Ultimate limit state design of retaining walls with numerical methods

G. Katsigiannis
Arup Geotechnics and University College London, London, UK
P. Ferreira & R. Fuentes
University College London, London, UK

ABSTRACT: This paper presents a comparison between the results obtained from simple empirical methods,
and full FE using Eurocode 7 (EC7) for the design of supported excavations. It shows the influence of the
current different EC7 implementation strategies for the use of partial material factors when using Design Ap-
proach 1, both for Combination 1, where the action effects are factored at the end of the calculations, and also
for Combination 2, where the soil strength parameters need to be factored and procedures are less clear. The
influence of different factors in the results is highlighted for structural forces in cantilever and supported
walls with increasing excavation depth and number of props. It shows the differences in prop loads obtained
from FEM and empirical approaches and illustrates the difficulties that designers face when using EC 7.

1 INTRODUCTION SLS. For ULS the designers used to simply factor

Nowadays, the increasing demand for underground the structural forces (bending moments, prop forces
infrastructure and basements in urban environments etc.) at the end of the calculations. However, the re-
highlights the need for achieving more economic quirements are now different in light of EC7. There
and safe design of retaining walls. The requirement are still a number of issues that need further re-
of limiting ground movements and movements of search before the ULS design can be routinely per-
adjacent structures and utilities (i.e. serviceability formed with FEM. Simpson (2012) and Simpson &
limit states) becomes an important factor; however, Junaideen (2013) give a good review of all these
there is also the need to ensure that no failure of the challenges associated with the ULS design with
support system occurs. Failures are rare but do occur FEM.
(Twine & Roscoe, 1999).
2.2 Material factoring strategies
Design of retaining walls in compliance with Eu-
rocode 7 Design Approach 1 (DA1), require Ulti- Two main different ways to apply partial factors in
mate Limit State checks for two Combinations or FEM have been suggested in the literature (Simp-
sets of partial factors. The merits of DA1 are dis- son, 2012). In Strategy 1, the partial factors are ap-
cussed in detail by Simpson (2007). Applying Com- plied to soil strength parameters from the beginning
bination 1 (DA1-1) with Finite Element Methods and the analysis is performed with factored values of
(FEM) requires factoring the action effects at the strength (design values in the EC7 terminology).
end of the calculations as the earth pressures cannot Conversely, in Strategy 2, calculations are per-
be factored from the beginning. Combination 2 formed with characteristic values and at critical
(DA1-2) is well suited to FEM as it requires factor- stages the material parameters are factored to their
ing the soil strength parameters which are the input design values to check that no ULS has been
parameter of commonly used constitutive models. reached. A good description of the two strategies has
However, Eurocode 7 gives no guidance on when to been given by Simpson (2012) (see Figure 1).
apply the material partial factors. Current practice While Strategy 1 has the obvious advantage of
suggests two different strategies that, quite often, simplicity, most authors prefer using Strategy 2. The
give different results in terms of design bending reason for this is that applying partial factors from
moments and prop forces (Simpson, 2012). This pa- the beginning results in unreasonable or optimistic
per compares different factoring methods, for canti- effects later. Moreover, in many cases, the designer
lever and multi prop walls, while the influence of can perform SLS and ULS checks with only one
different factors on the resulting discrepancies is in- analysis, by employing Strategy 2. The caveat being
vestigated. that this only applies when no changes in geometry,
water levels or surcharge loads are needed. Finally,
2 DESIGN METHODOLOGY another advantage of Strategy 2 is that it facilitates
the use of the automatic c-phi reduction technique
2.1 Finite Element Methods and Eurocode 7 (Simpson, 2012), which is well established among
FEM have been traditionally used in geotechnical the geotechnical engineering community.
engineering to obtain deformations and check for
 2.3 Automatic c-phi reduction
Many geotechnical engineering software packages
have included a facility called automatic c-φ reduc-
tion. By employing this technique the user can drive
the system to failure, at discrete stages, by reducing
stepwise the material strength properties and obtain
the corresponding material safety factor. If the fac-
tors are greater or lower than those required by the
code, the designer can revise the geometry and per-
form again the analysis (Simpson, 2012). Most
commercial packages use the automatic strength re-
duction with the simple Mohr Coulomb soil model,
while Potts & Zdravkovic (2012) have recently pro-
posed a method that can be applied with both simple
and more advanced soil models.
This is an interesting approach as it enables the
designer to get an estimate of economy at each criti-
cal stage. However, EC7 does not suggest dealing
Figure 1. Strategies for analysis of staged construction (after
with fully mobilised mechanisms but ensuring that
Simpson, 2012) failure is sufficiently unlikely to occur for the given
combinations of factors. The main disadvantage of
A few authors in the past compared the different the stepwise strength reduction is that it gives no
strategies of applying partial factors in staged exca- useful information about the design structural force
vation problems with FEM. Bauduin et al. (2000) values. EC7 requires that the designer should check
found that, factoring the material parameters only at against both GEO and STR Ultimate Limit States.
critical stages (Strategy 2) resulted in slightly higher Hence, it seems that the stepwise strength reduction
