equipment during the first phase of
testing ........................................................................................................................................... 9
Figure 2.6 Details of the experimental setup (Second phase of testing): (a) longitudinal
aspect of the model, (b) wall face instrumentation (Configurations No2 & No3), (c),(d)
pairs of accelerometers on shaking table and upper ring, (e) backfill accelerometer, (f) sand
pouring procedure, (g), (h) back - covered with rough sandpaper - and front view of the
wall (Configuration No1) ............................................................................................................. 9
Figure 2.7 Illustration of geometry and instrumentation of the shaking table model
(dimensions in mm) .....................................................................................................................10
Figure 2.8 Geometry and assembling of the retaining wall model (dimensions in mm) ............12
Figure 2.9. Strain gauges positions .............................................................................................13
Figure 2.10 Comparison of dynamic bending deformation recorded at the strain gauges at
the bottom of the cantilever (slab connection area) .....................................................................14
SERIES 227887 DYNCREW Project
x
Figure 2.11 Friction angle of Leighton Buzzard Sand as a function of mean effective stress
and relative density (from Bhattacharya et al., 2012) ..................................................................16
Figure 2.12 Empirical correlations for dynamic soil properties of LB sand derived from the
laboratory dynamic tests of Cavallaro et al. (2001) .....................................................................18
Figure 2.13 Dynamic soil properties of LB sand measured experimentally by shaking
table testing at EQUALS (from Dietz & Muir Wood, 2007) ......................................................19
Figure 2.14 Harmonic base excitation at the frequency of 7Hz and amplitude 0.05g ...............22
Figure 2.15. Scaled seismic signals imposed as base acceleration ............................................23
Figure 2.16 Comparison between the authentic and modified Sturno record .............................24
Figure 3.1 Accelerometer positions for white noise testing evaluation .....................................25
Figure 3.2 Transfer functions from the wall response in Configuration No1 – wall placed
on the foundation layer before backfill construction ..................................................................27
Figure 3.3 Transfer functions from the foundation layer response in Configuration No1 –
before backfill construction ........................................................................................................28
Figure 3.4 Transfer functions from the wall response in Configuration No1 – after backfill
construction .................................................................................................................................29
Figure 3.5 Transfer functions from the backfill response in Configuration No1 .......................30
Figure 3.6 Transfer functions from the wall response in Configuration No3 – wall placed
on the foundation layer before backfill construction (*refers to the soil prism behind the
wall) .............................................................................................................................................31
Figure 3.7 Transfer functions from the wall response in Configuration No3 after the
construction of the backfill ..........................................................................................................32
Figure 3.8 Transfer functions from the backfill response in Configuration No3. ......................33
Figure 3.9. Variation of earth pressure coefficient K0 at very low confining stress (from
Chu & Gan, 2004) .......................................................................................................................34
Figure 3.10. Increase of the soil shear modulus with depth, based on empirical relations .........35
Figure 3.11 Resonant frequency of the two-layered medium with properties: H1 = 0.4m,
Vs1 = 157m/s, H2 = 0.6m, Vs2 = 113m/s ......................................................................................36
Figure 3.12. Elastic parameters of the soil layers of the problem ...............................................37
Figure 3.13. Measurements of wall displacement and rotation for all configurations and
SERIES 227887 DYNCREW Project
xi
various base excitations. (a) Cumulative footing rotation versus sliding (LVDT-D1), (b)
Cumulative footing settlement (LVDT-D4) versus sliding, (c), (d) and (e) Incremental wall
displacement (LVDTs D1-D2-D3) for Configurations No1, No2 and No3, respectively ...........41
Figure 3.14. Comparison of typical experimental results for Configurations No1 and No3
under harmonic-sinusoidal excitation: (a) measured wall accelerations, (b) corresponding
wall displacements, (c) positive acceleration distribution (maximum inertial forces towards
the backfill), (d) negative acceleration distribution (maximum inertial forces towards the
wall) .............................................................................................................................................42
Figure 3.15. Comparison of typical experimental results for Configurations No1 and No3
under seismic excitation: (a) measured wall accelerations, (b) corresponding wall
displacements, (c) positive acceleration distribution (maximum inertial forces towards the
backfill), (d) negative acceleration distribution (maximum inertial forces towards the wall) .....43
Figure 3.16. Typical experimental results for Configuration No2: (a) measured wall
accelerations for harmonic sinusoidal and seismic excitation, (b) corresponding wall
displacement, (c) positive acceleration distribution (maximum inertial forces towards the
backfill), (d) negative acceleration distribution (maximum inertial forces towards the wall) ....44
Figure 3.17. Comparison of typical experimental results for Configurations No1 and No3
under harmonic-sinusoidal excitation: (a) measured wall accelerations, (b) corresponding
wall displacements, (c) negative acceleration distribution (maximum inertial forces towards
the wall), (d) increment of wall displacement (LVDTs D1-D2-D3) and (e) peak seismic
increment of bending moment for positive and negative acceleration. .......................................45
Figure 3.18. Comparison of typical experimental results for Configurations No1 and No3
under seismic excitation: (a) measured wall accelerations, (b) corresponding wall
displacements, (c) negative acceleration distribution (maximum inertial forces towards the
wall), (d) increment of wall displacement (LVDTs D1-D2-D3) and (e) peak seismic
increment of bending moment for positive and negative acceleration ........................................46
Figure 3.19. Backfill surface settlement distribution at failure: (a) initial grid geometry -
dimensions in mm, (b),(c) settlement distributions for Configurations No2 and No3,
respectively. .................................................................................................................................47
Figure 4.1. Gravity retaining wall on rigid base examined against (a) sliding according
Richards & Elms, (1979) and (b) rotation about the toe (point O) according to Zeng &
SERIES 227887 DYNCREW Project
xii
Steedman, (2000) .........................................................................................................................48
Figure 4.2. Comparisons of experimental permanent displacement and rotation with the
predictions of sliding/rotating block theory – Configuration No1 under harmonic loading .......51
Figure 4.3. Comparisons of experimental permanent displacement and rotation with the
predictions of sliding/rotating block theory – Configuration No3 under harmonic loading ......52
Figure 4.4.Comparisons of experimental permanent displacement and rotation with the
predictions of sliding/rotating block theory – Configuration No1 under seismic loading .........53
Figure 4.5. Comparisons of experimental permanent displacement and rotation with the
predictions of sliding/rotating block theory – Configuration No3 under harmonic loading ......54
Figure 4.6. Comparisons of time histories for experimental permanent displacement and
rotation with the predictions of sliding/rotating block theory – Configuration No1 under
harmonic loading 0.19g and 0.23g and under seismic loading 0.35g and 0.55g .........................55
Figure 4.7. Comparisons of time histories for experimental permanent displacement and
rotation with the predictions of sliding/rotating block theory – Configuration No3 under
harmonic loading 0.19g and 0.23g and under seismic loading 0.35g and 0.55g .........................56
Figure 5.1 Inertial and contact forces contributing dynamic bending moment on the
cantilever wall ..............................................................................................................................58
Figure 5.2 Stress variation in the backfill during dynamic loading ............................................58
Figure 5.3. Comparisons of experimental and theoretical dynamic bending moments for
Configuration No1 under harmonic loading of amplitude 0.15g, 0.19g and 0.23g ....................60
Figure 5.4 Comparisons of experimental and theoretical dynamic bending moments for
Configuration No1 under seismic loading of PGA 0.17g, 0.35g and 0.55g ................................61
Figure 5.5 Comparisons of experimental and theoretical dynamic bending moments for
Configuration No3 under harmonic loading of amplitude 0.15g, 0.19g and 0.23g ....................62
Figure 5.6 Comparisons of experimental and theoretical dynamic bending moments for
Configuration No3 under seismic loading of PGA 0.17g, 0.35g and 0.55g ................................63
SERIES 227887 DYNCREW Project
xiii
List of Tables
Table 2.1. Instrumentation summary........................................................................................... 8
Table 2.2 Soil properties .............................................................................................................17
Table 2.3. Pseudostatic critical accelerations and associated safety factors (SF) with
respect to sliding and bearing capacity ........................................................................................20
SERIES 227887 DYNCREW Project
xiv
SERIES 227887 DYNCREW Project
1
1 The DYNCREW Project
1.1 INTRODUCTION
The document reports the experimental study of the dynamic response of cantilever retaining
walls under 1-g shaking table testing, conducted at the Earthquake and Large Structures
Laboratory (EQUALS) which is part of the Bristol Laboratories of Advanced Dynamics
Engineering (BLADE), at the University of Bristol. The experiments were carried out during a
total testing period of five weeks, divided into two phases: a preparative phase of two weeks
(December 2010) and the main testing period of three weeks (June 2011), under the
collaboration of research teams from Universities of Naples and Sannio (Italy) and University of
Patras (Greece), within the SERIES project (Seismic Engineering Research Infrastructures for
European Synergies). The experimental program involves testing on scaled wall models founded
on compliant base, under different combinations of retaining wall geometries, soil configurations
and input ground motions. The initial motivation of this experimental study was the validation of
relevant stress limit analysis solutions (Evangelista et al, 2009, 2010; Evangelista & Scotto di
Santolo, 2011; Kloukinas & Mylonakis, 2011). The results aimed at shedding light onto salient
features of the dynamic response and can be useful for the better understanding and the
optimization of earthquake design of this particular type of retaining structure.
2 The experimental program
Reinforced concrete cantilever retaining walls represent a popular type of retaining system. They
are widely considered advantageous over conventional gravity walls as they combine economy
with ease in construction and installation. The concept is deemed particularly rational, as it
exploits the stabilizing action of the soil weight over the footing slab against both sliding and
SERIES 227887 DYNCREW Project
2
overturning, thus allowing construction of walls of considerable height. For walls of this type,
structural weight is not a predominant parameter as equilibrium depends mainly on backfill
action and resistance of foundation soil.
2.1 THEORETICAL BACKGROUND
The traditional approach for analyzing cantilever walls is based on the well-known limit
equilibrium analysis, in conjunction with a conceptual vertical surface AD (Fig. 2.1a,c) passing
through the innermost point of the wall base (vertical virtual back approach). A contradictory
issue in the literature relates to the calculation of active thrust acting on the virtual wall back,
under a certain mobilized obliquity ranging, naturally, from 0 (for a perfectly smooth plane) to φ
(for a perfectly rough plane). Efforts have been made by numerous investigators to establish the
proper roughness for the analysis and design or this type of structures as reported by Evangelista
et al (2009, 2010) and Kloukinas & Mylonakis (2011). Nevertheless, the issue of seismic
behaviour remains little explored. In fact many modern Codes, including the Eurocodes
(Eurocode 8-Part 5, 2004) and the Italian Building Code (NTC, 2008), do not explicitly refer to
cantilever walls. The current Greek Seismic Code (EAK, 2003) addresses the case of cantilever
retaining walls adopting the virtual back approach in the context of a pseudo-static analysis
under the assumption of gravitational infinite slope conditions (Rankine, 1857) and various
geometric constraints.
