Professional Documents
Culture Documents
Ppr0947r PDF
Ppr0947r PDF
Determination of Coefficient of
Subgrade Reaction
Jamshid Sadrekarimi
Department of Geotechnical Engineering, University of Tabriz, Tabriz, Iran
Email: jsadr@tabrizu.ac.ir
Maryam Akbarzad
Department of Geotechnical Engineering, University of Tabriz, Tabriz, Iran
Email: maryam_akbarzad@Yahoo.com
ABSTRACT
In this paper, different methods proposed for determination of the coefficient of subgrade
reaction, ks are discussed, compared and evaluated for their suitability and accuracy. The
geotechnical characteristics of a site on Tabriz Marl were selected as the base data and
settlement analysis results with different methods were compared with that of obtained from
analyses with advanced soil models using Safe and Plaxis software. It was discovered that for
Tabriz Marl, soft soil model is the best governing model and Vesic relation among the
methods of determination of ks leads to a negligible error in comparison to the soft soil model.
Also, in order to achieve more accurate results from these methods, it is proposed to use mean
elasticity modulus which takes into account the effect of geometric and mechanical properties
of sub-layers.
KEYWORDS: Winkler model, soft soil model, coefficient of subgrade reaction,
modulus of elasticity, settlement.
INTRODUCTION
Foundation-ground interaction has been one of the challenging problems in geotechnical
engineering since late nineteenth century. Because of the complexity of soil behavior, subgrade in
soil-foundation interaction problems is replaced by a much simpler system called subgrade
model. One of the most common and simple models in this context is Winkler hypothesis.
Winkler idealization represents the soil medium as a system of identical but mutually
independent, closely spaced, discrete and linearly elastic springs and ratio between contact
pressure, P, at any given point and settlement, y, produced by it at that point, is given by the
coefficient of subgrade reaction, ks (Dutta and Roy 2002).
Vol. 14, Bund. E 2
At first, this concept was introduced to use in analysis of rigid plates, but during the following
decades the theory was expanded to include the computation of stresses in flexible foundations
(Terzaghi 1955). In the area of soil-foundation interaction, lots of investigators have utilized this
model, such as Biot (1937), Terzaghi (1955), Vesic (1961), Horvath (1989), Daloglu and
Vallabhan (2000) and so on. Since 1920, the theory of subgrade reaction has also been used for
computing stresses in piles and sheet piles, which are acted on by horizontal forces above the
ground surface. In this case, the ratio between contact pressure and displacement of pile referred
to as the coefficient of horizontal subgrade reaction, kh (Terzaghi 1955). However, in this paper
only the coefficient of vertical subgrade reaction, ks, is taken into consideration.
The simplifying assumptions which Winkler hypothesis is based on causes some errors
(Terzaghi 1955; Stavridis 2000). One of its basic limitations lies in the fact that this model cannot
transmit the shear stresses which are derived from the lack of spring coupling. Several modified
models have been proposed to overcome these shortcomings in the middle of the twentieth
century by researchers such as Filonenko and Borodich, Heteny, Pasternak and Kerr (Dutta and
Roy 2002; Horvath 1989). In all the models, the connectivity of the individual Winkler springs is
accomplished by incorporating an elastic plate, which undergoes flexural or transverse shear
deformation. However, assigning the numerical values of flexural rigidity and shear modulus of
these plates, make the problem two-fold. Hence, these methods did not get enough popularity
among designers. Despite the flaws, which are attributed to Winkler hypothesis, using of this
method has led to useful results in studying the behavior of long flexible beams (Stavridis 2000).
Of course, this result is related to accurate estimation of kh. Therefore, designers already use this
approach widely and various computer soft wares are based on it.
Evaluation of the numerical values of ks is one of the most complex and sophisticated
problems in geotechnical engineering. Even, time and widespread use of ks have not eliminated
long-standing disagreement on the determination methods. In the other hand, this factor leads to
inaccuracy in the results of Winkler model and this aspect of the problem is scrutinized in this
paper by a case study. In the first half of the twentieth century, some articles that gave erroneous
values for ks have been published and it was assumed that this coefficient has a definite value for
any given subgrade. But Terzaghi (1955) dealt with the factors that influence the coefficient of
subgrade reaction in the comprehensive treatise and demonstrated that ks is not a fundamental soil
property and it is a problem-specific observed result and in addition to depending on elastic
characteristics of subgrade, it also relates to the geometry of the footing and loading scheme.
