Professional Documents
Culture Documents
and Devendra N. Singh the specimen is kept for some period of time in a closed container
to maintain its water content, an internal redistribution of water
April 2005, Vol. 131, No. 4, pp. 521–524.
is expected to occur because of an unbalanced total suction be-
DOI: 10.1061/共ASCE兲1090-0241共2005兲131:4共521兲
tween the micropores and macropores. Fig. 1 shows a possible
mechanism indicating how the total suction of a specimen
Setianto Samingan Agus1 and Tom Schanz2 changes with time under constant water-content conditions after
1
Research Assistant. Laboratory of Soil Mechanics. Bauhaus-Univ.
the specimen is subjected to a wetting process. In this case, the
Weimar. Coudraystrasse 11C, Weimar, 99421, Germany. E-mail:
agus-setianto.samingan@bauing.uni-weimar.de internal water redistribution is mainly a process of water move-
2
Professor. Laboratory of Soil Mechanics. Bauhaus-Univ. Weimar. ment from the macropores to the micropores. For a drying pro-
Coudraystrasse 11C, Weimar, 99421, Germany. E-mail: tom.schanz@ cess, the internal water redistribution mechanism decreases the
bauing.uni-weimar.de water volume in the micropores and hence decreases the size of
clay clusters although the total volume of the specimen may re-
main unchanged.
The authors have presented an interesting and fast technique 共i.e., On the basis of the information given by the authors that
the chilled-mirror hygrometer technique兲 in determining the soil– total suction of the specimen was measured every 30 min for 2 h
water characteristic curve 共SWCC兲 of fine-grained soils 共i.e., the it appears that, at longest, the specimen was allowed to equilibrate
silty soil and the white clay兲. The findings presented in the paper at a certain water content only for 2 h. Sometimes the true equi-
indicate that suction in fine-grained soils 共especially clay soils兲 is librium state of the specimen may not be able to be judged on the
mainly governed by the physicochemical interactions between basis of periodic measurement of specimen mass as per the nor-
water and the clay itself without significant contribution from the mal procedures. Total suction measurement data of bentonite-sand
soil pore geometry. This result is particularly true at high suction mixtures presented by Agus and Schanz 共in press兲 indicates that
values. The consequence of the dominant physicochemical as- total suction of the specimens changes with time after being kept
pects in SWCC of fine-grained soils is that the SWCC of fine- at constant water content for five weeks.
grained soils is not density-dependent, as also concluded by the Generally, the time period to reach the true equilibrium state
authors from the experimental data obtained. This discussion depends on the magnitude of the total suction gradient between
deals more with the clay soil tested by the authors 共i.e., the white each level of pores which, by and large, is controlled by the clay
clay兲. properties and permeability of the specimen at macro and micro
The authors have not mentioned the time period for which the scales. A period of 2 hours may be too short for the true equilib-
specimens were allowed to reach the “true” equilibrium state in rium to be attained by the specimens tested 共i.e., the white clay
the closed cup used, nor has information on the basic 共and physi- specimens兲, since the permeability of clay soils is generally low.
cochemical兲 properties of the soils been mentioned in the paper. In the discussers’ opinion, the almost identical values of total
The information is important, since most of clays exhibit a double suction measured in the four consecutive readings 共i.e., every
porosity structure and consist of intercluster 共or interaggregate or
30 min for 2 h兲 might not warrant that the true equilibrium state
macro-兲 pores and intracluster 共or intraaggregate or micro-兲 pores
has been reached, since normally after air-drying, the vapor space
共Gens and Alonso 1992; Yong 1999兲. In this case, the true equi-
above the specimen might still be influenced by the thermody-
librium state is consequently defined by the hydraulic equilibrium
namic nonequilibrium for a certain period of time.
between the pores of different levels 共Alonso 1998兲.
The previously mentioned aspects could have been described
in the manuscript and are, in fact, among the factors affecting the
measured soil-water characteristic curve of fine-grained soils.
Consequently, the fitting parameters obtained from the experi-
mental data will also be affected.