Recent theoretical findings obtained by means of stress limit analysis (Evangelista et al 2010;
Evangelista & Scotto di Santolo, 2011; Kloukinas & Mylonakis, 2011) indicate that a uniform
Rankine stress field can develop in the backfill, when the wall heel is sufficiently long and the
stress characteristics do not intersect the stem of the wall (wall). Given that the inclination
of the stress characteristics depends on acceleration level, a Rankine condition is valid for the
vast majority of cantilever wall configurations under strong seismic action. This is applicable
even to short heel walls, with an error of about 5% (Huntington, 1957; Greco, 2001).
Following the aforementioned stress limit analysis studies, closed-form expressions are derived
for both the pseudo-dynamic earth pressure coefficient KAE and the resultant thrust inclination, E
(Figure 2.1c), given by (Kloukinas & Mylonakis 2011):
11
1
sin sintan
1 sin cos
e e
E
e e
(2.1)
SERIES 227887 DYNCREW Project
3
where 1e = sin-1
[sin(+e)/sin] ande = tan-1
[ah/(1-av)] are the so-called Caquot angle and the
inclination of the overall body force in the backfill. The same result for E has been derived, in a
different form, by Evangelista et al (2009, 2010). In the case of gravitational loading (e = 0), the
inclination E equals the slope angle , coinciding with the classical Rankine analysis. In
presence of a horizontal body force component, E is always greater than increasing with e
up to the maximum value of , consequently improving wall stability. The robustness of the
above stress limit analysis becomes evident, since under Rankine conditions and mobilized
inclination E, the stress limit analysis and the Mononobe-Okabe formula results coincide. These
findings have been confirmed by numerical analysis results (Evangelista et al. 2010; Evangelista
& Scotto di Santolo, 2011).
Figure 2.1 Pseudo-dynamic analysis of cantilever retaining walls for seismic loading: (a)
Seismic Rankine stress field for the problem under consideration, (b) Seismic Rankine stress
tensor and stress, (c) Modified virtual back approach by stress limit analysis, (d) Stability
analysis according to Meyerhof (EC 7)
SERIES 227887 DYNCREW Project
4
A second key issue in the design of the particular type of retaining systems deals with the
stability analysis shown in Figure 2.1d. Traditionally, stability control of retaining walls is based
on safety factors against bearing capacity, sliding and overturning. Of these, only the first two
are known to be rationally defined, whereas the safety factor against overturning is known to be
misleading, lacking a physical basis (Greco, 1997; NTC, 2008; Kloukinas & Mylonakis, 2011).
It is important to point out that the total gravitational and seismic actions on the retaining wall
are resisted upon the external reactions H and V acting on the foundation slab. The combination
of these two actions, together with the resulting eccentricity e, determines the bearing capacity of
the wall foundation, based on classical limit analysis procedures for a strip footing subjected to
an eccentric inclined load (e.g. EC7, EC8). This suggests that the wall stability is actually a
footing problem and from this point of view, understanding the role of the soil mass above the
foundation slab and the soil-wall interaction is of paramount importance.
2.2 EXPERIMENTAL DESIGN AND SETUP
In the following paragraphs, the shaking table equipment, the experimental design and setup, the
material properties and the testing procedure are described in detail.
2.2.1 The Earthquake Simulator
The EQUALS-BLADE Earthquake Simulator, shown in Figure 2.2, consists of a 3m x 3m cast
aluminium platform weighing 3.8tons and capable of shaking a maximum payload of 15tons at
an operational frequency range of 0-100Hz. The platform has the shape of an inverted pyramid
consisting of four sections having a honeycomb-like network of stiffening diaphragms giving it
high strength and bending stiffness. The platform surface is an arrangement of 5 aluminium
plates with a regular grid of M12 bolt holes for attaching to the platform body and for mounting
of specimens. The platform sits inside a reinforced concrete seismic block that has a mass of
300tons. The block is located in a pit in the Earthquake Engineering Laboratory and is isolated
from the rest of the laboratory by a 20mm cork filled gap running between the block and the rest
of the laboratory. Hydraulic power for the ES is provided by a set of 6 shared variable volume
hydraulic pumps providing up to 900litres/min at a working pressure of 205bar. The maximum
flow capacity can be increased to around 1200litres/min for up to 16 seconds at times of peak
demand with the addition of extra hydraulic accumulators. The platform is attached to the block
by eight hydraulic actuators arranged so as to make best use of the available space. The
SERIES 227887 DYNCREW Project
5
horizontal actuators are attached to the block by concrete filled steel box sections and to the
platform by smaller closed triangular brackets. The vertical actuators are connected directly to
the block and platform. Each actuator has a dynamic capacity of 70kN and has a maximum
stroke of 300mm. The four vertical actuators each have a static section to carry the static loads of
the platform plus specimen.
Figure 2.2 The shaking table of EQUALS – BLADE (University of Bristol, UK)
2.2.2 The Equivalent Shear Beam container (shear “stack”)
The apparatus, shown in Figure 2.3, consists of eleven rectangular aluminium rings, which are
stacked alternately with rubber sections to create a hollow yet flexible box of inner dimensions
4.80m long by 1m wide and 1.15m deep (Crewe et al, 1995). The rings are constructed from
aluminium box section to minimize inertia while providing sufficient constraint for the K0
condition. The stack is secured to the shaking table by its base and shaken horizontally
lengthways (in the x direction). Its floor is roughened by sand-grain adhesion to aid the
transmission of shear waves; the internal end walls are similarly treated to enable complementary
shear stresses. Internal side walls are lubricated with silicon grease and covered with latex
membrane to ensure plane-strain conditions.
This type of containers should be ideally designed to match the shear stiffness of the soil
contained in it, as depicted in Figure 2.4. However, the shear stiffness of the soil varies during
shaking depending on the strain level. Therefore the matching of the two stiffnesses (end-wall
SERIES 227887 DYNCREW Project
6
and soil) is possible only at a particular strain level. The “shear stack” at the University of Bristol
is designed considering a value of strain in the soil close to the failure (0.01–1%). Therefore it is
much more flexible than the soil deposit at lower strain amplitudes and, as a consequence, the
soil will always dictate the overall behaviour of the container (Bhattacharya et al, 2012). Indeed
the shear stack resonant frequency and damping in the first shear mode in the long direction
when empty were measured prior to testing as 5.7Hz and 27% respectively, sufficiently different
from the soil material properties.
Figure 2.3 General and inside aspect of the Equivalent Shear Beam container (“shear stack”) (a)
side walls lubricated with silicone grease and covered with latex membrane, (b) wooden floor
and (c) transverse walls roughened by sand-grain adhesion
(a)
(b)
(c)
SERIES 227887 DYNCREW Project
7
Figure 2.4 Schematic diagram of the EBS container (Bhattacharya et al, 2012)
2.2.3 Instrumentation and recording hardware
Three basic types of instruments were used for the measurement of accelerations, displacements
and strains. An outline of the instrumentation employed is provided in Table 2.1, and some
characteristic aspects are shown in Figures 2.5 and 2.6. 21 1-D accelerometers were used to
monitor the shaking table, the shear stack and the wall-soil system, with the main area of interest
laying on the wall itself and the soil mass in its vicinity, as well as the response of the free field.
4 LVDT transducers were used to measure the dynamic response and permanent displacements
of the wall. 4 INDIKON type, non-contact displacement transducers, were used for monitoring
the settlement of the backfill surface during the first phase of testing, but were eliminated during
the second phase, because of restrictions in their operational range (~10mm). Alternatively, a
grid of coloured sand was used for this purpose. Additionally, 32 strain gauges were attached on
the stem and the base of the wall, on three cross sections, to monitor the bending of the wall.
Overall, apart from the INDIKON transducers, 57 data channels were employed. The number
and the positions of the instrumentation are shown in detail in the drawings of the sequent
paragraph, where the experimental model design and setup are analyzed.
An RDP 600-type modular electronics system will supply the excitation voltage for the
displacement transducers and the strain gauges. The signal conditioning of the LVDTs was made
via the RDP 611 amplifier modules. These amplifiers allow optimisation of both the excitation
voltage and gain and can impart DC offsets in order to zero signals. The completion of the bridge
and the excitation voltage for the strain gauges was made via RDP 628 strain gauge amplifier
modules. The SETRA accelerometer signals were amplified by a set of Fylde 245GA mini-
amplifiers. These have multiple gain, variable sensitivity and offset options. The amplified
SERIES 227887 DYNCREW Project
8
signals were supplied to a FERN EF6 multi-channel programmable filter that was set at a
common cut-off frequency of 80Hz on all channels (low pass Butterworth filter).
Signals generated by test instrumentation were supplied to ADC boards of type MSXB 028
manufactured by Microstar Laboratories, which are capable of simultaneous sampling and hold
from 64 channels of 16bit. Data collection was controlled with the SIMACQ v.3.00 software, at
a sampling rate of 1024Hz, except of the white noise tests sample at 256Hz. Data from the
seismic tests were stored in self processing MATLAB (.m) files. (MathWorks Inc., 1995).
Table 2.1. Instrumentation summary
Measured
Parameter Transducer Type Description Purpose
Acceleration Type
SETRA 141A
High output capacitance
type sensor with inbuilt
pre-amplifier.
Calibrated range: +/8g.
Operating frequency: 0-
3000 Hz
Used to measure the
horizontal and vertical
accelerations of the
shaking table, the shear
box, the wall and in the
backfill
Displacement
Τype
RDP DCTH LVDT
Linear variable
displacement
transformers
Range: +/ 12.5, +/ 50
and +/75mm
Used to monitor the
horizontal and vertical
displacement of the wall
Type
INDIKON
Non-contact displacement
transducer. Operation
based on eddy effect.
Range: 0-30 mm
Used to monitor the
settlement of the free
surface of the backfill. The
transducers will be
attached to plates resting
on a Perspex pad with
roughened underside
surface.
Strain
Strain gauge type
EA-13-120LZ-120
(Vishay Ltd)
Linear strain gauge
pattern, 3 mm length
Used to measure bending
strain of wall at various
elevations (on Y direction).
