E.

Cosenza (ed), Eurocode 8 Perspectives from the Italian Standpoint Workshop, 291-300, © 2009 Doppiavoce, Napoli, Italy

PERFORMANCE-BASED DESIGN OF EMBEDDED RETAINING WALLS

SUBJECTED TO SEISMIC LOADING

Luigi Callisto a, Fabio M. Soccodato b

a
University of Rome “La Sapienza”, Italy, [email protected]
b
University of Cagliari, Italy, [email protected]

ABSTRACT
In common practice, the seismic design of an embedded retaining wall is carried out using the
pseudo-static method. In this approach, constant forces are introduced in a limit equilibrium
calculation, and the seismic analysis of a retaining wall is treated similarly to the evaluation of
the safety against a collapse mechanism. This paper is aimed to propose a reconsideration of
the simple pseudo-static calculation: it shows that the method can be used within the context
of the performance-based design to predict the actual seismic performance of the wall, and
that concepts employed in the capacity design of structural members can be extended to the
design of embedded retaining walls. The paper also points to possible code prescriptions that
may provide guidance for the correct application of the pseudo-static method to the design of
retaining walls.

KEYWORDS
Seismic design, retaining walls, numerical analysis, nonlinear response, earthquake resistant
structures.

1 INTRODUCTION

In the performance-based design of a retaining wall, different degrees of seismic protection

can be prescribed, depending on the limit state being analysed. For instance, the current
Italian Construction Code (Decreto Ministeriale 14.1.2008) defines four different limit states,
each associated to a different seismic action, characterised by a given probability of
exceedance in the structure’s lifetime.
For a building, the seismic performance is usually expressed by the maximum instantaneous
displacement (e.g. the inter-storey drift) that occurs during the earthquake, and by the ductility
demand associated to this displacement. On the other hand, the seismic performance of a
retaining wall is more commonly expressed in terms of permanent displacements at the end of
the earthquake (Richards & Elms 1979, PIANC 2001). A possible reason of this difference is
that, unlike buildings, retaining structures undergo permanent displacements in one direction
only, and therefore the displacements increase monotononically, attaining a maximum at the
end of the seismic event.
In order for the displacements to be irreversible, they must stem from the instantaneous
development of a plastic mechanism; therefore these displacements are associated to quasi-
rigid body movements. For an embedded retaining wall, if one admits that only the soil
strength can be fully mobilised during the seismic event, while the retaining wall and the
292 L. Callisto, F.M. Soccodato

restraining system remain in an elastic state, then a plastic mechanism can form only if the
wall is cantilevered or singly restrained (propped or anchored). The present discussion is
largely devoted to the design of these wall categories. It is also assumed that, because of the
limited height of these wall types, the dynamic motion of the soil interacting with the
excavation is essentially synchronous. Some effects of asynchronicity on the design of the
retaining structures for deep excavations are discussed by Callisto & Aversa (2008).

2 EVALUATION OF THE SEISMIC PERFORMANCE

Upon instantaneous attainment of the available strength, cantilever and singly-restrained walls
can undergo rigid-body movements in a way which is qualitatively similar to the behaviour of
gravity retaining walls. This was shown, for instance, by the experiments of Richards & Elms
(1992) and Neelankatan et al. (1992), and by the results of some dynamic tests that were
recently carried out using the Cambridge Dynamic Centrifuge (Viggiani & Conti 2008): for
accelerations larger than a critical value, a progressive development of wall rotations was
observed. Figure 1 is taken from the results of a series of dynamic numerical analyses carried
out by Callisto et al. (2008): it shows the instantaneous distribution of the contact stresses σh
exerted by the soil against a pair of mutually propped retaining walls during the strong motion
phases of two different seismic events. During the earthquake, these stresses increase both at
the rear and in front of the walls, and permanent displacements occur as a consequence of full
mobilization of the soil strength. This phenomenon occurs in an alternate fashion to the two
facing walls: in Figure 1 it is happening to the left-hand wall. Figure 2 shows, for the two
walls, the progressive accumulation of the computed horizontal displacement of the toe
relative to the top.
For a given retaining wall, a critical value ac for the horizontal acceleration can be evaluated
by performing iteratively a limit equilibrium analysis, and finding the pseudo-static
acceleration for which soil strength is fully mobilised. Permanent displacements can then be
assumed to result from a Newmark-type integration of the relative motion, and are bound to
decrease as ac increases. For a given soil and a given excavation height H, the value of ac
depends essentially on the embedded length d. Therefore, the embedded length (or,
equivalently, the total length L = H + d) should be chosen on the basis of the maximum
displacement allowed for the seismic event (and therefore the limit state) under consideration.
Relationships between the permanent displacements u and the ratio ac/amax were recently
derived by Rampello & Callisto (2008) from a parametric integration of a database of Italian
Strong Motion Accelerograms (SISMA, Scasserra et al. 2008). Each recording was scaled to
different maximum accelerations amax, with a scale factor not exceeding the value of 2. Figure
3 shows the u-ac/amax relationships obtained for accelerograms recorded on rock and scaled to
amax = 0.35 g, that yielded the largest displacements. Also shown in the figure are regression
lines through the computed data points, of the form:

⎛ a ⎞
u = B exp⎜⎜ A c ⎟⎟ ; (1)
⎝ amax ⎠

specifically, the continuous line, computed with A = -7.4 and B = 1.8, is close to the upper
bound of the results and was used to develop a relationship between the expected
displacement (that define the requested seismic performance) and the pseudo-static horizontal
acceleration that was then adopted by Italian Construction Code (Decreto Ministeriale
14.1.2008) for the seismic design of embedded retaining walls.
Performance-Based Design of Embedded Retaining Walls Subjected to Seismic Loading 293

σ h (kPa) σ h (kPa)
100 75 50 25 0 -25 -50 -75 -100 -100 -75 -50 -25 0 25 50 75 100

2
z (m)

static
3 TM t = 5.0 s
AS t = 5.8 s
4
Kp = 6.162
5

Ka = 0.234

Figure 1. Contact stresses acting against a pair of propped retaining walls (Callisto et a. 2008).

0.02
right wall

0.01
u (m)

-0.01
left wall

-0.02
0 4 8 12 16 20
t (s)
Figure 2. Time histories of the top-toe relative displacements of the retaining walls shown in Figure 1.

Consider the example of Figure 4(a), taken from Soccodato & Callisto (2009), where an
excavation with H = 4 m in a homogenous coarse-grained soil is retained by a cantilevered
wall. The soil has a constant angle of friction ϕ = 35°, while the soil-wall angle of friction is δ
= 20°. For this retaining wall, the critical acceleration ac was found using the Blum (1931)
limit equilibrium method; the Mononobe-Okabe solution was used for the seismic active
pressure, while the coefficient of passive pressure was evaluated with the closed-form
solution developed by Lancellotta (2007).
Figure 4(b) shows, for the case at hand, the computed values of the critical acceleration ac
plotted as a function of the total length of the wall. As L varies from 7 to 9 m, ac increases
from 0.15 to 0.4 g. Figure 4(b) also shows the permanent wall displacements u computed
using equation (1) and assuming that the maximum horizontal acceleration in the soil
interacting with the wall is either amax = 0.5 g or amax = 0.75 g: as the length of the wall
increases, the permanent displacements u decrease rapidly.
Maximum bending moments were computed in the wall using the limit equilibrium method,
with a pseudo-static horizontal acceleration equal to ac: it can be expected that, as the
accelerations increase during an earthquake, so do the soil-wall contact stresses, until full
mobilisation of soil strength is attained, that is until the critical acceleration ac is reached. For
accelerations larger than ac, contact stresses cannot vary significantly, since soil strength is
294 L. Callisto, F.M. Soccodato

10
average trend
adopted function

u (m)
0.1

0.01

0.001
0.2 0.4 0.6 0.8
ac/amax

Figure 3. Relationships between permanent displacements and the ratio ac/amax obtained by Rampello &
Callisto (2008) for very stiff soils and amax = 0.35 g.

already fully mobilised, and the acceleration in excess of ac is spent to produce a movement
of the wall under quasi-constant contact stresses. Since the internal forces in the wall are a
consequence of the contact stresses, bending moments should increase as the acceleration
increases to up the critical value, and then remain constant while relative soil-wall
displacements occur.
The maximum bending moments computed using the critical acceleration are indicated as
M(ac) and are plotted in Figure 4(b) as a function of the wall length L. Since ac increases with
L, the bending moments increase as well. This means that if the wall is made longer in order
to improve its seismic performance (that is, to undergo smaller displacements) it will have to
sustain larger bending moments.
Of course, for a retaining wall with ac > amax the permanent displacements are negligible and
the internal forces in the wall are evaluated with a pseudo-static acceleration equal to amax.