The discussers are also surprised that no significant volume
change was observed in all specimens 共i.e., the silty soil and
white clay specimens兲. It is rather that the measurement of di-
mensions of the specimens was difficult since the specimens’ size
was probably too small. The SWCC data expressed as a relation-
ship between volumetric water content and suction are normally
used in computing, for instance, unsaturated permeability, shear
strength, and compressibility functions 共Fredlund and Rahardjo
1993兲. The discussers would like to suggest that in future re-
Fig. 1. Effect of internal water redistribution on total suction search, a separate experiment be performed to establish a relation-
magnitude ship between water content and void ratio during drying 共i.e.,
References
respectively.
Fig. 2 shows the stresses corresponding to the conventional
bearing capacity, quc, as a function of depth. The equations relat-
ing quc to depth are similar to those relating N60 to depth.
In Fig. 3, the normalized data from the load tests on footings
with diameters of 1.0, 2.0, and 3.0 m placed on the surface of
sands with relative densities of 50%, 70%, and 90% are put to-
gether. All data merge in a single curve, regardless of the footing
dimensions and relative densities of the sand.
The authors also state, “A modern design concept in geotech-
nical engineering is to treat serviceability and ultimate limit states
within the same framework.” The discusser fully agrees with
them. But obtaining the nonlinear normalized stress-settlement
curve does not require the use of complicated mathematical mod-
els, finite element methods, invariant stress tensors, and failure
criteria 共for example, the Drucker-Prager兲—concepts that usually
Fig. 2. qbL 共MPa兲 ⫻ depth, z 共m兲
are not familiar to practicing engineers.
Alternatively, this curve may be easily represented by the fol-
lowing equation 共Décourt 1994, 1999兲:
Fig. 1 shows the stress-settlement curves for the load tests
simulated by the author with the finite-element method. The data
were taken from Figs. 4共a兲, 4共b兲, and 4共c兲 of the paper; and their log q/quc = C + C log s/B
quality is obviously poor; however, one easily recognizes that qb where C is the coefficient of intrinsic compressibility. This coef-
is strongly influenced by the relative density of the sand and is ficient is approximately constant for a given family of soils. For
only slightly influenced by the footing diameter. example, for silica sands, C varies between 0.35 and 0.45, with
The discusser has adopted the practice of representing the soil C = 0.42 being a value that fits most of the results. For the much
resistance by establishing the variation of NSPT or qc, CPT values more intrinsically compressible calcareous sands, C is in the
with depth, by using statistical regressions. The variation of N̄60, range 0.60ⱕ C ⱕ 0.85. More-precise values may be obtained by
which is the average value of N60 for each depth, with depth for analyzing load test results.
sands of different densities is shown 共in Table 1兲. The value of quc may be correlated with penetration test re-
sults, either NSPT, as the discusser has done for the last 30 years,
or qc, as proposed by the authors. As far as sands are concerned,
disregarding the extreme conditions corresponding to very loose
and very dense states, the discusser finds no reasons to consider
separately different Dr and K0 conditions, especially because both
Dr and K0 are often not known in engineering practice. Some
correlations exist for assessing Dr on the basis of NSPT or qc; but
all of them are, at most, only valid for fresh pure sands and are
rarely found in real foundation problems, where sands are aged
and not necessarily pure.
The stress quc 共MPa兲 for natural, aged sands is given approxi-
mately by 0.095N60, 共Décourt 1995兲 or 0.25qc, if one assumes
that qc 共MPa兲 = 0.377N60.
If the values of NSPT or qc corresponding to the characteristic
depths are known and if a value for C is determined or assumed
computing footing settlements for any stress value becomes easy.
The main consequence of this approach is that settlements no
longer have to be limited to 25 mm 共or 1 in.兲. Settlements three or
four times bigger may be considered reasonable provided that
footings with different loads are dimensioned with different
stresses.
The idea is to compute the settlement of one of the largest
Fig. 3. Normalized plot footings and impose that value for all the others, obviously vary-
ture session.” X COBRAMSEF, Vol. IV, After-Congress, Foz de settlement equal to 20% of the foundation width, might not
Iguaçu 共in Portuguese兲, 179–206. correspond to the ultimate load computed from the BC equations.
