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.

The paper was published in the proceedings of the

20th International Conference on Soil Mechanics and
Geotechnical Engineering and was edited by Mizanur
Rahman and Mark Jaksa. The conference was held from
May 1st to May 5th 2022 in Sydney, Australia.
 Proceedings of the 20th International Conference on Soil Mechanics and Geotechnical Engineering– Rahman and Jaksa (Eds)
© 2022 Australian Geomechanics Society, Sydney, Australia, ISBN 978-0-9946261-4-1

The effect of the variation of the modulus of subgrade reaction on the design of
large shallow foundations
L'effet de la variation du module de réaction sur l’analyse des larges fondations superficielles

Grace Abou-Jaoude & Hassan Alzoaby

Civil Engineering Department, Lebanese American University, Lebanon, [email protected]

ABSTRACT: The design of shallow foundations represents one of the most common applications in soil-structure interaction. In
current design practices, the structure element (foundation) is modeled as either a rigid or a flexible plate, and the underlying founding
terrain is generally modeled by a coefficient of subgrade reaction (ks) introduced by Winkler in 1867. Designers commonly tend to
consider a uniform coefficient of subgrade reaction under shallow foundations, irrespective of their size or geometry. In 1995, the
American Concrete Institute (ACI) published a State-of-the-Art Review on Design and Performance of Mat Foundations (Special
Publication SP-152) in which various aspects of foundation modeling are addressed, including the coefficient of subgrade reaction.
An iterative procedure, known as the Discrete Area Method, was recommended to promote the consideration of variable k s values
along the mat foundation. This paper uses a new tool (Estephan et al. 2021) that automates the iterative exchange of the mod
ulus of subgrade reaction and the loading stresses between the geotechnical and structural engineers, respectively, in the
design of foundations. The aim is to expose the importance of adopting the exact and complete spatial variability of the
modulus of subgrade reaction in the semi-coupled simulations of soil/mat foundation interaction. The effect of the
variability of ks along the same raft on steel reinforcement design is investigated through a parametric study. The use of a variable
coefficient across the mat foundation led to a significant decrease in the steel reinforcement. The results show that the assumption of
a constant value of ks below the foundation has negative consequences on the accuracy of the design.

RÉSUMÉ : L’étude des fondations superficielles est l'une des applications les plus courantes dans l'interaction sol-structure. En pratique,
l’élément structurel (fondation) est construit sous la forme d’une plaque rigide ou flexible, et le sol est généralement conçu par un
coefficient de réaction (ks) introduit par Winkler en 1867. Normallement un coefficient uniforme est utilisé pour les fondations
superficielles, sans tenir compte de leur taille ou de leur géométrie. Cependant, cette hypothèse a des conséquences indésirables sur
l'exactitude des résultats. En 1995, l’American Concrete Institute (ACI) a publié une étude sur la Conception et la Performance des
Radiers Généraux (Special Publication SP-152) dans laquelle divers aspects de la modélisation des fondations sont adressés, y compris
le coefficient de réaction. Cet article utilise un nouveau outil (Estephan et al.2021) qui automatise l'échange itératif du module de réaction
de la fondation et des contraintes de chargement entre les ingénieurs géotechniques et structurels, respectivement, dans l’étude des
fondations. Le but est d'exposer l'importance d'adopter la variabilité spatiale exacte et complète du module de réaction de la fondation
dans les simulations semi-couplées de l'interaction sol/fondation. L'effet de la variabilité de ks le long d’une fondation sur la conception
des armatures en acier est éxaminé par une étude paramétrique. L'utilisation d'un coefficient variable à travers la fondation a conduit à
une diminution significative de l'armature en acier. Les résultats montrent que l'hypothèse d'une valeur constante de ks sous la fondation
a des conséquences négatives sur la précision de l’étude.

KEYWORDS: Mat foundation, modulus of subgrade reaction, Discrete Area Method, soil-structure interaction.

