Professional Documents
Culture Documents
Pseudo Static vs. Performance Based Seismic Bearing Capacity of Footings (Liquefaction)
Pseudo Static vs. Performance Based Seismic Bearing Capacity of Footings (Liquefaction)
Abstract
The effect of liquefaction on the post-shaking bearing capacity and settlement of strip footings is explored,
focusing on the case where a thin clay layer is overlying the liquefiable sand. Starting with a simple analytical
model of bearing capacity degradation developed earlier, both the use of a more realistic failure surface and a
more accurate simulation of the strength degradation in the liquefiable sand are explored and consequently
verified against pseudo-static numerical analysis. Furthermore, a parametric dynamic numerical analysis is
performed in order to study the coupled excess pore pressure-settlement response of the foundation and evaluate
the overall accuracy of the simplified analytical approach. Comparison of the analytical (pseudo-static) with the
numerical predictions shows that: (a) the former may provide reasonable estimates of the degraded bearing
capacity, however (b) the use of a reduced friction angle to model liquefaction-induced shear strength
degradation may prove overlay conservative, and (c) for common static factors of safety, settlements at the end
of shaking are an order of magnitude higher than initial static settlements and consequently they should form the
basis of design.
Clay H
Sand
uο+∆u
(a)
Fig. 2: Footing resting on liquefiable sand with a clay cap:
failure surface assumed by [3]
(a)
Sand
qu qu
Fig. 5: Comparison of analytically predicted failure surface in
W the clay cap with numerically computed strain increment
cuH cuH contours at failure.
cuL
W zero. To model this, the degraded in situ shear strength of
a sand τf was simply expressed as:
N
h h
τ f = (1 − U )σ VO tan φ + Uτ res
ι
(3)
qu΄
where, τres represents the residual shear strength of the
(b) liquefied sand. Among the various empirical correlations
which are available for the estimation of τres (e.g. [12],
Fig. 4: Footing resting on liquefiable sand with a clay cap: (a) [17], [18]) the one proposed by Stark and Mesri [18] was
Refined failure mechanism and (b) Limit equilibrium
used here. In average terms, this correlation can be
model
approximately written as:
This procedure proved adequately accurate, as it led ι
to deviation angles α fairly similar to these obtained τ res = 0.011Nl ,60σ VO (4)
numerically. This is shown in Fig. 5, where the
analytically computed refined failure surface in the clay where N1,60 is the corrected number of SPT blow count
cap is compared to the numerically predicted shear strain (e.g. [20]) and σ’vo is the initial vertical consolidation
rate contours at failure. Nevertheless, the more accurate stress.
failure mechanism had a rather minor effect on the There are two ways to introduce the above shear
degraded bearing capacity qULT,D. For instance, Fig. 6 strength degradation into conventional bearing capacity
shows a typical comparison between the degradation computations. The first is to define a reduced friction
factors ζ=qULT,D/ qULT,O derived from the basic and the angle φ*, in a way similar to that followed in the initial
refined failure mechanisms shown in Figs. 2 and 4 development of the method:
respectively. The observed differences are indeed
secondary and can be readily overlooked in view of other tan ϕ* = [(1 − U )tan ϕ + U tan ϕres ] (5)
more crucial problem uncertainties (e.g. the evaluation of
a representative excess pore pressure value for the whole where the value of the residual friction angle of the sand is
foundation area). given as:
EFFECT OF LIQUEFACTION-INDUCED SHEAR STRENGTH tan ϕ res = 0.011 N l ,60 (6)
DEGRADATION
The method of Cascone and Bouckovalas [3] assumes The second way is to define an apparent buoyant unit
the simplified degradation mechanism suggested by many weight γ*, in terms of the nominal buoyant unit weight γ of
seismic codes, namely that the friction angle of the sand sand and U:
degrades gradually with excess pore pressure build up,
according to Eq. 1, and finally diminishes to zero. γ * = γ ′(1 − U ) + Uγ res (7)
However, it is widely known that liquefiable soils attain a
residual strength even when effective stresses approach to with
0.011N l , 60 1
γ res = γ′ (8)
γ res
tan φ 0.8
=2
γ re s
.4 2
φ
=0
kN
Note that, using either Eqs. 5 & 6 or Eqs. 7 & 8 one 0.6 res =4 o
kN
/m
ζ
comes to essentially the same relation for the degraded φ
/m
3
3
insitu shear strength of the sand (Eq. 3). Still, the different 0.4 res =0 o
methods of simulating the degradation of shear strength
0.2
lead to considerably different values for the bearing
capacity degradation factor ζ. This is because bearing 0
capacity formulas are highly non-linear with respect to the 0 0.2 0.4 0.6 0.8 1
friction angle (φ or φ*), while they are linear with respect U
to the buoyant unit weight of the soil (γ or γ*).
