Professional Documents
Culture Documents
oilD
yna
m ic
sandE
arth
qua
k eE
ngin
eerin
g11
6(2
019
)57
0–57
9
ARTICLEINFO ABSTRACT
A double wedge model has been proposed to compute seismic sliding displacements of cantilever retaining w alls.
Keywords:
Cantilever retaining walls Experimental observations indicate formation of rupture planes in the form of an inverted triangular wedge in the
Permanent displacement backfill of retaining walls. Computation of critical yield acceleration at each time instant considering the v- shaped
Seismic design weakest rupture planes which evolve during the ground motion is the preliminary aim of the double wedge model.
Sliding displacements The model computes translational displacements considering the relative movement of soil wedge along these
Velocity compatibility rupture planes. It considers velocity compatibility along with the acceleration compatibility between the wall and
Depth of subsidence soil wedge. It also computes approximate depth of subsidence of the backfill soil during ground motion. A
simplified double wedge model has been also considered wherein the yield acceleration at a ll time instants is
computed with respect to a fixed wedge. The double wedge model and its simplified version have been compared
with different cases studies, which show the seismic sliding displacements matching well with the real
measurements.
1. Introduction values beyond 0.45 g as against M-O method wherein the earth pres- sures
increase exponentially beyond 0.45 g. However, selection of this coefficient
The seismic design of cantilever retaining walls is currently per- has been a challenge as it involves major uncertainties [15–17], and hence
formed on the basis of force based approach. These are designed usually the pressures computed using this approach have been proved to be over
as gravity walls considering that the soil above heel is a part of wall. The or under conservative. Also, failure can be measured and understood
earth pressure is computed on the back face of this soil mass above heel, better in terms of displacements rather than forces [6,18]. Developing
using Mononobe-Okabe approach (M-O) [1–5] considering soil to soil design philosophy with deformation based approach to understand the
interface angle as ‘ϕ’ and assuming a suitable horizontal earthquake performance of cantilever retaining struc- tures is the motivation of this
acceleration coefficient. Design guidelines suggest some value between paper.
PGA/g to 0.5PGA/g of the considered earthquake acceleration as this Richards and Elms [19] proposed a methodology to design a gravity
horizontal coefficient [6]. Numerical modelling and dynamic analysis have wall assuming a translational displacement that the structure can
shown M-O method to compute under-estimated pressures [7,8]. comfortably undergo causing distress to the adjacent structures in ac-
Experimental investigations have proved M-O method to overestimate the ceptable limits. The yield acceleration coefficient was determined using
pressures exerted on cantilever retaining walls [9,10]. Real-time case Franklin and Chang's [20] expression for the expected displacement. This
studies and experimental investigations [11,12] have witnessed the expression was proposed after studying 169 corrected horizontal
formation of v-shaped wedges in the backfill during ground motion. earthquake motions using Newmark's sliding block method [21]. Tri- cario
This evoked an idea of computing earth pressure on the rupture plane [22] experimentally observed Newmark's method to overestimate the
inclined towards the wall stem. Greco [13] proposed a methodology to accumulated displacements of flexible retaining walls as it does not
compute pseudo static active earth pressure using Coulomb's theory consider the vertical acceleration of soil wedge. Deyanova [23] nu-
considering v-shaped wedge formation in the backfill by minimizing the merically observed Newmark's method to underestimate the accumu-
factor of safety against sliding. This methodology, however, demanded lated displacements of gravity retaining walls as it does not consider the
selection of suitable critical horizontal coefficient of earthquake accel- deformation of soil at the base of wall. Also, Newmark's sliding block model
eration. Kolay et.al.[14] through dynamic analysis observed earth does not consider the relative movement between wall and the soil wedge.
pressures to become asymptotic for horizontal acceleration coefficient This over-simplifies the true behaviour of wall and the
⁎
Corresponding author.
