Finite

222
Chapter 6:
FINITE ELEMENT MODELLING OF THE
SEISMIC BEHAVIOUR OF DEEP
FOUNDATIONS
6.1 Introduction
For foundations built in seismic areas, the demands made to sustain load
and deformation during an earthquake will probably be the most severe in their
design life. It is therefore surprising that the reported number of foundation failures
during destructive earthquakes is so few. There are two main hypotheses that
attempt to explain the successful performance of foundations during seismic events.
The first is the positive rate effects associated with rapid cyclic loading which may
lead to a foundation having a higher capacity under dynamic loads than under static
loads. The second is the general over design of foundations to reduce settlement.
Due to the encouraging performance of foundations during seismic events, research
efforts have now moved away from trying to avoid foundation failures and towards
attempting to make financial savings with regard to their design. The aim of this
chapter is to use dynamic finite element analyses to investigate the behaviour of a
deep foundation subjected to seismic loading and to compare the results with those
predicted by the currently available design methods.
The first part of this chapter summarises the current understanding of the
behaviour of foundations during seismic events. The performance of foundations in
two seismic zones is presented, the first from Mexico City where the soil is mainly
clay and the second in Kobe, Japan where the soil is mainly loosely compacted fill.
The second part of the chapter looks at the current methods available for the
seismic design of foundations. The theory and assumptions that form the basis of
these methods are also presented and their limitations discussed.
223
The third part of the chapter presents the results of pseudo static finite element
analyses of a deep foundation first presented by Potts (2000) and in part reanalysed
for this thesis. The aim of this study is to check the underlying assumptions of the
limit equilibrium method that is used in the design of deep foundations subjected to
seismic loading.
The fourth part of the chapter presents the results of dynamic finite element
analyses of a deep foundation that use actual earthquake records for their input. The
results of this study are then compared qualitatively and quantitatively with the limit
equilibrium and pseudo static analyses.
6.2 The Behaviour of Foundations Subjected to Seismic Loads
6.2.1 Foundations in Clay
The lack of understanding of the fundamental seismic behaviour of deep
foundations led to a number of failures during the Mexico City earthquake of 1985.
The city is particularly vulnerable to damaging earthquakes due to it being built on a
thick clayey deposit with low shear strength and high compressibility. On
September 19
th
1985 a magnitude 8.1 M
g
earthquake struck Mexico City, in which
between 330 and 757 (depending on the source) buildings were seriously damaged
(Mendoza and Auvinet (1988)).
The buildings that were worst affected during the earthquake were those founded
on frictional piles and raft foundations. Several buildings on raft foundations
exhibited very large non-uniform settlements, which ultimately led to tilting of the
structures (Mendoza and Auvinet (1988)). One six storey building was reported to
have experienced 0.92 meters of earthquake induced settlement, despite having a
static factor of safety of 2.0. The cause of the excessive earthquake induced
settlement in this case was thought to be the over stressing of the ground beneath
the raft foundation under static loads. The stresses already present in the ground
due to the static loading caused large permanent deformations to develop when the
earthquake occurred. The passage of the seismic waves through the soil induced
large shear stresses which resulted in large permanent displacements. Buildings
224
founded on frictional piles performed worst of all. Mendoza and Auvinet (1988)
reported that 13.5% of all 9 to 12 story buildings, most of them founded on friction
piles, were severely damaged. Sudden differential settlements were reported, with
consequential tilting.
The study by Mendoza & Auvinet (1988) highlighted how significant the static
factor of safety is during dynamic loading. They reported that most of the instances
of ill behaviour of building foundations in Mexico City during the September 19,
1985 earthquake were close to yielding under static loads, which promoted the
generation of permanent deformations under the earthquake induced high cyclic
shear stress increments. End bearing piles and footing type foundations (i.e. strip
and raft footings) were reported to have been the most successful during the
Mexico City earthquake.
To better understand the seismic behaviour of frictional pile foundations during
seismic events, the foundations of a bridge pier in Mexico City were fully
instrumented (Mendoza et al. (2000a)). The foundations consisted of 77 square
reinforced concrete driven piles which were capped by a concrete raft. The
instrumentation consisted of thirteen pile load cells, eight soil-slab-contact pressure
cells, six piezometers, three sets of triaxial accelerometers and three vertical arrays of
magnetic extensometers extending down to a depth of 60 meters. The aim of the
exercise was to evaluate how the pile and raft system shared the static load during
and after construction and how this distribution was affected by an earthquake.
When long term conditions were reached, it was found that 85% of the static load
was taken by the piles and the remaining 15% by the slab. After the bridge was
opened to traffic the effects of two earthquakes were recorded at the bridge
foundation. The first hit on the 11
th
of January 1997 and had a magnitude M
e
= 7.3
and the second struck on the 19
th
July 1997 and had a magnitude M
a
= 6.3. Whilst
they came from different sources and had quite different magnitudes, their effects
were very similar and are summarised below. Due to their similarity only the larger
event is discussed:
Effect on pile loads:
225
The Fourier response spectra of the pile-box system variables (acceleration
and pore pressures) were very similar to that of the free field response,
implying that there is negligible dynamic pile-soil interaction for these
events.
The measured load on the piles initially dropped due to some load, caused
by a small vertical settlement being temporarily transferred to the slab.
The initial static load for the example pile given by Mendoza et al. (2000a)
was 530 kN. The maximum cyclic amplitude caused by the earthquake was
24.5 kN.
The load on the pile after the seismic event was unchanged, implying there
was no degradation of the soil-pile interface.
Effect on contact pressure:
At first the contact pressure went up due to the vertical displacement of the
foundation mentioned previously.
During the event the contact pressure increased, implying that the piles and
the slab were sharing the additional cyclic loading.
After some point the pile load stopped increasing, but the slab pressure
continued to rise. Mendoza et al (2000a) attributed this to some yielding of
the pile and any further increase in load being taken by the slab.
As mentioned previously, the pore pressure response oscillated in phase
with the foundation-soil system, but no residual pore pressure remained.
No settlement data was available for the two earthquakes reported by Mendoza et al.
(2000a) although measurements were presented by Mendoza et al. (2000b) for
another earthquake that occurred in June 1999 at the same site. The settlement
along one axis of the foundation is shown in Figure 6.1.
226

Figure 6.1: Settlements on the Impulsora Bridge foundation (after Mendoza et al.
(2000b))
The data indicates that the foundations were still settling under static loads,
although the earthquake clearly caused an accelerated movement of some 20
millimetres.
To over come the lack of instrumented piles subjected to earthquake loading some
research has focused on recreating the effect of earthquake loading on deep
foundations by using scale models in centrifuge machines (for example
Maheetheran (1990)). This work has concentrated on the frequency response of the
foundation and not the fundamental behaviour of the foundation during the
earthquake and is therefore not relevant to the work presented in this thesis.
Due to the lack of instrumented piles of the type mentioned previously, most
investigations into the seismic capacity of deep foundations have concentrated on
full scale tests subjected to cyclic loads. However, it is only possible to apply
relatively low frequency axial loading and this type of artificial scenario cannot
recreate the complex in-situ stress state induced during seismic events.
Despite the number of tests performed of this type, the effect of cyclic loading on
the capacity of a deep foundation is still not fully understood. Briaud and Felio
(1986) found that below some threshold load, continued cycling has no significant
influence on pile capacity. The value of this threshold loading was given as between
227
70 and 80% of the ultimate pile capacity for one way loading and 40% for two way
loading. It is generally accepted that for loads above the threshold value the action
of cyclic loading has two contradictory effects. The first is the beneficial increase in
undrained strength and soil stiffness due to the rapid application of the loading.
Increases in the ultimate capacity of a deep foundation of between 5% and 20% of
the static capacity have been reported for each order of magnitude increase in the
loading rate (Dunnavant et al. (1990)). Kraft et al. (1981) and Jaime et al. (1991)
reported an increase in pile-soil stiffness of between 10 and 25% per ten fold
increase of loading rate. This feature of soil behaviour could be captured by a soil
model in which the stiffness and/or strength parameters are related to strain rate,
although no such soil model is available in ICFEP at present. The second effect is
the degradation of the pile-load transfer mechanism due to a reduction of radial
stresses, the destruction of the soil particle bonds, or the generation of residual
failure surfaces (Saldivar (2002)).
It appears from full scale experience, mainly from off-shore piles, that of these two
effects, the beneficial gain in undrained strength and increase in stiffness induced by
the rapid application of the load is the most significant. Therefore generally the
design of piles to sustain cyclic loading in clays is not explicitly considered, provided
the piles possess adequate factors of safety for static loading (Jardine (1991), Focht
and ONeill (1985) and Grosh & Reese (1980)).
6.2.2 Foundations in Residual Soils and Sands
Extra considerations must be made when designing foundations to sustain
cyclic loading in sands. Generally, friction piles are avoided in this situation and end
bearing piles that extend down to a stiffer stratum at depth are used in preference.
In the case of sand, the strain rate effect mentioned previously for clay has very little
effect and there is the added danger of liquefaction. As mentioned previously,
during the seismic events recorded at the instrumented bridge pier in Mexico City
clay no pore pressures accumulated. The same result has been found for other piles
subjected to cyclic loading in clays (Chow (1997), McAnoy et al. (1982) and Puech &
Jezequel (1980)). Conversely, laboratory tests on clay have shown the accumulation
of pore pressures when subjected to cyclic loading (Van Eekelen and Potts (1978)).
228
The accumulation of pore pressures in clay is, however, not an issue when designing
foundation for seismic loading in clays because the rate of increase of pore
pressures in clay is very slow and is therefore not an issue in the time scale of an
earthquake. For foundations built in drained material, for example sand, the
accumulation of pore water pressures reduces the effective stress and may cause the
soil to liquefy. In this state the soil has zero shear strength and any foundation built
in the material will fail catastrophically. The most famous example of this type of
failure occurred during the Niigata earthquake in 1964. Figure 6.2 shows a
photograph of a residential building for which the foundations have failed due to
the liquefaction of the surrounding ground.