design values of the internal forces in structural is most useful when used as an extra analysis to
members for a singly supported sheet pile wall. identify the critical failure mode and the margin of
They also found that DA1-1 was less critical com- safety. However, it should be clear that this is not an
pared to both strategies of DA1-2. The authors also EC7 requirement.
considered a multi-propped diaphragm wall, obtain-
ing different design prop forces and bending mo- 2.4 Numerical and empirical methods for multi-
ments from the two strategies. Simpson and Yazdchi propped retaining wall design
(2003) studied a staged excavation for a multi-
propped wall and concluded that DA1-1 is more For multi-propped walls there are a number of em-
critical when the length of the wall has not been op- pirical graphs that can be used to obtain the prop
timised by carrying out a stability analysis. Expand- forces. Traditional methods (Terzaghi & Peck, 1967
ing on this, Simpson & Driscoll (1998) observed that and Peck, 1969) are simple to use and have been
for specific cases, Strategy 2 can give more onerous widely adopted in practice. They are based on field
results. Schweiger (2005) considered a single strut- measurements of prop loads and provide the design-
ted wall with a staged excavation and obtained quite er with conservative lateral earth pressure distribu-
similar results for both DA1-2 Strategies, where tions.
Strategy 2 resulted in only slightly higher design CIRIA C517 (Twine & Roscoe, 1999), enhancing
values. Repeating the calculations with a surcharge, Terzaghi's work and making it more relevant to UK
the same conclusions were drawn. In all cases DA1- practice, suggests the use of the Distributed Prop
1 gave lower design values. Simpson & Hocombe Load (DPL) method, based on 81 case histories and
(2010) used the case study of the Florence High field measurements of prop loads. Soils are classi-
Speed Station to compare the two strategies, con- fied in 4 classes named A, B, C and D, referring to
cluding that applying factors at discrete stages normally consolidated and slightly overconsolidated
(DA1-2 Strategy 2) mainly governs the design. clays, heavily overconsolidated clays, granular soils
While most authors seem to agree that Strategy 2 and mixed soils respectively. A distinction is also
might give more onerous results, this may not al- made between flexible (e.g. sheet pile) and stiff (e.g.
ways hold and therefore, there is a need for a better diaphragm and bored pile) walls. DPL is not the real
understanding of the resulting discrepancy. lateral stress distribution but gives values of charac-
teristic prop forces unlikely to be exceeded for any
temporary system in a similar excavation (Twine &
Roscoe, 1999). There are also a number of condi-
tions that the designer should check before using the
empirical graphs (geometry, surcharge, sufficient toe Hollow steel props with external diameter 406.4
embedment, etc). mm and width 12.5mm were simulated in all cases
BS8002 (1994) for multi-propped walls simplistical- with EA=3100000 kN/m.
ly recommends the use of Peck's diagrams without Three different geometries were analysed: (1)
mentioning how they should be used for ULS and cantilever with a retained height (H) of 6 m; (2) 4
SLS calculations. prop wall with H equal to 20m and embedment
depth of 6m and; (3) 5 prop wall with an H equal to
24m and embedment depth of 7.5m. The latter two
2.5 Water pressures and limit state design have the same system stiffness (same distance be-
tween prop levels) and therefore, allowed to verify
According to Simpson et al. (2011) the water pres- the impact of the rest of parameters if it is assumed
sures can have significant effect in the geotechnical that systems with similar stiffness present similar
design and they involve high uncertainties. Unex- strains (Long, 2001).
pected variation can drive the system to failure. The parameters investigated were K0 (values of
Hence, the designer should always consider the 1.0, 1.2 and 1.5) and the Eu/cu ratio (500 and 1,000)
worst water pressures that could occur and not just as the two more critical parameters for wall design
rely on the factors of safety. Engineering justice and when using Mohr-Coulomb (Yeow & Feltham,
expertise is essential especially in situations where 2008).
the water pressures might be critical for the design The following construction sequence was mod-
(Simpson et al., 2011). elled with an overdig of 0.5m considered:
BS8002 requires that “the water pressure regime  Stage 0 Initial state conditions
used in the design should be the most onerous that is  Stage 1 Wall installation and 20kPa surcharge
considered to be reasonably possible” (BS8002,
 Stage 2 Excavation to -4m (to –6m for the cantile-
1994).
EC7 (EN1997-1, 2004) suggests that a safety ver wall and end)
margin in terms of water pressures can be applied  Stage 3, Install strut 1
either by modifying the water table level or by fac-  Stage 4, 6 and 8 Excavation of 4m of soil
toring directly the water pressures. As water pres-
 Stage 5, 7 and 9 Install strut
sures cannot be factored with FEM, the designer
should consider the water table level as the most un-  Stage 10 Excavation to -20m (continue to -24m
favourable during the design lifetime of the structure for the 5 props case)
for ULS and as the most unfavourable under normal
conditions for SLS (EN1997-1, 2004).
EC7 Evolution Group 9 (EG9) which is working
on water pressures has recently proposed alternative 4 RESULTS
definitions, based on the concept of probability. Ac- The results of the analysis are presented below in
cording to this recent proposal, the water table for terms of prop loads for each of the walls modelled
SLS shall correspond to a recurrence period at least (Figures 2-7).
equal to the design lifetime of the structure, while
for ULS the design water table shall represent a rare
probability in the design lifetime of the structure. 4.1 Supported wall with 4 struts
The value of the probability may be specified in the
National Annex but a value of 1% is recommended.