3 DESIGN CRITERIA

In its essential terms, the design of an embedded retaining wall with no more than one restrain
could be performed through the following steps:
a. for a given limit state, define the required seismic performance by selecting the
maximum permanent displacement u;
b. evaluate the maximum horizontal acceleration amax expected for the limit state under
consideration;
c. from a relationship of the type shown in Figure 2, evaluate the critical acceleration ac
needed to meet with the desired seismic performance;
d. search iteratively, through the limit equilibrium method, the wall length L that gives the
required critical acceleration; if for any reason (e.g. for hydraulic needs) the chosen wall
length is larger than L, the critical acceleration ac must be recalculated using the actual
length;
e. compute the internal forces using the contact stresses evaluated with a = ac;
f. design the wall structure (and the eventual restraining system) on the basis of the
internal forces evaluated at the previous step.
Performance-Based Design of Embedded Retaining Walls Subjected to Seismic Loading 295

0.5 300

0.4 260

H=4m

ac (g), u (m)

M (kNm/m)
0.3 220
ac
0.2 u (amax=0.75 g) 180
u (amax=0.50 g)
L M (ac)
0.1 140

(a)
d (b)
0 100
7.0 8.0 9.0
L (m)
Figure 4. Layout of a cantilevered retaining wall in a coarse-grained soil (a) and plots of the critical
acceleration, the permanent displacement and the maximum bending moment as a function of the wall
length (b) (Soccodato & Callisto 2009).

Although the above sequence of activities is conceptually clear, some issues are believed to
necessitate further investigation.
Firstly, the evaluation of the maximum horizontal acceleration amax at step (b) poses some
problems: in principle, amax could be found using the simplified procedures based on
amplification coefficients, such as those prescribed by the Eurocode 8 (EN 1998-5) or by the
Italian Construction Code (Decreto Ministeriale 14.1.2008). Yet these procedures, being
largely based one-dimensional site response analyses, neglect the effect of two-dimensional
amplification and may underestimate amax. For instance, a parametric study on cantilevered
retaining walls, based on the results of two-dimensional dynamic numerical analyses (Callisto
& Soccodato 2009) showed that the maximum acceleration in the soil interacting with the
excavation may reach values larger than twice the corresponding maximum acceleration
computed in a one-dimensional analysis, and that this effect is not significantly related to the
soil or wall stiffness, but rather depends on the two-dimensional nature of wave propagation.
Step (c) implies the availability of relationships similar to the one shown in Figure 2, as
specific as possible to the geographic region and to the source mechanisms under
consideration.
Step (f) needs further discussion. It may be required that the wall structural strength should be
larger than the internal forces evaluated with the pseudo-static method using the critical
acceleration ac. This should be considered equivalent to a common practice used for the
capacity design of structural members: energy-dissipating elements of mechanisms are
chosen, and other elements are provided with a sufficient reserve strength capacity, to ensure
that the chosen energy dissipating mechanisms are maintained at their full strength throughout
the deformations that may occur. In the case at hand, a natural choice for the energy
dissipating element may be the soil interacting with the retaining wall, also considering that in
the initial conditions the strength of significant soil volumes located in the vicinity of the wall
is already fully mobilised.
However, a pseudo-static calculation of the bending moments in the walls with a = ac
typically assumes a linear distribution of the contact stresses, while the actual distribution of
the contact stresses may deviate from this simple distribution; this is visible in Figure 1, and is
further substantiated by Figure 5, that shows the instantaneous distribution of σh computed for
the cantilevered wall of Figure 4(a) during the development of a plastic mechanism
(Soccodato & Callisto 2009), together with the corresponding bending moments and
296 L. Callisto, F.M. Soccodato

σ h (kPa) M (kNm/m) u (m)

200 100 0 -100 -200 300 200 100 0 -0.01 0.01 0.03 0.05
0

t = 4.30 s
t = 4.41 s
t = 4.46 s
2
t = 4.55 s
ac - limit equilibrium
z (m)