Figure 6.2: Failure of building foundations due to liquefaction during Niigata
earthquake, Japan 1964 (after Zienkiewicz et al. (1999))
Further evidence of this kind of failure was apparent during the devastating
earthquake that struck Kobe, Japan on the 17
th
of January 1995. The areas that
suffered most damaged were those around the coastline that were reclaimed after
World War II. The material used to reclaim these areas was either decomposed
Granite or crushed sandstone from the Rokko mountains. Many of the buildings in
this area, built since the revisions to the Japanese buildings codes in 1974 and 1988,
did not exhibit any significant foundation damage. However, in some cases the
ground settlement caused by widespread liquefaction induced a large gap to form
between the new height of the street and the base of many engineered buildings
constructed on end bearing piles, which otherwise suffered no damage. Evidence of
this type of failure is shown in Figure 6.3.
229

Figure 6.3: Large gap induced by liquefaction-induced ground settlement in Port
Island, Kobe (after Tokimatsu et al. (1996))
Buildings founded on frictional piles or spread foundations experienced large
settlements, often accompanied by severe tilt, due to bearing capacity failure of their
foundations induced by soil liquefaction. Tokimatsu et al. (1996) gave several
examples of the magnitude of settlement that was induced by liquefaction around
piled foundations. These included a three storey building in Wadamiya-dori, Hyogo-
ku, which settled about 1.1 meters and tilted considerably; and a two story building
in Uosakihama that subsided more than 1.5 meters. However, not all foundation
damage was attributable to liquefaction. Some buildings which were founded in
material that did not liquefy still experienced significant amounts of settlement and
tilting. Examples of this include a building in Tamon-dori, Chuo-ku and the Uosaki
Junior High School in Uosaki that settled and tilted by 3 (Tokimatsu et al (1996)).
6.2.3 Summary
The effects of earthquake loading on deep foundations can be summarised
as follows.
The passage of the seismic wave through the soil surrounding the
foundation induces inertial forces that change the stress conditions in the
ground.
The shaking of the ground upon which the structure is built causes it to
rock back and forth on its foundations. This motion causes a cyclic loading
on the foundation which oscillates around the permanent static load.
230
The high frequency at which earthquake loading occurs changes some of the
fundamental properties of the soil (for example, the rate effects on stiffness
and strength).
The foundation failures mentioned previously in both Mexico City and Kobe can be
attributed to poor design and non-compliance with latest design codes. In Kobe this
was restricted to non-engineered buildings and those constructed before the latest
design codes were implemented. In Mexico City the worst damage occurred in
those buildings that were poorly designed for static loads. Due to the foundations
being overstressed, the cyclic shear stresses induced by the earthquake created large
permanent strains. In both events the most severely affected buildings were those
founded on frictional piles and, except for a few extreme cases, most did not fail by
creating a one sided slip surface, but rather settled, in some cases by more than 1
meter. Due to the heterogeneous nature of the building design and/or underlying
soil, the buildings that settled considerably also tended to tilt.
The generation of excess pore water pressures during cyclic loading does not appear
to be a problem for clays and dense sands. For loose fill material of the type used in
the reclamation of the land around the port of Kobe, liquefaction can pose massive
problems for foundations. The complete loss of bearing capacity as the soil losses
its shear strength can cause large settlements and in the most extreme cases the
foundation may fail catastrophically.
6.3 Current Design Practice for Foundations Subjected to
Seismic and Cyclic Loads
As mentioned previously, the two principal effects that an earthquake has
on a foundation are (i) the stress changes in the ground caused by the passage of the
seismic wave through the ground, and (ii) the second is the cyclic loading above the
permanent static loading caused by the shaking of the structure above. Generally,
the interaction between these two features is ignored and each is dealt with
separately.
231
6.3.1 The Design of Foundations for Cyclic Loading
The additional cyclic loading caused by the inertial forces from the structure
is generally only taken into account if either: (i) the transient component of the load
(Q
cyclic
) is equal to or higher than 30% of the permanent static load (Q
av
) or (ii) if the
permanent load (Q
av
) plus the cyclic load (Q
cyclic
) is greater than 90% of the static
capacity (Q
max.static
) for one way loading, or 75% in the case of two way loading
(Focht and ONeill (1985)). This is due to the widely held belief that the dynamic
capacity of a foundation is higher than the static capacity, for the reasons outlined in
Section 6.2. For large projects, like off-shore foundations, so called interaction
diagrams of the type shown in Figure 6.4 have been developed.

Figure 6.4: Cyclic interaction diagram from Haga tension pile tests (after Jardine,
(1991)).
These diagrams are based on full scale field tests (Karlsrud and Haugen (1985)) or
cyclic response analyses (Jardine (1991)) and are only applicable to the situation for
which they were developed. They can be used to determine how many cycles (N)
would cause failure for a given loading scenario or what level of cyclic loading
would be acceptable for a given number of cycles. An example of how such a
diagram was derived for piles in Mexico City clay can be found in Saldivar (2002).
6.3.2 The Seismic Design of Foundations
The static load bearing capacity of a foundation constructed in a frictional
material, with an angle of shearing resistance and cohesion intercept c, cannot be
232
found exactly. Over the years many analyses of the problem have been presented in
the literature and as early as 1973 Vesi listed 15 different solutions (Bowles (1988)).
It is therefore no surprise that an exact solution has not been found for the bearing
capacity of a foundation under seismic loading.
The first notable solution for a shallow foundation was proposed by Terzaghi in
1943. Following the classic work on plasticity by Prandtl, Terzaghi calculated the
capacity of a shallow foundation by assuming the failure mechanism shown on the
left hand side of Figure 6.5 and applying a modified version of the bearing capacity
theory developed by Prandtl (ca. 1920).

Figure 6.5: General footing-soil interaction for bearing capacity equations for strip
footings (after Bowles (1988))
The resultant capacity is given by Equation 6.1.
' ' 0.5
ult c q
q c N q N BN
= + + (6.1)
where N
is the bearing capacity factor for self weight

N
q
is the bearing capacity factor for surcharge
N
c
is the bearing capacity factor for cohesion
is the effective unit weight of the soil
q is the effective surcharge
B is the width of the foundation
The bearing capacity factors for cohesion and surcharge are given by Equations 6.2
and 6.3 respectively. The values for N
were given in graphical form although

Terzaghi never made it clear how they were calculated.
233

( )
( )
( )
2
0.75 2 tan
2
2cos 45 2
q
e
N

=
+
(6.2)

( )
1 cot
c q
N N = (6.3)
The framework proposed by Terzaghi was adopted by Meyerhof (1951, 1963) to
include the effects of foundation shape, foundation depth and load inclination. The
failure surface assumed by Meyerhof is shown on the right hand side of Figure 6.5.
The bearing capacity of the foundation is given by Equation 6.4, which is of the
same form as Equation 6.1 but includes the correction factors for load inclination
(i
c
, i
q
and i
) and depth (d
i
, d
q
and d
).
' 0.5
ult c c c q q q
q cN d i q N d i BN d i

= + + (6.4)
Where
tan
q p
N e K

=

( )
1 cot
c q
N N =

( ) ( ) 1 tan 1.4
q
N N
=
1 0.2
c p
D
d K
B
= +
1 0.1
q p
D
d d K
B
= = +
1
90
c q
i i
| |
= =
|
\ .
D
D

2
1 i
| |
=
|
\ .
D
D

( )
2
tan 45 2
p
K = +
= angle of resultant loading without a sign.
The approach used by Meyerhof was utilised by Hansen (1970) and Vesi (1973,
1974) to reanalyse the bearing capacity calculation. For both studies, the resulting
values of the bearing capacity factors for cohesion and overburden where found to
234
be the same as Meyerhofs. The N
terms differed and are given by Equations 6.5

and 6.6 by Hansen and Vesi respectively.

( )
1.5 1 tan
q
N N
= (6.5)

( )
2 1 tan
q
N N
= + (6.6)
The correction factors (i
c
, i
q
and i
) for inclined loading differed from those found by

Meyerhof and are given by Equations 6.7, 6.8 and 6.9 in the study by Hansen and
Equations 6.7, 6.10 and 6.11 in that of Vesi.

1
1
q
c q
q
i
i i
N
(6.7)
( )
5
1 0.5 tan
q b
i m = (6.8)
( )
5
1 0.7 tan
b
i m
= (6.9)
( ) 1 tan
m
q b
i m = (6.10)
( )
1
1 tan
m
b
i m

+
= (6.11)
where ( ) ( ) tan tan
b b b
m Q Q c = +
( ) ( ) 2 1 m B L B L = + +

b
Q = vertical load
B and L are the dimensions of the foundation
is the angle of the resultant load to the vertical
The early work by Terzaghi, Meyerhof, Hansen and Vesi did not explicitly include
the effects of seismic loading, although its effects can be included indirectly by using
some of the correction factors. Consider the forces applied to a foundation under
seismic conditions shown in Figure 6.6.

235

Figure 6.6: A deep strip foundation subjected to a horizontal acceleration k.g (after
Potts (2000))
Where E is the vertical component of loading at ground level
E
H
and X
H
are the normal and shear forces on the shaft base
V
P
, F
P
and V
A
, F
A
are the normal and shear forces on the active and
passive sides of the shaft, respectively
W
F
is the self weight of the shaft
W is the self weight of the soil
A
H
is the area of the shaft base
F
S
is the skin resistance of the shaft, F
S
= F
P
+ F
A

M is the seismically induced moment
g is the acceleration due to gravity
k is the applied horizontal acceleration as a fraction of gravitational
acceleration
The moment applied to the foundation due to the inertia forces in the structure can
be included into the foundation analysis by reducing it to an equivalent eccentric
load. If the applied vertical load is E, then the moment can be represented by an
eccentric force with an offset of e=M/E. Meyerhof (1953) recommended that in the
case of eccentric loading the effective width of the foundation B should be reduced
according to Equation 6.12.
236
' 2 B B e = (6.12)
If the inertia forces in the ground (W.k) are ignored (as they often are in foundation
design (Chen (1997))) then the problem shown in Figure 6.6 reduces to that of a
foundation on sloping ground. The ground surface and the foundation load are now
inclined at an angle given by ( )
1
tan i k
= , where k is the horizontal acceleration as

a fraction of gravity. This is illustrated in Figure 6.7.
2
1 E k +
2
1 W k +
1
tan i k
=

Figure 6.7: The inclined loading condition (after Potts (2000))
The foundation bearing capacity can now be calculated by using the solutions of
Hansen or Vesi with the appropriate correction factors for load inclination.
The importance of the inertial forces in the soil on the bearing capacity was
investigated by Sarma and Iossifelis (1990) for the case of a surface foundation. The
earthquake acceleration was represented by a constant horizontal force which is the
same for the structure and the surrounding soil. As indicated in Figure 6.8, the
lateral inertial force from the structure is represented by a shear force acting at
foundation level. The bearing capacity factors are obtained using the method of
inclined slices (Sarma (1979)) which is commonly applied to slope stability
problems. The asymmetrical failure mechanism assumed is shown in Figure 6.8.
237

Figure 6.8: Geometry of critical slip surface for general case (after Sarma and
Iossifelis (1990))
The geometries of the failure wedges are unknown at the start of the analysis. The
most critical failure mechanism is obtained by varying the angles
1
,
2
,
3
and
4

until the minimum bearing capacity is found. The failure mechanism is rejected if
any inter-slice force violates the soils failure criterion. This is a significant and more
theoretically sound assumption than that of previous researchers. Although the
blocks are assumed to behave rigidly, the material they are composed of must satisfy
the same failure criterion as the forces between the blocks. The resultant bearing
capacity factors for a range of horizontal accelerations and angles of internal friction
are shown in Figure 6.9.

Figure 6.9: Bearing capacity factors as a function of horizontal acceleration and
angle of internal friction (after Sarma and Iossifelis (1990))
238
Sarma and Iossifelis found that for horizontal accelerations below 0.1g the
differences between the solutions for an inclined load found by Meyerhof (1953)
and the new method were not significant and could be accommodated within the
factor of safety. For accelerations above 0.1g the differences became large and the
inclined load assumption was non-conservative.
The results presented by Sarma and Iossifelis are only applicable to shallow
foundation (D<B). The work was extended by Chen (1997) for deep foundations
subjected to seismic loads and shallow foundations situated at the top of sloping
ground. Additional assumptions had to be made for the deep foundation problem
to determine the soil pressures acting on the active and passive sides of the
foundation. The assumed failure mechanism and applied forces are shown in Figure
6.10.

Figure 6.10: Failure mechanism of a deep strip foundation in horizontal ground
(after Potts (2000))
Chen assumed that the horizontal forces V
a
and V
p
acting on the active and passive
sides of the foundation where given by Equations 6.13 and 6.14.