To show the potential effect of the different Fig. 7: Effect of sand degradation model on analytically
computed correction factor ζ (c*=1.0, φ=30o, H/B=0.5).
assumptions regarding shear strength degradation, Fig. 7
compares the variation of ζ with U for a loose sand
A total of ten (10) cases were examined, for various initial
(φ=30ο, N1,60 = 7), computed (a) for a degraded friction
loading conditions corresponding to average contact
angle with a φres=0 or φres=4o , and (b) for a reduced
pressure between 0 and 130 kPa. In all analyses, seismic
buoyant unit weight with γres=0 or γres=2.42kN/m3.
shaking was applied under undrained conditions. On the
Observe that degrading the friction angle of sand is by
other hand, static loading was applied under undrained
far more conservative than reducing the buoyant unit
conditions for the clay cap and under drained conditions
weight. The difference between the two approaches is
for the underlying sand. The analyses were performed with
remarkably large and cannot be simply overlooked for the
the Finite Difference computer code FLAC [11]
benefit of increased safety. Furthermore, note that taking
appropriately extended to account for excess pore pressure
into account the post-liquefaction residual strength of sand
build up during cyclic loading. A uniform grid of 1000
leads to reasonably lower degradation of the bearing
square 1mx1m elements was used. The first 2 m of depth
capacity (i.e. greater values for factor ζ), regardless of the
consisted of clay with c=40KPa and φ=0ο, while the rest
basic approach that was used to simulate liquefaction
18m consisted of sand with c=0KPa and φ=30o. The
effects.
width of the foundation was B=4m, giving an H/B ratio of
COUPLED DYNAMIC ANALYSES OF BEARING CAPACITY 0.5. The base of the sand layer was excited with the
DEGRADATION AND SETTLEMENT acceleration time history shown in Fig. 8.
The soil element response was taken as elastic-
In order to evaluate the validity and the accuracy of perfectly plastic with failure described by the Mohr-
the pseudo-static approach, a series of numerical analyses Coulomb criterion. Pore pressure build-up during seismic
was performed, where: shaking was simulated via a coupled effective stress –
a) The foundation was first loaded statically, with an plastic volumetric strain analysis. Namely, the rate of
initial uniform vertical load less than the ultimate. plastic volumetric strain due to cyclic loading was defined
b) Consequently, seismic shaking was applied, under by the following empirical relation (based on [6]):
constant static load, and a 2-D field of excess pore
b
pressures was computed in the liquefiable sand layer. ε
c) Following the end of shaking, the foundation load a
∆ε vol = bAγ cyc vol (8)
was further increased until failure. Aγ cyc
a
1
0.6
amax=0.42g
0.8
0.4
ζ=q ULT,D /q ULT,O
0.6
0.2
a (g)
Displacement (m)
(sand with clay cap)
where Ceijkl is the elastic stiffness matrix of the soil qST=0 kPa
0.03
skeleton. Under constant volume (i.e. undrained
conditions) this effective stress change gives the increase
of pore pressure ∆u due to cyclic loading. The same 0.04 coupled
procedure has been successfully used Byrne [2] for the static & seismic
implementation of the empirical model of Martin et al. [8] analyses
in the finite difference code FLAC [11]. 0.05
Typical results for the vertical load-deformation
response of the footing, obtained from this numerical
0.06
experiment, are shown in Figs. 9 and 10. Namely, Fig. 9
shows typical pore pressure contours in the foundation qST=50 kPa
soil, at the end of shaking, while Fig. 10 shows the load 0.07
qST=90 kPa
(average contact pressure)-displacement curves for:
a) One case of static loading with initial (pre-
shaking) soil conditions and a continuous soil profile 0.08
consisting of clay. This condition is used as reference for
Fig. 10. Results from static and coupled static-seismic
the definition of the degradation factor ζ.
