Sunteți pe pagina 1din 15

ARTICLE IN PRESS

Soil Dynamics and Earthquake Engineering 28 (2008) 577591


www.elsevier.com/locate/soildyn

Investigation of the dynamic response of rigid footings on tensionless


Winkler foundation
Ioannis N. Psycharis
Laboratory for Earthquake Engineering, School of Civil Engineering, National Technical University of Athens, Polytechnic Campus, 15780 Athens, Greece
Received 20 November 2006; received in revised form 23 July 2007; accepted 28 July 2007

Abstract
The dynamic response of a rigid footing resting on an elastic tensionless Winkler foundation is examined. A parametric investigation,
concerning the effect of the main parameters on the response, is performed for harmonic excitation. The parameters examined include the
stiffness and the damping of the foundation, the excitation frequency and the superstructure characteristics and loads. The maximum
rocking response, the minimum length of contact after uplift, the maximum stress developed at the soil and the factor of safety with
respect to the bearing capacity of the soil are used to measure the effect of each dimensionless parameter. An example for earthquake
excitation is also given for a plane frame. The results are compared to the ones of a simplied static approach based on the maximum
values of the applied loads, similarly to the procedure that is usually applied in practice. The results show that the static approach can
predict the response satisfactorily if resonance does not happen, if the stiffness of the foundation is not large compared to the stiffness of
the superstructure and if the dynamic part of the axial force of the column is not large; in these cases, it may underestimate or
overestimate the response signicantly, depending on the sign of the dynamic axial force that is considered.
r 2007 Elsevier Ltd. All rights reserved.
Keywords: Earthquake response; Harmonic response; Footings; Foundation; Winkler foundation; Uplift; Rocking

1. Introduction
The phenomenon of partial uplift of shallow foundations
is common during strong earthquakes and is due to the
inability of the soil to develop tensile stresses. As a result,
part of the base of a footing loses contact with the ground
whenever the vertical displacement of either corner of the
base becomes larger than a critical value, which depends
on the dimensions of the footing, the stiffness of the
foundation mat and the static loads. It is known that
foundation uplift affects the dynamic response of the
superstructure, usually in a benecial way, since it reduces
its response. For this reason, the effect of the base uplift on
the earthquake response of structures has been investigated
thoroughly, e.g. by Meek [1,2], Psycharis [3,4], Yim and
Chopra [5,6], Celep and Guler [7], Wang and Gould [8],
Olivet et al. [9], Harden and Hutchinson [10] among others.

Tel.: +30 2107721180; fax: +30 2107721182.

E-mail address: ipsych@central.ntua.gr


0267-7261/$ - see front matter r 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.soildyn.2007.07.010

On the contrary, the effect of partial lift-off on the


stresses developed at the soil has not been studied enough.
The seismic codes address this problem by static considerations, which are based on the forces and moments
applied to the footing during the seismic design situation
(Eurocode 7 [11], FEMA 356 [12], ATC 40 [13]). In the
approach of the American codes (for a thorough investigation see [14]), a relation is established between the
overturning moment and the footing rotation for each
possible state: elastic behaviour of the soil and full contact
of the base; inelastic behaviour of the soil and full contact
of the base; elastic behaviour of the soil and partial uplift;
inelastic behaviour of the soil and partial uplift. In
Eurocode 7 (EC7)Appendix D, the effect of the base
uplift is considered in the calculation of the bearing
capacity of the soil through a reduction of the effective
area of contact between the footing and the foundation
mat. The reduction in the dimensions of the area of contact
is proportional to the eccentricity of the loads applied,
which is measured by the ratio M/N, where M is the
applied moment and N is the axial force.

ARTICLE IN PRESS
I.N. Psycharis / Soil Dynamics and Earthquake Engineering 28 (2008) 577591

578

In both approaches, the applied loads M and N play an


important role to the design of the foundation. During an
earthquake, these loads vary with the dynamic response of
the superstructure. If a modal analysis is applied for the
calculation of the response of the superstructure, as it is
common in practice, the maximum values of the dynamic
components of M and N can be calculated through a
combination rule of the modal responses. However,
although new generation combination rules can handle
reasonably well the problem of the simultaneous values of
multiple response quantities, the sign of the axial force is
lost. Thus, it is not known whether the total axial force at
the base of the columns should be calculated by adding or
subtracting the dynamic component of N to the corresponding gravity loads. This dynamic component of the
axial force may be large, especially for columns located
close to the perimeter of the building and for coupled walls.
Thus, for strong earthquakes, forces larger than the static
ones can be developed. In such cases, the application of the
seismic codes with the most adverse combination of signs
for M and N may lead to a conservative design and
unnecessarily large dimensions of the foundation.
In this analysis, the dynamic response of a rigid footing
on elastic tensionless Winkler foundation is studied. A
parametric investigation of the effect of all the parameters
involved in the problem is performed for harmonic
excitation. Emphasis is given to the aforementioned
problem, concerning the sign of the axial force. Harmonic
wave forms are assumed for both the base motion and the
dynamic part of the loads imposed to the footing from the
superstructure. This type of excitation is chosen instead of
earthquake records in an effort to minimize the parameters
concerning the excitation and, also, because it gives a better
insight to resonance phenomena. However, an example for
earthquake excitation is also given.
The results of the dynamic analysis are also compared
with the ones of a static approach, which is based on
the maximum values of the dynamic components of the
loads, as it is the usual procedure in practice during the
application of the specications of the codes. In this
way conclusions concerning the accuracy of this approach
are drawn and a suggestion, concerning the value of the

axial force that should be used in these calculations, is


made.