2
0.5
a a
V D K = (6.13)

2
0.5
p p
V D K = (6.14)
where K
a
and K
p
are the active and passive coefficients of earth pressure respectively.
The actual values of these parameters cannot be calculated beforehand and
therefore maximum and minimum values were derived using the work by Okabe
239
(1924) and Mononobe and Matsuo (1929) on the seismic analysis of retaining walls.
The minimum value for K
a
is given by Equation 6.15 and the maximum value for K
p

is given by Equation 6.16.

( )
( )
( ) ( )
( )
2
2
cos cos
sin sin
cos cos 1
cos
a
i
K
i
i i
i

+
=
(
+ +
+ + (
+
(

(6.15)

( )
( )
( ) ( )
( )
2
2
cos cos
sin sin
cos cos 1
cos
p
i
K
i
i i
i

=
(
+
+ (
+
(

(6.16)
A parametric study was then undertaken to determine the influence this assumption
has on the calculated bearing capacity.
From the results of this parametric study, Chen concluded that the seismic bearing
capacity is strongly dependent on the assumed values of K
a
, K
p
and the angle of
shearing resistant between the soil and foundation on the active and passive sides,
a

and
p
respectively. As K
p
and
p
increased, the bearing capacity of the foundation
increased. Conversely, as K
a
and
a
increased the bearing capacity decreased.
Chen presented the results of his work in the form of seismic reduction factors that
can be used when calculating the bearing capacity of a deep foundation on level
ground subjected to an earthquake acceleration of k.g. An example of the reduction
factor for the N
term (i
e
) is shown in Figure 6.11 for a foundation with a D/B ratio
of 5. For comparison, the reduction factors for a surface foundation are also shown
in the figure.
240

Figure 6.11: Seismic reduction factor i
e
for a deep foundation for = 30, c = 0, D
= 5B (after Chen (1997))
As mentioned previously, it is clear that the results are heavily dependant on the
assumed values of K
a
and K
p
.
Potts (2000) used pseudo static finite element analysis to investigate the applicability
of the earth pressures assumed in Chens analysis. The term pseudo static applies to
static analyses that account for the earthquake induced inertia forces in the soil mass
by applying a body force horizontally. This arrangement is analogous to that
assumed in the limit equilibrium problem analysed by Chen. To give comparable
results, Potts analysed a 5 meter deep and 1 meter wide strip foundation in plane
strain assuming a Mohr-Coulomb failure criterion with the following material
properties.
E = Youngs modulus = 60000 kPa
v = Poissons ratio = 0.25
K
0
= earth pressure coefficient at rest = 1.0
= unit weight of soil = 19 kN/m
3

c = cohesion = 0 kPa
= angle of dilation = 0
The mesh used is shown in Figure 6.12.
241

Figure 6.12: Finite element mesh used in pseudo static analyses
To investigate the effect that (the angle of shearing resistance) and (the angle of
soil-shaft interface friction) have on the behaviour of the foundation, the problem
was analysed for 3 values of (20, 30 and 40) and for each , two values of
were chosen: = and =/2. One analysis was run for =0, however since this is
an unrealistic case no further such analyses were performed. For each combination
of and , the foundation was taken to a constant working load of 1000 kN. The
static capacity (found from displacement controlled analyses) and factors of safety at
working load for each foundation are given in Table 6.1.

Collapse Load
(kN)
Factor of Safety at
Working Load
20 20 1200 1.20
20 10 1100 1.10
30 30 3250 3.25
30 15 3000 3.00
30 0 2850 2.85
40 40 8250 8.25
40 20 7500 7.5
Table 6.1: Summary of bearing capacities and factors of safety for each combination
of and
242
Once the foundation had been taken to working load, the horizontal body force was
increased in increments of 0.01g until the foundation was seen to fail. This was
usually signified by the analysis failing to converge and/or excessive displacement of
the foundation. The value of horizontal acceleration as a fraction, k, of gravity, g, at
which failure occurred for each case is shown in Figure 6.13.

Figure 6.13: The effects of and on the limiting value of the seismic coefficient
(k) (after Potts (2000))
For the purpose of this thesis the case with = 30 and = 15 was reanalysed and
will form the basis of the discussion that follows. However, any conclusions that are
drawn from this analysis also apply to the other cases analysed by Potts (2000).
Figure 6.14 shows the vectors of incremental displacement at the last stable
increment of the analysis prior to failure and Figure 6.15 shows the degree of
strength mobilisation at failure for this case. The results are presented for the area
highlighted in Figure 6.12.
243

Figure 6.14: Vectors of incremental displacement at failure for =30 and = 15

Figure 6.15: Degree of strength mobilisation at failure for pseudo static analysis
The movement of the foundation shown in Figure 6.14 for the pseudo static
analysis is similar to that observed when a retaining wall rotates due to excessive
passive pressure. It is clear that the movement of the foundation is concentrated
near the surface and that no failure mechanism of the type assumed in the bearing
capacity analysis is present.
The plot shown in Figure 6.15 shows contours of stress level as a ratio of the shear
strength of the soil, making it possible to identify where failure has occurred (that is
the area enclosed by a contour with a value of 0.99). The way this ratio is calculated
is illustrated in Figure 6.16.
244

Figure 6.16: Calculation of degree of shear strength mobilisation (after Potts (2000))
Generally, the strength mobilisation on the passive side is higher than on the active
side, although the area that has yielded (enclosed by contour C) is only slightly
larger.
The horizontal stresses acting on the active and passive sides of the foundation at
failure are shown in Figure 6.17. For comparison also shown are the theoretical
minimum and maximum values given by the Mononobe-Okabe equations.

Figure 6.17: Active and passive pressures on foundation at failure ( = 30, = 15)
The pressures closely follow the limits given by the Mononobe-Okabe equations
and at no point exceed these limits. The same trend was also found in the remaining
cases presented by Potts (2000). Generally, the passive pressure was mobilised in all
cases, whilst the full active stress was not always reached.
245
Overall, the pseudo static analyses presented by Potts (2000) cast some doubt over
the limit equilibrium technique proposed by Chen (1997). The assumption of a
Prandtl type failure mechanism proved to be unrealistic for a deep foundation
subjected to a pseudo static horizontal acceleration. The failure mechanism was
found to depend on the factor of safety the foundation was operating at when taken
to working load. The mobilised values of K
a
and K
p
showed no clear pattern for all
cases, although they generally stayed within the limits specified by the Mononobe-
Okabe equations. Given the high level of dependence of Chens work on the values
assumed for K
a
and K
p
the finite element analysis demonstrates that the results
obtained using this method must be used with caution.
6.4 Summary
The behaviour of deep foundation under seismic loading is not yet fully
understood. The problem involves interaction between the seismic waves travelling
through the ground and the cyclic loading applied to the foundation from the
structure above. The rate at which these forces are applied can change some of the
fundamental properties of the soil (for example, the stiffness and strength of a soil).
From the case studies reviewed in this chapter, frictional piles suffer the most
damage during seismic events. Conversely, there are very few cases of reported
damage to buildings on footing type foundations.
The design of deep foundations subjected to seismic loading relies on the results
from limit equilibrium analyses. The most basic assumption is to ignore inertia
forces in the ground and treat the problem as an equivalent foundation with an
inclined load. More sophisticated design procedures where proposed by Sarma and
Iossifelis (1990) and Chen (1997) that include the inertia forces in the ground.
However, pseudo static finite element analysis by Potts (2000) demonstrated that
the failure mechanism assumed in these two studies may not be correct.
The pseudo static analysis approximates the earthquake induced inertia forces as a
constant horizontal body force applied throughout the mesh. This loading condition
then initiates a one sided failure mechanism. The number of one sided failure
mechanisms observed in the field is very few and can usually be explained by
246
building eccentricities. The most common form of damage sustained by
foundations during earthquakes is large amounts of settlement and in some cases,
tilting.
6.5 Finite Element Analyses
6.5.1 Introduction
The design procedures presented in the previous section for deep
foundations subjected to seismic loading are inconsistent with the behaviour
observed in the field. The assumption of a one sided failure mechanism appears to
be incorrect in the majority of cases and the limited number of recorded failures
implies that the majority of foundations may generally be over designed. Dynamic
finite element analyses will be used in this section to investigate the behaviour of
deep foundations under seismic conditions. The results obtained will then be used
to make an appraisal of the design methods presented in the previous section. To
make the analysis as realistic as possible the acceleration records from actual
earthquakes will be used and the elasto-plastic behaviour of the soil will be
modelled.
6.5.2 Description of analysis
To allow comparison to be made with the work by Chen (1997) and Potts
(2000), a foundation five meters deep and one meter wide will be analysed in plane
strain, using the same finite element mesh and material properties as in the pseudo
static analysis. As mentioned in Chapter 3, the choice of location for the romote
boundaries of the mesh poses extra problems for dynamic analyses. In addition to
making them far enough away so as not to affect the results in the areas of interest,
any boundary will also cause the reflection of waves back into the finite element
mesh. Placing the bottom boundary 20 meters from the ground surface implies the
presence of a very stiff stratum at this depth. Due to ICFEP not having any type of
absorbing boundary conditions this was the assumption made for this study. If the
mesh is not wide enough the stress bulb created by the loading of the foundation
will begin to be affected. For this study a 42 meter wide mesh was used and its
247
influence on the results investigated by performing an analysis with a wider mesh.
Vertical displacements were restricted on the bottom and side boundaries and the
earthquake record was applied to all nodes on the bottom and side boundaries of
the finite element mesh. This is illustrated in Figure 6.18.

Figure 6.18: Figure showing mesh and boundary conditions assumed in dynamic
analyses
Applying the acceleration to all nodes on the side and bottom boundaries of the
mesh is equivalent to the foundation and the surrounding material being contained
in a rigid box. Crewe et al. (1995) suggested that this assumption may affect the
overall response of the system by altering the shear stresses present at the boundary.
This was not felt to be an issue for this study as the remote lateral boundaries were
sufficiently far away so as not to affect the behaviour around the foundation. The
effect of the location of the lateral boundaries was investigated and is presented
later in the chapter. The material properties of the soil are the same as those used in
the pseudo static finite element analysis and are repeated below for completeness:
E = Youngs modulus = 60000 kPa
K
0
= earth pressure coefficient at rest = 1.0
= unit weight of soil = 19 kN/m
3

c = cohesion = 0 kPa
= angle of dilation = 0
248
The foundation is modelled as linear elastic with the following material properties:
E = Youngs modulus = 3010
6
kPa
= unit weight = 24 kN/m
3

To allow relative movement between the foundation and the surrounding soil,
interface elements were placed around this boundary. The material properties
assigned to these elements can have a large effect on the overall behaviour of the
analysis (see Potts and Zdravkovic (2001)). To determine the appropriate material
properties for the interface elements, a parametric study was undertaken. An angle
of shearing resistance of 30 was given to both the interface elements and the
surrounding soil for a range of interface element stiffness. The resultant load
displacement curves were then compared with that obtained from an analysis with
no interface elements. The results showed that for this arrangement the load-
displacement curves were not sensitive to the values given to the normal (K
n
) and
shear (K
s
) stiffness attributed to the interface elements provided their order of
magnitude was above 110
3
. For the remaining analyses the following values were
used.
K
s
= K
n
= 110
5
kN/m
3
The standard analysis used in the following discussion assumes the soil has an angle
of shearing resistance of 30 and an angle of interface friction between the
foundation and soil of 15 with a Mohr-Coulomb failure criterion. Although not
realistic, to allow direct comparison with the work of Chen (1997) and Potts (2000)
the presence of a pore fluid is not modelled.
6.6 Calculation of Foundation Capacity
To allow appraisal to be made of the currently available design methods, the
most popular will now be used to calculate the static and seismic capacity of the
standard foundation chosen for the finite element analysis.
249
6.6.1 Static Capacity
The solutions of Meyerhof, Hansen and Vesi presented earlier are now
used to determine the ultimate bearing capacity of the deep foundation described in
Section 6.5.2. The geometry and material properties of this foundation are
illustrated in Figure 6.19.