numerical analyses: typical load-displacement curves
b) One case of static loading with initial (pre- (for c*=1.0, φ=30o, H/B=0.5)
shaking) soil conditions and the actual soil profile
consisting of liquefiable sand with a clay cap. This this observation is disturbing as it suggests that there is no
condition is used as reference for the definition of the unique degradation factor ζ that can be defined in terms of
initial static factor of safety (F.S.), before the onset of the initial soil conditions, the foundation geometry and the
shaking. intensity of shaking. However, such an interpretation
c) Three typical cases of coupled static and seismic would prove premature, as the observed differences can be
loading, for uniform initial contact pressures qST=0, 50 adequately explained by the different values of the excess
and 90 kPa. pore pressure ratio U corresponding to each curve in Fig.
There are two main observations in this figure. First, 10.
note that the degraded bearing capacity of the footing To show this, Fig. 11 correlates the degradation
depends on the magnitude of the initial static load, i.e. the factors ζ obtained from all numerical analyses to the
design load of the foundation and the corresponding factor corresponding average values of the excess pore pressure
of safety against conventional static failure. At first glance
1
IIb
IIa
0.8 Ib
Ia
ζ (qult/qst,clay)
0
0 0.2 0.4 0.6 0.8 1
U (∆u/σ'vo)
Fig. 9: Pore pressure contours at the end of shaking for
qST=50kPa Fig. 11: Comparison between analytical and numerical
predictions of bearing capacity degradation factors ζ
ratios U at the end of shaking. The numerical predictions failure only, it is concluded that this branch of the load-
are compared to the ζ-U curves predicted analytically for displacement curves is attributed to instantaneous
the same soil and foundation conditions, assuming either foundation failure, each time ground accelerations exceed
degradation of the friction angle or reduction of the a critical level. This phenomenon, under drained soil
buoyant unit weight of the sand. In all cases, the average conditions, has been given detailed attention by Richard et
value of U was calculated on the basis of the numerically al. [16].
computed values along the respective failure surface (e.g. For the benefit of further insight, Fig. 12 shows the
see Fig. 9). Observe that the data points from the dynamic load displacements curves from the reference static
analyses define a more or less unique ζ-U relation which analysis (b), for uniform clay soil over sand, as well as
lays between the analytical predictions for degraded from all coupled static and seismic analyses. To highlight
friction angle and those for decreased buoyant unit weight the relative magnitude and the potential importance of the
of the sand. accumulated settlements, applied loads have been
The second observation of practical interest is that the normalized against the ultimate load for non-degraded
displacements at failure are considerably larger for the sand properties (q’ULT,O), while displacements have been
degraded soil conditions than for the reference static normalized against the corresponding diplacements at the
analyses for non-degraded soil. This difference is mostly onset of failure δ’ULT,O. Focusing upon the end of shaking,
due to the permanent settlements accumulated during denoted with a black star on the normalized load-
shaking (vertical branch of the load-displacement curves) displacement curves, it is realised that the corresponding
and increases rapidly with the magnitude of the initial displacements may become one order of magnitude higher
static load. Taking into account that seismic loading takes than those at conventional static failure while applied
place under undrained conditions, as well as that the static loads are still less than the degraded ultimate loads.
constitutive model used for the analyses allows for This is shown more clearly in Fig. 13, where the static
permanent shear and deviatoric strain accumulation upon factor of safety F.S.=q’ULT,O/qST is correlated to the
normalized displacements at three loading history instants:
q/qST,Cl-Sand the onset of shaking, the end of shaking and at post-
shaking failure. Observe that the normalised vertical load
0 0.2 0.4 0.6 0.8 1
-2
that will provide the same settlements at the end of
Static loading shaking as the the reference ultimate static loading ranges
(sand with clay cap) between 0.20 and 0.35, corresponding roughly to factors
0 of safety between 3.0 and 5.0. When the factor of safety is
7.1 reduced to about 2.0, liquefaction-induced settlements
4.2 3.5 3.0 increase by an order of magnitude.
2
In other words, there is a clear need to define a
performance based failure criterion of the foundation that
4 2.4 will depend upon settlements rather than on the bearing
capacity it self.
6 2.1
CONCLUSION
δ/δST,Cl-Sand
1.9
δULT REFERENCES
δ/δST,Cl-Sand