INTERNATIONAL SOCIETY FOR

SOIL MECHANICS AND

GEOTECHNICAL ENGINEERING

This paper was downloaded from the Online Library of

the International Society for Soil Mechanics and
Geotechnical Engineering (ISSMGE). The library is
available here:

https://www.issmge.org/publications/online-library

This is an open-access database that archives thousands

of papers published under the Auspices of the ISSMGE and
maintained by the Innovation and Development
Committee of ISSMGE.
FINITE ELEMENTS AND WINKLER MODEL
APPLIED TO RETAINING WALLS DESIGN
Lourenço D., Schnaid F. & Rocha M.M.
Department of Civil Engineering – Federal University of Rio Grande do Sul,
Porto Alegre, Rio Grande do Sul, Brazil.

ABSTRACT
This work presents a numerical study of the structural behavior presented by concrete retaining walls in frictional soils,
as well as results from a comparison among two computer programs based on complementary approaches: Finite
Element Method (FEM) and the Winkler Spring Model (WSM). The simulation using a comercial FEM program is the
reference for estimating maximum displacements and bending moments, which are used to validate the analysis with the
WSM. The works also aim at calibrating stiffness coefficients for the Winkler springs, enabling this approach to be
adopted as a simple and expeditious way to estimate moments and displacements of retaining structures.

RÉSUMÉ
Ce travail présente une étude décomposé en numéros du comportement structurel présenté par les rideaux de béton
face à des sols de frottement, aussi que, une comparaison des résultats entre deux programmes informatiques utilisés
pour chaque méthode basées en deux approches complémentaires: La méthode des Éléments Finis (MEF) et le modèle
des Ressorts de Winkler (MRW). La simulation à l'aide d’un programme FEM est la référence pour l'estimation des
déplacements et des moments de flexion maximales, ces valeurs sont utilisées dans la validation de l'analyse en utilisant
le MRW. Ce projet envisage aussi calibrer les coefficients de rigidité du modèle Winkler utilisés sur MATLAB, permettant
à cette approche d’être utilisé comme une forme simple et rapide d’estimer les moments et des déplacements de
structures de soutènement.

1 INTRODUCTION Simulations can use different constitutive models, from a

simple linear stress-strain relationships to any modelled
Conventional approaches to geotechnical design of rheological phenomena.
retaining structures consider soil stress fields that are In the present paper, Plaxis was used to evaluate the
analytically calculated and provide the basis for a limit stress distribution acting against a retaining wall illustrated
state equilibrium analysis. More sophisticated approaches in Figure 1(a), considering a Mohr-Coulomb model for
are nowadays available that are resourced with numerical sands with a dry weight volume of 16KN/m3. A total of 24
techniques, like the Finite Elements Method (FEM) to cases were simulated in Plaxis, twelve to a bending
assess the soil stresses loading retaining structures, as stiffness wall of 1.6×105kN/m2/m corresponding to a 0.4m
well as the stresses in the structural elements. These thickness, and twelve for a stiffness of 5.4×105kNm2/m,
methods are suitable for geometrically well defined corresponding to a wall thickness of 0.6m. In a parametric
problems, but sometimes lack practicality in a pre-desing approach, the friction angle (Φ) ranged from 30˚ to 42˚,
stage, where elements re-meshing may be too often with soil Young's Modulus (Es) of 4MPa, 20MPa and
required. Optionally, the Winkler Spring Model (WSM) 40MPa. The concrete Young's Modulus (Ec) was kept
may be accurate enough, and hence useful, for initial constant as 30GPa.
estimates of wall moments and displacements, even
though the entire stress distribution behind the wall may 2.2 WSM program
not be represented.
This paper also seeks to evaluating the soil pressures The WSM approach was computationally implemented as
and the wall structural response by means of a computer a Matlab script. The calculation process is iterative and
program developed on the basis of the WSM. The springs the soil is represented by one-dimensional elastic-plastic
are calibrated to enforce compatibility between WSM and spring elements, as shown in Figure 1(b). The yielding
the FEM, and some practical rules are derived to ensure a point of the elastic- perfectly plastic springs are related
consistent definition of their stiffnesses. The wall with the active and passive limit soil pressures. The wall is
geometry adopted for these calibrations is depicted in represented by linear beam elements, with two degress of
Figure 1(a), which dimensions are varied in a parametrical freedom per node, horizontal displacement and in-plane
study. rotation. Deformations by axial compression are not taken
into consideration. The discretization length is 5cm,
2 METHODOLOGIES ensuring an adequate accuracy level for the problem a
floating concrete wall behavior that serves as a restraint
2.1 FEM with Plaxis® for excavations.
The calculations for the wall deformation process are
The comercial program Plaxis® is based on the finite iteractive and summarized below.
element method and is a reliable tool for assessing
deformations and stability in geotechnical design.
Figure 1. Represention of a retaining wall an excavation:
a) real situation, b) Winkler model