Figure 6.19: Arrangement of standard deep foundation
The bearing capacity factors calculated for each method and the resultant limit load
are summarised in Table 6.2. Due to there being no cohesion in the soil, no N
c

bearing capacity factors have been calculated.
Method N
q
N
d
q
d
Q
ult
Meyerhof 18.401 15.668 1.866 1.866 3540 kN
Hansen 18.401 15.070 1.793 1.0 3277 kN
Vesi 18.401 22.402 1.793 1.0 3347 kN
Table 6.2: Bearing capacity of standard foundation by different methods
Potts (2000) found the capacity of the same foundation using a displacement
controlled finite element analysis to be 3000 kN. However, when the problem was
re-analysed for this thesis with smaller displacement increments and a tighter
convergence criteria, the capacity was reassessed to be 3315 kN. Details of this
analysis will be given later in this chapter. The capacities found by the Hansen and
Vesi method agree remarkably well with the finite element analysis, whilst the
bearing capacity obtained using Meyerhofs method is higher.
250
6.6.2 Seismic Capacity
The simplest way to account for seismically induced forces is to reduce the
problem to that of a foundation subjected to an inclined load. The equivalent load is
inclined from the vertical by an angle () given by Equation 6.18.

1
tan k

= (6.18)
The pseudo static finite element analysis presented in Section 6.3.2 illustrated that
the foundation shown in Figure 6.19 when operating at a factor of safety against
static failure of 3.315 (i.e. Q
w
= 1000 kN with a reassessed bearing capacity of 3315
kN), would fail when subjected to a constant horizontal acceleration of 0.21g. This
result can be compared with the seismic bearing capacities calculated using the
inclined load reduction factors of Meyerhof, Hansen and Vesi for an angle of ( =
tan
-1
0.21) 11.86. The bearing capacity factors and reduced ultimate limit load for
each method are presented in Table 6.3.
Method N
q
N
i
q
i

d
q
d
Q
ult
Meyerhof 18.401 15.668 0.754 0.366 1.866 1.866 2561 kN
Hansen 18.401 15.070 0.574 0.452 1.793 1.0 1865 kN
Vesi 18.401 22.402 0.624 0.493 1.793 1.0 2061 kN
Table 6.3: Seismic bearing capacity of standard foundation by different methods
The reduced bearing capacities, calculated assuming the earthquake loading
condition can be replaced by an equivalent inclined load, are much higher than the
working load used in the pseudo static finite element analysis (1000 kN), for which
the foundation failed at a horizontal acceleration of 0.21g.
Replacing the seismically induced inertial forces in a foundation by an inclined
loading does not include the effect of inertia forces in the ground on the bearing
capacity. As discussed previously, the work by Chen (1997) extended that of Sarma
and Iossifelis (1990) to include this additional force. Chen only presented his results
for a limited range of K
a
and K
p
values, and as these are not known quantities for the
limit equilibrium analysis, the values that most closely resemble those found from
the pseudo static analysis will be assumed. The finite element results showed that
the values of K
a
and K
p
at failure were close to the minimum and maximum values
251
given by the Mononobe-Okabe equations. For the soil parameters in this case, the
minimum and maximum values of K
a
and K
p
are 0.206 and 3.945 respectively. Chen
presented no result for values of this order and therefore the closest will have to be
assumed (K
a
= 1.0 and K
p
= 1.5). The bearing capacity is given by Equation 6.19.
0.5
ult e
q Bi N

= (6.19)
where i
e
is the seismic reduction factor and the value of N
includes the effect of

surcharge and skin friction on the bearing capacity. Chens results give the value of
N
as 342.67 and i
e
as 0.6344, resulting in a seismic bearing capacity of 2065 kN.
The results from Chens work are very similar to those obtained from the inclined
load analysis. One would intuitively think that including the inertia forces in the
ground would result in a lower bearing capacity than that obtained by analysing the
inclined loading case. These results suggest the opposite, although Chens results
include the contribution to the overall capacity from the shaft which is not
considered in the other cases. The calculation of this contribution is very difficult as
it relies on the assumed values of K
a
and K
p
which are not known at failure.
The large difference between the seismic bearing capacities calculated using an
equivalent inclined load and those obtained from pseudo static finite element
analysis, implies that the seismically induced inertia forces in the ground significantly
reduce the bearing capacity of a foundation. However, the method proposed by
Chen (1997) which includes these effects in a limit equilibrium approach predicts a
virtually identical seismic bearing capacity to that obtained from the inclined load
analysis. This may be due to the unrealistic values assumed for K
a
and K
p
in the
analyses. As noted by Potts (2000), although the software was developed at Imperial
College, it has not been possible to locate the program and re-analyse the problem
with realistic values for K
a
and K
p
.
There is clearly a discrepancy between the pseudo static finite element analysis and
the results of the limit equilibrium approach. However, both techniques assume that
seismically induced forces can be represented by a constant horizontal body force
equal to the mass multiplied by the maximum horizontal acceleration in the
earthquake record. This is unrealistic as the maximum acceleration only occurs once
during the earthquake and has much lower values for the rest of the record. The
252
only way to account for the time dependant behaviour is to model the problem with
dynamic finite element analysis and to use a real earthquake record as the input.
6.6.3 Finite Element Analysis of the Static Bearing Capacity
As mentioned previously, Potts (2000) carried out a displacement controlled
analysis of the foundation shown in Figure 6.19 to determine its ultimate bearing
capacity. The problem was re-analysed for this thesis to gain a better understanding
of the foundations behaviour under static loading. The displacement controlled
analysis was conducted by applying an incremental displacement of 0.01 meters at
the top of the foundation and then summing the reaction forces at the nodes with
this prescribed displacement. Horizontal displacements were allowed for nodes
directly under the foundation, and therefore the foundation was assumed to be
smooth and rigid. The test was continued until the reaction forces had plateaued
and the foundation had failed. The division of the applied load between the shaft
and the base was determined by summing the nodal forces along these boundaries
and then subtracting the nodal forces due to the initial stress conditions. The total
capacity and individual components for the standard foundation are shown in
Figure 6.20.

Figure 6.20: Load-displacement curve for standard foundation
The ultimate bearing capacity is 3315 kN which agrees well with the bearing
capacity found by the limit equilibrium techniques presented in Section 6.6.1 and is
slightly higher than that found by Potts (2000). This difference is probably due to
253
the smaller displacement increments and tighter convergence criteria used in the
analysis presented here.
6.6.4 Working Load
A factor of safety of 2.7 against static bearing capacity failure has been
chosen for the foundation. This gives a working load of 1228 kN which is similar to
that used by Potts (2000) for the equivalent pseudo static analysis. The working load
was applied to the foundation statically over 100 relatively small load increments to
ensure that no residual errors were carried over to the dynamic part of the analysis.
The displacement of the foundation at working load is illustrated on Figure 6.20.
The point lies on the load displacement curve and confirms that the load controlled
and displacement controlled analyses are equivalent. The degree of strength
mobilisation for the area around the foundation highlighted in Figure 6.18 at
working load is illustrated in Figure 6.21.
A
B

Figure 6.21: Degree of strength mobilisation around foundation at working load
There is a small bulb of soil under the foundation in which the strength has been
fully mobilised. In the area around the shaft of the foundation all the adjacent soil
has yielded. This is due to the shaft resistance being fully mobilised at working load
which is also illustrated in Figure 6.20 where the shaft component has plateaued and
begins to decrease slightly. This reduction is due to a small decrease in horizontal
stress around the foundation which in turn reduces the available shaft resistance.
254
6.7 Description Input Earthquake (Athens 7/9/1999)
The record chosen for the dynamic finite element analysis of the standard
foundation was the Athens earthquake of 1999. The earthquake was of magnitude
5.6 M
s
and struck at 11.56 on the 7
th
of September 1999, with its epicentre located
20 km North-West of Athens. The East-West record from station Athens3, which is
located in the basement of a building founded on rock, with a shear wave velocity of
1530 m/s in the Kallithea district of Athens (Ambraseys et al. (2000)) was chosen
from the many available. The original acceleration record and Fourier spectrum are
shown in Figure 6.22 and Figure 6.23 respectively.

Figure 6.22: Original East-West component of Athens earthquake (7/9/1999)

Figure 6.23: Fourier spectra of original East-West component of Athens earthquake
(7/9/1999)
255
The original record has a dominant frequency component at 4.5 Hz, a maximum
acceleration amplitude of 3.01 m/s
2
and a time step of 1/100
th
of a second. It is
important at this point to compare the predominant frequency component of the
earthquake record with the natural frequency of the finite element mesh. Assuming
the presence of the foundation will not alter the overall dynamic behaviour of the
finite element mesh, the fundamental natural frequency (T
1
) of a soil column is
given by Equation 6.20 (Kramer (1996)).

1
4H
T
V
= (6.20)
where H is the height of the soil column and V is the shear wave velocity of the
soil. The height of the soil column in this case is 20m and the shear wave velocity is
111.3 m/s giving a fundamental period of 0.7187 seconds and a fundamental
frequency of 1.391 Hz.
The pseudo static finite element analysis by Potts (2000) predicted a limiting
horizontal acceleration of 0.21g (2.06 m/s
2
) for the standard foundation chosen for
this study. To compare the dynamic analysis with the pseudo static finite element
analysis, the Athens record will be scaled to a maximum acceleration of 0.2g.
The earthquake record shown in Figure 6.22 is in its raw form and therefore has not
been filtered. The recording, digitisation and processing of the acceleration record
introduces noise into the signal, particularly in the low and high frequency ranges
(Shyam Sunder and Connor (1982)). The noise measured in the intermediate range
is relatively less noticeable (Menu (1986)). The problem associated with using non-
filtered input data is illustrated in Figure 6.24. The graph shows the unfiltered
acceleration record integrated twice using the Newmark time scheme ( = 0.6025
and = 0.3025) to give the equivalent displacement time history.
256

Figure 6.24: Equivalent displacement history from original East-West component of
Athens earthquake (7/9/1999)
The noise in the low frequency range causes the displacement time history, which is
dominated by the low frequency components, to drift away from the zero
displacement axis. This is unrealistic as it is known the accelerometer does not
experience any permanent displacement after the earthquake event.
To overcome this problem a filter is used to remove the extreme low and high
frequency components without affecting the important intermediate range. The
most accurate form of filter is an elliptic band-pass filter of the type developed at
Imperial College by Menu (1986). The filter works by completely removing
frequency components above and below maximum and minimum frequency limits
specified by the user. For a particular range below the minimum and above the
maximum frequencies the components are removed according to a linear variation.
This is illustrated in Figure 6.25.