2. System considered
The system considered is shown in Fig. 1(a). It consists
of a rigid footing connected to the superstructure by a
column. The footing is shown rectangular, with dimensions
b  a but the following analysis can also be applied to any
other shape that is symmetric about a vertical axis through
the middle point of the base, M, by assuming that b is the
width of the base, h is the height at which the centre of
gravity (CM) of the footing is located and a is the total
height of the footing, equal to the distance between the
bottom of the column and the foundation mat.
The soil is modelled as a tensionless Winkler foundation
that does not yield. For simplicity, the soil exibility in the
horizontal direction is not considered, because it does not
affect signicantly the vertical stresses that are developed at
the soil, which are of main concern in this analysis and
depend on the axial force and the rotation of the footing
only. It should be noted that, for tall buildings and slender
structures, the effect of the horizontal translation of the
base can be neglected in comparison to the effect of the
rocking even for the response of the superstructure [15].
Thus, the footing is allowed to rotate and move vertically
only.
The Winkler foundation is characterized by two parameters: the subgrade modulus, k0 and the corresponding
damping, c0. This simple model is widely used in
soilstructure interaction (SSI) problems, although it
cannot model precisely the dynamic response of footings
in which rocking and vertical vibrations occur simultaneously. This is due to the fact that these modes are not
independent from each other. In order to uncouple the
vibrations, a higher-order modelling should be applied, as
the one proposed by Hall et al. [16]. The present analysis,
however, aims to the qualitative investigation of the
dynamic response and not to the exact calculation of the
rotations and displacements. Taking under consideration
that the governing equations of motion after lift-off are

b
z

CM

CM
h
M

k0, c0

Kr, sup
M

k0, c0

Fig. 1. (a) Footing on elastic Winkler foundation, (b) equivalent model of analysis.

ARTICLE IN PRESS
I.N. Psycharis / Soil Dynamics and Earthquake Engineering 28 (2008) 577591

already very complicated, it was decided to use the simple


Winkler foundation mentioned above.
Fig. 1(b) shows the corresponding model that is used in
the analysis, which differs from the one of Fig. 1(a) only in
the absence of the column, which has been substituted by a
rotational spring placed at the centre of the base (point M).
This spring, which models the stiffness of the superstructure, has stiffness Kr,sup Kr,col+a2 Kx,col, with Kr,col
and Kx,col being the rotational and the horizontal stiffness
of the column (at its base), respectively and a being the
height of the footing. The spring is placed at point M and
not at the centre of gravity, CM, because the equations of
motion are derived about that point. There is no need to
add a horizontal spring at point M modelling the
horizontal stiffness of the column, because the base is not
allowed to move in this direction.
The motion of the footing is considered only on the
plane of the cross section shown in Fig. 1. In the direction
normal to this plane, a unit width is assumed for the
footing. The loads and the stiffness are considered per unit
length in this direction.
Since the horizontal movement of the base is restricted, the
motion can be described by two independent variables, namely
the rotation j and the vertical displacement z of the centre of
mass [see Fig. 1(b)]. Note that, although there is no horizontal
motion at the base, the rotation produces horizontal displacements at all other points, which, however, can be expressed in
terms of j. For example, for small rotations, the horizontal
displacement, x, of CM is equal to h j.
The superstructure loads, which in the real system act on
the footing through the column, are also transferred to
point M and are considered as external loads to the
footing, as shown in Fig. 2. If Mcol, Ncol and Vcol are the
moment, the axial force and the shear force at the base of
the column, respectively, the moment and the axial force at
point M are: M Mcol+a Vcol and N Ncol. The axial
force N is considered positive if it is compressive. No shear
force is considered at point M, because this point is not
moving in the horizontal direction.

CM

x
a

MG+ME

Kr,sup

M
k0, c0
..
zg

In this analysis, the response of the foundation is


uncoupled from the response of the superstructure. The
column loads Mcol, Ncol and Vcol can be approximately
calculated from the response of the superstructure assuming xed conditions at the base. However, when these loads
are applied to the footing, a rotation j occurs and thus, the
actual moment that acts at the base of the column is
smaller than Mcol, because it is relaxed by the term
Kr,sup  j. In this way, SSI effects are considered in an
indirect way. For a more accurate solution, the shear forces
of the columns could be calculated for the seismic loads of
the superstructure that correspond to the dynamic characteristics of the system on a exible foundation, while the
moments at the base of the columns are calculated from
these forces assuming that the base is xed. Such a
procedure, however, is not easily applied and, in general,
would not lead to the exact solution either, since the
redistribution of the seismic loads among the columns,
caused by the variation of their stiffness due to the rotation
at their base, is not considered. As an approximation, the
column loads can be directly calculated from the response
of the superstructure assuming xed base conditions. In
this case, the approximation concerns: (i) the earthquake
loads applied to the superstructure, because the exibility
of the foundation is not considered and (ii) the distribution
of these loads among the columns. The present analysis
focuses on the investigation of the response of the footing,
assuming known loads from the superstructure and,
thus, this matter is not examined further.
The superstructure loads consist of two terms: the static
term, which is due to the gravity loads of the superstructure
and will be denoted by the subscript G and the
earthquake term, which is caused by the dynamic response
of the superstructure and will be denoted by the subscript
E. The dynamic part of each load is time dependent.
Therefore, one can write:
Mt M G M E t,

(1)

Nt N G N E t.

(2)

3. Dynamic analysis

NG+NE

579

..
xg

Fig. 2. Base accelerations and superstructure loads.

The equations of motion can be derived from the


kinematics of the rocking block shown in Fig. 2. It is
assumed that the moment of inertia of the superstructure
does not affect the rocking response, because, in general,
the rotations of the footings produce only local effects and
are not related to the rotation of the structure as a whole.
Also, the mass of the superstructure that is moving with
the footing affects the vertical vibrations and must be
considered in the equations. Since a small part of the whole
structure is examined, the exact calculation of this mass is
not possible. For simplicity, it is assumed that this mass is
equal to NG/g, with g being the acceleration of gravity. In
other words, it is assumed that this mass is equal to the one

ARTICLE IN PRESS
I.N. Psycharis / Soil Dynamics and Earthquake Engineering 28 (2008) 577591

580

that corresponds to the static axial force at the base of the


column, although this is not generally true. Under these
assumptions, the following equations of motion can be
derived for small rotations (sin jEj and cos jE1) and for
j40:
Full contact:
1
1
12
_ 12
IMj
c 0 b3 j
k0 b3 j K r;sup j
mhx g M G M E ,


m N G =g z c0 b_z k0 bz


 m N G =g zg  N E .