3 ANALYSIS DESCRIPTION
The computer software PLAXIS V12.01 was used
for the analysis in its two-dimensional version. A
simple elasto-plastic Mohr-Coulomb constitutive
soil model was used in the simulations.
The soil chosen was typical stiff clay for which
traditional London Clay parameters were used. As
the analysis was performed for the short term an un-
drained soil strength profile equal to 60+8z was used
(depth z in metres and results in kPa). The water ta-
ble is assumed to be the most unfavourable at the
surface. Figure 2. Deriving strut loads from different factoring strategies
(Ko=1.2 and Eu=1000cu case)
4.2 Supported wall with 5 struts 4.3 Reducing the soil stiffness

Eurocode 7 does not suggest factoring soil stiffness.

As stiffness is, in most cases, a parameter of high
uncertainty; current practice suggests carrying out
parametric analyses using upper and lower bound
values. On the contrary, CIRIA C580 (Gaba et al.,
2003) recommends that soil stiffness should be fac-
tored by a factor of 2. As many designers in the UK
refer to CIRIA C580 for guidance on retaining wall
design, the effect of factoring the soil stiffness
should be better understood. The derived prop loads
are now compared for variations both in Ko and
Eu/cu ratio (see Figure 7).
Figure 3. Deriving strut loads from different factoring strategies
(Ko=1 and Eu=1000cu case)

Figure 4. Deriving strut loads from different factoring strategies

Figure 7. Comparing derived strut loads from FEM and CIRIA
(Ko=1.2 and Eu=1000cu case)
for variations in Ko and Eu/cu ratio.