Figure 6.25: Frequency components passed by filter
257
An initial guess is made from the users previous experience regarding what lower
frequency limit should be imposed. The earthquake record is passed through the
filter and the resultant acceleration is integrated twice to give the equivalent
displacement time record. If this record drifts away from the zero displacement axis,
the lower limit is increased and the process repeated. If, however, the displacement
remains on the zero displacement axis, the lower limit can be reduced until the
point is reached where the lowest possible frequency limit is used whilst ensuring
the displacement record remains on the zero displacement axis. The lower
frequency limit for the Athens record which had been scaled to a maximum
acceleration of 0.2g, was found to be 0.25 Hz. A standard maximum frequency of 25
Hz is used for all earthquake records. The resultant acceleration record and
equivalent displacement time history are shown in Figure 6.26 and Figure 6.27
respectively.

Figure 6.26: Athens earthquake record scaled to 0.2g maximum acceleration and
filtered
258

Figure 6.27: Equivalent displacement time history from Athens earthquakes after
filtering
It is clear that the important characteristics of the acceleration record have not been
changed whilst the displacement record now behaves in a much more realistic
manner.
6.8 Results of Dynamic Finite Element Analysis
6.8.1 Introduction
The static analysis of the working load presented in Section 6.6 represents
the initial conditions for the dynamic finite element analysis and the filtered
earthquake record presented in Section 6.7 represents the analysis input. The
earthquake is applied to the foundation by specifying a horizontal acceleration at all
nodes along the base of the finite element mesh. Using the specified time step of
0.01 seconds and the Newmark scheme, this acceleration is converted into an
equivalent incremental displacement which is specified in the list of nodal
displacements. The inclusion of inertia forces in the overall equilibrium equations
allows the shear wave induced by this displacement to propagate through the finite
element mesh. This section of the chapter will give details of the effects that this
shear wave has on the foundation. The Newmark parameters = 0.6 and =
0.3025 have been chosen for this analysis to give an unconditionally stable time
259
discretisation scheme with some added numerical damping. The analysis which
contained almost 4000 increments took approximately five days to run.
6.8.2 Results
The effects of the earthquake upon the foundation will be divided into the
effects during the earthquake and the permanent changes that remain after the
earthquake has finished. All results plotted for the mesh are for the area around the
foundation highlighted in Figure 6.18.
6.8.2.1 During the Earthquake
The most obvious difference between the pseudo static and dynamic finite
element analyses is the manner in which inertia forces are included. The pseudo
static analysis assumed the inertia forces could be included by applying a constant
horizontal body force which is equal to the acceleration multiplied by the soils mass.
This assumes the inertia force is constant and affects all parts of the soil equally at
all times. The dynamic finite element analysis on the other hand takes into account
the transient nature of the seismic wave. This is demonstrated in the following
strength mobilisation plots from various times during the earthquake analysis.

Figure 6.28: Degree of strength mobilisation around foundation after 2.0 seconds
260



261
The first plot shows the strength mobilisation around the foundation after 2.0
seconds of the earthquake. The area of soil that has yielded has not changed
significantly from the condition at working load shown in Figure 6.21. This is due to
the earthquake motion not being significant enough at this point to induce
additional plastic behaviour in the ground around the foundation.
Figure 6.29 shows the strength mobilisation after 4.1 seconds of the earthquake.
The area of ground that has yielded has clearly increased on the right hand side of
the foundation and decreased on the left hand side. One increment later (shown in
Figure 6.30) the large area of yielded ground has switched to the left hand side and
the right hand side now has less plasticity. The cause of this significant amount of
yielding is the greater intensity of the earthquake shaking which is illustrated by
Figure 6.26. The acceleration record reveals that the most intense period of the
earthquake shaking occurs around 4 seconds into the earthquake. The shape of the
contours during the most intense period of the earthquake is similar to that shown
in Figure 6.15 for the pseudo static analysis, although clearly during the dynamic
analysis the active and passive sides of the foundation change continuously during
the earthquake. The final figure shows the degree of strength mobilisation after 35
seconds of the earthquake. The total area of soil that has yielded is now less than
before the earthquake was applied and from this point onwards does not change
from increment to increment. The total earthquake record is almost forty seconds
long, although significant amounts of plasticity are only induced during the most
intense periods between 3 and 5 seconds into the earthquake. This behaviour and
the switching of plasticity from one side of the foundation to the other cannot be
recreated in pseudo static analyses and indicates that applying a constant and
homogenous horizontal body force to represent the earthquake induced inertia
forces may be over conservative.
Figure 6.32 shows an expanded view of the displacement time history during the
most intense period of shaking from the Athens 0.2g earthquake used in this
analysis.
262

Figure 6.32: Expanded view of the total displacement time history for Athens 0.2g
record, highlighting important analysis increments
Figure 6.33 to Figure 6.37 show the vectors of incremental displacement at the
increments highlighted in Figure 6.32. These time steps have been chosen to
represent snap shots of the foundations behaviour during the complete analysis.
They include increments where the input record is passing through the zero
displacement axis, as well as increments where the displacement is a maximum, both
in the positive and negative directions.

Figure 6.33: Vectors of incremental displacement at time 3.41 seconds
263



264

As the direction of the displacement time history changes, the predominant
direction of the displacement vectors change accordingly. This is noticeable between
Figure 6.34 and Figure 6.35, where the displacement time history, shown in Figure
6.32, changes from a positive to a negative gradient. The vectors of incremental
displacement correspondingly change from pointing predominantly to the right
hand side to pointing predominantly to the left. Figure 6.33 to Figure 6.36
demonstrate the complex interaction that exists between the foundation and the
surrounding soil during an earthquake event, although at no point does a failure
mechanism develop that could justify using a Prandtl type failure mechanism to
determine the bearing capacity of a deep foundation subjected to seismic loading.
To understand the overall behaviour, the stress paths experienced by two soil
elements located at different positions around the foundation will be investigated.
Figure 6.38 shows the stress path for a Gauss point located at a depth of 2.75
meters and 0.1 meters to the left of the foundation. Also shown are the failure
envelopes for the interface and solid elements. The stress paths are plotted in terms
of s and t which are given by the following relationships:

1 2
'
2
s
+
= and
1 2
'
2
t

=
where
1
and
2
are the major and minor principle stresses respectively.
265

Figure 6.38: Stress path for element adjacent to foundation
The initial part of the line represents the stress path associated with the application
of the working load. The stress path is for a solid element, however its close
proximity to the interface elements means that its behaviour will be heavily
influenced by their failure criterion, which is different to the rest of the soil. The
first part of the stress path is linear, moving towards the soils failure criterion as the
foundation load increases. After the interface elements have yielded, the stress
regime changes and the stress path alters its direction and begins to hook backwards
until it reaches the soil failure criteria. From this point onwards the stress path
moves down the failure envelope until the working load is reached. For the first two
to three seconds of the earthquake the stress point does not move significantly from
its position at the end of the working load. During the most intense period of
shaking both the values of s and t reduce until the shaking intensity reduces and
both stabilise. The horizontal and vertical stresses during the earthquake are plotted
for the same Gauss point in Figure 6.39 and Figure 6.40 respectively.
266

Figure 6.39: Horizontal stress history for Gauss point adjacent to foundation

Figure 6.40: Vertical stress history for Gauss point adjacent to foundation
During the most intense period of shaking the horizontal stress reduces significantly
whilst for the vertical stress, which experiences a similar transient component, the
permanent reduction is only slight. When the shaking intensity has reduced both
stresses stabilise to a residual value.
Figure 6.41 shows the stress path of a Gauss point 0.07 meters below the centre line
of the foundation and the failure criteria for the solid elements.
267

Figure 6.41: Stress path for element below foundation
The first part of the stress path, which represent the foundation being taken to
working load, is linear and shows the element tending towards the failure envelope
as the load increases. When the working load is reached the stress path for this
particular Gauss point had not quite reached the failure envelope, although Figure
6.21 shows that most elements at the base of the foundation had yielded at working
load. As the earthquake was applied, in a similar fashion to the element adjacent to
the foundation, the stress state did not change significantly. During the most intense
period of the earthquake record both the stress invariants s and t increased and the
stress path moved towards and up the failure envelope. The horizontal and vertical
stresses for the same Gauss point are shown in Figure 6.42 and Figure 6.43
respectively.

Figure 6.42: Horizontal stress below foundation during earthquake
268

Figure 6.43: Vertical stress below foundation during earthquake
The horizontal and vertical stresses below the foundation follow a very similar
pattern of behaviour during the analysis. At the start of the analysis neither changes
significantly until the most intense part of the earthquake motion. Between 3 and 5
seconds into the earthquake record the horizontal and vertical stresses increase by
15 and 25% respectively, causing the element to yield as shown Figure 6.41. After
the intense period of the earthquake both stresses stabilise to residual values.
One of the most significant observations made from the pseudo static analyses
presented by Potts (2000) was the active and passive pressures acting on the
foundation at failure. The results showed that the passive pressure is fully mobilised
whilst the active stress is not. All of the results presented lied within the limits given
by the Mononobe-Okabe equations. These results can now be compared with those
obtained from the dynamic finite element analysis. The horizontal stress acting on
the right hand side of the foundation after 4.17 and 4.31 seconds of the earthquake
is shown in Figure 6.44 along with the theoretical active and passive stresses given
by the Mononobe-Okabe equations. These increments were chosen as they
correspond to the minimum and maximum horizontal stresses shown in Figure 6.39
and therefore should be equivalent to the most extreme passive and active cases.
269

Figure 6.44: Horizontal stress on side of foundation during earthquake
The minimum and maximum stresses given by the Mononobe-Okabe equations
depend upon the acceleration applied to the soil around the foundation. In the
pseudo static analysis the horizontal acceleration applied to the mesh could be used
to calculate the available active and passive stresses. For the dynamic analysis the
earthquake record that was applied to the base of the mesh was scaled to have an
equivalent maximum acceleration value. However, the peak acceleration
experienced during the analysis will be different throughout the mesh due to the
complex interaction of the seismic wave. It is therefore difficult to determine what
should be taken as the available active and passive stresses. Despite this, Figure 6.44
illustrates that the stresses acting on the foundation during the earthquake do not
change significantly and whilst they generally remain within the limits given by the
Mononobe-Okabe equations, they are very different to those shown in Figure 6.17
for the pseudo static analysis.
6.8.2.2 After the Earthquake
The stress paths presented in the previous section show a decrease in
horizontal stress adjacent to the foundation and an increase in horizontal and
vertical stress below the foundation due to the earthquake motion. To confirm this
trend for the whole mesh Figure 6.45 and Figure 6.46 show the sub-accumulated
horizontal and vertical stresses from the beginning to the end of the earthquake
analysis.
270

Figure 6.45: Areas of mesh showing an increase in horizontal stress

Figure 6.46: Areas of mesh showing a significant increase in vertical stress
Figure 6.45 shows a band of soil adjacent to the foundation in which the horizontal
stress has reduced due to the earthquake analysis. The strip of soil below the
foundation shows an increase in horizontal stress. Figure 6.46 shows a narrow strip
of soil directly below the foundation in which the vertical stress has increased. The
soil surrounding the foundation has not experienced a significant change in vertical
stress. These results confirm that the trends observed in the stress paths shown in
Section 6.8.2.1 are representative for the whole mesh. Whilst the previous two
figures demonstrate the areas in which permanent stress changes had occurred, they
do not give an indication by what amount they had changed. A quantitative
assessment can be made by comparing the horizontal stresses acting on the side of
the foundation and the vertical stresses acting on the base of the foundation at the
271
end of the earthquake, with those present at working load. The results are shown in
Figure 6.47 and Figure 6.48.