3a

3b

With uplift:
"

 #
 #
 2 
1
b3
dz 3
1
b
dz 2
c0
_  c0
IMj

j
z_

3
j
2
j
8
4
"

 #
1
b3
dz 3

j K r;sup j
k0
3
j
8
"

 #
1
b2
dz 2

k0
d  z
2
j
4
mhx g M G M E ,


4a



 #
 2 
b dz
1
b
dz 2

m N G =g z c0
z_  c0
j

2
j
2
j
4
 #
 2 
1
1
b
dz 2
1
j  mg N G
k0 bz  k0

2
2
j
2
4


 m N G =g zg  N E
4b


in which m is the mass of the footing, IM is the mass


moment of inertia about point M, d is the static deection
due to the gravity loads, given by the relation d N G
mg=k0 b and x g ; zg are the base accelerations in the
horizontal and the vertical directions, respectively.
In the above equations of motion, second-order effects are
not included. These second-order effects concern the
horizontal displacement of the CM and the vertical
displacement of the corner of the base, caused by the
rotation of the footing. The former displacement produces
an extra moment of the vertical loads with respect to the
CM and the latter one an extra moment of the horizontal
loads. However, since the superstructure loads, which are
the most important external loads applied to the footing, are
considered in an approximate way, it is reasonable to neglect
second-order effects, which are of small importance.
Note that the above equations of motion can be seen as
an extension of the equations that govern the motion of a
free-rocking block [17], if the superstructure stiffness and
loads are included and the above-mentioned second-order
terms of the inertia forces are neglected.

The equations of motion for the full-contact regime [Eqs.


(3a) and (3b)] are valid for positive (j40) and negative
(jo0) angles of rotation. For the uplift regime, Eqs. (4a)
and (4b) are valid for j40, only. The equations for jo0
are similar to (4a) and (4b) but some terms appear with
opposite sign. One way to calculate the response for jo0 is
by using Eqs. (4a) and (4b) for this regime also, by
assuming that j is temporarily considered positive counterclockwise and the x-axis is positive to the left. In this case,
the sign of the terms x g , MG and ME in (4a) and (4b), that
are affected by this change, should be reversed. After
calculating the angle of rotation, one should change its
sign, in order to return to the initial notation of clockwise
positive angles (Fig. 1b).
Due to the gravity loads, an initial angle of rotation, j0,
exists in the system which must be considered in the solution
of the equations of motion. If Kr,tot is the total rotational
stiffness, the initial rotation is equal to MG/Kr,tot. The total
rotational stiffness is equal to the sum of the rocking stiffness
of the Winkler foundation, Kr,f and the stiffness of the
superstructure, Kr,sup, i.e. Kr,tot Kr,sup+Kr,f. Assuming that
uplift does not occur for gravity loads, Kr,f k0b3/12. Note
that z is measured from the equilibrium position and,
therefore, the initial value of z is zero.
The value of the static vertical displacement, d, should be
used to determine when uplift occurs or full contact is reestablished. Actually, uplift starts when the vertical
displacement of any corner of the base becomes greater
than d and full contact is re-established when this
displacement becomes smaller than d. Assuming that z is
positive upwards (Fig. 2), the criteria for the initiation and
the end of uplift can be written as:
Initiation of uplift:
 
z 12bjXd.

(5a)

End of uplift:
 
z 12bjpd.

(5b)

All the results of the present analysis were obtained by


integrating the equations of motion using RungeKutta of
the fourth order and a small time step, equal to 0.0002 s.
This time step is equal to 5% of the smallest eigenperiod
during full contact among all the cases of the parametric
investigation, while in most of them it is much smaller. The
criteria for uplift or contact were checked at each time step.
For all the cases examined, the maximum angle of rotation,
jmax, the minimum length of contact, smin, and the maximum corner stress developed at the soil, sc,max, (Fig. 3)
were determined from the corresponding time histories
using the following equations:
Full contact:
st b,

(6a)

ARTICLE IN PRESS
I.N. Psycharis / Soil Dynamics and Earthquake Engineering 28 (2008) 577591

Nf

Mf

Nf

CM
h

CM
s

Mf

581

c
c

Fig. 3. Length of contact and soil stresses: (a) full contact, (b) during uplift.




sc t k0 d  zt 12bjt .

(6b)

With uplift:
st

b
d
zt

,



2
jt
jt

sc t k0 stjt.

(7a)

(7b)

Note that complete separation of the footing from the


foundation is possible to occur for stiff foundations. If this
happens, s becomes zero and the analysis stops. In such
cases, the maximum value of jmax and sc,max up to the time
of complete separation was used in the plots that are
presented in the following.
In each case, the bearing capacity of the soil, Rd,
according to EC7 (Appendix D) was also determined and
compared to the total maximum axial force, Nf, at the
foundation level. Thus, the factor of safety was derived
from the ratio Rd/Nf. The bearing capacity depends on the
soil properties, the inclination of the load and the effective
length of the base, b0 . Here, the effective length was
assumed equal to 23 smin, as explained in the following
section. The calculation of the bearing capacity was based
on the maximum values of the loads, with the dynamic part
of the axial force obtaining positive (for compression) or
negative (for tension) values.
It should be noted that, in general, the factor of safety
should be interpreted differently for static and for dynamic
loading. For static loading, values of factor of safety below
unity are not allowed, while, for dynamic loading, values of
safety factor below 1.0 do not necessarily imply failure, as
they merely suggest initiation of plastic ow at the
foundation. However, in the static approach of this
analysis, the loads are pseudo-static, in the sense that they
simulate the maximum dynamic loads and, thus, their
action is not considered permanent. In this way, the factors
of safety between the dynamic analysis and the static
approach can be directly compared.