Kp = 6.162

Figure 5. Contact stresses, bending moments and horizontal displacements of a cantilevered embedded
retaining wall at several instants during a severe seismic event, compared with results of pseudo-static
calculation in critical condition (Soccodato & Callisto 2009).

horizontal displacements. The thick lines in Figure 5 show the limit equilibrium computations
with a = ac. It can be seen that the actual instantaneous distribution of the contact stresses can
be somewhat different from the simplified distribution, and this makes the maximum bending
moments Mmax larger than M(ac).
For the same cantilevered retaining wall in a coarse-grained soil, Callisto & Soccodato (2009)
showed that the ratio of Mmax to M(ac) increases with the stiffness of the wall, but is bounded
by a maximum of about 1.6. Therefore, this figure could be adopted as a multiplier of M(ac) to
account for the difference between the simplified and the actual distribution of contact
stresses, at least for the particular case examined. In order to protect the retaining wall from
bending yielding, a further over-strength factor might be required.
Soccodato & Callisto (2009) explored a different approach, in which no multipliers or over-
strength factors were used, and the walls were given a bending strength My about equal to
M(ac), allowing both the soil and the retaining walls to undergo plastic yielding during the
earthquake. Hence, in this approach yielding of the wall is called to compensate for the
inaccuracies of the assumed distribution of the contact stresses. The Authors carried out a
series of dynamic numerical analyses, in which two different seismic records were applied to
pairs of cantilevered retaining walls with H = 4 m and three different embedment depths. The
mechanical behaviour of the walls was described with a linearly elastic-perfectly plastic
moment-curvature relationship, in which, for the three different walls, the curvature at
yielding ψy was about constant. The soil behaviour was described by a non-linear hysteretic
constitutive model coupled with a Mohr-Coulomb failure criterion (Callisto & Soccodato
2007). Figure 6, taken from Soccodato & Callisto (2009), shows for the two seismic inputs
the maximum displacements and bending moments plotted as a function of the overall wall
length L. Results are also concisely reported in Table 1.
It can be seen from Figure 6 that the decrease of the displacements with L is qualitatively
similar to the results shown in Figure 4(b). The elastic-plastic walls undergo displacements uy
that are larger than those computed for the elastic walls (uel), but the difference becomes very
small with increasing L (see Table 1). Bending moments increase with L, as it was expected
Performance-Based Design of Embedded Retaining Walls Subjected to Seismic Loading 297

Table 1. Main results obtained from the numerical analysis presented by Soccodato & Callisto (2009).

L (m) uy/uel Mfin/My ψmax/ψy

7 1.59 0.96 43
8 1.39 0.87 20
9 1.13 0.71 2.7

0.6 0.6
Tolmezzo Assisi
left wall - elastic
left wall - yielding

right wall - elastic

0.4 0.4 right wall - yielding
ur (m)

0.2 0.2

(a) (b)
0 0
7.0 8.0 9.0 7.0 8.0 9.0

400 400
Tolmezzo Assisi

300 My 300 My
M (kNm/m)

200 200

left wall - elastic - max

left wall - elastic - final
100 100 left wall - yielding - final

right wall - elastic - max

right wall - elastic - final
(c) (d) right wall - yielding - final
0 0
7.0 8.0 9.0 7.0 8.0 9.0
L (m) L (m)
Figure 6. Permanent displacements and maximum bending moments in cantilevered embedded retaining
walls of different lengths (Soccodato & Callisto 2009).

(see Figure 4(b)). The final values Mfin to range from 70 % (L = 9 m) to more than 95 % (L =
7 m) of My.
The mobilised strength in the soil can be quantified by the ratio τ/τlim of the maximum
tangential stress acting at a point to the corresponding available strength. Figure 7 shows, for
the cantilevered wall, the contours of the mobilised strength computed by Callisto &
Soccodato (2007) before (a) and after (b) a severe earthquake. Before the earthquake, a
significant mobilisation of the shear strength is obtained behind the wall, where the soil is in
an active limit state, and right below the bottom of the excavation, where the soil is in a
298 L. Callisto, F.M. Soccodato