SERIES 227887 DYNCREW Project
9
Figure 2.5 Illustration of the model and experimental equipment during the first phase of testing
Figure 2.6 Details of the experimental setup (Second phase of testing): (a) longitudinal aspect of
the model, (b) wall face instrumentation (Configurations No2 & No3), (c),(d) pairs of
accelerometers on shaking table and upper ring, (e) backfill accelerometer, (f) sand pouring
procedure, (g), (h) back - covered with rough sandpaper - and front view of the wall
(Configuration No1)
SERIES 227887 DYNCREW Project
10
2.2.4 Shaking table model geometry and instrumentation
The model geometry adopted, shown in Figure 2.7, had to take under consideration the
dimensions of the shear box, the dynamic problem simulation feasibility and restrictions of the
shaking table equipment. Based on the forgoing, a maximum height of 1m of soil material was
selected, separated into the retained backfill of 0.6m (equal to the height of the wall, Η) and the
foundation soil layer of 0.4m (equal to the wall footing width, Β). The length of the retained
backfill was selected 5 times its height, whereas the corresponding free length in front of the wall
was 3 times the wall height. These dimensions were deemed sufficient to eliminate the boundary
effects and to ensure free field conditions in the middle of the backfill.
zx
4800
H
H
B
1150
shaking table
5H3H
B
H H H HHHH
backfill
"free field"
conditions
foundation layer
"free field"
conditions
Rankine prismatic
failure mechanismarea
foundation
failure
mechanismarea
zx
A1 A2
A7
A8A9A10A11
A3A4,5
A6
A13A14 A12
A15A16A17
A18
A19
A20 A21
D1
D3
D4
D2
D: LVDT
A: Accelerometer
horizontal
vertical
12001200200400900 900
300
300
300
600
400
250
250
(+)
(+)
shaking table
nodes of the surface deformation
measurment grid
Figure 2.7 Illustration of geometry and instrumentation of the shaking table model (dimensions
in mm)
SERIES 227887 DYNCREW Project
11
In Figure 2.7, the positions of the instrumentation (LVDT and accelerometers) in the model are
presented, which were kept the same during the whole testing procedure. The soil layer
parameters were also kept the same and will be presented in a next paragraph. The only
modification of the model geometry and setup relate to the geometry of the wall itself, which
was designed to exhibit different behaviour, between sequential testing cycles. The different wall
geometries adopted are presented in the table embedded in Figure 2.8. The retaining wall model
was made of Aluminium alloy 5083 plates with properties: unit weight γ = 27kN/m3, Young’s
modulus E = 70GPa, Poisson’s ratio = 0.3.
The material density of the model wall is similar to those of prototype concrete walls, while the
dimension proportions of the model followed the common construction analogies used in
practice (geometrical similarity). As shown in Figure 2.8, the 32mm thick plates are connected
with M12 bolts. The efficiency of this connection system, with respect to the total fixity
conditions demand, proved to be very good, as shown in the bending moment diagrams of Figure
2.10. The width of the stem of the wall is 970mm. A central wall segment of 600mm width was
created by two 1mm thick vertical slits penetrating 400mm down into the wall, to ensure the
response of the central segment under plane strain conditions, independently from possible
problems at the boundaries. The locations of the slits were 185mm from each side of the wall
stem. Following the same concept, the base of the wall is subdivided into four 240mm-wide
aluminium segments, that are each secured to the wall stem via three M12 bolts.
The positions of the 32 strain gauges are depicted in the drawings of Figure 2.9. The instruments
were placed symmetrically, at the same positions on both slab faces (1) on the central wall line,
and (3) on a second section close to the edge, for an extra control of the plane strain condition
requirement. Generally, the selection of the 32mm thick slab for the cantilever beam, followed
the criterion to keep the bending deformations close to the operational level of the instruments
( 2a d wM t D ~10-5
, where 3 212(1 )w w wD E t v ), in order to be measured, but at the same
time to simulate a relatively rigid wall. This is why no strain gauges were place at the upper
200mm of the wall stem, where strain lever due to bending is expected much smaller than the
instrumentation operational level (in the order of 10-6
).
Even under the design concept described above, the wall-soil relative flexibility parameter
3
w wd GH D (Veletsos & Younan, 1994, 2000; Giarlelis & Mylonakis, 2010), yields values
from 15 to 20, depending on the soil stiffness. These values correspond to a relatively flexible
SERIES 227887 DYNCREW Project
12
32
32
570
b2
b1
Confi
gura
tion Νο
1
b2
b1
298
250
250
70 0 0
Confi
gura
tion Νο
2
Confi
gura
tion Νο3
wal
l
cross
sec
tion
(zx
-pla
ne)
185
600
185
fron
t vie
w
(zy
-pla
ne)
1 m
m
slit
400
M12 b
olt
s
400
240
240
240
240
400
1 m
m
slit
12
34
5
602
Alu
min
ium
alo
y 5
083 p
late
s
14
- fo
undat
ion p
late
s, d
imen
sions:
240 x
400 x
32
to
5-
canti
lever
wal
l pla
te, dim
ensi
ons:
570 x
970 x
32
model
modif
icat
ions:
pla
n v
iew
(xy
-pla
ne)
970
Figure 2.8 Geometry and assembling of the retaining wall model (dimensions in mm)
SERIES 227887 DYNCREW Project
13
cantilever wall, not usually the case for the prototype retaining walls of this type. In practice, the
cantilever wall thickness increases with depth, from a minimum width at the load-free end on the
top, prescribed from constructional codes (t/H ratio equal to 0.05 or even smaller), to a notably
higher width at the base, designed to carry the shear force and the bending moment of the fixity
area. This usually leads to trapezoidal cantilever wall cross sections, providing high bending
stiffness.
y
185600185
970
back view
120 93
x
y
x
y
z
y
185 600 185
970
front view
12093
z
86
88
88
88
20
20
2050
50
plan view bottom
view
SG1
SG2
SG3
SG4
SG5
SG6
SG7
SG21
SG22
SG23
SG24
SG25
SG26
SG14
SG13
SG12
SG11
SG10
SG9
SG8
SG20 SG32 SG28 SG16
SG17
SG18
SG19
SG29
SG30
SG15SG31 SG27
Figure 2.9 Strain gauges positions
In Figure 2.10 time histories recorded at the strain gauges of the slab connection area, for the
typical case of harmonic excitation with PGA 0.1g, are presented. The efficiency of the plates
connection system to provide total fixity conditions is obvious, as the opposite strain gauges
provide the same bending moment readings, being on opposite faces (tension-compression) at
SERIES 227887 DYNCREW Project
14
the stem and the footing plate (see SG8-SG15 comparison). Minor exception is the strain gauges
pair SG16-SG20, placed at the short cantilever of the wall toe, which cannot bend but penetrates
the underlying soil. Additionally, the instrument SG1 exhibits an electrical anomaly, commonly
observed at the internal instruments being in contact to the soil mass.
Figure 2.10 Comparison of dynamic bending deformation recorded at the strain gauges at the
bottom of the cantilever (slab connection area)
SERIES 227887 DYNCREW Project
15
2.2.5 Soil material and soil layers properties
The required soil configuration consists of a dense supporting layer and a medium dense backfill.
The soil material is the same for all soil layers, consisting of dry, yellow Leighton Buzzard (LB)
sand BS 881-131 (silica sand with sub-rounded grain shape), Fraction B (Dmin = 0.6mm, Dmax =
1.18mm, D50 = 0.82mm, Gs = 2640Mg/m3, emin = 0.486, emax = 0.78), at different compaction
levels. This particular soil has been used extensively in experimental research at Bristol and a
wide set of density and stiffness data is available (Stroud, 1971; Tan, 1990; Cavallaro et al,
2001). Further information and references are provided in Bhattacharya et al (2012).
The base deposit was formed by pouring sand in layers of 150-200 mm from a deposition height
of 0.6m and then densifying by shaking. After densification, the height of the layer was reduced
to 390mm. The top layer was formed by pouring sand in axisymmetric conditions close to the
centre of the desired backfill region, without any further densification. The pouring was carried
out by keeping the fall height steady, approximately equal to 200mm in order to minimize the
densification effect of the downward stream of sand.
The packing density of the soil material is a parameter of major importance during 1-g shaking
table testing. It is known that the soil behaviour is non-linear, strongly depending on the isotropic
stress (first invariable of the stress tensor ij[ / 3]p trace ), which in small scale models is
ordinarily less than 5kPa. At this low stress level, soils exhibit an important dilative behaviour,
resulting to higher apparent friction angles, φpeak = φcrit + 0.8ψ (Bolton, 1986), compared to the
prototype. It is also known that the critical friction angle is not altered with density variations.
Thus, the only way for minimizing the error in the peak friction angle, is by controlling soil
dilatancy by reducing soil density. The specific technique has been proposed by various
researchers (Kelly et al, 2006; Leblanc et al, 2010). Figure 2.11 shows the variation of peak
friction angle with mean effective isotropic stress (p′) for silica sand of various relative densities
based on Eq. (2.1), following the stress-dilatancy work of Bolton (1986).
3 9.9 ln 1cv Dr p (2.2)
where φcv, is the critical friction angle, estimated at 34.3ο in the graph, and p’ is the isotropic
confining stress (in kPa). Based on this approach, if 120kPa of mean effective prototype stress at
50% relative density is to be modelled in a small scale laboratory model at 25kPa stress, the sand
SERIES 227887 DYNCREW Project
16
is to be poured at about 39% relative density ensuring that peak friction angle is the same. On the
other hand, this approach raises two major restrictions: First, there is a minimum density beyond
which sand cannot be poured. Second, a very loose soil mass is extremely sensitive to dynamic
densification during seismic excitation. This effect was noticed during the retaining wall
experiment, but its effect was kept generally small due to the following reasons: (1) the
foundation layer that was common in all cases, was relatively dense from the beginning, (2) the
backfill behind the wall was replaced after every cycle of strong shaking and failure of the model
and (3) the uniform, coarse-grained sand with rounded grain shape selected for the model,
provided a relatively stable structure, less sensitive to dynamic compaction.
Figure 2.11 Friction angle of Leighton Buzzard Sand as a function of mean effective stress and
relative density (from Bhattacharya et al, 2012)
Apart from soil mechanical properties, the low confining stress level of the shaking table model,
affects the dynamic soil properties, yielding lower stiffness modulus G, which obeys a parabolic
variation relatively to the isotropic stresses of the form n
G p , where the n is about 0.5. This
effect is taken under consideration during the shaking table model design through appropriate
scaling laws. Secondarily, the shear modulus G is also a function of soil density. These features
are presented and commented later on, at the paragraph where the investigation of the dynamic
properties of the model is explained and the results of the dynamic white noise testing are
SERIES 227887 DYNCREW Project
17
interpreted. This particular soil has been used extensively in experimental research at Bristol and
a wide set of density and stiffness data is available (Stroud, 1971; Tan, 1990; Cavallaro et al,
2001). Further information and references are provided in Bhattacharya et al, (2012). Laboratory
testing by Cavallaro et al (2001) provided the following empirical correlation between friction
angle φ and relative density Dr, which was used for a preliminary estimation of the soil strength
properties:
deg 0.238 % 28.4Dr
(2.3)
The packing density for each layer, determined from sand mass and volume measurements
during the deposition, and the corresponding predictions of Eq. (2.3) are summarized in Table
2.2 with the corresponding estimation of the peak friction angle.
Table 2.2 Soil properties
Soil layers Thickness
(mm)
Voids
ratio, e
Relative density,
Dr (%)
Unit weight
(kN/m3)
Friction angle, φ(o)
Cavallaro et al (2001)
Foundation 390 0.61 60 16.14 42
Backfill 600 0.72 22 15.07 34
Extremely important for the soil-wall system behaviour are the mechanical properties of the
interfaces, which are of two kinds: a) smooth interface with direct contact of sand on the
aluminum plate, and b) rough interface created by pasting rough sandpaper on the wall surface.