Figure 6.47: Horizontal stress distribution before and after earthquake

Figure 6.48: Vertical stress distribution before and after earthquake
Figure 6.47 shows a maximum reduction of 50.8 kPa at a depth 1.6 m which
represents 64% of the horizontal stress at working load. The largest increase shown
in Figure 6.48 occurs 0.6 meters across the foundation and represents an increase of
30% over the vertical stress present at working load.
To allow comparison with the pseudo static analysis presented in Section 6.3.2, the
displacement vectors accumulated due to the earthquake motion are shown in
Figure 6.49 for the dynamic analysis.
272

Figure 6.49: Displacement vectors due to earthquake motion
It is clear that no failure mechanism has developed and the predominant movement
of the foundation is a vertical settlement. The vertical displacement history of point
A shown in Figure 6.49 is plotted in Figure 6.50.

Figure 6.50: Vertical displacement history of point A
The initial displacement of 7.7 cm represents the vertical displacement of the
foundation at working load, which corresponds with the value shown in Figure 6.20.
The vertical displacement plot follows a similar pattern to the stress plots presented
in the previous section. The foundation only settles during the most intense period
of the earthquake, which is between the period of 3 and 5 seconds into the
earthquake record. The settlement stabilises very quickly after the most intense part
273
of the earthquake to a value of 13.9cm, which represents an 80% increase above the
displacement at working load.
6.8.3 Discussion
The principal effects of the earthquake on the behaviour of the foundation
can be summarised as follows:
The seismic wave has caused a reduction in horizontal stress around the
foundation and an increase in both horizontal and vertical stress below the
foundation.
The reduction in horizontal stress around the foundation has reduced the
load carrying capacity of the foundation shaft and consequently the base
capacity has been further mobilised to maintain equilibrium.
The vectors of incremental displacement during the earthquake show no
failure mechanism forming, however, they do illustrate the complex
interaction that occurs between the foundation and surrounding soil.
The displacement vectors accumulated during the period of the earthquake
around the foundation indicate large vertical displacements, but give no
indication of a failure mechanism.
The seismically induced shear stresses around the base of the foundation
have caused large plastic deformations to accumulate, resulting in
considerable foundation settlement.
The overall behaviour of the dynamic analysis is considerably different to that found
with the pseudo static analysis. The principal difference is the absence of a failure
mechanism in the dynamic analysis. The results of the pseudo static analysis
demonstrated that the assumption of a Prandtl type failure mechanism for limit
equilibrium calculations may not be realistic. The results presented by Potts (2000)
demonstrated that no consistent failure mechanism could be identified for all cases.
The results of the dynamic finite element analysis presented in Section 6.8.2.2
showed that the Athens earthquake of 1999 scaled to a maximum acceleration of
0.2g does not initiate any type of failure mechanism, but does however, cause the
foundation to settle a significant amount. This type of behaviour is more consistent
with that described for real events in Section 6.2. It was noted that the number of
274
bearing capacity type failures observed in the field are very few, whilst the most
common form of damage sustained during earthquakes is significant amounts of
settlement.
The reduction in horizontal stress caused by the earthquake, and the subsequent
reduction in shaft capacity illustrates why friction piles should be avoided in seismic
areas. The foundation used in this study is predominantly end bearing and therefore
did not fail when the load carried by the foundation shaft was transferred to the
base. If, however, a larger proportion of the working load where distributed to the
shaft, the results presented here demonstrate that a significant earthquake would
cause the load to transfer to the base and may cause catastrophic failure.
6.8.4 Analyses Using Wider Mesh
As mentioned in Chapter 3, the boundary conditions assumed in dynamic
finite element analyses introduce complications that are additional to those that
already exist with static analysis. At present ICFEP cannot model absorbing
boundary conditions and therefore all the analyses presented in this thesis assume
reflecting boundaries. With regard to the foundation analysis, the reflecting
boundary along the base of the mesh is equivalent to having a stiff rock at this
depth. This is a common scenario in reality and will be assumed for the foundation
analyses presented in this thesis. The boundary conditions on the side of the mesh
imply that an identical foundation is present on both sides of the original
foundation at a spacing of 42 meters. Therefore, because we are only interested in
the behaviour of a single foundation it is important to determine what effect the
location of the side boundaries has on the foundation analysis. For this purpose the
analysis using the Athens earthquake record scaled to a maximum acceleration of
0.2g was repeated using a wider mesh. The mesh shown in Figure 6.12 was used
with extra elements placed on both sides to make the total width of 82 meters. The
displacement history for point A is compared with the results from the standard
mesh in Figure 6.51.
275

Figure 6.51: Settlement history for point A subjected to 0.2g Athens earthquake
The settlement of the foundation at working load is slightly different, however for
the duration of the earthquake the difference between the two analyses did not
change. This implies that the side boundaries of the mesh are sufficiently far away
so that they do not influence the behaviour of the foundation significantly.
6.8.5 Analysis Including Structural Mass
The working load in the preceding analyses was reached by applying a
boundary stress on the top of the foundation. This was to isolate the effect of
inertia forces in the ground from the inertia forces in the building. This is of course
an unrealistic situation as the working load would in reality be applied by the
structure above the foundation, which would generate its own inertia forces when
an earthquake occurred. To investigate the influence that these inertia forces may
have on the behaviour of the foundation, the analysis was repeated with the working
load being applied by placing a concentrated mass on top of the foundation and
then gradually increasing the vertical body force in it until the correct working load
was achieved. This set up is illustrated in Figure 6.52.
276
Concentrated Mass
Representing Structure

Figure 6.52: Detail of mesh used in analysis including structural mass
The elements at the top of the foundation used to represent the concentrated mass
are extended above the ground surface to ensure that no moment is induced about
the top of the foundation. The displacement history of point A is compared with
the results from the analysis that did not include the structural mass in Figure 6.53.

Figure 6.53: Settlement history of point A subjected to Athens 0.2g earthquake with
and without structural mass
The result of including the structural mass in the analysis is a slightly higher
settlement at point A, although the difference is very small. Of course in reality the
structure will be much more complicated than the lumped mass used in this
analysis. The structure will vibrate in different modes and apply a complex
277
combination of forces and moments. This interaction problem must be investigated
further, although it is beyond the scope of this study where the primary interest lies
with the influence of seismically induced shear forces in the ground on the
behaviour of foundations.
It is generally accepted that whilst the Athens earthquake claimed the lives of 455
people (Pavlides et al. (2002)) it was not a particularly destructive earthquake. The
cause of the damage can be attributed to its close proximity to a major conurbation
rather than its energy content. In the following section the foundation analysis will
be repeated using the acceleration records from other earthquakes.
6.9 Dynamic Finite Element Analysis Results using Different
Input Earthquakes
6.9.1 Introduction
The analysis presented in the previous section used the acceleration record
from the Athens earthquake of 1999 scaled to a maximum value of 0.2g as its input.
The acceleration record was scaled to 0.2g because the pseudo static analysis of
Potts (2000), predicted that the standard foundation should fail when subjected to a
0.21g acceleration. The results demonstrated that that the foundation settled
considerably, but did not develop a failure mechanism. As mentioned previously,
the Athens earthquake of 1999 is not considered to be particularly destructive and
therefore the analysis was repeated using other earthquake records to investigate if
these can cause the foundation to fail. This study demonstrates the advantage of
dynamic finite element analysis, where the individual characteristics of each
earthquake record can be taken into account, over pseudo static analysis, in which
only a constant acceleration can be applied.
6.9.2 Description of Earthquake Records
The three additional earthquake records chosen for this study are described
in Table 6.4.
278
Name Country Station Component Date Magnitude
max
u
(m/s2)
f
min

(Hz)
Kocaeli Turkey Izmit E-W 17/8/99 Ms=7.8 2.193 0.5
Tabas Iran Dayhook E-W 16/9/78 Ms=7.33 3.317 0.475
Spitak Armenia Gukasian N-S 7/12/88 Ms=6.76 1.794 0.3
Table 6.4: Details of earthquake records used in study
Each record was scaled to a maximum acceleration of 0.2g and filtered in the same
fashion as described in Section 6.7 to eliminate the drift of the displacement time
history away from the zero displacement axis. The minimum frequency used in this
process for each record is given in the final column of Table 6.4. The filtered
acceleration and displacement records and the Fourier spectra of each earthquake
record are shown in the following figures.

Figure 6.54: Kocaeli earthquake acceleration record scaled to 0.2g
279

Figure 6.55: Kocaeli earthquake displacement record

Figure 6.56: Kocaeli earthquake Fourier spectrum

Figure 6.57: Tabas earthquake acceleration record scaled to 0.2g
280

Figure 6.58: Tabas earthquake displacement record

Figure 6.59: Tabas earthquake Fourier spectrum

Figure 6.60: Spitak earthquake record scaled to 0.2g
281

Figure 6.61: Spitak earthquake displacement record

Figure 6.62: Spitak earthquake Fourier spectrum
6.9.3 Results
Each analysis behaved in a similar fashion to the Athens earthquake. All
exhibited a reduction in horizontal stress around the foundation and an increase in
horizontal and vertical stress at the base of the foundation. However, the
earthquake induced settlement of the foundation varied considerably between each
earthquake. The settlement of point A shown in Figure 6.49 for each analysis is
shown in Figure 6.63.
282

Figure 6.63: Settlement of point A for different earthquakes scaled to 0.2g
All analyses started from the same displacement at working load (7.7cm), however
clearly the Kocaeli earthquake was the most damaging of all the records. The
earthquake induced settlement was 42.5cm which represents a 550% increase over
the displacement at working load. The rapid settlements observed in each plot
coincide with intense periods of shaking observed in their respective acceleration
records. For example, consider the settlement plot for the Kocaeli earthquake. The
graph shows three periods of rapid settlement. The first is between 3.0 and 5.0
seconds into the earthquake record and the second and third occur in the intervals
7.0 to 9.0 seconds and 38.0 to 41.0 seconds respectively. These periods of rapid
settlement can be correlated with periods of intense shaking identified in the
acceleration record shown in Figure 6.54. In between these periods the foundation
does not settle, implying that the foundation begins to settle when some threshold
acceleration is exceeded. The results for each analysis are summarised in Table 6.5.
Earthquake
Earthquake Induced
Settlement (cm)
Percentage Increase
Athens 6.2 80.5%
Spitak 9.6 124.7%
Tabas 20.6 267.5%
Kocaeli 42.5 551.9%
Table 6.5: Summary of results for 0.2g earthquakes
283
The results of this study clearly demonstrate that the peak ground acceleration,
which is commonly used in design codes, does not give an indication of the
destructive potential of an earthquake. The Athens earthquake of 1999 is clearly the
least destructive of all the records and produces 85% less settlement than the
Kocaeli earthquake, although they have the same peak ground acceleration.
To further investigate the behaviour of the foundation when subjected to
earthquakes with different energy content, the analysis was repeated with the four
earthquake records described previously scaled to different peak acceleration values.
The analyses performed are as follows:
Athens earthquake (7/9/1999): 0.1g, 0.2g, 0.3g, 0.4g, 0.5g
Kocaeli earthquake (17/8/1999): 0.05g, 0.1g, 0.125g, 0.175g, 0.2g, 0.3g
Tabas earthquake (16/9/1978): 0.1g, 0.166g, 0.2g, 0.25g, 0.3g, 0.4g
Spitak earthquake (7/12/1988): 0.1g, 0.2g, 0.3g, 0.4g, 0.5g
To ensure that the displacement record does not drift from the zero displacement
axis, each record was filtered using the same procedure as described in Section 6.7
and the same frequency limits described in Table 6.4. The earthquake induced
settlement for each analysis is plotted against their respective peak ground
accelerations in Figure 6.64.