passive limit state. At the end of the earthquake the stress state in most of the soil interacting
with the walls is quite far from a plastic limit state and the corresponding distribution of the
contact stresses produces the large post-seismic bending moments of Figure 6. However,
Callisto & Soccodato (2009) showed that a further small excavation causes the soil behind the
wall to reach once more a limit active state and therefore produces a decrease of the contact
stresses, with an ensuing reduction of the bending moments. Hence, the relatively high
internal forces that remain locked into the wall after the earthquake are not deemed capable to
endanger the overall safety of the system.
In order to judge the performance of a wall that undergoes plastic yielding during the seismic
event, it is of interest to quantify the curvature ductility demand, that is the ductility required
for the wall sections to undergo the computed plastic curvatures without a significant strength
degradation. Table 1 reports, for the cases considered, the values of the curvature ductility
factor, defined as the ratio of the maximum curvature ψmax to the curvature at yield ψy. It
appears that for the walls with L = 8 and 9 m the curvature ductility demand can easily be
satisfied by a properly detailed r.c. section.
0
1 1
0.95 0.95
0.90 0.90
0.85 0.85
0.80 0.80
0.75 0.75

-4

-8

(a) (b)
-12
10 12 14 16 18 20 22 24 10 12 14 16 18 20 22 24

Figure 7. Contour plots of the mobilised strength before (a) and after (b) a seismic event (Callisto &
Soccodato 2007).

4 DEVELOPMENT OF TECHNICAL CODES

The Eurocode 8 (EN 1998-5) prescribes that the pseudo-static design of flexible retaining
walls be carried out using the maximum expected acceleration amax. From the above
discussion, it follows that this prescription is equivalent to the requirement that even under a
severe earthquake no permanent displacements should be tolerated.
On the contrary, the Italian Construction Code provides a relationship, obtained directly from
Figure 3, between the maximum permanent displacement (that is, the seismic performance)
and a factor β = ah/amax < 1. The factor β multiplies the expected maximum acceleration amax
to yiled the horizontal acceleration ah that the code requires for use in a pseudo-static
calculation. For a cantilevered or singly restrained wall, ah coincides with the critical
acceleration ac if the wall is designed on the basis of a pseudo-static limit equilibrium
calculation performed with unit partial or global safety factors. In this case, if the relationship
of Figure 3 holds true, and if the maximum acceleration is evaluated accurately, the wall will
not move more than required, and the maximum bending moments will be evaluated as M(ac),
consistently with the approach discussed in the previous section; this value might be
amplified by an over-strength factor to protect the structural elements from yielding.
Performance-Based Design of Embedded Retaining Walls Subjected to Seismic Loading 299

It must be stressed that the seismic internal forces in the retaining wall are proportional to the
critical acceleration, and therefore to the soil strength available for the specific plastic
mechanism considered. If for any reason this strength is larger than anticipated, the retaining
wall will be subjected to larger internal forces. For instance, if the strength parameters of the
soil are underestimated, the computation does not err on the safe side: the pseudo-static
analysis will lead to an underestimation of the critical acceleration and therefore the
maximum bending moments evaluated as M(ac) will be too low.
Currently, the Italian Construction Code requires that the pseudo-static design of a retaining
wall be carried out applying partial coefficients to the strength properties of the soil. This
leads to the design of walls with ac > ah; hence, the embedment depths are larger than those
strictly required for the desired performance, and the seismic displacements are likely to be
smaller than expected. But these same walls will have to sustain bending moments
proportional to ac, larger than those computed with ah < ac. A possible, concise way to
overcome this difficulty would be to require unit partial coefficients for the seismic pseudo-
static calculations. This would be consistent with the real nature of the seismic pseudo-static
calculations, than only apparently deal with the safety with respect to a collapse mechanism,
but rather serve the purpose to asses the actual seismic performance of a structure.
However, since the length of a retaining wall may be dictated by requirements other than its
seismic performance, it may well be that the actual length of a wall is larger than what would
be strictly needed for the desired seismic performance: a technical code should therefore
explicitly prescribe that the internal forces in the wall be always computed in the hypothesis
that the available soil strength is fully mobilised.
The above concepts are yet to be extended to the seismic design of retaining walls with
multiple restrains. In these cases, mobilisation of soil strength may not be sufficient for the
development of a plastic mechanism, and several different mechanisms may be possible,
depending of the choice of the energy-dissipating elements. For this wall types, significant
efforts are still needed in order to identify the more convenient plastic mechanisms and to
compute the corresponding values for the critical acceleration. In spite of this, it is believed
that the concepts exposed herein still hold their validity: relationships like the one shown in
Figure 3 may be used to evaluate the seismic displacements, while the maximum internal
forces must be calculated considering the full strength of the energy-dissipating elements of
the chosen plastic mechanism.

