Vous êtes sur la page 1sur 20

EARTHQUAKE ENGINEERING AND STRUCTURAL DYNAMICS

Earthquake Engng Struct. Dyn. 2003; 32:871890 (DOI: 10.1002/eqe.252)

A unied formulation of the piecewise exact method for


inelastic seismic demand analysis including the P-delta eect
M. N. Aydnoglu; and Y. M. Fahjan
Department of Earthquake Engineering; Bogazici University; Kandilli Observatory and Earthquake Research
Institute (KOERI); 81120 C
 engelkoy; Istanbul; Turkey

SUMMARY
The non-linear analysis of single-degree-of-freedom (SDOF) systems provides the essential background
information for both strength-based design and displacement-based evaluation=design methodologies
through the development of the inelastic response spectra. The recursive solution procedure called the
piecewise exact method, which is eciently used for the response analysis of linear SDOF systems,
is re-formulated in this paper in a unied format to analyse the non-linear SDOF systems with multilinear hysteresis models. The unied formulation is also capable of handling the P-delta eect, which
generally involves the negative post-yield stiness of the hysteresis loops. The attractiveness of the
method lies in the fact that it provides the exact solution when the loading time history is composed
of piecewise linear segments, a condition that is perfectly satised for the earthquake excitation. Based
on simple recursive relationships given for positive, negative and zero eective stinesses, the unied
form of the piecewise exact method proves to be an extremely powerful and probably the best tool for
the SDOF inelastic time-history and response spectrum analysis including the P-delta eect. A number
of examples are presented to demonstrate the implementation of the method. Copyright ? 2003 John
Wiley & Sons, Ltd.
KEY WORDS:

piecewise exact method; P-delta eect; non-linear analysis; seismic demand; inelastic response spectra

1. INTRODUCTION
Despite the fact that single-degree-of-freedom (SDOF) systems are the simplest systems of
structural dynamics, they provide much of the background information for the practical seismic response analysis of multi-degree-of-freedom (MDOF) systems. In this respect, elastic
and inelastic response spectra developed in various forms are the main tools of both the
strength-based design and the displacement-based evaluation and design methodologies.

Correspondence to: M. N. Aydnoglu, Department of Earthquake Engineering, Bogazici University, Kandilli Observatory and Earthquake Research Institute (KOERI), 81120 Cengelkoy, Istanbul, Turkey.
E-mail: aydinogn@boun.edu.tr

Copyright ? 2003 John Wiley & Sons, Ltd.

Received 12 April 2002


Revised 27 June 2002 and 22 August 2002
Accepted 22 August 2002

872


M. N. AYDINOGLU
AND Y. M. FAHJAN

In the analysis of non-linear SDOF systems under earthquake excitation, the hysteretic
behaviour is generally represented by multi-linear hysteresis loops [1]. The solution of the
equation of motion can be achieved by various numerical methods, details of which can be
found in the standard textbooks [2; 3]. However, it is rather surprising that the so-called
piecewise exact method based on a simple recursive solution of equation of motion has
attracted relatively little attention in the literature with emphasis rather given to the solution
of linear systems [3]. As a matter of fact, the inception of the method for linear systems
by Nigam and Jennings dates back to the late 1960s [4; 5] followed by the extension to
inelastic systems by Gates and Iwan a decade later [6; 7]. In the 1980s Nau [8; 9] developed
the recurrence formulae for bilinear hysteretic model without the P-delta eect, which are
also adopted by Gupta [10]. It is also reported [11] that the method has been applied to the
elastic-perfectly plastic hysteretic model with P-delta eect included.
The attractiveness of the method lies in the fact that it is based on a closed-form exact
solution of equation of motion when the loading time history is composed of piecewise linear
segments, a condition that is perfectly satised for the earthquake excitation. The accuracy of
the solution is easily controlled with a suciently small time step size.
Since the basis of the method is well known and it has already been implemented by a
number of researchers as referred above, the present paper has no claim of originality for the
basics of the method. However it is believed that a general formulation is still needed to cover
all possible situations encountered in the solution of non-linear SDOF systems. Recognizing
this need, the piecewise exact method is re-formulated in this paper in a unied format,
which is applicable to any type of multi-linear hysteretic models including the P-delta eect.
The formulation does not require dierent equations of motion to be written for each linear
segment of the hysteresis loops. With a single general expression, it is capable of handling
the positive, negative or zero stiness of any linear segment.
As long as the response is developed along the linear segments of the hysteresis loops, it can
be determined by simple recurrence relationships, for which recurrence coecients are given
in the present paper in tabulated form. However, an iterative procedure is necessary to capture
the transition points at the intersection of the linear segments associated with the specied
displacements (e.g. yield points) and the maximum displacements (unloading points) where
the relative velocity becomes zero. When such points are detected, the time step is iteratively
reduced to obtain the solution with an acceptable error margin, for which simple and ecient
algorithms can be applied.
With the use of simple recurrence relationships coupled with the precise evaluation of
transition points, the unied form of the piecewise exact method given in this paper proves
to be an extremely powerful and probably the best tool for the non-linear SDOF time-history
analysis and in particular for the response spectrum analysis applicable to all types of multilinear hysteresis models including the P-delta eect.

