Professional Documents
Culture Documents
V
b
2Dt; and the pile head dynamic forces, F
p
Dt: The pile
head acceleration,
V
p
2Dt; is evaluated by solving the
response of the foundation subsystem. The pile head
forces, F
p
2Dt; are equivalent to the column base forces
on the structure at the second time step, and are then
obtained by solving the response of the structure subsystem
for the corresponding support excitations. This successive-
coupling incremental procedure is repeated until the entire
response history is determined. It is seen that the
continuous response history is well approximated by the
discrete step approach using a sufciently small time step
(i.e. Dt T=80; where T is the period of the excitation).
The substructuring technique coupled with a successive-
coupling scheme is though approximate for a nonlinear
analysis but quite effective. The effect of this approximation
on the accuracy of the results depends on the size of the time
step considered. For a very small time step T=80 used in
the present analysis, the results converged and no
accumulation of errors was noticed. Further, a fully coupled
system would require enormous computation. Thus, the
additional accuracy, which may not be signicant, obtained
at the expense of large computational cost, may not be
justied.
4. Structure subsystem
The structure subsystem may be subjected to non-
uniform foundation motion (support excitation) that is
equal to the pile heads motion. The non-uniform foundation
motion is obtained from the coupling kinematicinertial
interaction of the SPS system (Fig. 5).
Fig. 4. Finite element mesh for the pile group system: (a) top plan, (b) front
elevation with initial pressure distribution.
Fig. 5. Schematic of successive-coupling scheme.
B.K. Maheshwari et al. / Soil Dynamics and Earthquake Engineering 24 (2004) 343356 346
The incremental equations of motion for a structure
subjected to uneven support excitations are derived assum-
ing the principle of superposition to be valid within each
incremental time step, provided that the time step is
sufciently small. For a structure subsystem with n active
DOF and m support DOF, the multiple-support excitation is
obtained by the superposition of the dynamic responses of
the subsystem due to each independent support input. The
total dynamic equilibrium equations of the structure
subsystem with all m support DOF having individual
motions can be written as [19]:
M
s
U C
s
_
U K
s
U 2M
s
R
s
U
g
1
where M
s
; C
s
; and K
s
are n n matrices of the mass,
damping and stiffness of the structure subsystem, respect-
ively; U;
_
U and
U are n 1 vectors of displacement,
velocity and acceleration of the subsystem, respectively; R
s
is an n m matrix that contains the pseudo-static response
inuence coefcients and may be updated at each time step
for a nonlinear structure;
U
g
is an m 1 vector that contains
the non-uniform support motion and is equivalent to the pile
head motion,
V
p
:
Eq. (1) is constructed in an incremental form using the
Newmark time integration scheme (constant average
acceleration method [20]). The incremental equations of
motion of the subsystem at time t Dt can be written as:
4M
s
Dt
2
2C
s
Dt
K
s
DU
2M
s
R
s
tDt
U
g
2K
s
t
UM
s
4
t
_
U
Dt
t
U
!
C
s
t
_
U 2
where Dt is the time step; superscript t indicates the time;
and DU are increments of the dynamic response vector U;
such that
tDt
U
t
UDU:
For the nonlinear response of the structure subsystem,
Eq. (2) is formulated using the modied NewtonRaphson
iteration scheme. The stiffness and damping matrices, K
s
and C
s
; are replaced by the corresponding tangent stiffness
and damping matrices,
t
K
s
and
t
C
s
; which are updated at
each time step t: The stiffness and mass matrices of the
structure subsystem are constructed using normal nite
element method procedures. The damping is assumed to be
proportional to the stiffness [8].
For a structure subjected to non-uniform support
motions, the support reactions of the structure depend on
the total displacements of the active DOF U
t
as well as the
relative displacements of the support DOF U
g
: Therefore,
the support forces for non-uniform support motion cases are
different from those for rigid ground motion cases and
should be calculated as shown by [19]:
F
g
K
gs
U
t
K
gg
U
g
K
gs
U R
s
U
g
K
gg
U
g
3
where F
g
is an m 1 support force vector that is equivalent
to the dynamic pile head force vector F
p
; K
gs
is an m n
matrix that represents the coupling of the support forces and
the motions of the active DOFs; and U
t
represents the total
displacement vector. K
gg
is an m m matrix that represents
the coupling of the support forces and the motions of the
support DOFs. For each time step, the support forces F
g
are
calculated and used as the pile head inertial force input for
the pile foundation subsystem. For nonlinear seismic
response of the structure subsystem, the support force-
active DOFs motion coupling matrix K
gs
and the support
force-support DOF motion coupling matrix K
gg
are updated
at the beginning of each time step t:
Although the whole structure subsystem is considered,
only a quarter of the foundation subsystem is used.