E-MAIL ADDRESSES: prajakta.ramesh.jadhav@iitgn.ac.in (P.R. Jadhav), ap@iitgn.ac.in (A. Prashant).
https://doi.org/10.1016/j.soildyn.2018.09.047
Received 27 February 2017; Received in revised form 16 August 2018; Accepted 29 September 2018
Av ailableon lin
e1 4N ovem ber2 018
02 67 -7
2 61/© 20
1 8ElsevierL td
.A llrig
h tsre
serv ed.
P.R. JADHAV, A. S
oilD
yNm
A icsA
NDErtH
A QUk
AeEn
gin
eerin
g11
6(20
19)5
70–
5 79
PRASHANT
A C
adjacent soil wedge. The multiblock model proposed by Stamatopolous
et al. [24] accounts for the change in geometry with time due to the (a)
khy'
outward displacements in the computation of yield acceleration unlike β
conventional sliding block model for slopes. Stamatopolous et al. [25] also k hg khy''
kvy' kt'g
proposed two-body model to compute coseismic displacements which is a
combination of gravity as well as seismic forces for gravity walls in kv
kvy'' θ2 kn'g
accordance with multiblock model for slopes. Zarrabi's model
θ1
[26] had also taken the amplitude of ground motion into consideration
B
along with change in geometry for gravity walls. This model considers
acceleration compatibility between wall and soil wedge at all time in-
stants during the ground motion, however, velocity compatibility was not P1
taken into account. For cantilever retaining walls, Kolay et.al., [14] (b)
the sliding displacements were highly overestimated by Richards and Ws (1- kv')
Elms approach [19] and slightly underestimated by the approach of Wong Ws (khy')
[27] at lower PGA and highly overestimated at higher PGA va- lues.
A A C
In this paper, double-wedge approach has been proposed to com- pute
(c)
sliding displacements of wall considering relative movement be- tween
wall and soil wedge along the rupture planes developed at that time
instant such that full contact is maintained between wall and soil wedge
throughout the motion. Velocity compatibility has been ensured along Ws
with the acceleration compatibility between the wall and soil wedge. The θ2
θ2
soil wedge tends to undergo subsidence in order to main- tain this contact
P1
by applying thrust on the back face of the wall [25]. The proposed Wkhy''
formulation also computes the depth of subsidence of soil wedge for a θ2 θ2
finite horizontal displacement by considering a simple and approximate
deformation profile. A simplified double wedge model has also been B B
proposed considering movement of soil wedge to occur pre- ferably along
Bntan
the already formed failure plane.
B
instant. Different positions
n
of outer rupture planes will have
2. Description of double wedge model
There are two possible cases depending upon the geometry of the
developed soil wedge, i.e., the case when rupture plane exists only in
backfill throughout and the other case when it intersects with the back face
of the wall. The intersection of outer rupture plane with the back face of
wall has been observed; for instance, Huang et al. [31] showed occurrence
of this phenomena in their experiments on cantilever re- taining walls
under geostatic conditions itself which can persist during seismic loading. Fig. 2. Double wedge approach: when outer rupture plane is intersecting wall
Hence, it is imperative to simulate this case. (a) Accelerations resolved along inner rupture plane and outer rupture plane (b)
force triangle (c) Free body diagram of soil wedge and wall with locked soil mass.
Case 1. Non-intersection of outer rupture plane with the back face of the
wall
with normal to the plane BC in order to resist the movement of soil
Referring to the free body diagrams of both wall with locked soil wedge along BC as shown in Fig. 1(a). The angles subtended by vectors,
mass and soil wedge shown in Fig. 1(c), and substituting the expres- N, P1 and Ws are ∠ʊ = 90 - (θ2 +ϕ+ α), ∠τ = θ1-ϕ+ α and ∠µ = 180-
sions for P1, N and Bn, the expression for yield acceleration coefficient
of w ual to 1 is as follows. (∠ʊ +∠τ), respectively, as shown in Fig. 1(b). Using sine rule, the
J tan magnitude of P1 is determined by Eq. (4).