2. EQUATION OF MOTION INCLUDING P-DELTA EFFECT


The seismic response of a non-linear SDOF system is governed by the following equation of
motion:
mu(t)
 + cu(t)
+ fS (t) = mu g (t)
Copyright ? 2003 John Wiley & Sons, Ltd.

(1)

Earthquake Engng Struct. Dyn. 2003; 32:871890

873

UNIFIED FORMULATION OF THE PIECEWISE EXACT METHOD

fs(t)

fs(t)
k

fy

fy

uy

uy

u(t)

(a)

u(t)

(b)

Figure 1. Standard hysteresis models: (a) bilinear model. (b) peak-oriented model.

in which m and c represent the mass and linear viscous damping coecient, respectively,
u(t)
 and u(t)
refer to the relative acceleration and velocity, and u g (t) denotes the earthquake
ground acceleration. The third term on the left-hand side represents the restoring force, which
can be expressed as
fS (t) = f(t) fG (t)

(2)

where f(t) refers to the non-linear internal force and fG (t) represents the equivalent P-delta
force. The latter is expressed as
fG (t) = kG u(t)

(3)

where u(t) refers to relative displacement and kG represents the geometric stiness coecient.
On the basis of linear displacement approximation [2], which is generally adopted to express
the P-delta eect, the geometric stiness coecient is dened as
kG =

N mg
=
h
h

(4)

in which N represents the weight, g denotes the acceleration of gravity and h is the height
of the SDOF pendulum.
In the seismic demand analysis, the hysteresis curves representing the non-linear behaviour
are generally assumed to be composed of linear segments as shown in Figure 1. The gure
depicts two basic standard hysteresis models, namely, the bilinear model and the peak-oriented
model with positive or zero post-yield stiness. The model shown in Figure 1(b) is known as
the CloughJohnston model without stiness degradation in unloading [1] modied by Mahin
and Bertero [12] such that the response in reloading is oriented to the immediately preceding
unloading point. The bilinear version of the Takeda model [1; 13] is similar, but it considers
the strength degradation during unloading in the outer as well as in the inner hysteresis loops.
Copyright ? 2003 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2003; 32:871890


M. N. AYDINOGLU
AND Y. M. FAHJAN

874

fs(t)

fs(t)

k
(1-)k

fy

(1-)k

fy

tan k

tan k

-1

-1

e2 k

e2 k

uy
uy

u(t)

u(t)

tan-1k

(a)

(b)

Figure 2. Deviation of standard hysteresis models due to the P-delta eect:


(a) bilinear model, (b) peak-oriented model.

Figure 2 shows the deviations of the hysteresis loops due to the P-delta eect characterized
by the stability coecient , which is expressed as
=

k G mg
=
k
kh

(5)

where k refers to the initial stiness coecient. In Figures 1 and 2, the post-yield stinesses
are indicated with k and e2 k for the cases excluding and including the P-delta eect,
respectively, the latter of which is dened as,
e2 k = ( )k

(6)

which may be negative when the P-delta eect is signicant, as indicated in Figure 2.

3. RECURSIVE SOLUTION OF NON-LINEAR SDOF SYSTEM


The restoring force fS (t) in Equation (1) may be appropriately expressed as
fS (t) = fS (ti ) + ej k[u(t) u(ti )]

(7)

where u(ti ) and fS (ti ) refer to the displacement and the restoring force including the P-delta
eect, respectively, determined previously at a given time station t = ti (Figure 3). ej k is the
eective stiness, representing the slope of any linear segment j of the hysteresis loop in
terms of the initial stiness k. In the initial elastic range, e1 = 1 and e1 = (1 ) and in the
post-yield range, e2 =  and e2 =   for the cases where the P-delta eect is excluded
Copyright ? 2003 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2003; 32:871890

875

UNIFIED FORMULATION OF THE PIECEWISE EXACT METHOD

fs(t)

fs(t)

fs(t)

e2k

e3k

fs(ti)

fs(ti)
e1k =(1-)k
e1k =(1-)k

fs(t)

u(ti)

u(t)

u(t) u(ti)

u(t)

u(t)

Figure 3. Displacements and restoring forces at time t and ti .