1 INTRODUCTION engineering software. These simulations are based,

fundamentally, on the very old, yet simple Winkler’s hypothesis
Large shallow foundations (mats) are commonly adopted for (Winkler 1867). According to this hypothesis, the soil is
structures whenever isolated foundations are found not capable idealized as a system of identical yet independent linear elastic
of meeting the bearing capacity and settlement criteria, and/or springs, and the soil’s stiffness is represented by the modulus of
whenever a closed foundation system is required such as a subgrade reaction (ks). Determining the most suitable value of
planned foundation below water table. The design of a mat the modulus of subgrade reaction is, however, highly dependent
foundation involves both structural and geotechnical engineers, on the stress-deformation properties of the subgrade and the
who work in close coordination to determine the required dimensions of the foundation (Terzaghi 1955). In early practice,
geometry, thickness, and steel reinforcement of this concrete due to the lack of modern computational tools, the modulus of
element. The layout and amount of steel reinforcement is directly subgrade reaction was taken as a uniform value obtained using
dependent on the interaction between the foundation and the empirical methods (Fraser & Wardle 1976, Bowles 1982, Holtz
underlying soil/rock and, specifically, from the straining actions 1991). In an attempt to reach more accurate representations of
within the foundation generated from the applied loads and soil the subgrade behavior, various models that take into
stiffness. consideration the coupling effect among the soil springs were
In current design practices, semi-coupled simulations are later developed (Kerr 1965, Horvath 1983, Banavalkar & Ulrich
widely adopted by the practicing engineers where, in the case of 1984, Horvath 1989, Ulrich 1991, Colasanti & Horvath 2010,
shallow foundations, the structure (mat) and the soil are modelled Jeong & Cho 2014, Loukidis & Tamiolakis 2017). These models
independently in structural and geotechnical engineering were all deemed crucial since failing to predict the correct
software but with two important boundary conditions: 1) the settlement profile was found to drastically influence the
modulus of subgrade reaction in the structural engineering calculated maximum differential settlement and the maximum
software and 2) the loading stresses in the geotechnical