4. Static approach
The results of the dynamic analysis were compared to the
ones of a simplied static approach. In this procedure, the
maximum values of the forces applied to the footing were
used to calculate the angle of rotation, the length of
contact, the soil stress and the safety factor.
First, the total axial force and the total moment applied
to the footing were calculated by adding the weight of the
footing and the inertia forces to the loads from the
superstructure. Assuming that the superstructure is moving
in phase with the base, these forces are:
M tot;max M G M E;max mAg h,

(8)

N tot;max=min N G  N E;max mg,

(9)

where m is the mass of the footing, Ag is the amplitude of


the base acceleration and MG, ME,max, NG, NE,max, are the
static and the maximum dynamic components of the
moment and the axial force of the superstructure,
respectively. In Eq. (9), the plus sign in front of NE,max
corresponds to a compressive dynamic axial force leading
to the maximum value of Ntot, while the minus sign
corresponds to the case of a tensile dynamic force leading
to the minimum value of Ntot. If the seismic response of the
superstructure creates signicant axial forces to the
columns, it is possible that Ntot,min becomes negative,
meaning that the total axial force applied to the footing is
upward.
If uplift does not occur, the maximum angle of rotation
would be:
jmax

M tot;max
.
K r;tot

(10)

If uplift occurs, the total rotational stiffness should be


calculated using a reduced value for Kr,f, in order to take
under consideration the reduced length of contact. This
would require an iterative procedure, since this length
depends on jmax. In practice, SSI is usually applied using
the stiffness of the foundation that corresponds to the fullcontact state, since consideration of uplift would require a

ARTICLE IN PRESS
I.N. Psycharis / Soil Dynamics and Earthquake Engineering 28 (2008) 577591

582

non-linear analysis. In this way, the loads applied to the


foundation are calculated considering the full-contact
regime only and any unfavourable effect of uplift is taken
under consideration later, during the calculation of the
bearing capacity of the soil. Following this procedure, jmax
was calculated without reducing Kr,tot in Eq. (10). This
approximation leads to an underestimation of the maximum rotation.
After the maximum angle of rotation was estimated, the
moment, Mf, applied to the foundation was calculated by
the relation: Mf Mtot,maxKr,sup  jmax. Since the corresponding axial force is Nf,max/min Ntot,max/min, the minimum length of contact and the maximum soil stress were
determined by static equilibrium (see Fig. 3) using the
following relations:
Full contact:
smin b,
sc;max

N f 6M f
2 .
b
b

(11a)
(11b)

With uplift:
smin

3b 3M f

,
2
Nf

sc;max

2N f
.
smin

(12a)
(12b)

If the minimum value of the axial force, Nf,min, which


corresponds to the minus sign of NE,max in Eq. (9), is used
in the equations after uplift, Eq. (12a) may lead to very
small or even negative values of the contact length, smin,
implying that complete separation occurs. In such cases,
large stresses are developed at the soil, as can be seen from
Eq. (12b).
It should be noted that the length of contact after uplift,
as dened by Eq. (12a), is different than the effective length
of contact, b0 , used for the calculation of the bearing
capacity of the soil according to EC7. The latter is dened
as b0 b  2M f =N f . Using Eq. (12a), the effective length
according to Eurocode can be written as: b0 23 smin. This
relation was also used in the dynamic analysis for the
calculation of the bearing capacity.
5. Parametric investigation for harmonic excitation
The parametric investigation was performed for an
orthogonal footing with base width b 4.00 m, made of
concrete with specic weight g 25 kN/m3. Nine dimensionless parameters, related to the stiffness and the
damping of the foundation, the stiffness of the superstructure, the dimensions of the footing, the frequency of
the excitation and the loads from the superstructure were
examined. The effect of each parameter was investigated by
varying its value while all the other parameters were kept

constant. The dimensionless parameters examined and


their typical values that they attained when the effect of
another parameter was investigated are given in Table 1.
The terms that appear in these quantities are: the rocking
eigenfrequency of the footing (Fig. 2) for the full-contact
p
state, which can be expressed as oj K r;tot =I M ; the
corresponding eigenfrequency for vertical vibrations durp
ing full contact, dened as oz k0 b=m N G =g; the
characteristic frequency p, which depends on the geometry
p
of the footing and is dened by the relation p mgh=I M
similarly to the one for a block rocking on a rigid
foundation [18], with the difference that the pole of
rotation is located at the middle of the base and not at
the corner; the coefcient of critical damping of the
foundation, which is measured either by the damping
corresponding to the rocking vibrations during full
p
1
contact: zj 24
c0 b3 = I M K r;tot or by the damping corresponding to the vertical vibrations: zz 12c0 b=
p
m N G =gk0 b; the weight of the footing WF mg;
the static eccentricity eG M G =N G mg and the dynamic eccentricity eE M E;max =N G mg. Note that oj
and oz are partially coupled, since both depend on the
stiffness of the soil. However, oz changes with the axial
load of the column, while oj is independent of this load
and also depends on the stiffness of the superstructure. The
relation between oj and oz for an orthogonal footing is
given in the following (see Eq. (13)).
As mentioned above, the effect of each parameter on the
response of the system was measured by means of the
maximum angle of rotation, jmax, the minimum length of
contact, smin, the maximum corner stress of the soil, sc,max
and the factor of safety with respect to the bearing capacity
of the soil. The length of contact was normalized with
respect to the width of the base, b and the stress of the soil
was normalized with respect to the corresponding stress for
gravity loads only, sc,G, which is equal to N G mg=
b 6M G =b2 . For the calculation of the bearing capacity,
drained conditions were assumed for the soil and the
method proposed in Section D4 of Appendix D of EC7 was
applied. The following effective properties of the soil were
Table 1
Typical values of the dimensionless parameters considered in the analysis
Parameter