4.4 Discussion

4.4.1 Comparison of different factoring strategies

In all cases under consideration DA1-1 gives the

highest values of prop forces and bending moments.
The soil strength seems not to be critical for design.
The divergence between DA1-2 strategies 1 and 2 of
material factoring becomes more apparent as the ex-
Figure 5. Deriving strut loads from different factoring strategies
cavation depth (and hence the number of excavation
(Ko=1.5 and Eu=1000cu case)
stages and props) increase with DA1-2 strategy 2
giving higher values of prop loads in all simulations
(Figures 2, 3, 4 and 5). Another interesting thing is
that the bending moment distribution (not presented
here) is the same regardless of the number of props
and the excavation depth. Unlike sheet pile walls,
concrete walls are not necessarily reinforced equally
on both sides or uniformly along their depth, there-
fore more than one bending moment might be criti-
cal to the design. Overall, as EC7 requires checks
for both combinations, DA1-1 and DA1-2 strategy 2
govern the design for the material properties consid-
ered.
Figure 6. Comparing derived design prop loads from FEM and A further advantage of using Strategy 2 is that
CIRIA for variations in Ko Ultimate Limit States are extreme and unlikely cas-
es. As such, driving the system to ULS from the be- 5 CONCLUSIONS
ginning, by creating an unrealistic initial stress field While a much broader study is needed, some useful
(Strategy 1), might be confusive. It seems, there- conclusions can be drawn from the work done in this
fore, more appropriate and intuitive for the designer article:
to understand the concept behind the EC7 ULS de-  DA1-2 Strategy 2 gives higher values of prop
sign, by performing the analysis from the character-
loads than Strategy 1 in all these cases.
istic initial stress field and checking for these unlike-
ly states only at discrete critical stages (Strategy 2).  Increasing Ko values will make DA1-2 Strategy 2
On the other hand, it could be attractive to use DA1- to become more onerous, in terms of prop load-
2 strategy 1, as it complies with the code and gives a ing, than DA1-1.
more economic design. Also, it only requires one  DA1-1, in most cases, governs the design of the
sequence of calculations whereas DA1-2 strategy 2 props for the material and geometries considered,
requires an extra set of separate calculations steps especially for lower degrees of overconsolidation
for each construction stage. (exemplified by the lower value of K0) and where
the strength of the soil is not critically important
4.4.2 Numerical vs empirical methods for the design.
Empirical and FEM calculations provide different  The authors favour factoring at discrete stages
prop force values for the geometries and materials (i.e. DA1-2 Strategy 2) for two main reasons: (1)
considered. Is the apparent discrepancy because the it seems to be more critical in most cases and (2)
numerical methods are inadequate, or because of the it seems a more intuitive way of using the code
situations assumed in design (surcharge, excavation than Strategy 1 where it may lead to confusion
level, ground water levels, length of open excavation but also because it can be easily be applied in
etc.) do not always occur in practice? Most of the conjunction with SLS checks.
analysis done to date is 2D, i.e. it assumes that a  The choice of Strategy, for deriving wall bending
complete level of excavation is opened before any moments, is less important than for strut forces:
props are inserted at that level. Usually this does not this is because walls can usually display some
occur in practice. So if 2D analysis over-predicts
plasticity and redistribute bending moment if it is
strut forces, would it also be true of 3D?
How good is the C517 assessment that the force high at some point, whereas DA1-2 strategy 2 is
in the top strut will be equal to that in lower struts? attractive for deriving strut forces because struts
Again, sequence of excavation, as discussed before, can be brittle, and DA1-2 strategy 1 might under-
would be important. If you do a very non-2D exca- estimate them.
vation at an early stage, you can get the top strut in
with very little wall displacement, and it then picks  Empirical and FEM calculations provide different
up a lot of load. prop force values: Reasonable variations in Ko
This leads to the following questions relevant to and soil stiffness can significantly affect design
EC7: prop loads derived from FEM.
 Is it appropriate to use results of FEM to de-
rive design strut loads?  In all these cases, FEM results in lower values of
 Why (if it is true) are computed values strut forces at the top of the wall, increasing with
(sometimes/usually/always) greater than excavation depth. Note that there is not much dif-
measured values? Can/should anything be ference in the total force supporting the wall in
done to amend this? (For example, Roscoe the two cases – about 8.5%.
and Twine (2010) found that they had to re-
duce strut stiffness by about 50% to recover
the measured values on one project. Is there
any justification for this?) 6 ACKNOWLEDGEMENTS
 What sort of soil models must be used to get The authors gratefully acknowledge the support
results appropriate for design? of the project partners, EPSRC, Arup and European
 Are there any dangers of under-estimating Institute, University College of London.
strut loads, which could be dangerous? A
particular issue for struts is that there behav-
iour is quite often brittle, so it could be dan- 7 REFERENCES
gerous to under-design them.
BS EN 1997-1, Eurocode 7 – Geotechnical design,
Part 1 – general rules. (2004). London: British
Standards Institution.
Bauduin, C., DeVos, M., & Simpson, B. (2000). tures. Proc ER2010, ASCE Earth Retention
Some Considerations on the Use of Finite Ele- Conference 3, Seattle, Aug 2010.
ment Methods in Ultimate Limit State Design.
LSD 2000 International Workshop on Limit Simpson, B., & Junaideen, S. M. (2013). Use of
State Design in Geotechnical Engineering Numerical Analysis with Eurocode 7. 18th
Melbourne, Australia. Southeast Asian Geotechnical & Inaugural
AGSSEA Conference, Singapore Leung, Goh &
Code of practice for earth retaining structures, BS Shen (eds).
8002. (1994). London: British Standards Insti-
tution. Simpson, B., Vogt, N., & Van Seters, A. J. (2011).
Geotechnical safety in relation to water pres-
Gaba, A. R., Simpson, B., Powrie, W., & Beadman, sures. Proc 3rd Int Symp on Geotechnical Safe-
D. R. (2003). Embedded retaining walls - guid- ty and Risk, Munich.
ance for economic design. CIRIA Report C580,
(p. 390). London: CIRIA. Simpson, B., & Yazdchi, M. (2003). Use of finite
element methods in ultimate limit state design.
Long, M. (2001). ”Database for Retaining Wall and LSD 2003: International workshop in Limit
Ground Movements due to Deep Excavations.” State Design in Geotechnical Engineering
J. Geotech. Geoenviron. Eng., 127(3), 203–224. Practice, Cambridge, Massachusetts.