581
bending moments in the structural element leading to inaccuracy modeled as a 2D dimensional thick slab element supported on
and erroneous results in steel reinforcement calculations. discrete elastic springs defined by the modulus of subgrade
However, and even with the powerful computational tools at reaction.
hand, practitioners using the semi-coupled simulations nowadays Settle3 is a three-dimensional software that computes
still have to assign manually the values of the modulus of immediate and long-term settlements under surface loads such as
subgrade reaction and the loading stresses in the structural and structural foundations and embankments. Complex soil profiles
geotechnical software, respectively. Since soil-structure and loading conditions can be created using this software. Settle3
interaction is, routinely, an iterative process, convergence in allows for inserting 3-dimensional load distribution, while
design can only be obtained after multiple analyses are conducted vertical settlements are calculated in one dimension assuming
(ATC 2020). This time-consuming procedure often represents that only vertical settlements can occur. Strains at each element
another source of human error and uncertainty in design. are calculated from the 1D modulus of elasticity and effective
Additionally, when the design schedule is tight, the interaction stresses which in turn are obtained using the Boussinesq method
between the geotechnical and structural engineers is sometimes in a homogeneous, semi-infinite half space (Rocscience 2009).
cut short and practitioners tend to revert to the simplified The mat-subgrade contact area in CSI SAFE is divided into
Winkler’s model and adopt a uniform value of the modulus of small rectangular areas where the contact pressure can be
subgrade reaction. practically considered constant (Horvilleur & Patel 1995). Each
This paper uses a new tool (Estephan et al. 2021) that rectangular element is assigned its corresponding modulus of
automates the iterative exchange of the modulus of subgrade subgrade reaction as obtained from the Settle3 analysis.
reaction and the loading stresses between the geotechnical and The soil is modeled with a layer thickness that is at least two
structural engineers, respectively, in the design of foundations. times the foundation width. Immediate settlement is addressed in
The aim is to expose the importance of adopting the exact and this study without consideration of water effect (long term
complete spatial variability of the modulus of subgrade reaction deformation, pore water pressure).
in the semi-coupled simulations of soil/mat foundation The new tool in Settle3 is used to automate the iterative
interaction. The effect of the variability of ks along the same raft analysis. The soil reaction pressure at the base of the foundation
on steel reinforcement design is investigated through a generated by SAFE is introduced as a flexible load in the Settle3
parametric study. model. Then, the immediate settlement at the middle of each
previously defined element is calculated, and the modulus of
subgrade reaction (ks) is deduced by dividing the pressure by the
2 METHODOLOGY resulting settlement. The new set of ks values is then generated
and conveyed to the SAFE model for the next iteration. This
2.1 Parametric study iterative process remains ongoing until the settlement profiles
A parametric study is proposed to study the effect of varying the from SAFE and Settle3 converge similarly to what is shown in
applied loads on the foundation and the modulus of elasticity of Figures 1 and 2 for the case where Q = 12000 kN and E = 2000
the underlying soil/rock on the steel reinforcement design when MPa.
the exact and complete spatial variability of the modulus of
subgrade reaction is adopted. The focus of this work is on high
rise buildings and towers that would be supported by large
shallow foundations on soft/strong rock.
The foundation in this study is characterized by 15-m long
spans with 3 spans per direction. The spacing between the
columns is kept constant in both directions in all studied cases.
Two values of vertical loads (Q) are investigated (6000kN and
12000kN). This load is applied on each column based on the
tributary area served. This is a realistic assumption by which the
load on the interior columns is exactly Q, the one on the edge
columns is Q/2, and the load on the corner columns is Q/4. For
every load/column, the analysis is performed using three values
of the modulus of elasticity of the soil/rock (E), 100, 1000 and
2000 MPa.
In a typical design, the mat foundation acts either as a rigid or Figure 1. Settlement profile obtained from SAFE after finalizing the
as a flexible plate governed by its thickness. The thickness is iterative process.
defined for each case based on the punching shear criteria at the
critical load transferred by the columns of the superstructure.

2.2 Adopted procedure

The steps described in the ACI’s special publication (SP 152) for
the Discrete Area Method were adopted in this work (Horvilleur
& Patel 1995, Horvath 1995). Two commercially available
software are used to conduct the analysis. SAFE (CSI 2016) is
used to obtain the soil reaction pressure resulting from the
applied loading. Settle3 (Rocscience 2009) is used to obtain the
settlement profile and the corresponding modulus of subgrade
reaction due to the loading stresses obtained from SAFE.
SAFE is a finite element software that is used to calculate the Figure 2. Settlement profile obtained from Settle3 after finalizing the
shear forces and moments of the concrete mat foundation. It can iterative process.
be also utilized to get the distribution of the soil pressure acting
on the base of the foundation (CSI 2016). The foundation is