Figure 6.64: Earthquake induced settlement for the four earthquakes scaled to
different peak ground accelerations
284
Of the 22 earthquakes analysed the most destructive of all was the Spitak record
scaled to a maximum acceleration of 0.5g. This earthquake caused 82.5cm of
settlement, however the Kocaeli earthquake scaled to 0.3g gave only slightly less
settlement (79.5 cm) and is therefore overall the most destructive record. The
results demonstrate that for each record the peak ground acceleration is related to
the settlement of the foundation, however, between records there is a large
discrepancy between records with similar peak acceleration values. The problem of
finding a single parameter that can characterise an earthquake record has been the
focus of attention for many seismologists since earthquakes have been studied.
Many parameters have been proposed that characterise the amplitude, frequency
content and duration of strong ground motion, although it is regarded as impossible
to combine all three in a single parameter (Jennings (1985) and Joyner and Boore
(1988)). The results of the study presented in this chapter demonstrate that the
maximum acceleration values from an earthquake record cannot be used to assess
the destructive potential of an earthquake. The strong motion parameter proposed
by Arias (1970) attempts to include the total duration of the earthquake motion and
a measure of the prolonged intensity of shaking by integrating the acceleration
squared over the entire earthquake motion. The Arias intensity (I
a
) is defined by
Equation 6.21.
( )
2
0
2
d
T
a
I a t dt
g
( =

(6.21)
Where g is the acceleration due to gravity and a(t) is the earthquake acceleration
record. The cumulative absolute velocity, given by Equation 6.22, has also been
proposed as a ground motion parameter and has been found to correlate well with
structural damage potential (Kramer (1996)):
( )
0
d
T
CAV a t dt =

(6.22)
where T
d
is the duration of the strong motion. The main difficulty in assessing
different strong motion parameters is the lack of earthquake records close to
recorded damage. The results of the finite element analyses presented in this chapter
will now be used to try and correlate structural damage (in this case foundation
settlement) with strong motion parameters calculated from the input earthquake
285
records. The following parameters will be used in this investigation; acceleration
squared (AS), velocity squared (VS), displacement squared (DS), absolute acceleration (AA),
absolute velocity (AV) and absolute displacement (AD), which are defined by the following
equations:
( )
2
0
d
T
AS a t dt ( =

(6.23)
( )
2
0
d
T
VS v t dt ( =

(6.24)
( )
2
0
d
T
DS d t dt ( =

(6.25)
( )
0
d
T
AA a t dt =

(6.26)
( )
0
d
T
AV v t dt =

(6.27)
( )
0
d
T
AD d t dt =

(6.28)
where v(t) and d(t) are the velocity and displacement time histories of the earthquake
record respectively found by integrating the filtered acceleration records using the
Newmark scheme ( = 0.6, = 0.3025). The earthquake induced settlement is
plotted against each parameter for each earthquake record in the following figures.
286

Figure 6.65: AS parameter against earthquake induced settlement

Figure 6.66: VS parameter against earthquake induced settlement

Figure 6.67: DS parameter against earthquake induced settlement
287

Figure 6.68: AA parameter against earthquake induced settlement

Figure 6.69: AV parameter against earthquake induced settlement

Figure 6.70: AD parameter against earthquake induced settlement
288
The most successful of all these parameters at normalising the earthquake induced
settlement is the AA parameter, which is equivalent to the cumulative absolute
velocity defined previously. For AA values below 8.0 m/s*s the parameter
normalises the settlements well, although for more damaging earthquakes the lines
appear to be diverging. Further investigation of this is clearly needed, although it is
important to note that the parameter AS, which is linearly equivalent to the Arias
intensity, did not perform a great deal better than the peak ground acceleration at
normalising the earthquake induced settlement.
The most damaging of all the earthquake records used for this study was the Spitak
record scaled to a maximum acceleration of 0.5g. Figure 6.71 shows an enlarged
view of the displacement time history for this record.

Figure 6.71: Expanded view of the displacement time history for Spitak 0.5g record,
highlighting important analysis increments
The incremental displacement vectors for the two time increments shown in Figure
6.71 and the final earthquake induced displacement vectors for this analysis are
shown in Figure 6.72 to Figure 6.74.
289

Figure 6.72: Vectors of incremental displacement at time = 12.43 seconds


Figure 6.74: Earthquake induced displacement vectors for Spitak 0.5g earthquake
290
The vectors of incremental displacement shown in Figure 6.72 and Figure 6.73 do
not show any failure mechanism forming during the analysis. The displacement
vectors shown in Figure 6.74 illustrate the settlement of the foundation due to the
application of the earthquake motion. This again illustrates that seismically induced
forces in the ground to not initiate a bearing capacity type failure, although they do
cause the foundation to settle a significant amount.
6.10 Analyses on Foundations at Different Working Loads
6.10.1 Introduction
The results presented in Section 6.9.3 highlighted the importance of
including the individual characteristics of each earthquake record in a foundation
analysis. The results also illustrated that some of the traditional parameters for
defining an earthquake record, namely peak ground acceleration and Arias intensity,
do not correlate well with potential earthquake damage.
Potts (2000) demonstrated using pseudo static finite element analysis that the factor
of safety (FOS) a foundation is operating at dictates the failure mechanism that
develops when the horizontal forces are applied to the foundation. To investigate
what effect the factor of safety has on the behaviour of the foundation when an
earthquake is applied, the analysis of the foundation subjected to the 0.3g Athens
earthquake was repeated with the standard foundation taken to different working
loads.
6.10.2 Description of Analyses
The working loads and respective factors of safety that were analysed are
summarised in Table 6.6. The factors of safety are calculated assuming the bearing
capacity of the foundation is 3315kN, which was found from the original finite
element analysis.
291

Analysis Working Load (kN) Factor of Safety
1 3013.0 1.1
2 2550.0 1.3
3 1657.5 2.0
4 1228.0 2.7
5 975.0 3.4
6 473.6 7.0
Table 6.6: Summary of analyses undertaken at different working loads
The load-displacement curve of the foundation is shown in Figure 6.75, with the
displacement at the different working loads illustrated.

Figure 6.75: Load-displacement curve with different working loads illustrated
The displacement history of point A is shown in Figure 6.76 for each analysis. Due
to each analysis starting from a different displacement at working load, only the
earthquake induced settlement is shown. After the first 20 seconds of the
earthquake record no further settlement is observed and therefore only the first 20
seconds of the displacement curves are shown.
292

Figure 6.76: Settlement history for the standard foundation at different working
loads
As one would intuitively expect, the foundation which is closer to failure at working
load experiences more earthquake induced settlement than the foundation which
has a lower working load. Whilst the shape of the settlement curve remains
approximately the same and the rapid settlement starts at approximately the same
time, the foundation with a lower factor of safety experiences significant amounts of
settlement for a longer period of time. This implies that when a foundation is closer
to failure, the threshold acceleration that causes the foundation to settle is lower.
This is a logical result as the plastic strains around the foundation at higher working
loads are more wide spread and hence it would take less seismically induced shear
strains to create additional plastic deformations. The earthquake induced settlement
against static factor of safety is shown in Figure 6.77 with a line of best fit through
the data points.
293

Figure 6.77: Earthquake induced settlement for Athens 0.3g record with different
working loads
As the factor of safety approaches unity, the earthquake induced settlement tends to
infinity. The line of best fit is given by Equation 6.29:

1.0431
0.161( 1) d FOS

= (6.29)
where d is the earthquake induced settlement for the 0.3g Athens record. This
relationship can now be used to predict the earthquake induced settlement that the
standard foundation would experience if the limit equilibrium techniques described
in Section 6.6 had been used to design it for seismic loading. The magnitude of
reduced bearing capacity for the standard foundation subjected to a 0.3g earthquake
are given in Table 6.7, using the methods proposed by Meyerhof, Hansen and Vesi.
Method N
q
N
i
q
i

d
q
d
q
ult
Meyerhof 18.401 15.668 0.663 0.197 1.866 1.866 2218
Hansen 18.401 15.070 0.444 0.308 1.793 1.000 1435
Vesi 18.401 22.402 0.490 0.343 1.793 1.000 1609
Table 6.7: Seismic bearing capacity of standard foundation subjected to 0.3g
earthquake
BS 8004 (1986) suggests that a relatively high factor of safety, of 2-3, should be used
when calculating acceptable loads for foundations to account for uncertainties in
soil conditions and to ensure settlement does not become excessive. For the
294
purpose of this study a factor of safety of 2.5 will be used. The resultant overall
factor safety and earthquake induced settlements predicted by Equation 6.29 are
summarised in Table 6.8.
Method q
ult
Working Load (kN) Overall FOS Settlement (cm)
Meyerhof 2218 887.36 3.74 5.635
Hansen 1435 573.89 5.78 3.151
Vesi 1609 643.51 5.15 3.647
Table 6.8: Predicted earthquake induced settlement for a foundation designed using
limit equilibrium methods
The result of using the limit equilibrium techniques is a reduced ultimate bearing
capacity and in the event of an earthquake the lower stresses in the ground causes
the foundation to settle less. For the example given here the earthquake induced
settlement is around 3-6 cm, which would not cause serious damage to the building
constructed on the foundations. The results presented here only relate to the Athens
earthquake of 1999 scaled to a maximum acceleration of 0.3g. If a more destructive
earthquake, like the Kocaeli event were to happen, the results shown in Section 6.9
demonstrate the potential settlement that could occur.
Therefore, it can be concluded that whilst the finite element results demonstrate
that the approach of using a limit equilibrium technique for calculating the bearing
capacity of a foundation subjected to seismic loading is incorrect, the result is
conservative because reducing the working load reduces that amount of settlement
the foundation experiences in the event of an earthquake.
6.11 Large Displacement Analyses
6.11.1 Introduction
The analyses presented in the previous sections assumed small displacement
finite element theory, which implies the deformation of the mesh remains small
during the analysis. This is equivalent to a total Lagrangian frame of reference which
is illustrated in Figure 6.78.
295
x
y
1
2
1
2
Initial undeformed
configuration
Current unknown
configuration
v u

Figure 6.78: Total Lagrangian system (after Izzuddin (1991))
Clearly, the displacements induced in the foundation by some of the more
destructive earthquakes are large and may invalidate this assumption. To investigate
the effect that this assumption has on the results of the foundation analysis, each
one was repeated using large displacement theory. The fundamental assumption of
small displacement finite element theory is that any displacement of the mesh
during the analysis is small compared to the dimensions of the mesh and therefore
the mesh can be assumed not to change during the analysis. Large displacement
analyses which use an updated Lagrangian system redefine the mesh at the end of
each increment according to the calculated displacements. This is illustrated in
Figure 6.79.
x
y
1
2 1
2
Initial undeformed
configuration
Current unknown
configuration
v
u
Last known
configuration