The critical friction angles of the interfaces were measured directly on the model with static pull
tests. The resulting values are:
Smooth soil-wall interface (Aluminum - soil): δ = 23.5ο
Rough soil-wall interface (sandpaper - soil): δ = 28.5ο
It can be observed that the roughened interface yields a friction close to the soil critical state
angle of Eq. (2.3), as expected for a totally rough surface. More information referring to the LS
sand behaviour on interfaces, as well as the corresponding dilatancy angles, can be found in the
experimental work of Lings & Dietz (2005) and Dietz & Lings (2006).
At last, the dynamic properties of the soil material is well described through empirical curves for
the G/G0 ratio degradation and the increase of damping ratio D (%), derived from the
experimental study of Cavallaro et al (2001):
SERIES 227887 DYNCREW Project
18
0
( ) 1
1 (%)
G
G
(2.4)
0
( )
( )(%)
G
GD e
(2.5)
where: α = 20, β = 0.9, η = 134 and λ = 4.6. The specific curves, plotted in Figure 2.12, have
been derived from laboratory resonant column tests under confining stress levels ranging from
50 - 150kPa, which is much higher than those corresponding to the shaking table conditions.
This leads to an overestimation of the damping ratio D, as shown by Dietz & Muir Wood (2007),
who measured the dynamic soil parameters directly on the shaking table model. Their
experimental curves are presented in Figure 2.13. These experimental studies were conducted at
the same soil material (LB sand, 14-25) and at the same density Dr close to 60%.
Figure 2.12 Empirical correlations for dynamic soil properties of LB sand derived from the
laboratory dynamic tests of Cavallaro et al (2001)
SERIES 227887 DYNCREW Project
19
Figure 2.13 Dynamic soil properties of LB sand measured experimentally by shaking table
testing at EQUALS (from Dietz & Muir Wood, 2007)
2.3 EXPERIMENTAL PROCEDURE
2.3.1 Model Configurations
The experimental procedure was repeated for three different Configurations (No1, No2 and No3)
for the wall model presented in Figure 2.8 to provide different response in sliding and rocking of
the base. These configurations are summarized in Table 2.3. In Configuration 2, the wall heel
was shortened by 50mm and the toe was totally removed. In Configuration 3, the geometry from
Configuration 2 was adopted, after increasing the frictional resistance of the base interface from
23.5o to 28
o (approximately equal to the critical state angle), by pasting rough sandpaper. The
interface friction angles were measured by means of static pull tests on the wall. The differences
between these three configurations, in terms of a pseudo-dynamic stability analysis according to
EC7 and EC8, are summarised in Table 4, ranging from a purely sliding-sensitive wall
(Configuration 1), to a purely rotationally sensitive one, mobilizing a bearing capacity failure
mechanism (Configuration 3). Configuration No2, represents a more complex, intermediate
mechanism, yielding simultaneously to sliding and bearing capacity.
SERIES 227887 DYNCREW Project
20
Table 2.3. Pseudostatic critical accelerations and associated safety factors (SF) with respect to
sliding and bearing capacity
Test configuration
Critical
acceleration for
SFsliding = 1
SFBearing capacity
at critical sliding
acceleration
Critical
acceleration for
SFBearing capacity = 1
SFsliding
at critical
bearing capacity
acceleration
Configuration No1 0.18g 7.45 0.35g 0.68
Configuration No2 0.14g 1.46 0.17g 0.93
Configuration No3 0.23g 0.44 0.17g 1.14
2.3.2 Iterative testing procedure
The experimental procedure described in this paragraph corresponds mainly to the more
systematic second phase of testing (June 2011), whereas the tests of the first phase were mainly
useful to the calibration and better preparation of the testing model of the second phase. Every
testing Configuration was subjected to the same series of dynamic tests, described below step-
by-step:
1) Placement of the wall on the foundation layer and investigation of its dynamic
behaviour by means of white noise testing.
2) Backfill construction and repetition of white noise testing to investigate both the
backfill and the combined soil-wall system response.
3) Harmonic-sinusoidal dynamic testing at various excitation frequencies and increasing
amplitude, until yielding of the wall and sufficient plastic deformation of the system is recorded.
4) Removal of the backfill, reposition of the wall on site and repetition of steps 1 and 2.
5) Dynamic excitation of the system with real earthquake signals, scaled in frequency
and increasing amplitude, until yielding of the wall and sufficient plastic deformation of the
system.
6) Removal of backfill and wall, placement of the next wall Configuration and repetition
of the whole procedure (steps 1 - 6)
The most time consuming part of the procedure described above is the removal and
reinstallation of the backfill and the wall, demanding the handling of a great amount of sand
(approximately 2.6tons) and the dismantling-reconnection-check of a great number of
instruments and cable connections. Given the time restrictions of the 3 weeks available for
testing, some modifications of the program were necessary: (a) In the case of Configuration No
1, the plastic deformation of the system during the harmonic excitation was kept relatively small
SERIES 227887 DYNCREW Project
21
(1/100 of wall height) and the earthquake loading continued to the same specimen. The
elimination of step 4 was deemed reasonable since harmonic testing had been performed
extensively during the first testing phase (December 2010). (β) The backfill was totally removed
only once, at the end of Configuration No1. At the end of all testing cycles, only the soil prism
behind the wall was removed and replaced, with special care for the temporary support of the
rest backfill. These modifications provided significant time economy for the completion of the
whole experimental program.
2.3.3 White noise testing
During white noise exploratory testing, a random noise signal of bandwidth 1-100Hz and RMS
acceleration = 0.005g was employed. During each exploratory series test, and simultaneous data
acquisition, system transmissibility was monitored using a two-channel spectrum analyser
(Advantest 9211C). The analyzer computes the frequency response function (FRF) between the
input and the output signals of interest (A1 accelerometer on the shaking table and A6 on the top
of wall stem respectively). The frequency response function is applied to the product between the
signal data and a Hanning window function (rectangular). Natural frequency and damping values
for resonances up to 40Hz (i.e. within the seismic frequency range) were determined for well-
defined resonances using the output of the analyzer’s curve fitting algorithm by means of a least-
squares error technique. Some characteristic outputs of the curve fitting algorithm are provided
in the Appendix.
For the interpretation of the dynamic behaviour of the system, more transfer functions between
any selected pair of channels need to be calculated using a Matlab code. These results are
presented in the ensuing, together with an interpretation of the resulting resonant frequencies.
2.3.4 Dynamic testing with harmonic, sinusoidal excitation
This type of input acceleration was imposed by sinusoidal excitation consisting of 15 steady
cycles. To smoothen out the transition between transient and steady-state response, the excitation
comprises of a 5-cycle ramp up to full test level at the beginning of the excitation, and a 5-cycle
ramp down to zero at the end. An example of the harmonic excitation signals is presented in
Figure 2.14. With reference to frequency and acceleration level, a wide range of excitation
frequencies (from 1 to 58Hz, every 3Hz) and various amplitudes was tested during the first
testing phase (December 2010). At the second phase of testing (June 2011) the excitation
SERIES 227887 DYNCREW Project
22
frequencies were restricted to a set of 5 frequencies (4, 7, 13, 25 and 43Hz), used initially at a
low acceleration amplitude of 0.05g, for studying the dynamic response of the system. An
excitation frequency of 7Hz was then selected for a series of harmonic excitations with
increasing amplitude, until failure. These conditions are essentially pseudostatic, as the above
frequency is much smaller than the resonant frequencies of the system.
Figure 2.14 Harmonic base excitation at the frequency of 7Hz and amplitude 0.05g
2.3.5 Dynamic testing with real earthquake records
Three earthquake records from the Italian and American database were selected for the
earthquake testing:
a) The Sturno record from Irpinia, 1980 earthquake (Mw= 6.9, PGA=0.321g), characterized by a
long strong motion duration of 16.2s and energy transfer to a wide range of frequencies from
0.25 to 10Hz (predominant frequency of 0.44Hz).
b) The Tolmezzo record from Friuli, 1976 earthquake (Mw=6.5, PGA=0.315g) with a smaller
duration of 4.92s, frequency range from 0.8 to 5Hz and predominant frequency of 1.5Hz.
c) The Northridge record from Los Angeles, 1994 earthquake (Mw = 6.7, PGA =0.47g), with
duration of 10.3s and narrow frequency range of 0.14 to 0.7Hz
The reproduction of these signals at the shaking table demanded an iterative matching process,
for calibrating of the displacement input of the table actuators. This procedure was performed
after the first phase of testing (December 2010), just before the removal of the model from the
shaking table, in order to have the same mass and geometrical conditions. The reproduction of
the signals was not 100% accurate, due to the large weight and eccentricity of the model.
SERIES 227887 DYNCREW Project
23
The authentic signals were scaled by a frequency scale factor of 5, assuming a geometrical scale
factor of n = 9 corresponding to a prototype of 5.4m high and applying the scaling law n0.75
,
which is valid for 1-g modelling (Muir Wood et al., 2002). The frequency-scaled signals were
applied at a low acceleration amplitude of 0.05g to measure the dynamic response of the model
and then the Sturno record was selected for carrying out increasing amplitude dynamic testing,
until failure in sliding or tilting of the retaining wall. The specific signal was selected because of
the long duration and wide frequency range. Time histories and corresponding Fourier spectra
are presented in Figure 2.15, whereas in Figure 2.16 the differences between the authentic (red
dashed line) and the modified Sturno record used as excitation signal (continuous black line) are
presented, in terms of acceleration, velocity and displacement time histories and Fourier spectra.
Figure 2.15 Scaled seismic signals imposed as base acceleration
SERIES 227887 DYNCREW Project
24
Figure 2.16 Comparison between the authentic and modified Sturno record
3 Experimental results
3.1 DYNAMIC PROPERTIES OF THE MODEL
Estimation of the dynamic model parameters was based on the interpretation of the white noise
testing results and their validation through empirical and theoretical relations from the literature.
In Figures 3.2 to 3.8 the transfer functions calculated at the positions shown in Figure 3.1, for
Configurations No1 and No3, during various phases of model construction, are presented. The
resulting resonant frequencies are noted on the graphs.
SERIES 227887 DYNCREW Project
25
A1
A8A9A10A11
A5
A6
A18
A20
A15
Figure 3.1 Accelerometer positions for white noise testing evaluation
From Figure 3.2, referring to dynamic response of the wall without the backfill in Configuration
No1, the frequencies 25.5Hz and 23.5Hz clearly appear. The first one corresponds to the
predominant frequency of a fixed cantilever beam with length L = 0.57m according to Eq. (3.1)
(Clough & Penzien, 1993) for Ε = 70GPa, m = 0.864kN/m (mass of the cantilever beam per m of
length) and Ι=2.73*10-6
m4 the moment of inertia for the 32mm thick cross section, of 1m width.