Value

oj/p
Kr,sup/Kr,f
zj (zz)
a/b
eG/b
eE/b
NG/WF
NE,max/NG
o/oj (o/oz)
Ag/g
b (m)
g (kN/m3)

30.00
2.00
0.15 (0.05)
0.20
0.05
1.00
7.50
0.20
0.50 (2.50)
0.50
4.00
25.00

ARTICLE IN PRESS
I.N. Psycharis / Soil Dynamics and Earthquake Engineering 28 (2008) 577591

assumed: internal friction j0 301; cohesion c0 40 kPa;


specic weight g0 18 kN/m3 and overload pressure at
foundation level q0 30 kPa. These soil properties do not
represent any particular soil type and are used here just as a
numerical example. A strip foundation was considered by
assuming that the length of the footing in the normal
direction is equal to 10.00 m. Also, the shear force acting
on the footing from the superstructure was found assuming
that the column is clamped at both ends and the basestorey height is 5.00 m.
The parametric study was performed for the horizontal
component of the base excitation only. It was assumed that
the ground motion, x g and the dynamic components of the
superstructure loads, ME(t) and NE(t) are harmonic
functions of frequency o, specically, x g Ag sin ot,
ME(t) ME,max sin (ot) and NE(t) NE,max sin (ot). This
is an approximation, since it is assumed that the superstructure is moving with the same frequency with the base
excitation, which is valid for the steady-state response only.
Also, the assumed 1801 phase difference is not, in general,
true. It should be noted, however, that the base excitation
affects the impulsive forces of the footing only, which are
generally small compared to the loads from the superstructure. In this sense, these approximations affect slightly
the results and are adopted here in order to minimize the
number of the parameters concerning the excitation.
In Fig. 4, the effect of the ratio of the stiffness of the
superstructure to the stiffness of the foundation, Kr,sup/Kr,f,
is illustrated. On each plot, the results obtained from the
dynamic analysis and the static approach are shown for
positive and negative values of NE,max. It is evident that the
ratio Kr,sup/Kr,f affects signicantly the length of contact,

583

the soil stress and the factor of safety. Note that these
results were obtained for constant total rotational stiffness,
Kr,tot, since the ratio oj/p remains constant (see Table 1).
Thus, when Kr,sup/Kr,f decreases, the superstructure becomes softer and the foundation stiffer. For a given
rotation, a reduction in Kr,sup leads to a larger portion of
the total moment transferred to the foundation and,
therefore, for small values of Kr,sup/Kr,f the length of
contact obtains small values and the corner stress of the
soil becomes large.
The angle of rotation, jmax, depends on the total
rotational stiffness and not on the ratio Kr,sup/Kr,f. For the
plots of Fig. 4, the rotational stiffness during full contact,
Kr,tot, is not changing and thus jmax is constant according to
the static analysis and slightly changing according to the
dynamic analysis (in this case, it is affected by the rotational
stiffness after uplift, which is smaller than Kr,tot). In the
contrary, jmax is changing signicantly when Kr,tot is
varying. This is shown in Fig. 5, which plots the variation
of the rotation and the soil stress with the ratio oj/p, which
is proportional to the square root of Kr,tot. These results
correspond to a constant value of Kr,sup/Kr,f (Table 1) and
thus the stiffness of the superstructure is varying similarly
with the stiffness of the foundation; the latter depends on the
subgrade modulus, k0, the variation of which is also shown
on the plots of Fig. 5 (top horizontal axis). It is evident that
jmax decreases exponentially as the stiffness of the soil
increases, while the soil stress is practically unaffected.
The results depicted on Figs. 4 and 5 show that the sign
of NE,max is not important according to the dynamic
analysis. In the contrary, the static approach is signicantly
affected by the sign of NE,max. Thus, unrealistically small

Fig. 4. Effect of the ratio of the stiffness of the superstructure to the stiffness of the foundation on the response of the footing.

ARTICLE IN PRESS
584

I.N. Psycharis / Soil Dynamics and Earthquake Engineering 28 (2008) 577591

Fig. 5. Effect of the ratio oj/p on the response of the footing.

values for the length of contact and the safety factor and
quite large ones for the soil stress were obtained for
NE,maxo0, while for NE,max40 the soil stress was close to
the one of the dynamic analysis, in spite of the fact that the
approximate method overestimates the length of contact
and the bearing capacity, especially for medium to large
values of Kr,sup/Kr,f (soft foundation). Note that the value
of NE,max used for the derivation of these results was
relatively small, equal to only 20% of NG (see Table 1).
The foundation damping does not seem to affect the
results signicantly, as it is shown on the top-row diagrams
of Fig. 6. In these plots, the coefcient of damping for
rocking vibrations during full contact, zj, was varying
from 0.05 to 0.50 and the results show a small decrease in
the rotation and the soil stress as the damping increases, as
expected.
Similarly, the dimensions of the footing are not
important to the stresses that are developed at the soil, as
illustrated on the bottom-row diagrams of Fig. 6. Note that
in all the results, the width of the footing, b, was assumed
constant, equal to 4.00 m and thus, the height of the
footing was varying from 0.4 to 2.0 m. The small inuence
of the ratio a/b on the stresses of the soil was expected,
since the dimensions of the footing affect the inertia forces,
which are usually small compared to the loads from the
superstructure. Note that the angle of rotation is decreasing exponentially as a increases. This happens because the
rotational stiffness increases with a2 in order oj/p to
remain constant.
It is known that, for a free rocking block on an elastic
foundation, the size of the block is important to the
response (e.g. [19]) with smaller blocks experiencing larger
rotations than ones with larger dimensions of the same
aspect ratio. For the problem that is examined here, the
footing is not free, since it is connected to the superstructure. The size enters in the analysis through the
characteristic frequency p. In order to check whether it is
important, the plots of Fig. 5 were reproduced for the same
values of the dimensionless parameters and for one half of

the width of the base, i.e. for b 2.00 m instead of 4.00 m.