and included, respectively, as indicated in Figure 2. Substituting Equation (7), Equation (1)
can be rewritten as
mu(t)
 + cu(t)
+ fS (ti ) + ej k[u(t) u(ti )] = mu g (t)

(8a)

which is valid for tti . Dividing both sides by m, the above equation of motion can be
re-arranged as
u(t)
 + 2!u(t)
+ ej !2 u(t) = u g (t)

fS (ti )
+ ej !2 u(ti )
m

(8b)

in which ! = (k=m)1=2 represents the initial natural frequency and  = c=(2m!) is the linear
damping factor of the SDOF pendulum.
On the other hand, the ground acceleration is expressed as a linear function between the
time stations t = ti and t = ti+1 as
u g (t) = u g (ti ) +

u g

t

(9a)

where
u g = u g (ti+1 ) u g (ti )

(9b)

t = ti+1 ti

(9c)

 = t ti

(9d)

Thus, Equation (8b) can be written in the time span ti 6t 6ti+1 , i.e. 066t, as
u()
 + 2!u()
+ ej !2 u() = u G (ti )
Copyright ? 2003 John Wiley & Sons, Ltd.

u g

t

(10)

Earthquake Engng Struct. Dyn. 2003; 32:871890

876


M. N. AYDINOGLU
AND Y. M. FAHJAN

where the eective pseudo ground acceleration u G (ti ) at t = ti is expressed as


u G (ti ) = u g (ti ) +

fS (ti )
ej !2 u(ti )
m

(11)

It is worth noting that the sum of the second and third terms is constant along each linear
segment. Therefore it needs to be calculated only once at the beginning of the segment. In
particular, for linear systems the sum of those terms vanishes and the right-hand side of
Equation (11) reduces to the rst term only.
The solution of Equation (10) is the combination of a complementary solution corresponding
to the free vibration response and the particular solution satisfying the right-hand side of
the same dierential equation. Considering positive, negative and zero eective stinesses,
the general solutions combining the complementary and particular solutions are given in the
following:
(a) For positive eective stiness of linear segment j (ej 0) with ej 2 0:
u() = [C1 cos(!D ) + C2 sin(!D )] exp(!) + C3  + C4

(12a)

in which the complementary solution corresponds to the underdamped free vibration [9]. !D
is dened as
!D = !(ej 2 )1=2

(12b)

Note that in the special case of ej = 2 , the complementary solution is associated with the
critically damped free vibration [9] for which a general solution is not given here. The relevant recurrence coecients can be derived from the underdamped case through Taylor series
expansion as given below by Equations (17).
(b) For negative eective stiness of linear segment j (ej 0) with 0:
u() = [C1 cosh(!D ) + C2 sinh(!D )] exp(!) + C3  + C4

(13a)

in which the complementary solution corresponds to the overdamped free vibration. In this
case !D is dened as
!D = !(ej + 2 )1=2

(13b)

Note that Equations (13) are also applicable to positive eective stiness of linear segment j
with 2 ej [9].
(c) For zero eective stiness of linear segment j (ej = 0) with 0:
u() = C1 exp(2!) + C2 + C3 2 + C4 

(14a)

(d) For zero eective stiness of linear segment j (ej = 0) with  = 0:


u() = C1  + C2 + C3 3 + C4 2

(14b)

The above expressions for the general solutions are also given in Table I. The corresponding
velocity expressions are omitted for the sake of brevity. Integration constants C3 and C4 are
determined rst from the particular solution and they are then substituted into the general
solution to obtain constants C1 and C2 by imposing displacement and velocity conditions
Copyright ? 2003 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2003; 32:871890

877

UNIFIED FORMULATION OF THE PIECEWISE EXACT METHOD

Table I. Integration constants of piecewise exact method for linear segments of hysteresis models with
positive, negative or zero slopes.
General expression for integration constants:

!: initial natural frequency

i ) + Ck3 u G (ti )
Ck = Ck1 u(ti ) + Ck2 u(t
+Ck4 (u g =t)

: linear viscous damping factor


ej : eective stiness ratio of linear segment j

where u G (ti ) = u g (ti ) + fS (ti )=m ej !2 u(ti )

u(ti ): displacement at t = ti
u(t
i ): velocity at t = ti
fS (ti )=m: Pseudo-acceleration at t = ti
u g (ti ): ground acceleration at t = ti
u g : ground acceleration increment
t: time increment
Integration constants for positive and negative eective stinesses with 0
Negative eective stiness: ej 0; 0

Positive eective stiness: ej 0; ej 2 0


!D = ! (ej 2 )1=2
u() = [C1 cos(!D ) + C2 sin(!D )]
exp(!) + C3  + C4

!D = ! (ej + 2 )1=2
u() = [C1 cosh(!D ) + C2 sinh(!D )]
exp(!) + C3  + C4
 = ej !2