2
1 4
1.87525.6
2
EIf Hz
mL (3.1)
This frequency (estimated both theoretically and experimentally), reduces to 23.5Hz when the
response of the wall is examined, as the cantilever beam is fixed to a rectangular footing that is
not completely rotationally constrained. The same frequency further reduces to 17Hz, in the case
of Configuration No3, when the footing dimensions decrease and the system is even more free to
rotate (Figure 3.6). In that specific case, the rocking frequency of the foundation is predominant
for the system - not of the cantilever beam. This explains why the frequency of 25.5Hz is not
measured. At the foundation soil layer, before the backfill construction, the frequencies of 72 Hz
(free field) and 77Hz (under the wall) were measured. When the backfill is added, the
frequencies of 36, 47 and 52-55Hz, appear at all cases and have to be interpreted in terms of the
system vibrating mode. Finally, from the accelerometers on the rings of the shear box, an extra
frequency of 12 – 13.5Hz was recorded, corresponding to the resonant frequency of the container
full of soil, which naturally exhibits a stiffer behaviour.
Beginning from the foundation layer and the resonant frequencies of 72 and 77Hz, the relation
G0 = 16ρH 2
f connecting the resonant frequency with the shear wave propagation velocity Vs and
accordingly with the shear modulus at low strains G0, yields the values of 20.5GPa and 23.5GPa
SERIES 227887 DYNCREW Project
26
respectively. The soil under the wall is naturally slightly stiffer than at the free field, due to the
additional confining stress imposed from the self weight of the wall. There are many empirical
equations available in soil mechanics literature, describing the parabolic relation of G0 with the
isotropic confining stress. Suitable formulas derived for rounded silicate sands that can be used
for comparisons have been proposed from Hardin & Black (1966), Hardin & Drenvich (1969)
(Eq. 3.2) και Iwasaki et al. (1978) (Eq. 3.3) among others.
2
0
(2.973 )3230
1
eG p
e
(3.2)
20.4
0
(2.17 )9000
1
eG p
e
(3.3)
In the above equations, 01 2 / 3vp K . A more detailed analysis of this topic can be found
in Ishihara (1996).
SERIES 227887 DYNCREW Project
27
Figure 3.2 Transfer functions from the wall response in Configuration No1 – wall placed on the
foundation layer before backfill construction
SERIES 227887 DYNCREW Project
28
Figure 3.3 Transfer functions from the foundation layer response in Configuration No1 – before
backfill construction
SERIES 227887 DYNCREW Project
29
Figure 3.4 Transfer functions from the wall response in Configuration No1 – after backfill
construction
SERIES 227887 DYNCREW Project
30
Figure 3.5 Transfer functions from the backfill response in Configuration No1
SERIES 227887 DYNCREW Project
31
Figure 3.6 Transfer functions from the wall response in Configuration No3 – wall placed on the
foundation layer before backfill construction (*refers to the soil prism behind the wall)
SERIES 227887 DYNCREW Project
32
Figure 3.7 Transfer functions from the wall response in Configuration No3 after the construction
of the backfill
SERIES 227887 DYNCREW Project
33
Figure 3.8 Transfer functions from the backfill response in Configuration No3.
SERIES 227887 DYNCREW Project
34
Empirical Eqs (3.2) and (3.3) for the foundation layer properties, yield the same stiffness values
(calculated at the middle of the layer) to those recorded, if the vertical confining stress is used
instead of the isotropic stress corresponding to an earth pressure coefficient Κ0~1. This earth
pressure coefficient is much higher than the values 0.445 to 0.46 that have been estimated
experimentally for dense to loose Leighton Buzzard sand 14-25, from triaxial tests (Stroud,
1971). On the other hand, the triaxial tests were conducted at much higher confining stresses p’,
than in the shaking table conditions. Recent experimental findings (Chu & Gan, 2004) indicate
that the earth pressure coefficient Κ0 in loose silicate sands converges to 1 when the confining
stress converges to zero (Figure 3.9). In this light, it is a rational hypohesis to use the vertical
stress in Eqs (3.2) and (3.3) (that is the isotropic p’ for K0 = 1).
Figure 3.9 Variation of earth pressure coefficient K0 at very low confining stress (from Chu &
Gan, 2004)
In the same way, if the calculations from Eqs (3.2) and (3.3) are repeated for the middle of the
backfill, the value of 20GPa is derived for G0, corresponding to the resonant frequency around
47Hz. With the backfill present, the foundation soil underneath becomes stiffer, with G0
calculated at 40GPa. If the frequency of 47Hz, recorded also during white noise testing analysis,
is the dominant frequency of the 0.6m high backfill, the next step is the calculation of the
resonant frequency of the two-layered profile of Figure 3.11, by means of the known
amplification factor of Eq. (3.4) (Kramer, 1996).
2 21 2 1 2
* * * *
1 2 1 1 1 2
1AF =
cos cos sin sins
s s s s s
VH H H H
V V V V V
(3.4)
SERIES 227887 DYNCREW Project
35
Where * 1 2s sV V i . Indeed, for the given soil properties of each layer and for an
approximate damping ratio β = 0.05, the calculated resonant frequency of 36Hz is exactly the
same as the one measured during white noise testing. The only frequency coming from white
noise testing, remaining to be indentified is that of 52 -55Hz.
Figure 3.10 Increase in soil shear modulus with depth based on empirical relations
SERIES 227887 DYNCREW Project
36
H1
H2
ρ2, Vs2
ρ1, Vs1
u f f
u (t) = u0 e -iωt
Figure 3.11 Resonant frequency of the two-layered medium with properties: H1 = 0.4m, Vs1 =
157m/s, H2 = 0.6m, Vs2 = 113m/s
It is known that the resonant frequency f1 of a soil layer of finite length, is related to the
corresponding free field frequency f1,ff of an infinite layer through a geometrical coefficient Φ,
as shown in Eq. (3.5), which is a function of Poisson ratio v and the backfill’s height to length
ratio, H/L.
1 1, fff f (3.5)
Analytical expressions for Φ have been proposed by numerous researchers (Matsuo & Ohara,
1960; Wood, 1973; Veletsos et al, 1995; Wu & Finn, 1999), under appropriate simplifications
and approximations during the elastodynamic equations solving, as is the assumption of zero
vertical strain or zero vertical dynamic stress. If these expressions from literature are applied to
the shaking table model, that is for f1ff = 47Hz, H = 0.6m, L = 3m και ν = 0.3, the resonant
frequency of 56 – 58Hz is derived for the soil-wall system. This is slightly higher than the
frequencies actually recorded, but on the other hand, the retaining wall at the experimental
conditions is not rigid, as the analytical solutions assume. Given that the retaining wall can yield
and rotate, the soil-wall system is naturally softer, and from this point of view, the recorded
frequencies of 52 - 55Hz are in total agreement with the theoretical predictions. Summarizing the
SERIES 227887 DYNCREW Project
37
conclusions of the dynamic tests, the elastic parameters of the model are presented in Figure
3.12. Additionally, the Poisson’s ratio ν can be taken equal to 0.3 for all layers, and the damping
ratios were measured from 2.5% to 5%, as shown in the results of the white noise testing in the
Appendix.
Figure 3.12 Elastic parameters of the soil layers of the problem
3.2 EXPERIMENTAL RESULTS FOR DISPLACEMENTS, SEISMIC LOADS AND
FAILURE MECHANISMS
The measurements of system displacements for all tested configurations are summarized in
Figure 3.13. The total, cumulative settlements and rotations of the wall, for each series of
sequential input motions are presented in Figures 3.13a and 3.13b, and the incremental
displacements for each input motion are presented in Figures 3.13c-3.13e, indicating different
behaviour of the wall models under the same input. The measurements confirm the predictions
for the expected failure modes and the levels of critical acceleration. The sliding failure is clearly
visible in Configuration No1, as is the bearing capacity failure in Configuration No3.
Configuration No2 although designed to be weaker in sliding, also exhibits significant rotational
deformations caused by the high eccentricities induced by the seismic thrust. Rotational
deformations are also observed in Configuration No1 for high acceleration levels, revealing that
walls resting on a compliant base exhibit local bearing capacity failure near the toe, due to high
SERIES 227887 DYNCREW Project
38
compressive stress concentration. This observation elucidates the importance of properly
designing retaining structures to avoid developing significant rotational response.
Some characteristic experimental measurements are presented in Figures 3.14 to 3.18, relative to
failure mechanisms, accelerations, dynamic and permanent displacements and bending moments,
organized in sets of graphs suitable for direct comparisons. Most of the herein presented results
relate mainly to Configurations No1 and No3, for they exhibit yielding near similar acceleration
conditions but in different modes. Limited results are presented for Configuration No2, which is
significantly weaker compared to the others, failing almost simultaneously in foundation sliding
and rotation (i.e., bearing capacity) mode, thus is less important for comparison reasons.
Typical results from Configurations No1 and No3, for two cases of harmonic loading (0.19g και
0.23g) are presented in Figures 3.14 and 3.17. The following observations can be made: First the
response of each configuration is as expected. A translational response mode is evident in
Configuration No1 and a rocking one in Configuration No3. Sliding discontinuities are obvious
on the acceleration time histories of Configuration No1 (Fig. 3.14a), at a critical acceleration
slightly higher than that of Table 2.3. Note that the translational yield acceleration is not steady,
but always increases after every successive yielding, as even a small rotation causes penetration
of the wall toe into the foundation soil thus increasing passive resistance. On the other hand,
Configuration No3 starts rotating at initiation of yielding, without any evidence of sliding
discontinuities on the recorded accelerogram. Second, in both cases the wall stem appears to
have an amplified response, mainly because of foundation rocking and secondarily of pure
bending of the stem. Naturally, this is more evident in Configuration No3. Third, both models
exhibit a consistent, repeatable, behaviour with respect to yielding.
The same results for earthquake loading on Configurations No1 and No3 are presented in Figures
3.15 and 3.18. In this case, the input motion contains higher effective peak accelerations, but the
number of important strong cycles (half cycle pulses) is only three. The sliding failure is again
clearly visible in Configuration No1, as is the bearing capacity failure in Configuration No3
caused by the high eccentricities induced by seismic thrust. An important notice about the failure
modes arising from the combination of the two comparisons, is that the bearing capacity failure
is more affected by the input acceleration level, whereas the behaviour of pure sliding
mechanisms is mainly controlled by the time interval of the strong motion, as known from
sliding block theory (Newmark & Rosenblueth, 1971; Kramer, 1996). Accordingly, rotational
mechanisms appear to be more critical under strong earthquakes, even though they are
SERIES 227887 DYNCREW Project
39
sufficiently resilient against sliding (Fig 3.17d and Fig 3.18d). Moreover, some rotational
deformation is also observed in Configuration 1 for high acceleration levels, revealing that a wall
resting on a compliant base exhibits local bearing capacity failure near the toe, due to
concentration of high compressive stresses. This observation elucidates the importance of
properly designing retaining structures to avoid development of significant rotational response.