b (1 kv) 1+G
khy = Ws sin
1+G P=
(3) 1
sin µ (4)
G tan 1 tan 2
wall
For this case, the tangential movement of the soil wedge is con-
R (sin 1 cos 1 tan )(kv cos 1 kh sin 1 cos 1) sidered only along the interfaces AB and BC shown in Fig. 2(a). Along
The value of θ2 is varied from 0 to tan−1(L/h). For a particular value of θ2, the value of θ1 can be computed using Coulomb's force triangle
approach
P.R. JADHAV,by
A. maximizing the earth pressure force on the locked in mass. The equations of equilibrium
SoilDy
N Amicshave
ANDE been
rtH
A QUAsolved
keEngin
efor
erinsoil
g116 wedge
(20
19)5and
70–wall
579
In Fig. 1(a), AB forms outer rupture plane and BC forms inner rupture
PRASHANT
with locked soil mass shown in Fig. 2(c). Eq. (5) represents yield
acceleration equation for wall with locked soil mass at factor of safety
equal to 1, which is derived in Appendix A.
plane. Ws is the weight of soil wedge in equilibrium. Considering the
J1 tan b (1 kv) kh G P1T1 cos 1
+ J2 + R
free body diagram of this soil wedge, active earth pressure force P1 acts 1+G Ws
at an angle ϕ to the normal of plane AB. This ϕ represents the friction khy = 1
1+G
+ J1 (5)
between soil particles along the inner rupture plane, which resists
sliding along AB. Reaction force from the soil mass acts at an angle ϕ Here,
P.R. JADHAV, A. S
oilD
yNm
A icsA
NDErtH
A QUk
AeEn
gin
eerin
g11
6(20
19)5
70–
5 79
PRASHANT
this khy' and kvy', the value of θ1 is recalculated for which P1 is
W cos T1 2 maximum in step A3. Steps A2 to A4 are repeated until the absolute
J1 =
W (sin( + )tan b
2 cos( + )) difference between θ1 values for consecutive iteration is less than a
threshold value (less than 1°). Usually, convergence has been
P1T2 T4 cos 1
J2 = achieved within 15 iterations.
Ws T3
G tan 1 tan 2
A6. The velocities are then computed for both wall as well as soil
wedge at that time step in such a way that movement of wall to-
R (sin 1 cos 1tan )(kh sin 1 kv cos 1 cos 1) wards the backfill is not allowed.
A7. The displacements are computed for the exceedance of earth-
T1 sin( 1)tan cos( 1)
quake acceleration from the yield acceleration of the wall by nu-
T2 sin( 2 1)tan cos( 2 1) merical integration which may lead in approximate analysis [34].
However, it has been observed from the parametric studies on dif-
T3 sin ( 2)tan b cos ( 2) ferent time steps by Stamatopolous et al. [25] that acceleration
For this case, the failure wedge, as shown in Fig. 2(a), will exert earth owing to the simplicity involved, numerical integration can be re- liably
pressure on wall at part AD and on the soil mass at part AB and thus adopted in the implementation of double wedge model. Also, it is
resulting in total seven unknowns unlike Case 1 with six unknowns and recommended that applied acceleration be obtained by using
six equations available for analysis. In order to take care of this un- equivalent linear or non-linear site-specific response analysis instead
derdetermined system, the maximum earth pressure on the AD part of of directly using bedrock motion to realistically simulate the field [25].