The results obtained were exactly the same with the ones
shown in Fig. 5 and thus, it can be concluded that the
parametric investigation presented here is practically
independent of the absolute size of the footing.
The effect of the frequency of the excitation, o (same for
the base motion and the dynamic loading from the
superstructure) on the response is illustrated on Fig. 7.
The frequency o is shown normalized with respect to the
rocking eigenfrequency during full contact, oj (bottom
horizontal axis) and with respect to the corresponding
vertical eigenfrequency, oz (top horizontal axis). This is
possible because, for an orthogonal footing, oj and oz are
related according to the equation:
s
oj
1 N G =mg1 K r;sup =K r;f
,
(13)

oz
1 4a=b3
which leads to a constant relation between oj and oz for
constant values of the dimensionless terms of Table 1.
It is evident from the plots of Fig. 7 that resonance
phenomena occur when oEoj and oEoz, which result to
complete separation of the footing from the soil (the length
of contact and the factor of safety become zero in this case)
and a signicant increase in the rotation and the soil stress.
The maximum values of jmax and sc,max are observed for
values of oj/o and oz/o a little greater than unity, because
the effective rocking frequency and the effective vertical
frequency after uplift are shorter than oj and oz,
respectively. The term effective is used here to denote
the apparent frequencies of the response, since uplifting
systems do not possess natural frequencies in the classical
sense.
The response of the system when o is close to oj is
further examined in Fig. 8, in which the time-histories of
the maximum soil stress (at the most unfavourable corner),
the rotation and the vertical displacement of CM are given
for oj/o 0.8, 1.0 and 1.1 and for NE,max40 and
NE,maxo0. Note that a signicant increase in the vertical

ARTICLE IN PRESS
I.N. Psycharis / Soil Dynamics and Earthquake Engineering 28 (2008) 577591

585

Fig. 6. Effect of the foundation damping, zj, (top row) and the aspect ratio of the footings dimensions (bottom row) on the response.

Fig. 7. Effect of the excitation frequency, o, (same for the base motion and the dynamic loads from the superstructure) on the response of the footing.

oscillations occurs for oj/o 1.0 and 1.1, compared to the


case of oj/o 0.8, which is larger than the corresponding
increase in the rotations, although the resonance concerns

the rotational vibrations. As a result, complete separation


happens in these cases. This shows that a transfer of energy
from the rocking mode to the vertical mode occurs. This

ARTICLE IN PRESS
586

I.N. Psycharis / Soil Dynamics and Earthquake Engineering 28 (2008) 577591

Fig. 8. Time-histories of the maximum soil stress (top row), the rotation (middle row) and the vertical displacement of CM (bottom row) for oj/o 0.8,
1.0 and 1.1 and for NE,max40 and NE,maxo0.

phenomenon, which happens after uplift, has also been


observed in the case of a free rocking block on elastic
foundation [19].
The static approach is independent of the frequency of the
excitation. The results shown in Fig. 7 were obtained for
constant superstructure loads and thus, the curves corresponding to the static approach are horizontal lines. Note
that the static approach shows a signicant dependence of all
the quantities, except of the rotation, on the sign of NE,max.
Thus, for NE,maxo0 the length of contact is about one third
the value corresponding to NE,max40, the soil stress is almost
double and the bearing capacity according to EC7 is nil,
giving a safety factor equal to zero. In the contrary, the
dynamic analysis is much less dependent on the sign of
NE,max, as it is evident from the plots of Figs. 7 and 8.
The last parameters examined are related to the loads
from the superstructure. The results concerning the static
loads (not presented here) show that jmax and sc,max
increase as MG increases, while smin and the factor of safety
decrease, as expected. The dynamic part of the applied
moment was measured by the eccentricity eE M E;max =
N G mg, which was normalized with respect to the
width of the base, b. Note that eE is not the real eccentricity
of the dynamic part of the moment, because the dynamic

part of the axial force, NE,max, is not included in the


denominator. This is done in order to normalize ME,max
with respect to a quantity that is independent of the other
dynamic loads of the superstructure. The dynamic axial
force was measured by the ratio NE,max/NG, which attains
both positive and negative values, since NE,max can be
compressive or tensile.
The results concerning the dynamic eccentricity, eE, are
given in Fig. 9. The dynamic analysis shows that the angle
of rotation and the soil stress increase almost linearly with
eE while the length of contact and the safety factor decrease
rapidly and become zero for values of eE greater than
approximately 1.5b. The sign of the axial force does not
seem to be important, especially for small values of eE. The
static approach gives relatively good results for NE,max40
and for eE/bo1.25, except of the expected underestimation
in the angle of rotation. For eE/b41.25, however, the soil
stress increases exponentially. Much worse are the results
for NE,maxo0: in this case, the static approach largely
underestimates the length of contact and the factor of
safety and overestimates the soil stress, even for eE/b less
than unity. It should be noted that complete separation
occurs for eE/b greater than about 1.75 for both the
dynamic and the static analyses.

ARTICLE IN PRESS
I.N. Psycharis / Soil Dynamics and Earthquake Engineering 28 (2008) 577591

587

Fig. 9. Effect of the dynamic eccentricity of the superstructure loads on the response of the footing.

Fig. 10. Effect of the dynamic axial force from the superstructure on the response of the footing.