From the acceleration distributions of Figs 3.17c and 3.18c, it can be observed that the
earthquake loading results to conditions closer to the assumptions of pseudostatic analysis, as a
soil mass moving in phase with the wall is evident, especially for the rotational mode of
Configuration No3. Contrary to the case of earthquake excitation, wall and soil under harmonic
loading appear to respond in a quite different way. Finally, the peak seismic increment of
bending moments is compared in Figs 3.17e and 3.18e. A noteworthy observation is that the
earth pressure on the wall stem increases when the system moves towards the backfill, that is for
an acceleration not critical for overall stability. On the other hand, at yielding acceleration, earth
pressure on the stem is minimum. This is in agreement with the findings of the analysis
presented by Green et al (2008) on a full scale numerical model and the experimental and
numerical results of Al Atik & Sitar (2010). By comparing Configurations No1 and No3, it can
be clearly identified that rotational modes induce lower earth pressures on the wall (due to
rotational flexibility of the foundation) but different distributions leading to a higher point of
application of the thrust.
Finally, the response of Configuration No2 to harmonic and earthquake excitation with PGA =
0.17g presented in Figure 3.16, exhibits a similar behaviour to both Configurations No1 and
No3, resulting to simultaneous sliding and rotational failure. As seen from both the acceleration
time histories (3.16a) and the failure mechanisms (Figure 3.16d) the sliding failure mode
prevails, which is consistent to the critical accelerations estimated in Table 2.3.
In Figure 3.19, measured settlement profiles at the state of failure are plotted together with the
assumed failure mechanisms for Configurations No2 and No3. The following are worthy of note:
First, these failure mechanisms were observed only for transient earthquake loading, whereas in
the case of harmonic excitation, the settlement profile could not clearly reveal the emergence of
the main failure planes, as it had more-or-less a smooth parabolic shape. This can be explained in
view of a non-uniform settlement and deformation mechanism and stronger dynamic effects
imposed by earthquake loading. Second, the progressive, cumulative development of plastic
deformation strongly affected the clear emergence of the failure mechanism. For example, in
SERIES 227887 DYNCREW Project
40
Configuration No1, deformation took place in small increments for sequential strong shaking. In
Configuration No3, the deformation progress was also incremental – the failure mechanism
presented in Figure 3.19 was observed at a very high deformation level, not reached in
Configuration No1. Only in the case of Configuration No2, when important plastic deformation
was instantly developed, a clear failure mechanism was noticed on the soil surface. Although
there is actually not a “rigid block” response in the retained soil mass, the experimental findings
show that the earthquake excitation induces a more uniform acceleration distribution within the
retained soil mass, which corresponds to a more uniform Rankine stress field, as assumed in the
pseudo-static analyses. Third, the assumed failure mechanisms confirm the estimations based on
the material properties, the stability analysis and the yield accelerations presented in Tables 2.2
and 2.3 respectively. This comparison is only indicative, as these are velocity characteristics, not
exactly coinciding with the corresponding stress characteristics.
SERIES 227887 DYNCREW Project
41
Figure 3.13. Measurements of wall displacement and rotation for all configurations and various
base excitations. (a) Cumulative footing rotation versus sliding (LVDT-D1), (b) Cumulative
footing settlement (LVDT-D4) versus sliding, (c), (d) and (e) Incremental wall displacement
(LVDTs D1-D2-D3) for Configurations No1, No2 and No3, respectively
SERIES 227887 DYNCREW Project
42
Figure 3.14. Comparison of typical experimental results for Configurations No1 and No3 under
harmonic-sinusoidal excitation: (a) measured wall accelerations, (b) corresponding wall
displacements, (c) positive acceleration distribution (maximum inertial forces towards the
backfill), (d) negative acceleration distribution (maximum inertial forces towards the wall)
SERIES 227887 DYNCREW Project
43
Figure 3.15. Comparison of typical experimental results for Configurations No1 and No3 under
seismic excitation: (a) measured wall accelerations, (b) corresponding wall displacements, (c)
positive acceleration distribution (maximum inertial forces towards the backfill), (d) negative
acceleration distribution (maximum inertial forces towards the wall)
SERIES 227887 DYNCREW Project
44
Figure 3.16. Typical experimental results for Configuration No2: (a) measured wall
accelerations for harmonic sinusoidal and seismic excitation, (b) corresponding wall
displacement, (c) positive acceleration distribution (maximum inertial forces towards the
backfill), (d) negative acceleration distribution (maximum inertial forces towards the wall)
SERIES 227887 DYNCREW Project
45
Figure 3.17 Comparison of typical experimental results for Configurations No1 and No3 under
harmonic-sinusoidal excitation: (a) measured wall accelerations, (b) corresponding wall
displacements, (c) negative acceleration distribution (maximum inertial forces towards the wall),
(d) increment of wall displacement (LVDTs D1-D2-D3) and (e) peak seismic increment of
bending moment for positive and negative acceleration.
SERIES 227887 DYNCREW Project
46
Figure 3.18 Comparison of typical experimental results for Configurations No1 and No3 under
seismic excitation: (a) measured wall accelerations, (b) corresponding wall displacements, (c)
negative acceleration distribution (maximum inertial forces towards the wall), (d) increment of
wall displacement (LVDTs D1-D2-D3) and (e) peak seismic increment of bending moment for
positive and negative acceleration.
SERIES 227887 DYNCREW Project
47
Figure 3.19 Backfill surface settlement distribution at failure: (a) initial grid geometry -
dimensions in mm, (b), (c) settlement distributions for Configurations No2 and No3,
respectively.
4 Interpretation of displacements in light of
sliding block theory predictions
A key issue related to the performance based design of retaining walls is the control of the
allowable displacements, involving sliding and tilting, to ensure the safety and good performance
of retaining systems and nearby structures. Various simplified methods are available in the
literature following the pioneering work of Newmark (1965), for the prediction of permanent soil
deformation under a given base excitation motion. These include the simplified Richards & Elms
(1979) method for sliding displacement prediction and the Zeng & Steedman (2000) method for
the corresponding tilting mode. These methods are schematically presented in Figure 4.1. Their
predictions are compared to the experimental results, for validation and better understanding of
the retaining wall behaviour and the corresponding failure mechanisms.
In order to apply the sliding block methodology in the case of the cantilever wall, the soil mass
above the foundation slab shall be considered as part of the “rigid block”. In addition, the
retaining walls of Figure 4.1 are founded on a rigid base so the rotational displacement estimated
by the Zeng & Steedman (2000) methodology, arises from the tendency of the wall to rotate
SERIES 227887 DYNCREW Project
48
about the toe (point O). As mentioned in the introduction, this is not the critical failure
mechanism when the wall is founded on a compliant base. In this case, the tilting deformation is
mainly an effect of bearing capacity failure at the toe area, something that is not taken into
consideration by Zeng & Steedman (2000), nor by any other simplified method in the literature.
Figure 4.1. Gravity retaining wall on rigid base examined against (a) sliding according Richards
& Elms, (1979) and (b) rotation about the toe (point O) according to Zeng & Steedman, (2000)
The permanent deformation of the wall according to the sliding/rotating block theory, is
calculated by double time integration, of the relative wall-base linear acceleration or the relative
radial acceleration, for which a good estimation of the corresponding “critical accelerations” is
necessary. Apart from performing the time integration, Richards & Elms (1979) proposed the
simple formula given in Eq. (4.1), which provides an upper-bound estimation of permanent
displacement based on true records (Franklin & Chang, 1977)
2 3
max max
40.087perm
y
a
a
vd (4.1)
In Eq. (4.1) amax and vmax is the maximum base acceleration and velocity and ay is the sliding
critical acceleration. The prediction of the above equation has to be multiplied with the number
of cycles for the case of a harmonic excitation.
The results of the theoretical methods are compared to the corresponding experimental data of
SERIES 227887 DYNCREW Project
49
Configurations No1 and No3, in Figures 4.2 to 4.6, separately for harmonic and seismic loading.
In the first graph of each figure, the permanent displacements (increments) recorded along the
wall by the 3 LVDTs D1-D3, are presented as functions of the motion PGA. On the graph the
critical acceleration values against sliding and tilting (bearing capacity) are also noted. It has to
be mentioned at this point, that the critical acceleration that can cause uplift and overturning of
the wall (provided the existence of a rigid base) are much higher (about 0.62g for Configuration
Νο1 and 0.43g for Configuration Νο3). In the next two graphs, the experimental data are
compared to the predictions of the theoretical methods, both for wall sliding (LVDT - D1) and
wall rotation, calculated from the measurements of LVDTs D1 and D2. In Figures 4.7 and 4.8,
theoretical and experimental time histories for sliding and tilting are compared, for
Configurations No1 and No3 under harmonic loading of 0.19g and 0.23g and under seismic
loading with PGA of 0.35g and 0.55g, respectively.
The following can be concluded from the graphs: First, yielding accelerations estimated
in Table 2.3 are confirmed judging from the behaviour of the experimental curves. Second, the
predictions of Richards & Elms (1979) method for the wall sliding are in relatively good
agreement with the experimental measurements. In the case of Configuration No1, which is a
mainly sliding mechanism, Richards & Elms (1979) method significantly overestimates the
permanent deformation, with increasing seismic excitation. In the case of Configuration No3,
there is a better agreement than in Configuration No1. For accelerations lower than critical the
permanent deformation is underestimated and it is overestimated for higher acceleration values.
These deviations can be explained from the point of view that Equation 4.1 is an upper bound of
the permanent deformation and considers the critical acceleration as a constant value. In reality,
the critical acceleration always increases, as the wall toe penetrates the underlying soil layer,
resulting to increasing sliding resistance. This has been already observed at the experimental
results (see Figures 3.14 and 3.16), as well as in numerical simulations (Green et al, 2008).
Additionally, the sliding block technique neglects any possible wall sliding back towards the
backfill, which in the case of strong shaking may not be negligible. Finally, the sliding block
technique initially underestimates permanent displacement in the case of Configuration No3, as
this deformation is mainly a product of the tilting mode of failure mobilized first.
Third, the Zeng & Steedman (2000) method fails to predict the experimentally measured
permanent rotation of the wall at low accelerations, whereas at higher accelerations, a notable
overestimation is observed. This can be explained taking under consideration the compliance of
SERIES 227887 DYNCREW Project
50
the foundation, and the corresponding bearing capacity failure mechanism, which is the
predominant rotational failure mode. It appears that the compliance of the base and the
development of plastic deformation under the wall toe absorbs energy and is rather helpful for
the stability of the wall during strong earthquakes, not allowing wall uplift and overturning. On
the other hand, comparisons of the two experimental Configurations show that the rotational
mechanism of Configuration No3 systematically yields greater deformation at strong earthquake
loading, whereas the sliding mechanism of Configuration No1 is more affected from harmonic
loading. This may indicate that the bearing capacity failure happens more rapidly, compared to
the sliding block mechanism mobilization, that is purely a time integration effect, depending on
the time interval within which the critical accelerations is been exceeded. Generally the
behaviour of sliding mechanisms is more stable and predictable, plus the fact that the wall can
sustain important sliding without damage, contrary to tilting structures.