wall has been computed using the expression proposed by Mononobe-
Okabe [32,33]. It has been assumed that the assumptions of Mononobe- The values of khy'' values are calculated firstly for Case 1 only
Okabe [32,33] are valid in the region around AD and the force has been wherein all the values of θ2 lie in (0-tan−1(L/h)) range. Case 1 is suf-
computed considering the acceleration coefficients of the soil wedge, ficient only if the minima is obtained for plot of khy'' v/s θ2 for the value
i.e., khy' and kvy'. Computation of force, P2, using Mononobe-Okabe of θ2 ranging between 0-tan−1(L/h) signifying that the outer rupture
[32,33] along AD implies the formation of small failure wedge, ADE, plane is not intersecting wall as shown in Fig. 3(a) for a typical example
which would still traverse with the parent wedge ABCD. This is in ac- of cantilever retaining wall of height 5.625 m and heel length 1.95 m.
cordance with the formation of multiple shear bands in the parent Otherwise, Case 2 is also considered to obtain khy'' values for the values
failure wedge observed by Huang et al. [31]. Thus, in order to know the
value of P1, a force polygon is drawn for the known value of P2 as
shown in Fig. 2(b). The force polygon is divided into 2 parts as shown in (a)
0.23
Fig. 2(c) by dotted line. Solving the force polygon, the active earth
pressure exerted on part BC of wall is as shown in Eq. (6).
=
1
Xsin 0.225
P
sin (6)
Where, 0.22
ky
For three unknowns, two compatibility equations (Eqs. 1 and 2) and A iteration and corresponding cri-
4 tical θ1 in Eq. (3) for Case 1 and Eq. (5) for Case 2, the value
one equilibrium equation depending upon the case (Eqs. 3 or 5) have of khy''
been formulated for the assumed values of θ1 and θ2. In order to obtain .
U
the yield acceleration coefficients for the appropriate θ1 and θ2 the
s
model can be adopted by implementing the following steps. i
n
3. Implementation of model g
t
A1. Initially a value of θ2 is assumed and other parameters are h
calculated as given in following steps. This value of θ2 is then iter- e
ated from 0 to tan−1(L/h) for Case 1 and beyond tan−1(L/h) for Case v
a
2 to achieve minimum value of khy".
l
A2. The values for kh and kv are obtained from the considered ac-
u
celeration time history. For each θ2, α is computed as tan−1(khy'/ 1- e
kvy') with initial values of khy' as zero and kvy' as vertical ground s
acceleration coefficient value. o
A3. θ1 is varied from ϕ to suitable angle less than 90°. Active earth f
pressure force is calculated for computed α and each value of θ1 θ
using Eqs. (4) and (6) for Cases 1 and 2, respectively. To obtain 2
0.2
1
0.20
5
0
(b)
5
10
15
20
2
(
)
0.1
6
0.1
5
0.1
4
ky
0.1
3
0.1
2
0.1
1
0.
1
0
1
0
2
0
3
0
4
0
2
(
)
is calculated. Fig. 3. Double wedge approach: Variation of yield acceleration coefficient with
A5. Using Eqs. (1 and 2), khy' and hence kvy' can be calculated. Using θ2 for a wall with same geometrical configuration. (a) When outer rupture plane
is not intersecting wall (b) When outer rupture plane is intersecting wall.
P.R. JADHAV, A. S
oilD
yNm
A icsA
NDErtH
A QUk
AeEn
gin
eerin
g11
6(20
19)5
70–
5 79
PRASHANT
mass and soil wedge. Let this time instant be known as crossing time
(a) instant, x, as shown in Fig. 4(b). Let Vvw1 and Vvs1 represent the vertical
velocity of the wall and soil wedge, respectively, at the time instant T1.
Let Vvw2 and Vvs2 represent the vertical velocity of wall and soil wedge,
respectively, at the time instant T2 at the end of the time interval ∆t. Let
khy1' and khy2' represent the horizontal yield acceleration of the soil
wedge wedge at the beginning and the end of the time interval. The following
steps are followed to implement the required velocity compatibility.
B1. Identify the time interval, ∆t during which the vertical velocity
of wall is observed to exceed that of soil wedge in step A6.