After that, particularly between the 1950s and 1980s, this concept has been scrutinized and
investigators have proposed numerous relations. Some of these relations were empirical and some
of them were derived by using the elastic continuum theory.
Nevertheless, there is not enough information in technical literatures about the computational
validity and accuracy of comprehensive application of these relations in engineering practice and
in some cases; relations of determination of horizontal coefficient of subgrade reaction are
utilized for evaluation of vertical coefficient (Okeagu and Abdel-Sayed 1984). In reality, a unique
suggestion with respect to variety of mechanical properties of soils seems to be impossible. Also,
need for more research on this topic have been emphasized (Daloglu and Vallabhan 2000).
Hence, in this paper, different methods, proposed for determination of ks, are compared and
evaluated for their suitability and accuracy. The geotechnical parameters of a site on Tabriz Marl
were selected as the base data and settlement analysis results with these methods are compared
with that of obtained from analysis with advanced soil models.
Vol. 14, Bund. E 3
In general, the methods of determination of ks can be classified as: 1- Plate load test (Dutta and
Roy 2002; Bowles 1998), 2- Consolidation test (Dutta and Roy 2002; Bowles 1998), 3- Triaxial
test (Dutta and Roy 2002), 4- CBR test (Nascimento and Simoes 1957) and 5- Empirical and
theoretical relations that are proposed by researchers (Bowles 1998; Elachachi et al. 2004).
Among these methods, approaches 1 and 5 are utilized more than the others. Due to lack of
enough data about plate load test, only empirical and theoretical relations are taken into
consideration. As mentioned formerly, various relations of ks have been proposed by researches
and some of them are represented in Table 1; wherein, Es = modulus of elasticity, υs = Poisson’s
ratio, B = width of footing, EI = flexural rigidity of footing, ks1 = the coefficient of subgrade
reaction for a plate 1 ft wide, µ = non-dimensional soil mass per unit length, B' = least lateral
dimension of footing, IS and IF = influence factors which depend on the shape of footing and
parameter m takes 1, 2 and 4 for edges, sides and center of footing, respectively.
1 Biot 0 .9 5 E s B4 Es
ks = [ ]0 . 1 0 8
B ( 1 − ν s ) ( 1 − ν 2s ) E I
2
2 Terzaghi B + 1 2
k s = k s1 ( )
For sands 2B
1
For clays k s = k s1
B
3 Vlassov E s (1 − ν s ) μ
ks = ( )
(1 + ν s )(1 − 2 ν s ) 2 B
4 Vesic 0 .6 5 E s EsB4
ks = 12
B ( 1 − ν 2s ) EI
7 Selvadurai 0 .6 5 Es
ks = .
B 1 − ν 2s
8 --- Es
ks =
B ′(1 − ν s2 ) mI s I F
Eq. (1) and (4) are defined for infinite beams resting on an elastic soil continuum (Biot 1937;
Vesic 1961), but application of them in mat footings is observed widely in technical literatures
(Bowles 1998). Eq. (2) when the quantity of the coefficient of subgrade reaction beneath a plate
of 1 ft wide is defined only can be used. This equation is also relevant in analysis of plate load
test results by substituting width of loading plate with 1ft, but some of the researchers instead of
using these equations in plate load test suggest using of those modified by Arnold (Al-sanad et al.
1993). Eq. (3) is introduced for beams and plates resting on elastic half space (Elachachi et al.
2004), but ambiguities of estimating µ make the problem more complex. Eq. (5), (6) and (7) are
proposed for computing the coefficient of horizontal subgrade reaction in buried circular conduits
(Okeagu and Abdel-Sayed 1984) and are employed for evaluation of ks in few limited cases
(Elachachi et al. 2004). Also, ks can be determined using the theory of elasticity. By rewriting the
Vol. 14, Bund. E 4
relation of settlement of rectangular plates resting on elastic half space, ks can be expressed as Eq.
(8) (Bowles 1998).
From the above description, it can be concluded that the equation obtained from elasticity, Biot
relation and Vesic relation are more adequate for evaluating of ks. Also using of these three
equations has been proposed more than others in technical literature and many investigators
compared the results of their proposed relations and methods with those obtained from these
equations for suitability and merits (Daloglu and Vallabhan 2000). Hence, in this paper accuracy
and precision of these relations in predicting settlement and contact pressure are considered in
detail.