SERIES 227887 DYNCREW Project
51
Figure 4.2 Comparisons of experimental permanent displacement and rotation with the
predictions of sliding/rotating block theory – Configuration No1 under harmonic loading
SERIES 227887 DYNCREW Project
52
Figure 4.3 Comparisons of experimental permanent displacement and rotation with the
predictions of sliding/rotating block theory – Configuration No3 under harmonic loading
SERIES 227887 DYNCREW Project
53
Figure 4.4 Comparisons of experimental permanent displacement and rotation with the
predictions of sliding/rotating block theory – Configuration No1 under seismic loading
SERIES 227887 DYNCREW Project
54
Figure 4.5 Comparisons of experimental permanent displacement and rotation with the
predictions of sliding/rotating block theory – Configuration No3 under harmonic loading
SERIES 227887 DYNCREW Project
55
Figure 4.6 Comparisons of time histories for experimental permanent displacement and rotation
with the predictions of sliding/rotating block theory – Configuration No1 under harmonic
loading 0.19g and 0.23g and under seismic loading 0.35g and 0.55g
SERIES 227887 DYNCREW Project
56
Figure 4.7 Comparisons of time histories for experimental permanent displacement and rotation
with the predictions of sliding/rotating block theory – Configuration No3 under harmonic
loading 0.19g and 0.23g and under seismic loading 0.35g and 0.55g
SERIES 227887 DYNCREW Project
57
5 Interpretation of dynamic bending moments
Given that no direct measurements of earth pressures were made, some estimates of earthquake
loading on the retaining wall can only be inferred from the bending moment measurements on
the wall stem. In fact, the strain gauges performed adequately in recording the dynamic variation
of the dynamic component of bending moment, but not its static counterpart due to gravitational
action. The only available information arises from the seismic component of the bending
moment, ΔΜ. As has already been mentioned, an increase in the wall bending moment (positive
ΔΜ) is observed when the wall moves towards the backfill and a corresponding decrease when
the wall is in an active failure condition. The same remarks have been based on numerical
simulations of a full scale cantilever retaining wall by Green et al (2008).
In the present section, an interpretation of the measured dynamic bending moments is presented,
based on simplified assumptions for the variation of inertial forces and earth pressures behind the
wall. This assumptions are presented in Figures 5.1 and 5.2. First, it is a rational hypothesis to
assume that the inertial force acting on the soil mass moving with the wall (hatched gray area in
Figure 5.1), exerts no earth pressure to the wall stem, as indicated by the AASHTO - NCHRP
Report 611 (2008). As a result, the actions that have to be taken under consideration are the
variation of earth pressures ΔPE on the vertical virtual back and the inertial force Fw acting on the
cantilever itself, at the time of minimum and maximum bending moments on the wall, as
recorded from the strain gauges. The first action is calculated by means of the acceleration
recordings at backfill accelerometers A13 and Α16 and the second from the relative top to base
acceleration of the cantilever beam Α6 - Α5, under the assumption of linear with depth
accelerations distribution (Al Atik & Sitar, 2007).
The calculations of earth pressure variations follow a kinematically mobilized assumption shown
in Figure 5.2, suggesting that when the inertial action drives the system towards to the backfill,
stress condition moves from K0 to passive, so contact stresses increase by an amount (ah/g)γz.
SERIES 227887 DYNCREW Project
58
Respectively, when the inertial forces drive the system towards active failure, there is a
corresponding decrease in contact stresses. If the seismic action is strong enough to cause active
failure, the contact stresses are described by an active stress field (see paragraph 2) and the
variation of stresses is equal to (ΚΑΕ – Κο)γz. A value of Κο~1 can be used in the calculations as
previously.
A5
A6
A13
A16
Wall inertial forces(A6A5) γw τw
Dynamic earth pressures
a(z) γs z ή (Κ0 ΚΑ) γs z
acceleration distribution
a(z)
A13
A16Fw
ΔPE
Figure 5.1 Inertial and contact forces contributing dynamic bending moment on the cantilever
wall
γz
10
KAE γz K0 γz
(ah/g) γz (ah/g) γz
ahah
( ) ( )kh kh
Figure 5.2 Stress variation in the backfill during dynamic loading
SERIES 227887 DYNCREW Project
59
In Figures 5.3 to 5.6, comparisons of the experimentally measured bending moments with the
predictions of the simplified theoretical approach are presented, for various cases of harmonic
and seismic loading and Configurations No1 and No3. Specifically, the cases of strong harmonic
(0.19g and 0.23g) and seismic (0.35g and 0.55g) loading, as well as two cases of weaker
excitation causing negligible yielding to the system (harmonic 0.15g and seismic 0.17g), are
presented. The experimental time history records for these cases are provided in the Appendix.
In all the above cases the experimental recordings show an increasing variation of the dynamic
bending moment, with increasing excitation level. The predictions of the simplified theoretical
model consistently fit to the experimental data, with higher deviations observed in the case of
Configuration No3, probably due to the characteristic rocking response of the later. The fitting
could be better, if a non-linear earth pressure distribution, suitable for rotational yielding modes
is applied (Kloukinas, 2012).
With respect to the wall stem inertial force effect on the dynamic bending moment, the
theoretical model estimates range between 2.5% and 10% of the total dynamic moment. This is
in agreement with the findings of Al Atik & Sitar (2007), who estimated that the contribution of
the inertial force of the wall stem to the bending moment ranges between 5 to 26 percent of total.
The contribution is minor at the point of the active failure and maximum at the opposite
direction.
Finally, it must be noted that for a safer interpretation of the dynamic loads on the system, more
information is needed through rigorous numerical simulations of the experimental model.
SERIES 227887 DYNCREW Project
60
\
Figure 5.3 Comparisons of experimental and theoretical dynamic bending moments for
Configuration No1 under harmonic loading of amplitude 0.15g, 0.19g and 0.23g
SERIES 227887 DYNCREW Project
61
Figure 5.4 Comparisons of experimental and theoretical dynamic bending moments for
Configuration No1 under seismic loading of PGA 0.17g, 0.35g and 0.55g
SERIES 227887 DYNCREW Project
62
Figure 5.5 Comparisons of experimental and theoretical dynamic bending moments for
Configuration No3 under harmonic loading of amplitude 0.15g, 0.19g and 0.23g
SERIES 227887 DYNCREW Project
63
Figure 5.6 Comparisons of experimental and theoretical dynamic bending moments for
Configuration No3 under seismic loading of PGA 0.17g, 0.35g and 0.55g
SERIES 227887 DYNCREW Project
64
6 Conclusions
In the present report, the results of a suite of shaking table tests on model cantilever retaining
walls conducted in the BLADE laboratory at the University of Bristol, were presented. The
initial motivation of this experimental study was the validation of recent stress limit analysis
solutions for the seismic design of this type of retaining structures (Evangelista et al, 2009, 2010;
Kloukinas and Mylonakis, 2011) in conjunction with the absence of any specific, relative
regulations in established seismic codes, including EC-8. Special issues related to stability design
and response of walls founded on compliant base, were also studied. The experimental design
and procedure, as well as the main outcomes were described in detail and can be summarised to
the following conclusions:
1) The experimental results confirm the predictions of the theoretical stress limit analysis,
with reference to the failure mechanisms and the critical yield accelerations of the system.
Pseudo-static stability analysis proves to behave adequately for both harmonic and seismic
excitation, although important dynamic effects are evident in the first case, with reference to the
response of the backfill and the wall stem. On the other hand, earthquake loading results to
conditions closer to the assumptions of the pseudo-static analysis, namely the uniform
distribution of the acceleration and the “rigid block” response of the backfill.
2) The response of the various experimental configurations confirm the predictions of the
stability analysis of the retaining wall in terms of an equivalent footing and the relationship
between wall tilting and bearing capacity failure. Wall tilting is commonly observed at the vast
majority of retaining walls founded on compliant base, related to bearing capacity failure due to
high eccentricities and inclination of the reaction transferred to the wall foundation. This remark
highlights the importance of a proper design of walls founded on compliant base with respect to
sliding and rocking.
SERIES 227887 DYNCREW Project
65
3) The experimental measurements of permanent plastic deformation of the soil-wall
system fit quite well to the predictions of the simplified sliding block technique of Richards &
Elms (1979), but not to the predictions of the corresponding rotating block model of Zeng &
Steedman (2000). The significant deviations observed are explained in terms of tilting failure
related to wall foundation settlement and bearing capacity failure, instead of uplift and
overturning mechanism adopted in Zeng & Steedman (2000) analysis.
4) With respect to the effect of seismic loads on the structure, it is proved that the soil
thrust maximizing the bending moment of the stem, do not coincide to the critical earth pressure
for overall stability, but they appear in opposite phase. This indicates that different load
combinations have to be used for the static design of the wall and for the stability analysis of the
system against sliding and bearing capacity. This is in agreement with numerical analysis results
by Green et al (2008). Moreover, the dynamic bending moments interpretation is in agreement
with the findings of centrifuge testing by Al Atik & Sitar (2007), relative to the effect of the
inertial force on the wall stem.
5) Finally, the suitability of modern codes and specifications (ΕC7, AASHTO) in design
of retaining walls compliant to sliding is confirmed. The specific behaviour appears to be less
sensitive to seismic excitations compared to rotationally sensitive mechanisms. The rotational
compliance of wall foundations result in smaller bending loads on the cantilever beam, but also
on greater permanent rotational deformations, that can cause induce failure on the structure.
SERIES 227887 DYNCREW Project
66
References
AASHTO (Association of State Highway and Transportation Officials) (2009). ― "LRFD Bridge Design
Specifications, Customary U.S. Units", 4th
Edition (2007) with 2008 and 2009 Interims, AASHTO, Washington,
D.C.
Al Atik, L. and Sitar, N. (2007). "Development of Improved Procedures for Seismic Design of Buried and Partially
Buried Structures", Final Report 2006/06, Pacific Earthquake Engineering Research Center- PEER, University
of California, Berkeley.
Al Atik, L. and Sitar, N. (2010). "Seismic Earth Pressures on Cantilever Retaining Structures", Journal of
Geotechnical and Geoenvironmental Engirineering, ASCE, Vol. 136, No. 10, pp. 1324-1333.
Bhattacharya, S., Lombardi, D., Dihoru, L., Dietz, M.S., Crewe, A.J. and Taylor, C. (2012). "Role of Seismic
Testing Facilities in Performance-Based Earthquake Engineering Geotechnical", Geological and Earthquake
Engineering, V. 22, pp. 135-158.
Bolton, M.D. (1986). "The Strength and Dilatancy of Sands", Géotechnique, 36(1):65–78
Cavallaro, A., Maugeri, M., Mazzarella, R. (2001). "Static and Dynamic Properties of Leighton Buzzard Sand from
Laboratory Tests", Proc. of 4th int. conf. on recent adv. in geotech. earthquake engrg. and soil dyn. and
symposium in honour of Prof. WD Liam Finn, San Diego, California.