Therefore, only one fourth of the column base force (for
the single pile system) or one fourth of the sum of all four
column base forces (for the pile group system) is transferred
from the structure to the foundation, i.e. F
p
F
g
=4:
5. Foundation subsystem
5.1. Governing equation and solution
The foundation subsystem is composed of piles and the
surrounding soil. The equation of motion for this subsystem
is written as:
M
F
V C
F
_
V K
F
V 2M
F
R
F
V
b
F
p
4
where M
F
; C
F
; and K
F
are matrices of the mass, damping
and stiffness of the foundation subsystem, respectively; V;
_
V
and
V are the vectors of displacements, velocity and
acceleration of the subsystem, respectively; R
F
is the
pseudo-static response inuence coefcients matrix that is
updated at each time step;
V
b
is the vector of bedrock
acceleration due to seismic excitations and is assumed to
consist of vertically propagating shear waves; and F
p
is the
pile head force vector that is calculated from the support
force vector F
g
(Eq. (3)). Eq. (4) is formulated in the same
incremental form as for the structure subsystem (Eq. (2)).
For the nonlinear soil model, the incremental dynamic
equilibrium of the foundation subsystem for the ith iteration
at time t Dt can be written as:
4M
F
Dt
2
2
t
C
F
Dt
t
K
F
DV
i
2M
F
R
F
tDt
V
b
t
F
p
2
tDt
F
i21
2
t
C
F
2
Dt
tDt
V
i21
2
t
V 2
t
_
V
2M
F
4
Dt
2
tDt
V
i21
2
t
V 2
4
Dt
t
_
V 2
t
V
5a
where superscripts t and i indicate the time t and ith
iteration, respectively;
tDt
F
i21
is the vector of nodal
forces in the i 21th iteration of the current time step.
B.K. Maheshwari et al. / Soil Dynamics and Earthquake Engineering 24 (2004) 343356 347
The stiffness and damping matrices, K
F
and C
F
; are replaced
by the corresponding tangent matrices,
t
K
F
and
t
C
F
; which
are updated at each time step t: DV
i
are increments of the
dynamic response vector V at the ith iteration, such that
tDt
V
i
tDt
V
i21
DV
i
5b
For the foundation subsystem, the mass matrix M
F
is
diagonal because all masses are lumped at the nodal points.
The global damping matrix,
t
C
F
; includes the contributions
of both material damping and radiation damping (including
dashpots along the boundary). The stiffness matrix,
t
K
F
; is
symmetric and is determined assuming full coupling in all
three directions of motion and includes the stiffness of
springs at the boundary nodes.
5.2. Boundary conditions
The boundary conditions are shown in Fig. 4a,b. The
presence of the springs provides stiffness, giving this
boundary a distinct advantage over the standard viscous
boundary [21,22]. The constants of the Kelvin elements in
the two horizontal directions are calculated using the solution
developed by Novak and Mitwally [22], and the constants of
the vertical Kelvin elements are calculated using the solution
developed by Novak et al. [23]. These constants are
frequency dependent. For transient excitation, the constants
are determined based on the predominant frequency of the
excitation. The stiffness and damping of the Kelvin elements
are evaluated using the area of the element face (normal to the
direction of loading). Further details on the evaluation of
these constants are described in Maheshwari et al. [17,18].
All the nodes along the base are xed in all three
directions. The nodes on the axis of symmetry are free to
move in the vertical direction and along the direction of the
axis of symmetry, and are xed in the perpendicular
horizontal direction (Fig. 4a). The nodes on the axis of anti-
symmetry are constrained in the vertical direction and along
the direction of this axis and are free to move in the
perpendicular horizontal direction (Fig. 4a,b).
It should be noted that the boundary conditions at the axis
of symmetry and anti-symmetry are developed with due
consideration of waves and loading patterns and thus they
reect mirror images. This means they should reproduce
exactly the same effects of the missing part (including other
piles). The effects of pilesoil pile interaction would be
reproduced through these boundaries. For the foundation
subsystem only (without the superstructure), the results of
the quarter model were compared with that obtained using
the full model for (2 2) and (3 3) pile groups. The
results from both cases were exactly the same.
5.3. Damping matrix C
F