Peck, R. B. (1969). Deep excavations and tunnelling Terzaghi, K., & Peck, R. B. (1967). Soil mechanics
in soft ground. 7th Int. Conf. SMFE. Mexico in engineering practice (2nd edition). New
City. State-of-the-art vol., 225–290. York: J. Wiley and Sons.

Potts, D.M, & Zdravkovic, L. (2012). Accounting Twine, D., & Roscoe, H. (1999). Temporary prop-
for partial material factors in numerical analy- ping of deep excavations-guidance on design,
sis,. Géotechnique, 62, 1053–1065. CIRIA C517. London: CIRIA.
Roscoe, H., & Twine, D. (2010). Design and per-
formance of retaining walls. Proceedings of the Yeow, H., & Feltham, I. (2008). Case histories back
ICE - Geotechnical Engineering, Volume 163, analyses for the application of the Observational
Issue 5. Method under Eurocodes for the SCOUT
project. 6th International Conference on Case
Schweiger, H. F. (2005). Application of FEM to Histories in Geotechnical Engineering.
ULS design (Eurocodes) in surface and near
surface geotechnical structures. Proc. 11th Int.
Conf. Computer Methods and Advances in Ge-
omechanics, Patron Editore, Bologna, 4, 419–
430.

Simpson, B. (2007). Approaches to ULS design -

The merits of Design Approach 1 in Eurocode
7. ISGSR2007 First International Symposium
on Geotechnical Safety & Risk pp 527-538.
Shanghai Tongji University, China.

Simpson, B. (2012). Eurocode 7 – fundamental is-

sues and some implications for users, Keynote
Lecture. Proc Nordic Geotechnical Meeting
2012, DGF Bulletin 27.

Simpson, B., & Driscoll, R. (1998). Eurocode 7 - a

commentary. Watford, UK: Construction Re-
search Communications Ltd.

Simpson, B., & Hocombe, T. (2010). Implications of

modern design codes for earth retaining struc-