For analyzing based on Winkler model and advanced soil models, Safe v. 8.06 and Plaxis v.
7.2 soft wares are used, respectively. In Plaxis soft ware, advanced soil models consisting soft
soil, creep soft soil, hardening soil and Mohr-Coulomb models may be applied. So, the model
which shows better coincidence to the mechanical behavior should be employed.
Soft soil model, as it is obvious from its name, is suitable for soft soils. The special feature of
these materials is their high degree of compressibility. This is best demonstrated by oedometer
test data. Another characteristic of soft soils is the linear stress-dependency of soil stiffness. All
of these features are taken into consideration in soft soil model (Manual of Plaxis). In Soft soil
creep model, in addition to the features of soft soils, creep (secondary compression) is considered.
This model is adequate for soils that secondary compression is a significant percentage of
ultimate settlement (Manual of Plaxis).
Hardening soil model supersedes the hyperbolic model by using the theory of plasticity rather
than the theory of elasticity. In Plaxis soft ware, this model is able to simulate the behavior of
both soft soils and stiff soils. Hardening soil model provides stress dependent stiffness according
to power law (Manual of Plaxis).
different places of the site, by rotary drilling. Sampling and standard penetration tests were
carried out at 3 m intervals to a depth of 25 m. Corrected N values are plotted for four of the
boreholes in Fig. 1 and they are in proportion to Er70.
In a laboratory, Index and physical and mechanical tests were performed on the undisturbed
samples. Referring to the consolidation test results, it was observed that the soil is over
consolidated up to 900 kPa and 950 kPa for yellow marl and gray marl, respectively. Also, some
of the samples showed swelling. Hence, soil mass is heterogeneous, but it is assumed to be
homogeneous in analyses.
N
0 20 40 60 80 100
0
9
Depth (m)
12
15
18
21
24
27
obtained from modeling of the consolidation test in Plaxis with different soil models. For all the
models, identical average values of density (γwet, γdry) and failure parameters (C', φ') were defined.
The angle of dilation was assumed to be zero, however.
For hardening soil and Mohr-Coulomb models, variations of modulus of elasticity, Es, with
effective stress level are based on the data obtained from consolidation test, which are plotted in
Fig. 2. It should be noticed that in oedometer test, restriction of lateral stain influences the
modulus value Eoed, whereas in real soil acted upon loads, lateral strains develop in all directions.
Hence, Eoed can be related to Es as following:
(1 +ν s )(1 − 2ν s ) (9)
Es = Eoed
(1 −ν s )
1
E oed = (10)
mv
in which υs = Poisson’s ratio and mv = coefficient of volume compressibility. Table 3 shows the
strength parameters of hardening soil and Mohr-Coulomb models which are employed for
modeling.
30000
Yellow Marl
Gray Marl
25000
20000
Es (kN/m )
2
15000
10000
5000
0
0 100 200 300 400 500 600 700 800
2
Effective stress (kN/m )
Figure 2: Modulus of elasticity-stress level diagrams in yellow marl and gray marl
In hardening soil model, E 50ref = secant modulus of elasticity in standard drained triaxial test,
ref
E oed = tangent modulus of elasticity for primary oedometer loading at reference stress level and
p ref represents reference stress for stiffness. Numerical values of E oed
ref
are defined from oedometer
test results. Because there is not any data about triaxial test, so numerical values of E 50ref are
evaluated from Fig. 2 at initial stress level ( σ 0′ ) which samples have in soil mass before coring
which is defined as pref . Parameter m is expressed as:
Vol. 14, Bund. E 7
m
C cot ϕ − σ 1′ (11)
E s = E sref
C cot ϕ + p
ref
wherein C = cohesion, φ = angle of internal friction, E sref = modulus of elasticity at p ref and Es =
modulus of elasticity at σ 1′ . Substituting appropriate values from Fig. 2, m value is obtained 1 for
both marl soils. In Plaxis, in the case of soft soil, it is realistic to use m = 1 and soft soil model
can be used instead of this model. Therefore, hardening soil model is eliminated from analysis
and the most suitable model is gathered among Mohr-Coulomb, soft soil and soft soil creep
models.
ref
(kPa) E oed 6846 8673 -- --
ref
(kPa) E 50 6484 8239 -- --
(kPa) p ref 212 240 -- --
m 1 1 -- --
Es (kPa) -- -- 6484 8239
νs -- -- 0.3 0.3
With Mohr-coulomb model, Es is also estimated using Fig. 2. In soft soil creep and soft soil
models, considering linear dependency of Es to effective stress level (Eq. (11)) stress-strain
relation is expressed using modified compression index, λ*, modified swelling index, κ* and
modified secondary compression index, µ*. These parameters can be determined from
compression index, swelling index and secondary compression index, from Terzaghi
consolidation test results, respectively. The relevant relations are available in Manual of Plaxis
soft ware. Numerical values of these parameters are given in Table 4 and default setting is used
for other advanced parameters.