5 ACKNOLEDGEMENTS

The work presented in this paper is part of the ReLUIS research project, funded by the Italian
Department of Civil Protection. The contribution of the Italian Ministry of University,
through the research project Effetto delle azioni sismiche sulla stabilità dei pendii e sul
comportamento di scavi e gallerie (PRIN 2007) is also acknowledged.

6 REFERENCES

Blum, H. (1931). Einspannungsverhältnisse bei Bohlkwerken. Wil. Ernst und Sohn, Berlin.
Callisto L. & Soccodato F.M. (2009). “Seismic design of flexible cantilevered retaining walls”. Paper
submitted for publication in the Journal of Geotechnical and Geoenvironmental Engineering.
Callisto L. e Aversa S. (2008). “Dimensionamento di opere di sostegno soggette ad azioni sismiche”. In:
Opere geotecniche in condizioni sismiche, MIR 2008. Pàtron, Bologna, 273-308.
300 L. Callisto, F.M. Soccodato

Callisto L., Soccodato F.M. & Conti R. (2008). “Analysis of the seismic behaviour of propped retaining
structures”. Proc. Geotechnical Earthquake Engineering and Soil Dynamics IV Conference,
Sacramento.
Callisto, L. and Soccodato, F.M. (2007). “Seismic analysis of an embedded retaining structure in coarse-
grained soils.” Proc. 4th Int. Conf. on Earthquake Gotechnical Engineering. Thessaloniki.
Decreto Ministeriale 14.1.2008 del Ministero delle Infrastrutture. “Nuove norme tecniche per le
costruzion”i. S.O. n. 30 alla G.U. del 4.2.2008, n. 29.
EN 1998-5. “Eurocode 8: Design of structures for earthquake resistance – Part 5: Foundations,
retaining structures and geotechnical aspects”. CEN European Committee for Standardization,
Bruxelles, Belgium.
Lancellotta, R. (2007). “Lower-bound approach for seismic passive earth resistance.” Géotechnique
57(3), 319-321.
Neelakantan, G., Budhu, M., and Richards, R. (1992). “Balanced seismic design of anchored retaining
walls.” Journal of Geotechnical Engineering, 118(6), 873-888.
Newmark, N.M. (1965). “Effects of earthquakes on dams and embankments.” Fifth Rankine lecture.
Géotechnique 15(2), 139–193.
PIANC (2001). Seismic Design Guidelines for Port Structures. Working Group n. 34 of the Maritime
Navigation Commission, International navigation association, Balkema, Rotterdam, pp. 474
Rampello S. e Callisto L. (2008). “Stabilità dei pendii in condizioni sismiche”. In: Opere geotecniche in
condizioni sismiche, MIR 2008. Pàtron, Bologna, 241-271.
Richards, R., and Elms, D.G. (1979). “Seismic behavior of gravity retaining walls.” Journal of the
Geotechnical Engineering Division, ASCE, 105(GT4), pp. 449-464.
Richards, R., and Elms, D.G. 1992. Seismic passive resistance of tied-back walls. Journal of
Geotechnical Engineering, 118, No. 7, 996-1011.
Scasserra, G., Lanzo, G., Stewart, J.P., and D'Elia B. (2008). “SISMA (Site Of Italian Strong-Motion
Accelerograms): a web-database of ground motion recordings for engineering applications.”
Proceedings of 2008 Seismic Engineering International Conference commemorating the 1908
Messina and Reggio Calabria Earthquake July 8-11, 2008 – Messina and Reggio Calabria.
Soccodato, F.M., and Callisto, L., (2009). “An approach to the seismic design of embedded retaining
walls based on the results of dynamic numerical analysis.” Proceedings of the International
Conference on Performance-Based Design in Earthquake Geotechnical Engineering- IS Tokio 2009.
In print.
Viggiani, G.M.B., and Conti, R., (2008). Personal communication.