Chu, J. and Gan. C.L. (2004). "Effect of Void Ratio on K0 of Loose Sand", Géotechnique, Vol.54, No 4, pp. 285-
288.
Clough, R. W. and Penzien, J. (1993). "Dynamics of Structures", 2nd
Edition, McGrawHill, Inc., New York.
Crewe, A.J., Lings, M.L., Taylor, C.A., Yeung, A.K. & Andrighetto, R. (1995). "Development of a Large Flexible
Shear Stack for Testing Dry Sand and Simple Direct Foundations on a Shaking Table", European seismic
design practice, Elnashai (ed), Balkema, Rotterdam.
Dietz, M.S., Lings, M.L. (2006). "Post Peak Strength of Interfaces in a Stress-Dilatancy Framework", Journal of
Geotech. and Geoenvironmental Engrg., ASCE, vol. 32, n. 11.
Dietz, M.S. and Muir Wood D. (2007). "Shaking Table Evaluation of Dynamic Soil Properties", Proceedings of the
fourth international conference on earthquake geotechnical engineering, Paper 1196, Thessaloniki, Greece,
June 24–28.
ΕΑΚ 2000 (2003), Greek Seismic Code, Earthquake Planning and Protection Organization, Athens.
EN 1997-1 (2004). "Eurocode 7, Geotechnical Design, Part 1: General Rules", CEN, E.C. for Standardization,
Bruxelles.
EN 1998-5 (2004). "Eurocode 8, Design Provisions for Earthquake Resistance of Structures, Part 5: Foundations,
Retaining Structures and Geotechnical Aspects", CEN E.C. for Standardization, Bruxelles.
SERIES 227887 DYNCREW Project
67
Evangelista A., Scotto di Santolo A., Simonelli A.L. (2009). “Dynamic response of cantilever retaining walls. Proc.
of International Conference on Performance-Based Design in Earthquake Geotechnical Engineering — from
case history to practice — Tokyo, June 15 - 18, 2009.
Evangelista, Α., Scotto di Santolo, Α. and Simonelli, Α.L. (2010). "Evaluation of Pseudostatic Active Earth Pressure
Coefficient of Cantilever Retaining Walls", Soil Dynamics and Earthquake Engineering, Vol. 30, Issue 11, pp.
1119–1128.
Franklin, A.G. and Chang, F.K. (1977). "Permanent Displacements of Earth Embankments by Newmark Sliding
Block Analysis", Report No. 5, Earthquake Resistance of Earth and Rockfill Dams; Miscellaneous Paper s-71-
17, U.S. Army Corps of Engineers Waterways Experiment Station, Vicksburg, Miss.
Giarlelis, Ch. and Mylonakis, G. (2011). "Interpretation of Dynamic Retaining Wall Model Tests in Light of Elastic
and Plastic Solutions", Soil Dynamics & Earthquake Engineering, 31(1), 16-24.
Greco, V.R. (1997). "Stability of Retaining Walls Against Overturning", Journal of Geotech. & Geoenv.
Engineering, ASCE, Vol. 123, No 8, pp. 778 – 780.
Greco, V.R. (1999). "Active Earth Thrust on Cantilever Walls in General Conditions", Soils and Foundations, Vol.
39, No 6, pp. 65–78.
Green, R.A., Olgun, C.G. and Cameron, W.I. (2008). "Response and Modeling of Cantilever Retaining Walls
Subjected to Seismic Motions", Computer-Aided Civil and Infrastructure Engineering, No 23, pp 309–322.
Hardin, B.O. and Black, W.L. (1966). "Sand Stiffness under Various Triaxial Stresses", Journal of the Soil
Mechanics and Foundations Division, ASCE, Vol. 92, No. SM2, pp. 27-42.
Hardin, B.O. and Drenvich, V.P (1972). "Shear Modulus and Damping in Soils: Design Equations and Curves",
Journal of the Soils Mechanics and Foundation Division, ASCE, Vol. 98, No. SM7, July, pp. 667-692.
Huntington, W. C. (1957). "Earth Pressures and Retaining Walls", John Wiley and Sons, New York.
Ishihara, K. (1996). Soil Behaviour in Earthquake Geotechnics, Oxford Science Publications
Iwasaki, T., Tatsuoka, F., Tokida, K. and Yasuda, S. (1978). "A Practical Method for Assessing Soil Liquefaction
Potential Based on Case Studies at Various Sites in Japan", Proc. 2nd Int. Conf. on Microzonation, San
Fancisco, pp. 885-896.
Kelly, R.B., Houlsby, G,T. and Byrne B.W. (2006). "A Comparison of Field and Lab Tests of Caisson Foundation in
Sand and Clay", Géotechnique, 56(9):617–626
Kloukinas, P. (2012). "Contributions to Static and Seismic Analysis of Retaining Walls by Theoretical and
Experimental Methods", Ph.D. Dissertation, University of Patras, Greece.
Kloukinas, P., Mylonakis, G., Papantonopoulos, K. and Atmatzidis, D. (2007). "Seismic Earth Pressures on Gravity
Walls by Stress Limit Analysis", 4ICEGE, Paper No.1671, Thessaloniki, Greece, June 25-28.
Kloukinas, P. and Mylonakis, G. (2011). "Rankine Solution for Seismic Earth Pressures on L – Shaped Retaining
Walls", 5ICEGE, Santiago, Chile, January 10-13.
Kramer, S. L. (1996). "Geotechnical Earthquake Engineering", Prentice Hall, Upper Saddle River, New Jersey.
Leblanc, C., Byrne, B.W. and Houlsby, G.T. (2010). "Response of Stiff Piles to Random Two-Way Lateral
Loading", Géotechnique, 60(9):715–721.
Lings, M. and Dietz, M. (2005). "The Peak Strength of Sand Steel Interfaces and the Role of Dilation", Soil and
Foundations, Vol. 45, No. 6, pp. 1-14.
MathWorks Inc. (1995). "MATLAB v.4 User’s Guide", Prentice-Hall Inc., New Jersey.
SERIES 227887 DYNCREW Project
68
Matsuo, H. and Ohara, S. (1960). "Lateral Earth Pressure and Stability of Quay Walls During Earthquakes"
Proceedings of the second World Conference on Earthquake Engineering, Tokyo and Kyoto, Japan, Vol. 1, pp.
165-181.
Meyerhof, G.G. (1953). "The Bearing Capacity of Foundations Under Inclined and Eccentric Loads", Proceedings,
3rd
International Conference on Soil Mechanics and Foundation Engineering, Vol. 1 , 16-26.
Muir Wood, D., Crewe A. and Taylor C.A. (2002). "Shaking Table Testing of Geotechnical Models", International
Journal of Physical Modelling in Geotechnics, Vol. 2, pp..01-13.
NCHRP Report-611 (2008). "Seismic Analysis and Design of Retaining Walls, Buried Structures, Slopes and
Embankments", National Cooperative Highway Research Program, Tranportation Research Board, Washington
DC.
Newmark, N. M. (1965). "Effects of Earthquakes on Dams and Embankments", The Fifth Rankine Lecture,
Geotechnique, Vol. 15, No.2, pp. 137-161.
Newmark, N.M. and Rosenblueth, E. (1971). "Fundamentals of Earthquake Engineering", Prentice Hall, Englewood
Cliffs, New Jersey.
NTC (2008). Italian Building Code, DM 14 Jan., G.U. n. 29, 4 Feb., n. 30.
Rankine, W.J.M. (1857). "On the Stability of Loose Earth", Philosophical Transactions of the Royal Society of
London, Vol. 147, pp. 9 – 27.
Scotto di Santolo, Α. and Evangelista, Α. (2011). "Dynamic Active Earth Pressure on Cantilever Retaining Walls",
Computers and Geotechnics, Volume 38, Issue 8, pp 1041-1051.
Stroud, M.A. (1971). "The Behaviour of Sand at Low Stress Levels in the Simple Shear Apparatus", PhD Thesis,
Cambridge University, UK
Tan, F.S.C. (1990). "Centrifuge and Theoretical Modelling of Conical Footings on Sand", PhD Thesis, University of
Cambridge, U.K..
Veletsos A. S. and Younan A. H. (1994). "Dynamic Modeling and Response of Soil – Wall Systems". J. Geotech.
Engrg., 120(12), 2155-2179.
Veletsos A.S., Parikh, V.H. and Younan A.H. (1995). "Dynamic Response of a Pair of Walls Retaining a
Viscoelastic Solid", Earthquake Engrg. and Struct. Dyn., 24(12), 1587-1589.
Wood, J.H. (1973). "Earthquake-Induced Soil Pressures on Structures", EERL 73-05, Earthquake Engineering
Research Laboratory, California Inst. of Technology, Pasadena, CA.
Wu, G. and Finn, W.D.L. (1999). "Seismic Lateral Design of Rigid Walls", Canadian Geotechnical Journal, Vol.
36(3), pp. 509-522.
Younan, A.H. and Veletsos, A.S. (2000). "Dynamic Response of Flexible Retaining Walls", Earthquake
Engineering and Structural Dynamics, Vol.29, pp. 1815-1844.
Zeng, X. and Steedman, R.S. (2000). "Rotating Block Method for Seismic Displacement of Gravity Walls", Journal
of Geotechnical and Geoenviromental Engineering, ASCE, 126, 709-717.
SERIES 227887 DYNCREW Project
69
Appendix
SERIES 227887 DYNCREW Project
70
Appendix A.1 – Resonant frequencies and damping ratios from white noise testing
Configuration No1 – Wall on foundation layer – without backfill
SERIES 227887 DYNCREW Project
71
Configuration No1 – After placement of the backfill
SERIES 227887 DYNCREW Project
72
Configuration No2 – Wall on foundation layer – without backfill
SERIES 227887 DYNCREW Project
73
Configuration No2 – After placement of the backfill (before harmonic loading)
SERIES 227887 DYNCREW Project
74
Configuration No2 – After placement of the backfill (before seismic loading)
SERIES 227887 DYNCREW Project
75
Configuration No3 – Wall on foundation layer – without backfill
SERIES 227887 DYNCREW Project
76
Configuration No3 – After placement of the backfill (before harmonic loading)
SERIES 227887 DYNCREW Project
77
Configuration No3 – After placement of the backfill (before seismic loading)
SERIES 227887 DYNCREW Project
78
Appendix A.2 – Resonant frequencies and damping ratios from white noise testing
In the following set of figures, time histories of strain gauges measurements interpreted into
bending moments (Nm) are presented, for the same cases presented in Section 5 of the report.
The measurements correspond to the strain gauges arrays being situated on central line of the
wall stem. The internal array SG1-SG5 is plotted with black line and the external array SG8-
SG12 with red. As it can be seen from the graphs, the internal instruments work properly only at
Configuration No1. As a result, the bending moments estimation is mainly based on the external
array of strain gauges.
SERIES 227887 DYNCREW Project
79
SERIES 227887 DYNCREW Project
80
SERIES 227887 DYNCREW Project
81
Ο