Loading steps are applied same as the laboratory tests. The results of these analyses together
with the average of the measured laboratory data are plotted in Fig. 3 for comparison.
Based on Fig. 3, for the both marl soils, the soft soil model shows better coincidence to the
mechanical behavior in comparison to other advanced soil models. Also, it corresponds to
empirical observations, because the examined soil mass, in common with soft soils, exhibits
comparatively high degree of compressibility (Sadrekarimi and Kia, 2005). Therefore, the soft
soil model is used as the basis for methods of determination of ks in the subsequent analyses.
Vol. 14, Bund. E 8
Table 4: Strength parameters of soft soil and soft soil creep models
Parameters of soft soil model Parameters of soft soil creep model
Mohr-Coulomb
-0.001
Settlement (m)
-0.002
-0.003
-0.004
-0.005
10 100 1000 10000
2
log P (kN/m )
0.001
Test Data
(b) Soft-Soil
0
Soft-Soil-Creep
Mohr-Coulomb
-0.001
Settlement (m)
-0.002
-0.003
-0.004
-0.005
10 100 1000 10000
2
log P (kN/m )
Figure 3: Comparison between stress-settlement curves obtained from advanced soil models and
results of consolidation test on (a) Yellow marl, (b) Gray marl
Vol. 14, Bund. E 9
In order to consider stress-history caused by excavating, mat was modeled 6 m below the
original ground level. Interface elements are used to model soil-foundation interaction. Layering
and soil properties were defined referring to Table 2. Since properties of soil down to the depth of
25 m are available, texture and engineering properties of ground down to the influence depth of
super structure, regarding the local information on Tabriz subsoil zonation, is assumed the same
as the layer No. 6. The soil mass from 23 m to 106 m deep was divided into several layers based
on the modulus of elasticity profile in Fig. 2, in a manner that the error due to assuming linear
variation of Es with stress level became negligible. For all layers, except sandy layer, soft soil
model is selected and strength parameters of both marl soils are introduced from Table 4. The
geometry configuration is illustrated in Fig. 4-a.
For sandy layer, Mohr-Coulomb model was employed. Es and its increase with depth are
assigned using the results of SPT test (Fig. 1), which is given by (Bowles 1998)
E S = 1200 ( N + 6 ) (12)
wherein, N is SPT number for Er55. After modification of N and substituting in Eq. (12), Es value
beneath the foundation level was estimated as 116040 kPa and its increase per meter of depth
obtains 3300 kPa.
The finite element model is generated automatically and irregular 6–node triangle medium-
sized elements are used for the soil and plane-strain analysis is chosen. The deformed mesh is
shown in Fig. 4-b.
With settlement of layered soil systems, numerous investigation have been carried out and
various relation have been proposed, but in all these methods only effect of geometric
characteristics of layers on Es were taken into consideration (Enrico and Giovanni 1993), whereas
it is obvious that the effect of external load decreases with depth (Boussinesq theory) (Bowles
1998). Hence, moduli of elasticity of upper layers are more effective on settlement than the lower
layers. This issue was named depth factor, IDi, in this paper.
Evaluation of the equivalent modulus of elasticity consists of two steps: assigning the effect of
geometric properties of layers and characterizing the value of depth factor. For the first one,
thickness of each layer is selected and depth factor is defined as a ratio of settlement at mid-point
of thickness of each layer to total settlement of the geometry modeled in Plaxis software (Fig. 4-
a) and equivalent modulus of elasticity, Ese, can be estimated by the suggested relation given by
Vol. 14, Bund. E 10
n
E si I Di H i
(13)
i =1
E se = n
I Di H i
i =1
in which Esi = modulus of elasticity at mid-point of thickness of each layer, Hi = thickness of each
layer and n = number of layers. Relevant values of these parameters are represented in Table 5.