A thorough investigation of the effect of the axial force is


illustrated on the plots of Fig. 10, on which the response is
plotted versus the ratio NE,max/NG. It is evident that the
static approach is extremely conservative for NE,maxo0,
leading to zero values of the safety factor (which,

practically, means that the footing dimensions are insufcient) while it is against safety for NE,max40, especially for
large values of NE,max/NG. In the contrary, the dynamic
analysis is practically independent of NE,max/NG, even for
large tensile values of NE,max (e.g. for NE,max/NG 0.8).

ARTICLE IN PRESS
I.N. Psycharis / Soil Dynamics and Earthquake Engineering 28 (2008) 577591

588

The response is further examined in Fig. 11, in which the


time histories of the soil stress and the vertical displacement
of CM are given for NE,max/NG 0.5 and NE,max/
NG 0.5: although the change of sign of NE,max causes
a signicant phase shift to the vertical displacement, its
effect on the soil stress is small.
In practice, the correct sign of NE,max that corresponds
to ME,max is not known, since these values are usually
derived from a modal analysis after an SRSS or a CQC
combination of the modal responses. For this reason, a

negative value of NE,max (tension) is usually used in favour


of safety. This procedure, however, may lead to a very
conservative design according to the aforementioned
results. In the contrary, the use of a positive value for
NE,max (compression) may be against safety. The results for
harmonic excitations show that neglecting the dynamic
part of the axial force would be more realistic. This
proposal is further investigated in the following section
through an example for earthquake excitation.
6. Example for earthquake excitation

Fig. 11. Time-histories of the maximum soil stress (top row) and the
vertical displacement of CM (bottom row) for NE,max/NG 0.5 and
NE,max/NG 0.5.

An example for the plane frame shown on Fig. 12(a),


which is subjected to an earthquake excitation, is presented
here. The width of the footings in the normal direction is
2.00 m. In Fig. 12(b), the theoretical model for xed base
conditions is given. First, the static loads (self-weight and
load q on the beam) were applied to the xed base model,
in order to determine MG and NG; also, a unit rotation at
point A in order to determine Kr,sup. Next, this model was
subjected to the El Centro, 1940 earthquake magnied 1.20
times and ME(t) and NE(t) were calculated.
Using these forces and moments as the superstructure
loads, the equations of motion (3) and (4) were solved and
jmax, sc,max and smin were derived for each footing. The
maximum values of ME(t) and NE(t) were used to calculate
the response according to the static approach. The
procedure was repeated for eight values of the subgrade
modulus, ranging from 2.5 to 20 MN/m3. In all cases, zj
was taken equal to 0.15.
An example of the results obtained is given in Fig. 13, in
which the time histories of the maximum soil stress (at the
most unfavourable corner of the footing) and the length of
contact for k0 15 MN/m3 are shown for footings A and
B. On the right column, a zoom for t 2.04.0 s is
presented. Note that when one footing is in full contact
the other is rocking in the uplift regime, since the dynamic
axial forces from the superstructure that act on them have
opposite signs. Another interesting observation is that the
maximum soil stress occurs at t 2.63 s for footing A
(right top diagram), when this footing is in the full-contact

7.00
q=80 KN/m

7.00

C 40/40

2.00

2.00

4.00

C 40/40

0.60

4.00

B 30/60

Fig. 12. (a) Geometrical data of the plane frame considered in the analysis, (b) equivalent model for xed base conditions.

ARTICLE IN PRESS
I.N. Psycharis / Soil Dynamics and Earthquake Engineering 28 (2008) 577591

589

Fig. 13. (a) Time histories of the maximum soil stress (at the most unfavourable corner of the footing) and the length of contact for k0 15 MN/m3, (b)
zoom for t 2.04.0 s (El Centro, 1940 earthquake).

Fig. 14. Variation of the normalized minimum length of contact and the maximum soil stress with the subgrade modulus according to the dynamic
analysis and the static approach (El Centro, 1940 earthquake).

regime (right bottom diagram) and not during uplift, as


one would expect. This happens because the soil stress
depends not only on the rotation but also on the vertical
displacement of the footing.
The length of contact according to the static approach is
also shown on the plots of Fig. 13 for NE,max40 and
NE,maxo0. The latter assumption leads to a very small
value for smin/b, equal to 0.11, which would not be
acceptable in a design procedure. The corresponding soil

stress is very large (1266 kPa) which exceeds the scale of the
vertical axis on the top-row diagrams and for this reason it
is not shown. In the contrary, the assumption NE,max40
leads to an underestimation of uplift to about one half of
its value; the soil stress, however, is predicted reasonably
well, except of a couple of spikes, in the case of footing A,
that exceed the value of the static approach.
The variation of the minimum length of contact and
the maximum soil stress with the value of the subgrade

ARTICLE IN PRESS
590

I.N. Psycharis / Soil Dynamics and Earthquake Engineering 28 (2008) 577591

modulus is shown in Fig. 14 for both the dynamic analysis


and the static approach. According to the dynamic
analysis, almost nil values of smin occur for k0 5 MN/
m3 showing some kind of resonance. The approximate
method cannot capture this phenomenon and predicts
signicantly smaller uplift (larger values of smin than the
dynamic analysis) for subgrade modulus less than 10 MN/
m3; the soil stress, however, is not affected by this decrease
in the length of contact and both methods give similar
results.
For stiffer soils with k0 greater than 10 MN/m3, the static
approach leads to unrealistically small values for the length
of contact and extremely high soil stresses, if the dynamic
part of the axial force from the superstructure is considered
tensile (NE,maxo0). In the contrary, for NE,max40 (compressive), the soil stress obtained is slightly underestimated
in spite of the fact that uplift of the footings is signicantly
underestimated. The results corresponding to NE,max 0
are also plotted on these diagrams, for comparison:
concerning the soil stress, the results are similar to the
ones for NE40 and rather on the safety side; concerning
the minimum length of contact, the values obtained for
NE,max 0 are close to the average values for NE,max40
and NE,maxo0.