Décourt, L. 共1999兲. “Foundations under working load conditions.” Proc., Additionally, the use of numerical models tends to overestimate
XI Panamerican Conf. on Soil Mechanics and Geotechnical Engineer- the BC of a foundation resting on sand with a depth of founda-
ing, Foz do Iguaçu, Vol. 4, 453–488. tion, D f , equal to zero, if tensile stresses, usually developed
Décourt, L. 共2002兲. SPT, SPT-T Brazilian practice: Advantages, limita- close to the soil surface, are not adequately eliminated
tions and critics, ABMS 共in Portuguese兲. 共Rodríguez-Roa 1977, 2000兲. In this discussion, the validity of
Fellenius, B. H. 共1999a兲. “Bearing capacity of footings and piles—Does the BC equations studied by the authors is analyzed, on the basis
it really exist?” Symp. on Southern Ontario Glacial Soils State of of the aforementioned tests at Texas A&M, as well as other pub-
Practice and Recent Developments, Toronto. lished experimental data.
Fellenius, B. H. 共1999b兲. “Bearing capacity of footings and piles—A
delusion?” DFI Annual Meeting, Dearborn, Michigan.
Terzaghi, K. 共1942兲. “Discussion on pile driving formulas.” Proc., ASCE,
68共2兲, 311–323. Analyzed Experimental Data
Terzaghi, K. 共1943兲. Theoretical soil mechanics, Wiley, New York.
Small and full-scale footing load tests were selected from the
literature to cover a wide range of footing widths.
This analysis involved nine tests performed on five different
sands: Texas A&M, Berlin, Mol, Chattahoochee River, and Maipo
Discussion of “Estimation of Bearing River, as shown in Table 1 of this discussion. All the analyzed
Capacity of Circular Footings on Sands tests, except the tests done at Texas A&M, were selected so that a
Based on Cone Penetration Test” clearly defined ultimate load could be observed in the load-
by Junhwan Lee and Rodrigo Salgado settlement curve.
April 2005, Vol. 131, No. 4, pp. 442–452.
DOI: 10.1061/共ASCE兲1090-0241共2005兲131:4共442兲
Bearing-Capacity Equations
Fernando Rodríguez-Roa1
1
Professor of Geotechnical Engineering, Pontificia Universidad Católica The formulations proposed by Brinch Hansen 共1970兲 and Vesic
de Chile, Casilla 306, Santiago 22, Chile. E-mail: frroa@ing.puc.cl 共1973兲 are among the most commonly used BC equations at
present. The Bolton and Lau 共1993兲 and Michalowski 共1997兲
solutions are less frequently employed because they predict
The authors should be congratulated for their excellent work on substantially different ultimate loads, depending on the degree
the bearing capacity 共BC兲 of circular footings on sands, based on of roughness of the foundation base. However, experimental
the use of a simplified nonlinear constitutive model to simulate evidence has shown that foundation roughness has little effect on
the pre- and postfailure behavior of sands. Load-settlement curves the BC for vertical loads 共Vesic 1973; Ko and Davidson 1973;
obtained from footing load tests performed at Texas A&M Uni- Villalobos 2000兲.
versity in 1994 were used for comparison with FE predictions to Eq. 共12兲 of the original paper is only valid for ratios
validate their numerical model. L / B ⱕ 7.0 共Perkins and Madson 2000兲. For L / B = 7, the plane-
The authors used this model to carry out a series of FE strain condition is assumed to be reached.
analyses for different circular footings resting on the ground The iterative procedure employed by the authors to compute
surface, assuming the Ottawa sand properties for the supporting the predicted ultimate load, qu1t, from Eqs. 共10兲, 共11兲, and 共12兲
soil. The values of the limit unit BC, qbL, obtained from the of the original paper, is limited to analysis of axisymmetric and
numerical analyses, were compared in Fig. 6 of the original paper plane-strain problems. In this discussion, a similar iterative pro-
the soil is weightless 共␥ = 0兲 and if there is a surcharge q0 on the computed by using these equations and values measured at
ground surface, the bearing-capacity equation can be derived rig- s / B = 0.1 or any other value is not indicated.
orously 共e.g., Bolton 1982, Lyamin et al. 2006, Salgado 2006兲. As to the point regarding the incorrectness of superposition of
The limit unit resistance qbL for that case is given by the c, q0, and ␥ terms to build the triple-N formula, Lyamin et al.