Substituting these values, Ese is obtained equal to 21021 kPa. Whereas, if one disregards layering,
Es along soil-foundation interface equals to 116040 kPa. It is evident that the significant
difference between these values will lead to a remarkable error in predicted settlement. This
indicates the importance of layering in determination of ks, especially in the project under
consideration, wherein the modulus of elasticity varies considerably from sand layer to marl.
Substituting Es = 21021 kPa, υs = 0.3, B = 20 m, EcIc =3.1×108 kN.m2 and relevant values for
IS, IF and m in Eq. (1), (4) and (8) the ks values were computed and summarized in Table 6. It
should be noticed that the relation obtained from the theory of elasticity (Eq. (8)) gives various
quantities for edges and center of the footing. Therefore, for estimating the average coefficient of
subgrade reaction the suggested method by Bowles (1998) is used.
(a) (b)
Esi
127590 6226 6846 7673 8408 9077 10387 12593 16155 23386
(kPa)
Hi
3 3 3 2.7 3.3 6. 61 11.86 28.81 15.25 20.47
(m)
IDi 0.99 0.93 0.85 0.78 0.71 0.61 0.47 0.27 0.13 0.04
Vol. 14, Bund. E 11
Interpreting the results of contact pressure is a little more complex. In Winkler model,
elasticity, Biot and Vesic relations lead to approximately equal values of contact pressures. But
the values obtained from Winkler model have great difference with that of the soft soil model.
Winkler approach gives the maximum contact pressure 35 % greater than the soft soil model
does. The difference is derived from ignoring the lateral pressure of soil in Winkler model. In
reality, lateral pressure of soil elements on the soil-foundation interface reduces the vertical
pressure. This feature is considered in the soft soil model (Manuel of Plaxis).
accuracy of the results significantly depends on the equivalent modulus of elasticity. Hence,
variations of Es against effective stress should be determined.
-0.22
-0.24
-0.26
-0.28
Settlement (m)
Plaxis
-0.3
Elasticity
Biot
-0.32
Vesic
-0.34
-0.36
-0.38
-0.4
0 5 10 15 20
-330
Elasticity
-335 Biot
Vesic
-340
Contact Pressure (kN/m2)
-345
-350
-355
-360
-365
-370
0 5 10 15 20
Figure 6: Contact pressure diagrams obtained from elasticity, Biot and Vesic relations
Nevertheless, in this article, there is not any plate-load test result, but evidently in this test only
mechanical properties of the layers placed within the influence depth of the loading plate, which
is too small in comparison with the actual size of a foundation, affect ks value. It can be concluded
that if the rate of the variation of Es with respect to depth is considerable, results of plate-load test
cannot be reliable.
Vol. 14, Bund. E 13
In summary, in order to minimize the error associated with the relations introduced formerly,
it is proposed that enough information about theories and accuracy of the chosen relation on the
similar projects (if possible) should be sought.
CONCLUSION
1- The coefficient of subgrade reaction is a concept that is valid only at soil-foundation
interface, but in this article, in order to increase the accuracy of the results, the effect of
layering and mechanical properties of the subsurface soil on ks are dealt with.
2- Among the methods for determination of ks value, Vesic relation leads to acceptable
accuracy in evaluating settlement in comparison to the soft soil model. Accordingly, this
relation is suggested as a governing relation for estimating ks for the given soil mass.
3- Winkler relation gives contact pressure greater than actual values and it is derived from
disregarding the effect of lateral pressures of soil mass.
4- In common practice, in order to minimize inaccuracy of ks relations, two items should be
considered. At first, one should have vast study and awareness on the basic theories of
these relations; and secondly, in addition to geometric properties of layers, variation of the
mechanical properties with depth is also considered in evaluation of the equivalent modulus
of elasticity.
REFERENCES
1. Al-sanad, H. A., N.F. Ismael and R.P. Brenner (1993) “Settlement of Circular and Ring
Plates in Very Dense Calcareous Sands,” Journal of Geotechnical Engineering, ASCE,
119(4): 622-638.
2. Biot, M. A. (1937) “Bending of Infinite Beams on an Elastic Foundation,” J. Appl. Mech.
Trans. Am. Soc. Mech. Eng., 59: A1-7.
3. Bowles, J. E. (1998) “Foundation Analysis and Design,” 6th ed., McGrow-Hill
International press.
Vol. 14, Bund. E 14
© 2009 ejge