Firstly, the soil pressures along the wall are estimated, as

well as the resulting horizontal forces and the active and Figure 2. Example of Plaxis FEM output
passive limits. Pore pressures are not considered. The
soil stiffness, represented by the Winkler springs, is added
to the diagonal of the wall stiffness matrix and the
horizontal loads added to corresponding degree of
freedom in the load vector. The system is solved and the
wall displacements are calculated. Once these
displacements are known, the elastic-plastic constitutive
relations of the springs are checked out. If the plastic limit
is reached, the spring stiffness is decreased in order to
keep the soill pressure within the active or passive limits.
Hence the system in solved again, iteractivelly, until
convergence is reached. If the convergence does not
occur, equilibrium is not possible and the system is
unstable.
The main input parameters for a parametrical study
are the wall length and thickness, the soil parameters, the
elevation of the bottom of each soil layer, excavation
depth, specific weight, friction angle, cohesion intercept,
soil elastic stiffness per area, KEs, and the dimensionless
coefficients KwA (Winkler active coefficient) and KwP
(Winkler passive coefficient).

3 RESULTS
Figure 3. Matlab script for WSM output
Horizontal displacements at the top of the retaining wall
obtained with FEM are analysed in conjunction with
maximum bending moments. Calculated values are hence
compared with those estimated by the WSM.
Plaxis results from the parametrical analysis are
presented as diagrams describing axial forces, shear
forces and bending moments. An example of Plaxis
output for the case of sand friction angle of 30˚, soil
Young Modulus 4GPa, 5m excavation, and wall thickness
40cm, is presented in Figure 2. The equivalent results
from the Matlab script are presented in Figure 3.
Having Plaxis as reference, horizontal displacements
and maximum bending moments for the top of the
retaining wall are compared with those from Matlab script
such that at each run the parameters KwA, KwP and KEs are
adjusted to enforce compatibility between both methods
(FEM and WSM). Physically, the coefficients KwA, KwP for
the Winkler model are the same as the active and passive
coefficients for the Mohr-Coulomb model, because with
these two factors the program calculates acting pressures Figure 4. Soil stiffness decay during iteration, represented
on the retaining structure. along the concrete wall, passive and active sides
 Figure 4 shows the soil stiffness evolution and its [3]
variation with depth. The fast convergence of the
iteractive process can be observed by the dotted lines With this set of parameters, further cases were
reaching a well defined limit at both passive and active analyzed in Matlab, allowing comparisons for sand with
sides of the wall. The convergence process may also be different friction angles to predict horizontal displacements
observed in Figure 3, in the dotter lines showing the wall and bending moments.
displacement evolution as the soil stiffness is decreased
by plastification. Usually, it has been observed that less
than 20 iteractions are requided for convergence, and the
computational time is really not a matter of concern.

4 INTERPRETATION OF THE RESULTS

Interpretation of the results is focused on the observed

relationship between the limit state of active and passive
pressures calculated in FEM using a Mohr-Coulomb
model and the coefficients adopted in the WSM.
From the analysis, it is possible to observe that the
Winkler active coefficient (KwA) depends on soil friction
angle, as the Rankine active pressure coefficient (KA)
does. Figure 5 shows that KwA also depends on the ratio
between excavation depth (Hesc) and total wall length
(Htotal). It is once again emphasized that KwA was obtained Figure 5. KwA and KA coefficients
after fitting Matlab to Plaxis data.
The Rankine passive pressure coefficient is shown to
depend solely on the friction angle. The Winkler passive
coefficient seems to depend only on the ratio between the
excavation depth and the total wall length, as shown in
Figure 6.
Database from Figure 5 and Matlab data output was
used to establish a set of equations to guide the choice of
coefficients in the Winker model. Coefficient KwA was
expressed as a function of the friction angle and the “Hesc
/ Htotal“ ratio with the equation:

[1]

Figure 7 makes clear that, for different sands and Figure 6. KwP and KP coefficients
different excavation levels, the ratio “KwA (adjusted) /
KwA(equation)” exhibits little variation and is close to
unity.
Similarly, an expression for the Winkler's passive
coefficient, KwP, was proposed from the observation of
Figure 6 as:

[2]

Figure 8 also indicates that the relationship

“KwP(adjusted) / KwP(equation)” is close to unity, giving a
perfect setting for the Winkler passive coefficient, which is
not dependent on the friction angle.
From the sole observation of fitted data, it was found
that KEs is of the same order of magnitude as the Young's
Figure 7. KwA(adjusted) and by KwA(equation) coefficients
soil modulus (Es) and is affected by the excavation depth.
As a result, the following equation was proposed:
Figure 8. KwP(adjusted) and KwP(equation) coefficients