共2006兲 have proposed an alternative bearing-capacity equation for
qbL = cNc + q0Nq 共1兲 sands that does not require separate q0 and ␥ terms. We believe
that a form without the artificial superposition of effects presents
1 + sin tan considerable advantages.
Nq = e 共2兲
1 − sin A last point that we need to make concerning the bearing-
capacity equation is that it is derived, even in its most evolved
Nc = 共Nq − 1兲cot 共3兲 and up-to-date forms, for perfectly plastic materials, typically for
materials that follow an associated flow rule. The perfectly plastic
This outcome means that the expressions that we have histori- material has a single regardless of the strain level, which elimi-
cally used for Nc and Nq are exact. nates the possibility of simulating the progressive mobilization of
If there is no surcharge and if the soil is purely frictional 共i.e., different friction angles along the slip mechanism. Progressive
c = 0兲, then the following equation is used: failure should therefore be handled at the design stage, when se-
1 lecting a proper method to arrive at a value to use with the
qbL = ␥BN␥ 共4兲 bearing-capacity equation. The assumption of an associated flow
2
rule deviates from the reality of sands, which follow a nonasso-
It is not widely recognized that N␥ as expressed by the ciated flow rule. The implication of this deviation is the possibil-
equation ity of overpredicting the limit bearing capacity of real sands by
using the bearing-capacity equation if its terms are developed on
N␥ = 1.5共Nq − 1兲tan 共5兲 the basis of the assumption of an associated flow rule.
due to Brinch-Hansen 共1970兲 is nearly exact for up 35° 共Sal- In summary, we know that Brinch-Hansen’s formulation for
gado 2006兲. A superior solution for N␥, rigorous up to values of N␥ is correct for up to 35° 共and we know that there is a newer
that far exceed practical needs, is given by Martin 共2005兲. Sal- solution in Martin 2005 that is correct for all values that we
gado 共2006兲 proposes the following equation to fit that solution: might need兲, and we know that Vesic’s is not; but the use of all
other modification factors that go into the bearing-capacity equa-
N␥ = 共Nq − 1兲tan共1.32兲 共6兲 tion obscures the problem considerably. Additionally, we agree
which is, for all practical purposes, exact for values observed in with Décourt’s assertion that the bearing-capacity formula in its
the entire range of practical problems engineers will face on Earth current form has considerable limitations but believe that
共and, indeed, in low-gravity environments as well!兲. Rodríguez-Roa is correct in wishing to retain the concept of the
Although the bearing-capacity equation is exact for strip foot- bearing-capacity equation as a tool for the geotechnical and foun-
ings on the surface of either weightless soil or soil with both dation engineer. In that regard, in the near future, geotechnical
nonzero weight and c = q0 = 0, when we combine these two sepa- engineers will have access to a version of the bearing-capacity
rate results in a single equation, as traditionally done, we lose equation that will be nearly exact 共or at least whose level of
rigor. When we then try to account for the fact that foundations uncertainty will be small and known precisely兲 and that will thus
come in all shapes 共square, circular, rectangular, etc.兲, are embed- be extremely useful to them.