582
 The generated design outcomes after the implementation of a Tables 1 and 2 show that for a foundation resting on a
variable ks are compared to those of three cases (A, B and C) in relatively weak soil (E = 100 MPa), savings in top reinforcements
which each foundation is modeled with uniform ks values. are observed at the edges and the center for both values of vertical
In case (A), the uniform ks value is chosen as the weighted load Q.
average of the first generated set of the variable ks values.
In case (B), the uniform ks value is chosen as the rigid modulus Table 1. The variation in percentages of steel reinforcement for
of subgrade reaction kR, that can be calculated assuming the Q = 6000 kN.
foundation is infinitely rigid using equations 1 and 2 (Fraser &
E=100 MPa
Wardle 1976, Holtz 1991)
Top Bottom
𝐸𝐸
𝑘𝑘𝑅𝑅 = (1)
𝐶𝐶𝑓𝑓 𝐵𝐵(1−𝑣𝑣𝑠𝑠2 ) ks,uniform (kN/m3) Edge Center Edge Center
𝐿𝐿 2909 27.90% 12.50% 0.00% -20.40%
0.85 ( )0.45 A
𝐶𝐶𝑓𝑓 = 𝐵𝐵
5𝑣𝑣3
(2)
𝐿𝐿 𝐵𝐵 1+𝑒𝑒 𝑠𝑠 B 3250 28.14% 11.39% 0.00% -19.80%
[1+0.1(2+ ) ]
𝐵𝐵 𝐻𝐻
C 918 30.54% 22.90% 0.00% -31.40%
where E is the soil’s Young Modulus; B and L are the width E=1000 MPa
and length of the mat respectively; vs is the subgrade’s Poisson’s
ratio; H is the height of the soil layer beneath the mat and Cf is a Top Bottom
correction factor.
ks,uniform (kN/m3) Edge Center Edge Center
In case (C), the uniform ks value is chosen as the simplest
value of modulus of subgrade reaction obtained by dividing the A 29874 33.60% 0.00% 0.00% 14.10%
average uniform distributed load on the foundation by the
maximum settlement, as shown in equation 3. B 32495 32.80% 0.00% 0.00% 13.70%

𝑞𝑞𝑎𝑎𝑣𝑣𝑒𝑒𝑎𝑎𝑎𝑎𝑎𝑎𝑒𝑒
C 4116 43.00% 8.80% 0.00% 14.80%
𝑘𝑘𝑠𝑠 = (3)
∆𝑚𝑚𝑎𝑎𝑚𝑚 E=2000 MPa
where qaverage is the sum of all column loads divided by the
total mat area and Δmax is the maximum observed settlement. Top Bottom
3
ks,uniform (kN/m ) Edge Center Edge Center
3 RESULTS AND DISCUSSION
A 56476 28.00% 0.00% 0.00% 25.40%
This section discusses the variation of the resulting percentages B 64990 26.80% 0.00% 0.00% 24.30%
of steel reinforcement differences (calculated in reference to the
same foundation designed with a uniform modulus of subgrade C 5317 42.70% 5.89% 0.00% 30.00%
reaction ks) for the top and bottom steel reinforcement. The
percentages presented in Tables 1 and 2 are calculated using The percentages in decrease, and accordingly the areas of top
equation 4 steel reinforcement savings, are almost identical for cases A and
B compared to that of case C. This is because designing the
(𝐴𝐴𝑠𝑠,𝑚𝑚𝑎𝑎𝑚𝑚 )𝑘𝑘 − (𝐴𝐴𝑠𝑠,𝑚𝑚𝑎𝑎𝑚𝑚)𝑘𝑘 foundation with a very low uniform value of ks calculated based
𝑠𝑠,𝑢𝑢𝑢𝑢𝑢𝑢𝑓𝑓𝑢𝑢𝑎𝑎𝑚𝑚 𝑠𝑠,𝑣𝑣𝑎𝑎𝑎𝑎𝑢𝑢𝑎𝑎𝑣𝑣𝑣𝑣𝑒𝑒
× 100 (4) on the maximum settlement (case C) results in an extremely
(𝐴𝐴𝑠𝑠,𝑚𝑚𝑎𝑎𝑚𝑚 )𝑘𝑘
𝑠𝑠,𝑢𝑢𝑢𝑢𝑢𝑢𝑓𝑓𝑢𝑢𝑎𝑎𝑚𝑚
overdesigned foundation, and this applies for the three values of
where As,max is the maximum area of steel reinforcement
E. For cases A and B, as the modulus of elasticity E increases