when this ratio decreases the amount of uplift and the


stresses that are developed at the soil increase while the
factor of safety decreases.
The static approach can, in general, predict the response
satisfactorily. Its accuracy, however, decreases signicantly
in two cases: (a) if resonance happens and (b) if the
dynamic part of the axial force of the superstructure
becomes large. In case (a), which can happen when the
excitation frequency is close to either the rotational or the
vertical eigenfrequency of the system for the full-contact
regime, the static analysis underestimates, in general, the
response. In case (b), the static analysis greatly depends on
the sign of the dynamic part of the axial force of the
superstructure: for tensile forces, it overestimates the
response and can be conservative; for compressive forces
it may give results against safety. It seems that neglecting
the dynamic axial force of the superstructure is not, in
general, against safety while it leads to signicantly more
economical design. These results are veried for the
earthquake response of the frame, but more research is
needed on this subject.

7. Conclusions

The work presented in this paper was performed in the


framework of the research program Ultimate bearing
capacity of shallow foundations under seismic actions,
coordinated by Professor M. Kavvadas of the National
Technical University of Athens and nanced by the
Greek Earthquake Planning and Protection Organization
(OASP).

The dynamic response of footings sitting on an elastic,


tensionless Winkler foundation is examined in this paper.
The superstructure stiffness is considered by an equivalent
rotational spring, modelling the stiffness of the column.
The equations of motion are derived for each regime of
the response (full contact and uplift) and are integrated
numerically.
A parametric investigation of the effect of a number of
dimensionless terms, involving all the main parameters, as
the foundation stiffness, the foundation damping, the
frequency of the excitation and the superstructure loads, is
performed for harmonic excitation. The investigation
concerns the effect of each dimensionless parameter to
the maximum angle of rotation, the minimum length of
contact after uplift, the maximum stress developed at the
soil and the factor of safety with respect to the bearing
capacity of the soil according to Eurocode 7, assuming that
all the other parameters remain constant. The results of the
dynamic analysis are compared to the ones of a simplied
static approach, which is based on the maximum values of
the loads. An example for a plane frame subjected to the El
Centro, 1940 earthquake is also given for several values of
the subgrade modulus. The most important conclusions
can be summarized as follows:
The investigation showed that the stiffness of the
foundation practically does not affect the length of contact
and the soil stress, as far as resonance does not occur.
Similarly, the damping of the foundation does not seem to
be signicant. In the contrary, the ratio of the superstructure stiffness to the foundation stiffness is important:

Acknowledgements

References
[1] Meek JW. Effect of foundation tipping on dynamic response. J Struct
Div (ASCE) 1975;101:1297311.
[2] Meek JW. Dynamic response of tipping core building. Earthquake
Eng Struct Dyn 1978;6:43754.
[3] Psycharis IN. Dynamics of exible systems with partial lift-off.
Earthquake Eng Struct Dyn 1983;11:50121.
[4] Psycharis IN. Effect of base uplift on the dynamic response of DSOF
structures. J Struct Eng (ASCE) 1991;117:73354.
[5] Yim CS, Chopra AK. Earthquake response of structures with partial
uplift on Winkler foundation. Earthquake Eng Struct Dyn 1984;12:
26381.
[6] Yim CS, Chopra AK. Simplied earthquake analysis of structures
with foundation uplift. J Struct Eng (ASCE) 1985;111:90630.
[7] Celep Z, Guler K. Dynamic response of a column with foundation
uplift. J Sound Vibration 1991;149(2):28596.
[8] Wang X-F, Gould PL. Dynamics of structures with uplift and sliding.
Earthquake Eng Struct Dyn 1993;22:108595.
[9] Oliveto G, Calio` I, Greco A. Large displacement behaviour
of a structural model with foundation uplift under impulsive
and earthquake excitations. Earthquake Eng Struct Dyn 2003;32:
36993.
[10] Harden C, Hutchinson T. Investigation into the effects of foundation
uplift on simplied seismic design procedures. Earthquake Spectra
2006;22(3):66392.
[11] European Committee for Standardization (CEN). Eurocode 7
geotechnical design. Part 1: general rules. EN 1997-1, Brussels, 2004.

ARTICLE IN PRESS
I.N. Psycharis / Soil Dynamics and Earthquake Engineering 28 (2008) 577591
[12] American Society of Civil Engineers (ASCE). Prestandard and
commentary for the seismic rehabilitation of buildings. FEMA 356.
Washington, DC: The Federal Emergency Management Agency; 2000.
[13] Applied Technology Council (ATC). Seismic evaluation and retrot
of concrete buildings. ATC 40, Redwood City, CA, 1996.
[14] Allotey N, El Naggar MH. Analytical moment-rotation curves for
rigid foundations based on a Winkler model. Soil Dyn Earthquake
Eng 2003;23:36781.
[15] Jennings PC, Bielak J. Dynamics of buildingsoil interaction. Bull
Seismol Soc Am 1973;63:948.

591

[16] Hall Jr JR, Constantopoulos IV, Michalopoulos AP. Higher order


Winkler model for soilstructure interaction. In: Proceedings of the
third international conference on numerical methods in geomechanics, Aachen, Germany, 1979. p. 9338.
[17] Psycharis IN, Jennings PC. Rocking of slender rigid bodies allowed
to uplift. Earthquake Eng Struct Dyn 1983;11:5776.
[18] Housner GW. The behaviour of inverted pendulum structures during
earthquakes. Bull Seismol Soc Am 1963;53(2):40317.
[19] Psycharis IN, Jennings PC. Upthrow of objects due to horizontal
impulse excitation. Bull Seismol Soc Am 1985;75(2):54361.

S-ar putea să vă placă și