ded in the ground 共and embedment cannot be modeled by using
an equivalent surcharge兲, may have inclined bases, may be sub- Proper Definition of Ultimate Load
jected to inclined loads, and may be installed in sloping ground,
all hope is lost for a solution that retains rigor and has an uncer- Some confusion exists in the foundations’ literature—and is par-
tainty that we know and can control, unless modern analytical ticularly obvious in the case of pile foundations—regarding the
tools are used 共as done by Salgado et al. 2004 for clays and meaning of the term ultimate load and regarding what exactly is
Lyamin et al. 2006 for sands兲. The correction factors in current calculated by the bearing-capacity equation. Let us start by ad-
use in practice to account for all these deviations from the original dressing the term ultimate load. This term is intimately related to
analyses of the pioneers of plasticity theory are semiempirical and another term, ultimate limit state. An ultimate load is a load that,
can be improved upon. once applied to the foundation, will cause it to move in ways that
So, although we understand Rodriguez-Roa’s desire to attempt will lead to an ultimate limit state. An ultimate limit state, in turn,
to use experimental results to gauge whether the Vesic 共1973兲 or is a potentially dangerous condition, typically associated with
Brinch-Hansen 共1970兲 factor is best, we believe that the future damage to the superstructure. Structural damage is likely to fol-
final solution to the bearing-capacity problem will follow from low if a footing completely fails 共plunges兲 under an applied load,
first establishing the proper analytical basis, which fortunately is which obviously happens if and only if the applied load is equal
ments. A catch-all criterion could therefore be excessively conser- cient knowledge to adjust footing sizes and have different bearing
vative for many structures. Nonetheless, a quick calculation may pressures if needed to minimize differential settlement 共although,
aid in assessing what a reasonable criterion would be. obviously, zero differential settlement is unachievable兲, it can be
Let us consider a case with footings with B of the order of 2 m helpful. It is easier to achieve in the context of Brazilian practice,
spaced at approximately 5 m 共a short span, to be conservative兲 in which the geotechnical engineer functions more completely as
and a structure that would be damaged by an angular distortion ␣ a foundation engineer, sizing also the footings, than in North
of the order of 1 / 150. Let us assume further that the differential American practice, where there has traditionally been a divide
settlement between the two footings would be 75% 共again, on the between the geotechnical and structural engineering components
high and thus conservative side兲 of the total settlement. Let us see of foundation design. Although some change has occurred re-
what this criterion produces in terms of angular distortion if cently in North American practice, the geotechnical engineer’s
s / B = 0.1 is used as an ultimate load criterion work has traditionally ended at supplying an “allowable bearing
pressure” to the structural engineer, who then does the footing
0.75 ⫻ 0.1 ⫻ 2 m 0.15 m 1 sizing. Another impediment to implementing this approach is eco-
␣= = ⬇ 共7兲 nomics. Ultimately, if the perceived economic advantage of the
5m 5m 33 approach 共which requires more time from both the structural en-
Even the 0.1B criterion corresponds to too large a settlement to gineer and the geotechnical engineer兲 is small, it will not be
be a catch-all criterion, so anything greater that 0.1B is certainly adopted. Last, optimization of differential settlements at the de-
excessive. Again, an ultimate load, by definition, is associated sign stage should not be confused with eliminating differential
with an ultimate limit state, a state associated with potential struc- settlements, for the source of these differential settlements is
tural damage. In other words, if adjacent footings are subjected to mostly the uncertainties in the loads and in the soil-foundation
loads that would produce settlements of 10% of B in each footing systems.
and if the difference of these settlements 共resulting not only from
any difference in footing width, but, more important, from the
spatial variability of the soil兲 divided by the span is equal to or Calculation of Settlements and Possible Existence
less than the 1 / 150 required to cause structural damage, then that of a Unique, Nondimensionalized Load-Settlement
definition would be appropriate. We have shown that it is not Curve
appropriate by developing an example in which an s = 0.1B crite-
rion would clearly lead to an angular distortion approximately 5 Décourt argued the uniqueness of the load-settlement curves for
times greater than 1 / 150. Note also that reducing the angular footings so long as load and settlement are normalized as follows:
distortion to acceptable levels is not an appropriate use of the qb is divided by qb,ult = qb,10% 共the unit load corresponding to 0.1B
factor of safety. The factor of safety is supposed to account for relative settlement兲 and settlement s is divided by B. In addition
uncertainties in load and resistance; it reduces an appropriately to the difficulty of estimating values of qb,ult 共Décourt suggests
defined ultimate load by a sufficient amount so that events that that it is roughly equal to 10% of the standard penetration test
would cause the ultimate load to be less than the applied load are 共SPT兲 blow count兲, there is a conceptual shortcoming of this ap-
sufficiently unlikely. The function of the factor of safety is to proach: the fact that we are normalizing with respect to something
address probabilistic processes, not physical ones. Last, neither that does not mean the same thing in different settings. For ex-
the ultimate load nor the factor of safety is supposed to address ample, for very loose sands, qb,10% ⬇ qbL; but for dense sands, it
serviceability limit states, which should be checked independently may be much less than qbL. For true normalization of qb, one
of ultimate limit states. Serviceability would involve comparing should therefore use qbL 共or, as an alternative, a quantity com-
the angular distortion with an already reduced tolerable angular pletely independent of the footing, such as cone resistance qc兲,
distortion 共typically taken as 1 / 500兲. Although the factor of safety which has the same meaning no matter what conditions are in
is occasionally used to indirectly account for serviceability limit place. Normalization with respect to qc does not lead to a unique
states, this use is inappropriate, as pointed out by Becker 共1996兲. curve, as shown in Fig. 7 of the paper.