calculated in the models with uniform or variable ks values.
and the subgrade is stronger, the minimum reinforcement
Design strips are drawn over the foundation covering the entire
criterion governs in the central portion of the mat, but significant
footprint. The percentages of steel saving/addition resulting from
top steel reinforcement savings are still observed at the edges. In
the implementation of a variable ks are extracted at the edges and
case C, top steel reinforcement savings are observed at the center
the middle of the critical design strips where the maximum
as well, again due to the very safe design generated from the
required areas of steel reinforcement are observed. The mat at
uniform assumption.
hand is symmetrical i.e. the results are the same at both edges.
As for bottom areas of steel reinforcements, the impact of
All the calculations are performed while taking into
consideration the minimum steel reinforcement required by ACI implementing a variable ks seems to be highly sensitive to the
318-19 (ACI 2019). The value of (0.00%) in Tables 1 and 2 value of the modulus of elasticity of the soil/rock. Varying the
indicates that both areas of the required steel reinforcement, for modulus of subgrade reaction under the mat shows that a
the uniform and the variable ks cases, fall below the required localized increase in bottom steel reinforcement occurs below the
minimum reinforcement. interior columns. For all the studied cases, the minimum
The results confirm that adopting a uniform ks value in design reinforcement criterion governs at the edges of the mat.
leads to constant over design of top reinforcement irrespective of For weak soil/rock conditions (E = 100 MPa) and both values
the uniform ks value assigned (Case A, B, or C). Assigning higher of Q, assuming a uniform ks value results in a significantly under
loads to the interior columns induces higher settlements in the designed bottom steel reinforcement detail, requiring an addition
center of the foundation and, thus, leads to the generation of of 20% to 64% of steel reinforcement when compared to the
tensile stresses at the bottom and compressive stresses on the top exact variable model. This under design is significantly evident
of the foundation section. Therefore, the maximum bottom areas in case C where the uniform ks value is the lowest.
of steel reinforcement are observed below the interior columns. For stronger soil/rock conditions, we observe that assuming a
On the other hand, maximum top areas of steel reinforcement are uniform ks results in an over designed bottom steel reinforcement
observed in the critical design strips at the edges of the detail, similarly to top steel reinforcement. Savings are larger for
foundation. The reduction in the top reinforcement is, thus, Q=6000 kN than Q=12,000kN for all cases.
highest at the edges and it decreases closer to the center.