B2. The acceleration is assumed to vary linearly between two con-
secutive time instants as the time interval is small. The crossing time
instant in the time interval, x, as shown in Fig. 4(b), at which both wall
and soil wedge have same velocities should be determined using Eq.
(7). At this time instant, there will be no relative move- ment between
wall, soil wedge and backfill mass.
x= (Vvs1 V ) t
(b) (Vvw2 + Vvs1
Vvs2
(7)
Velocity
Fig. 7. Schematic illustration of wall geometry with soil backfill properties and acceleration time histories for four case studies. (a) Case study 1: Ground acceleration
time history of Parkfield, California Earthquake (CIT [38]). (b) Case study 2: Ground acceleration time history of Takatori, Kobe Earthquake (PEER [42]). (c) Case study
3: Ground acceleration time history of Takatori, Kobe Earthquake (PEER [42]). (d) Case study 4: Schematic illustration of wall and digitized ground accel- eration time
history of Sturno-Irpinia Earthquake [40].
P.R. JADHAV, A. S
oilD
yNm
A icsA
NDErtH
A QUk
AeEn
gin
eerin
g11
6(20
19)5
70–
5 79
PRASHANT
Table 1
Displacements computed by four methods for all the four case studies.
Case study No. Parameters Actual Newmark Zarrabi (gravity walls) Double wedge model Simplified double wedge model
Case study 1
300 analysis assuming slip surface as the vertical face passing through heel
Case study 2 of wall. From the result obtained, it can be observed that this
Difference from actual values (%)
250
200 Case study 3 methodology either underestimate or highly overestimate the com-
150 puted displacement.
•
Case study 4
100 For computation of results using Zarrabi's model, the soil mass above
50
heel has been assumed to be a part of wall and the soil to soil friction
0
angle has been considered as the interface angle. This model has been
observed to generally underestimate the obtained dis- placements.
-50
The displacements computed using double wedge model can be
•
-100
observed to be much closer to the actual displacements in all case
-150
Newmark Double wedge Simplified Double Zarrabi studies. Double wedge model, being generalized model, can be adopted
wedge comfortably for all types of ground motions.
Methods
•
Fig. 8. Variation of computed displacements from four methods with respect to It can be observed that there is not much deviation in results com-
actual values for four case studies. puted by double wedge model and simplified double wedge model.
Simplified double wedge model gives approximately close results to
The weight of the wall was computed as 34.69 kN/m. The ground the measurements and hence can be directly adopted by engineers
motion data was extracted from PEER data base [42] as shown in Fig. owing to its simplicity.
7(b). • It can be also observed that the results computed using the proposed
Case study 3: A series of centrifuge experiments were performed on the models slightly underestimates the computed displacements in most
model cantilever retaining wall shown in Fig. 7(c) to develop a better cases. This needs consideration while prescribing design require-
ments. This will also need to give due care to the uncertainties in-
understanding of the distribution of magnitude of seismic earth
volved in the choice of parameters like interface friction angle be-
pressure at the Centre of Geotechnical Modelling UC Davis [10]. The
tween base of wall and soil.
geometrical and material properties were taken from the report [40]
and shown in Fig. 7(c). The structure was subjected to Kobe (1994)
earthquake record from Takatori site, as sourced from PEER database
7. Conclusion
[42] and shown in Fig. 7(c). The interface friction angle at the base was
chosen as 30°. Corresponding to PGA of 0.64 g, the rigid body
A double wedge model for estimating permanent displacement is
translation (δ/H) according to the measurements was 0.0165 m for
proposed for realistic design of cantilever retaining walls subjected to
5.67 m high wall.