Figure 10. Bending moments calculated for all excavation

Figures 9 and 10 show a sound aggrement between phases with the two wall stiffnesses and friction angles of
both approaches, corroborating the suggested equations 30 ˚, 34 ˚, 38 ˚ and 42 ˚
for parameters KwA, KwP, e KEs. From this comparative
analysis, it is possible to suggest that results expressed in
terms of horizontal displacements and bending moments 5 CONCLUSIONS
are similar in WSM and FEM analysis, in spite of the
simplifications adopted in the former. The lines of linear Geotechnical and structural engineers often adopt
approximation of each graph gives R2 in the range different approaches to the design of retaining structures.
between 0.987 and 0.998, confirming the accuracy of Whereas the Finite Element Method became a routine in
comparison between the two methods. geotechnical practice, mainly due to the complexity of soil
stress fields, from the structural engineer point of view the
Winkler Spring Model remains an appealing alternative to
predict displacements, shear and bending moments.
In this context, the present paper attempts to evaluate
the compatibility between these two design approaches.
Throughout a set of calibrations it has been demonstrated
that a correct choice of Winkler coefficients might produce
results which are in agreement to finite element
calculations, provided that soil shear strength and
stiffness are properly taken into account.

REFERENCES

Alonso, U.R. 2003. Dimensionamento de Fundações

Profundas. 3rd ed., Edgard Bluche LTDA, Brasil, 66-
70.
Atkinson, J. H.; Coop, M.R.; Stallebrass, S.E.; Viggiani, G.
Figure 9. Displacements calculated for all excavation 1990. Measurement of Stiffness of Soils and Weak
phases with the two wall stiffnesses and friction angles of Rocks in Laboratory Tests, In Proceedings of the 25th
30 ˚, 34 ˚, 38 ˚ and 42 ˚ Annual Conference of the Engineering Geology
Group, Leeds, U.K. British Geology Society, 21-27.
Atkinson, J.H. and Sällfors, G. 1991. Experimental
Determination of Stress-Strain-Time Characteristics in
Laboratory na in-situ Tests, General report, In
Proceedings of the 10th European Conference on Soil
Mechanics and Foudation Engineering, Florence, Italy,
3: 915-956.
Bolton, M.D. 1986. The strength and dilatancy of sands,
Geotechnique, 36(1): 65 - 78.
Bowles, J.E. 1998. Foundation Analysis and Design, 4th
ed., McGraw Hill, Peoria, Illinois, USA, 400-410.
Burland, J.B.; Booms, B.B.; De Mello, V.F. 1979.
Behaviour of foundation and structures, State of the
Art Report. Proc. 9th Intl Conf. on Soil Mech. and
Found. Eng., II, Tokyo, Japan, 495-546.
Clayton, C.R.I.; Milititsky, J.; Woods, R.I. 1993. Earth
Pressure and Earth-Retaining Structures, 2nd edition,
Chapman & Hall, London, UK, 272-288.
Fernandes, M.M. 1995. Mecânica dos Solos, Faculdade
de Engenharia da Universidade do Porto, Portugal,
2:7.6-7.17.
Lim, Y. and Briaud, J.L. 1996. Three dimensional nom
linear finite element analysis of tieback walls and soil
nailed walls under piled bridge abutment, Rep. To the
Federal Highway Administration and the Texas
Departament of Transportation, Departement of Civil
Engineering, Texas A & M University, College Station,
Texas, USA.
NBR-6118 – Projeto e Execução de Obras de Concreto
Armado. NBR. Associação Brasileira de Normas
Técnicas. Brasil
Ng, C.W.W.; Lings, M.L.; SImpson, B.; Nash, D.F.T. 1995
An Approximate Analysis of the Three-Dimensional
Effects of a Diaphragm Wall Installation,
Geotechnique, 45(3): 497-507.
Reffatti, M. 2002. Análise Numérica de uma Escavação
de Grande Porte em Porto Alegre/RS. Porto Alegre,
Dissertação de Mestrado em Engenharia-PPGEC,
Universidade Federal do Rio Grande do Sul, Porto
Alegre, RS, Brasil.
Rocha, M.M. 2009. Um Modelo Elastoplástico para o
Cálculo Prático de Paredes Diafragmas, Porto Alegre.
Documento não publicado-PPGEC, Universidade
Federal do Rio Grande do Sul, Porto Alegre.
Terzaghi, K. 1955. Evaluation of Coefficients of Subgrade
Reaction, Géotechnique 5(4): 297-326.