583
Table 2. The variation in percentages of steel reinforcement for 6 REFERENCES
Q = 12000 kN.
ACI (American Concrete Institute) (2019). Building code requirements
E=100 MPa for structural concrete and commentary. ACI 318-19. Farmington
Hills, MI: ACI.
Top Bottom
ACI Committee 336 (1988). Suggested analysis and design procedures
ks,uniform (kN/m3) Edge Center Edge Center for combined footings and mats. Structural Journal, 85(3), pp.304–
324.
A 3124 26.00% 19.20% 0.00% -49.32% Applied Technology Council, ATC. 2020. A Practical Guide to Soil-
Structure Interaction. FEMA P-2019. Federal Emergency
B 3250 25.90% 18.90% 0.00% -48.71% Management Agency, Washington, D.C.
Banavalkar, P.V. and Ulrich, E.J. (1984). Structural and geotechnical
C 1121 28.00% 27.33% 0.00% -64.00% features of Republicbank center. In: Council on Tall Buildings and
Urban Habitat. Singapore, pp.363–370.
E=1000 MPa
Colasanti, R.J. and Horvath, J.S. (2010). Practical subgrade model for
Top Bottom improved soil-structure interaction analysis: software
implementation. Practice Periodical on Structural Design and
ks,uniform (kN/m3) Edge Center Edge Center Construction, 15(4), pp.278–286.
CSI (Computers and Structures, Inc.). (2016). CSI analysis reference
A 30209 35.10% 0.00% 0.00% 2.59% manual for SAP2000, ETABS, SAFE and CSIBridge, 534. Berkeley,
CA:CSI.
B 32495 34.70% 0.00% 0.00% 2.61% Estephan R., Malek C, Abou-Jaoude G., Yacoub T., Skeini H., Turk M.,
and Valiulin D. (2021). A new design tool for shallow foundations
C 2365 42.31% 17.40% 0.00% 0.00% analysis offering enhanced accuracy, reduced design time, and
E=2000 MPa optimized construction cost. Submitted to the Proceedings of the
International Society for Soil Mechanics and Geotechnical
Top Bottom Engineering, Sydney 2021.
Fraser, R.A. and Wardle, L.J. (1976). Numerical analysis of rectangular
3
ks,uniform (kN/m ) Edge Center Edge Center rafts on layered foundations. Géotechnique, 26(4), pp.613–630.
Holtz RD (1991). Stress distribution and settlement of shallow
A 52808 35.14% 0.00% 0.00% 14.00% foundations. In: Fand H-Y, editor. Foundation engineering
handbook.
B 64990 33.60% 0.00% 0.00% 13.39% Horvath, J.S. (1983). New subgrade model applied to mat foundations.
Journal of Geotechnical Engineering, 109(12), pp.1567–1587.
C 5866 43.50% 13.58% 0.00% 10.00% Horvath, J.S. (1989). Subgrade models for soil-structure interaction
analysis. In: Foundation Engineering: Current Principles and
Practices. New York: American Society of Civil Engineers, pp.599–
612.
4 CONCLUSIONS Horvath, J.S. (1995). Subgrade modeling for mat foundations: a review
and critique of analytical methods. ACI Special Publication, 152,
In common practice, practitioners may still be using a uniform pp.117–160.
Horvilleur, J.F. and Patel, V.B. (1995). Mat foundation design - a soil-
value of the modulus of subgrade reaction in the design of structure interaction problem. ACI Special Publication, 152, pp.51–
shallow foundations to avoid the complexities associated with 94.
implementing the exact spatial variability of the modulus of Jeong, S. and Cho, J. (2014). Proposed nonlinear 3-D analytical method
subgrade reaction ks. This paper presents the drawbacks of for piled raft foundations. Computers and Geotechnics, 59, pp.112–
implementing such a simplistic approach and shows that using 126.
the new tool developed in Settle3 and with proper cooperation Joseph Edward Bowles (1982). Foundation analysis and design. 4th ed.
New York: Mcgraw-Hill.
between structural and geotechnical engineers, a more accurate
Kerr, A.D. (1965). A study of a new foundation model. Acta Mechanica,
representation of the real behavior of shallow foundations can be 1(2), pp.135–147.
achieved. The Discrete Area Method is adopted to produce a Liao, S.S.C. (1995). Estimating the coefficient of subgrade reaction for
variable set of the modulus of subgrade reaction for a large mat plane strain conditions. Proceedings of the Institution of Civil
foundation while varying the load applied and the modulus of Engineers - Geotechnical Engineering, 113(3), pp.166–181.
elasticity of the subgrade. If the mat is resting on a weak Loukidis, D. and Tamiolakis, G.-P. (2017). Spatial distribution of
foundation soil, implementing a variable ks was shown to lead to Winkler spring stiffness for rectangular mat foundation analysis.
Engineering Structures, 153, pp.443–459.
a significant drop in the top steel reinforcements and an increase Rocscience, Inc. (2009). Settle3 Theory Manual. Toronto, Canada:
in the bottom steel reinforcements at only localized areas around Rocscience.
the interior columns. On the contrary, for stronger foundation Terzaghi, K. (1955). Evaluation of coefficients of subgrade reaction.
soil, the results showed that designing a large shallow foundation Géotechnique, 5(4), pp.297–326.
with a uniform ks value results in an over designed top and Ulrich Jr., E.J. (1991). Subgrade reaction in mat foundation design.
bottom steel reinforcement. Concrete International, 13(4), pp.41–50.
Winkler, E. (1867). Die Lehre von Elastizität und Festigkeit: 1. teil.
Prague: H. Dominicus.
5 ACKNOWLEDGEMENTS

This work was developed during the COVID-19 pandemic as a

collaboration between the Lebanese American University and
geotechnical and structural practicing engineers from Dar Al
Handasah (Lebanon office).
Special gratitude and appreciation go to Julien Sassine for his
commendable contribution and efforts in performing the
preliminary analyses in this work.

584