Case study 4: A series of shaking table tests on cantilever retaining seismic excitation, which considers v-shaped rupture planes in the backfill
wall at Earthquake and Large structures laboratory (EQUALS) of as observed in earlier experimental investigations. This model considers
University of Bristol, UK [12,41]. The model retaining wall was made relative movement of soil wedge along these rupture planes. The proposed
from concrete (γ = 24 kN/m3) and of the dimensions shown in Fig. model accounts for velocity compatibility of the soil wedge with respect
7(d). Both the backfill and foundation soil consisted of dry, yellow to adjacent wall and backfill. Depth of subsidence of backfill soil has also
been computed in this approach. To reduce the computing effort, a
Leighton Buzzard silica sand of unit weight 13.73 kN/m3 and friction
simplified double wedge approach is also proposed which considers fixed
angle as 34°. The interface angle of foundation soil was 23°. The model
shape of rupture planes during earthquake shaking. The formulations are
was subjected to an earthquake excitation of Sturno re- cord from
Irpinia-Italy obtained from [12]. computed differently for outer rupture plane intersecting and not
intersecting the wall, and these models were validated by four different
6.2. OBSERVATIONS from CASE studies case studies. Double wedge model was ob- served to compute closer
displacements to actual values. Simplified
double wedge model computed approximately close results to that of
The displacements computed by all the methods for all the four case
double wedge model. Owing to the simplicity of the simplified double
studies are as shown in Table 1 and as represented in Fig. 8.
wedge model, it can be easily implemented by engineers.
Assuming that the vertical ground acceleration is not that strong and can be resisted by the weight of the wall,
kvy kv (A-9)
Solving for kt' in (A-4) and substituting in (A-5), and using Eq. (A-9),
kvy kv tan 1 (kh khy) (A-10)
Thus, from Eq. (A-10), it is evident that k v k v, as the wedge ruptured at θ1 can undergo vertical movements due to its horizontal inertial
acceleration and horizontal and vertical ground motion.
Substituting (A-10) in (A-8) and rearranging,
khy khy (khy kh)tan 1 tan 2 (A-11)
Eq. (A-11) gives the relationship between all the horizontal acceleration coefficients.
Firstly, the intersection of rupture planes with walls are considered, the earth pressure force, P2 exerted through part of soil wedge on the back
face of wall can be directly calculated from M-O formulation.
Consider the free body diagram of the soil wedge as shown in Fig. 2(c). The equilibrium equations along normal and tangential direction to plane
BC are as given by Eqs. (A-12) and (A-13).
N Ws cos 1 P1 sin( 2 1) P2 sin( 1) Ws kn (A-12)
P1 P2
cos 1 (sin( + 2 1)tan cos( + 2 1))
cos 1 (sin( + 1)tan cos( + 1))
Ws Ws (A-15)
Considering free body diagram of wall with locked soil mass, the equilibrium equations along horizontal and vertical direction with respect to
base are as follows,
Bn P1 =
Wkhy
(A-16) –
PRASHANT2)
WP.R. P2 sin(
JADHAV, A.
) S
oilD
yNm
A icsA
NDErtH
A QUk
AeEn
gin
eerin
g11
6(20
19)5
705 79
(A-17)
kv W
P1
sin(
(A-18)
Bn tan b P1 cos( 2) P2 cos(
P2 T1 cos 1
khy = R + J1 ((1 k v)tan b khy) + J2
Ws (A-19)
khy'' can be obtained substituting Eq. (A-19) in Eq. (A-11) and rearranging,
J1 tan b (1 k v) kh G P2 T1 cos 1
+ J2 + R
1+G Ws
khy = 1
1+G
+ J1 (A-20)
P.R. JADHAV, A. S
oilD
yNm
A icsA
NDErtH
A QUk
AeEn
gin
eerin
g11
6(20
19)5
70–
5 79
PRASHANT
The values for J1, J2, G, R, T1 and T2 are explained in the section.
When outer rupture plane is not intersecting wall, then following the steps similar to the above case, equilibrium equations with respect to free body
diagram can be deduced for both wall with locked soil mass and soil wedge and the interface angle along outer rupture plane making an inclination of θ2
will be considered ϕ.