Figure 6.79: Updated Lagrangian system (after Izzuddin (1991))
296
Large displacements can also be accommodated by adopting an Eulerian reference
frame, in which the material moves through the finite element mesh. This approach
is more commonly adopted for fluid mechanics problems.
This option is not available in ICFEP and therefore large displacements will be
accommodated by employing an updated Lagrangian system. Using the new mesh
configuration the stiffness matrix can then be recalculated and the next analysis
increment performed on the new mesh. In addition to the stiffness matrix, the mass
and damping matrices must also be recalculated for each increment during dynamic
large displacement analyses. The Jaumann stress tensor and velocity strain tensor
have been implemented into ICFEP to deal with the constantly changing geometry
of the finite element mesh. This method is commonly preferred to the 2
nd
Piola-
Kirchoff stress tensor for non-linear materials with path dependant constitutive
models. The stresses in the material can be calculated directly from the incremental
strains when using the 2
nd
Piola-Kirchoff stress tensor, whilst the Jaumann stress
state is related to the strain rate and therefore the incremental strains must be
obtained via an integration process (Bathe (1982)). This implies that the material
non-linearity is more accurately accounted for by the Jaumann stress tensor.
6.11.2 Description of Analyses
In theory, the ultimate bearing capacity of the standard foundation should
be different when analysed using large displacement analyses, due to the increasing
overburden pressure as the foundation is pushed into the ground. This, however,
will not be taken into account for these analyses and the same working load of 1228
kN will be used for the large displacement analyses. The displacement of the
foundation at working load was practically the same for the large displacement
analysis and the small displacement results (7.63mm and 7.66mm respectively),
which is to be expected as the soil stiffness does not vary with depth.
The displacement history of point A from the large displacement analysis when
subjected to the Spitak earthquake record scaled to a maximum acceleration of 0.5g
is compared with equivalent small displacement analysis in Figure 6.80.
297

Figure 6.80: Displacement history of point A for Spitak 0.5g earthquake for large
and small displacement analyses
The two plots follow a very similar path, with the large settlements occurring at the
same points in the earthquake record, although the large displacement analysis
predicts significantly less earthquake induced settlement (57.6cm compared to
82.5cm for the small displacement analysis).
The vectors of incremental displacement for the time steps highlighted in Figure
6.71 are shown in Figure 6.81 and Figure 6.82 for the large displacement analysis.

298

Figure 6.83 shows an enlarged view of the original and final mesh configurations
around the foundation.

Figure 6.83: Final mesh configuration around the foundation for large displacement
analysis of the Spitak 0.5g earthquake
The amount by which the foundation has settled is clearly demonstrated by the final
mesh configuration. The displacement vectors around the foundation due to the
application of the earthquake are shown in Figure 6.84.
299

Figure 6.84: Displacement vectors around foundation for Spitak 0.5g earthquake
The vectors of incremental displacement and the mode of settlement are similar to
those observed from the small displacement analysis and clearly no failure
mechanism has been induced by the application of the earthquake.
The earthquake induced settlement for each of the Athens, Kocaeli, Tabas and
Spitak earthquake records described in Sections 6.7 and 6.9 are compared for large
and small displacement analyses in the following figures

Figure 6.85: Earthquake induced settlement for Athens earthquake
300

Figure 6.86: Earthquake induced settlement for Kocaeli earthquake

Figure 6.87: Earthquake induced settlement for Tabas earthquake

Figure 6.88: Earthquake induced settlement for Spitak earthquake
301
For all cases the large displacement analyses predicts less earthquake induced
settlement than the equivalent small displacement analysis. As the earthquake
becomes more destructive, the difference between the two analyses becomes
greater. The reduced earthquake induced settlement observed in the large
displacement analysis is due to the increased normal stresses in the ground at greater
depths. As the foundation settles, the horizontal and vertical stresses increase as the
amount soil above the base of the foundation increases. This has the effect of
moving the stress path away from the failure envelope and therefore greater shear
stresses must be applied to induce permanent plastic deformations. This is
illustrated in Figure 6.89.

Figure 6.89: Figure illustrating the different amounts of shear stress required to
induce plastic strains at different foundation depths
This phenomenon is demonstrated in Figure 6.80 where the amount of settlement
predicted by the large displacement analysis initially follows the small displacement
analysis, but then the rate at which settlement occurs reduces, resulting in less
overall settlement. The foundation in the small displacement analysis does not
experience the increase in normal stresses due to settlement because the mesh stays
constant during the analysis and therefore the foundation does not change its
location with respect to the mesh.
6.11.3 Summary
The results presented in this section for large displacement analyses have
demonstrated the importance that large displacements can have on the behaviour of
foundations subjected to seismic forces. The application of small displacement finite
302
element theory to geotechnical problems is normally valid, because the
displacements experienced in problems like embankments and tunnels are often
small compared to the size of the problem. To take a foundation to failure often
requires applying large displacements, for example the 0.9m of displacement used in
the capacity analysis of the foundation presented in Section 6.6.3. However, the
displacements at working load are usually small and the assumption of small strains
is normally valid. The examples of the behaviour of foundations during seismic
events presented in Section 6.2 demonstrated that under these conditions the
displacements are often large and the finite element results presented in the sections
that followed reinforced this fact. It is therefore important to model seismic events
which may induce large movements with large displacement theory.
6.12 Summary
The results of the dynamic finite element results presented in this chapter
predict some important features of the observed seismic behaviour of foundations
that other analytical techniques, such as limit equilibrium and pseudo static finite
element analyses, fail to predict. The main differences between the currently
available analytical tools and dynamic finite element analysis are summarised below.
The use of a Prandtl type failure mechanism to predict the reduced bearing capacity
of a foundation when subjected to seismic loading has been shown to be unrealistic
by the results of the pseudo static finite element results first presented by Potts
(2000) and partially repeated for this thesis. The failure mechanism induced by the
application of a pseudo static horizontal force was found to depend on the static
factor of safety the foundation was working at before the application of the
horizontal force. However, none of the cases presented by Potts (2000) resembled a
Prandtl type failure mechanism and most where dominated by a lateral movement
of the foundation.
The results of the dynamic finite element analysis demonstrated that the pseudo
static analysis was overly conservative. The peak ground acceleration of an
earthquake only occurs once during an earthquake event and therefore to apply a
constant horizontal body force equivalent to this inertial loading is over
conservative. Four earthquake records (Athens (1999), Spitak (1988), Tabas (1978),
303
and Kocaeli (1999)) where scaled to a peak ground acceleration of 0.2g and applied
to a foundation that the pseudo static analysis predicted would fail at this
acceleration. No failure of the foundation was observed, although significant
amounts of settlement were predicted. The fact that each earthquake record,
although scaled to the same peak ground acceleration, predicted different amounts
of settlement illustrates that the magnitude of the peak acceleration is not a good
indication of earthquake damage potential.
An overall reduction in horizontal stress was observed around the shaft of the
foundation, which resulted in the load carried by the shaft being redistributed to the
base. This in turn caused an increase in vertical stress under the base of the
foundation.
The first set of analyses was undertaken at a constant working load which gave a
reasonable factor of safety against static failure. To investigate the influence that the
working load has on the behaviour of a foundation subjected to earthquake loading
the analysis of the foundation subjected to the Athens 0.3g earthquake was repeated
for different working loads. The results demonstrated that the practice of reducing
the static capacity of a foundation, and therefore increasing its factor of safety,
indirectly resulted in a more satisfactory seismic performance of the foundation.
The smaller shear strains induced around the foundation at lower working loads
result in less permanent plastic deformations when the earthquake loading was
applied.
As mentioned previously the peak ground acceleration of an earthquake record did
not correlate well with the amount of earthquake induced settlement. To investigate
if a parameter could be identified which could be used to predict potential
earthquake damage, the four earthquake records used previously (Athens (1999),
Kocaeli (1978), Spitak (1988) and Tabas (1978)) were scaled to a range of peak
ground acceleration and the foundation analyses repeated for one working load. A
single peak value, such as peak acceleration or peak velocity cannot be used to
differentiate between earthquake records and therefore a range of parameters were
used that integrated different values over the whole earthquake record. Of all the
parameters tested none satisfactorily normalised the foundation settlement results,
although the most accurate was the absolute velocity integrated over the entire
304
record. The commonly used Arias intensity was found not to perform any better
than the peak ground acceleration.
The standard finite element approximation of small displacement theory is not valid
for the magnitude of settlements observed in some of the more destructive
earthquakes presented. To investigate the influence that large displacements may
have on the behaviour of the foundation, all the analyses were repeated using large
displacement analysis. The general behaviour of the foundation was found to be the
same as the small displacement analysis. No failure mechanism was developed for
any of the earthquake records, although large settlements were still predicted.
However, the settlement predicted by the large displacement analysis was always less
than that predicted by the small displacement analysis. As the earthquake record
became more destructive, the difference became greater. This was found to be due
to the normal stresses acting on the foundation increasing in the large displacement
analysis as the foundations depth increased. In the small displacement analysis the
foundations depth did not change with respect to the mesh and therefore the
normal stresses did not change as the foundation settled. At present the density for
each element remains constant during the analysis. This implies that the mass of an
element will vary during the analysis as its geometry changes. This assumption may
affect the results as the elements compressed under the foundation during the
application of the earthquake will effectively reduce their mass, and hence their
contribution to the inertial forces.
Of course any conclusions drawn from the analyses presented here are limited by
the assumptions made. The first and most obvious assumption is the non-inclusion
of the building structure in the analysis. As mentioned previously, the cyclic loading
caused by the rocking of the structure above will significantly influence the
behaviour of the foundation during a seismic event. However, the effects of loading
rate on soil behaviour cannot at present be replicated by any soil model in ICFEP
and therefore the influence of the cyclic loading would only be negative. It was
therefore decided not to include the structure above the foundation in the analysis
and to therefore isolate the influence of the seismic shear wave on the foundation
from the cyclic loading from the structure above.
To allow comparison of the results with the pseudo static analysis presented by
Potts (2000) and the limit equilibrium work by Chen (1997), the presence of a fluid
305
in the pores of the soil has not been included in the analysis. This is of course
unrealistic and would have a major influence of the overall behaviour of the
foundation. The analysis presented here is equivalent to a foundation resting on a
dry sand which is not a common scenario. Future research should move away from
this ideal situation which is a requirement for limit equilibrium analyses and towards
modelling the pore pressure build up during a seismic event and the possibility of
liquefaction.
The use of a Mohr-Coulomb failure criterion for the analyses was necessary to allow
comparison of the results with those obtained by Potts (2000) and Chen (1997). As
mentioned in Chapter 4, the Mohr-Coulomb model is not ideally suited to dynamic
finite element analyses. The lack of plasticity below the yield surface and the soil
becoming elastic upon unloading are not realistic features of the dynamic behaviour
of soil which can be recreated by using sophisticated kinematic soil models of the
type developed by Al-Tabbaa and Wood (1989) and Stallebrass and Taylor (1997).
The behaviour of foundations subjected to earthquake loading should be
investigated using this type of model in the future. Despite these limitations the
finite element results presented have many similarities with the observed behaviour
of foundations described at the beginning of this chapter.

Finite

Uploaded by

Copyright:

Available Formats

Finite

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Finite

Uploaded by

Copyright:

Available Formats

222

is the bearing capacity factor for self weight

were given in graphical form although

terms differed and are given by Equations 6.5

) for inclined loading differed from those found by

= , where k is the horizontal acceleration as

includes the effect of

You might also like