It does not appear helpful, therefore, to define ultimate loads Let us explore what happens if we normalize the unit load
for footings as a fraction of B, since the purpose is to associate versus settlement curve with respect to qbL instead of with respect
them with ultimate limit states, unless much smaller numbers— to qb,10%. For that, consider a fixed value of B 共so, the same
possibly in the 0.02B to 0.05B range—are used. Additionally, the footing兲 resting on soils with different DR. Since B is always the
large variations in footing size and spans would lead a single, same, qbL increases with DR because increases. As we go from
catch-all criterion to be too conservative in some cases. Proper a relative density DR1 to a greater relative density DR2, the origi-
interpretation of a load test on a footing should be to establish the nal curve therefore tends to translate downward because of the
limit load 共if deemed useful and allowed by the reaction system兲 increase in bearing-capacity and then leftward because of the re-
and to define settlements that would be associated with ultimate duction in settlement with the greater stiffness of a higher-DR
and serviceability states for the structure under consideration. A sand. For the curve to be unique, the leftward move because of
mination of suitable parameters to use in limit-bearing capacity which we discussed here in response to points raised by the dis-
and settlement computations. As previously indicated, the limit cussers. Providing charts that would be of some assistance in the
bearing-capacity equation was developed for perfectly plastic design of footings that are based on CPT results was one of the
soils following an associated flow rule. Such a soil would, upon objectives of the paper.
failure, develop the same friction angle throughout. In reality, we
must deal with the following in arriving at suitable values of to References
plug in the bearing-capacity equation:
• Soil is not perfectly plastic. Dilative soils 共far more common Becker, D. E. 共1996兲. “Eighteenth Canadian Geotechnical Colloquium:
than contractive soils in practice兲 will have a peak friction Limit states design for foundations. Part I: An overview of the foun-
angle, which is followed by a gradual drop until a critical-state dation design process.” Can. Geotech. J., 33共6兲, 956–983.
friction angle is reached at very large strains. Additionally, this Bolton, M. 共1982兲. “A guide to soil mechanics.” M. D. & K. Bolton.
peak friction angle depends on DR and on the mean effective Brinch Hansen, J. 共1970兲. “A revised and extended formula for bearing
stress at failure. capacity.” Bulletin No. 28, Danish Geotechnical Institute, Copen-
• Sandy soils do not follow an associated but rather a non- hagen, Denmark.
associated flow rule, with the result that there is less constraint Lyamin, A. V., Salgado, R., Sloan, S. W., and Prezzi, M. 共2006兲. “Three-
on the slip mechanism in reality than is assumed in deriving dimensional bearing capacity of footings in sand.” 具http://www.ecn.
the bearing-capacity equation. purdue.edu/~rodrigo/papers/典.
Martin, C. M. 共2005兲. “Exact bearing capacity calculations using the
• When we assume perfect plasticity with a set value of , we
method of characteristics.” Proc. 11th Int. Conf., IACMAG, Turin, Vol.
also ignore the effects of the loading path on the value of . In
4, 441–450.
reality, we know that the s in triaxial compression, triaxial
Salgado, R. 共2006兲. The engineering of foundations. McGraw-Hill.
extension, plane-strain compression, simple-shear loading, and Salgado, R., Lyamin, A., Sloan, S., and Yu, H. S. 共2004兲. “Two- and
other loading paths are all different and that the stress paths three-dimensional bearing capacity of footings in clay.” Geotech-
followed by elements on different portions of the slip mecha- nique, 54共5兲, 297–306.
nism are different. Vesic, A. S. 共1973兲. “Analysis of ultimate loads of shallow foundations.”
Although we currently have the knowledge to deal with every J. Soil Mech. and Found. Div., 99共1兲, 45–73.