Ck

Ck1

Ck2

Ck3

Ck4

C1
1
0
1=
C2
!=!D
1=!D
!=(!D )
C3
0
0
0
C4
0
0
1=
Integration constants for zero eective stiness with 0 and  = 0
Zero eective stiness: ej =0 0
 = 2!
u() = C1 exp() + C2 + C3 2 + C4 

2!=2

( 22 !2 )=(2 !D )
1=
2!=2

Zero eective stiness: ej =0;  = 0


u() = C1  + C2 + C3 3 + C4 2

Ck

Ck1

Ck2

Ck3

Ck4

Ck1

Ck2

Ck3

Ck4

C1
C2
C3
C4

0
1
0
0

1=

1=2

1=3
1=3
1=(2)
1=2

0
1
0
0

1
0
0
0

0
0
0
1=2

0
0
1=6
0

1=
0
0

1=
0
1=

Also applicable to 0ej 2 .

at time station t = ti , i.e. as initial conditions at  = 0. Integration constants C1 C4 can be


expressed in a general form as
i ) + Ck3 u G (ti ) + Ck4
Ck = Ck1 u(ti ) + Ck2 u(t
Copyright ? 2003 John Wiley & Sons, Ltd.

u g
t

(k = 1; 2; 3; 4)

(15)

Earthquake Engng Struct. Dyn. 2003; 32:871890

878


M. N. AYDINOGLU
AND Y. M. FAHJAN

Ck1 to Ck4 are listed in Table I for all integration constants of Equations (12)(14). Substituting the integration constants C1 to C4 , displacement and velocity expressions are evaluated
for the time station t = ti+1 , i.e. for  = t. Thus the following recursive relationships are obtained, through which the displacements and velocities at the time station t = ti+1 are directly
calculated in terms of those obtained at the previous time station t = ti :
i ) + B11 u G (ti ) + B12 u g
u(ti+1 ) = A11 u(ti ) + A12 u(t

(16a)

u(t
i+1 ) = A21 u(ti ) + A22 u(t
i ) + B21 u G (ti ) + B22 u g

(16b)

The coecients of recursive relationships are presented in Table II, which are applicable to
positive, negative or zero eective stiness of any segment of the hysteresis loop. It is worth
noting that the expressions given for Ck1 to Ck4 in Table I and the expressions given for
the recurrence coecients in Table II are common for both positive and negative eective
stinesses.
As noted above, the general solution given by Equation (13a) for the negative eective
stiness (ej 0) and the resulting recurrence coecients are equally applicable to the special
case where 0ej 2 , which is indicated in Table II as a footnote. Another special case arises
when ej is just equal to 2 , which is mentioned above as corresponding to critically damped
free vibration response. In this particular case, !D 0 in Equation (12b) and the recurrence
coecients A11 , A12 and A22 can be obtained from the Taylor series expansion as
A11 = (1 + !t) exp(!t)

(17a)

A12 = t exp(!t)

(17b)

A22 = (1 !t) exp(!t)

(17c)

which are also given in the footnote of Table II. Other recurrence coecients can be determined in terms of the above coecients as dened in Table II.
It is clear that the unied formulation of the piecewise exact method described above for the
earthquake excitation also applies to a general dynamic load, p(t), provided that the loading
time history is discretized as piecewise linear segments. In this case, the right-hand side of
Equation (7) is replaced with p(t), accordingly u g (ti ) and u g in Tables I and II are replaced
with p(ti )=m and p=m, respectively, where p represents the dynamic load increment
between the time stations ti and ti+1 .
4. IMPLEMENTATION OF THE METHOD
In the implementation of the piecewise exact method, the time step t can be taken equal to
the time step of the earthquake record for the response along the linear segments. However,
as a general rule, its maximum value should not exceed (t)max = T=10, where T refers to the
initial natural period of the system (T = 2=!). It means that the time step is to be reduced
accordingly for short-period systems.
The precise determination of peak displacements and intersection points of linear segments
of hysteresis loops is essential for the accuracy of the solution. Therefore zero-velocity control
Copyright ? 2003 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2003; 32:871890

879

UNIFIED FORMULATION OF THE PIECEWISE EXACT METHOD

Table II. Recurrence coecients of piecewise exact method for linear segments of hysteresis models
with positive, negative or zero slopes.
!: Initial natural frequency

Recurrence relationships:
i ) + B11 u G (ti ) + B12 u g
u(ti+1 ) = A11 u(ti ) + A12 u(t

: Linear viscous damping factor

i ) + B21 u (ti ) + B22 u


u(t
i+1 ) = A21 u(ti ) + A22 u(t

ej : Eective stiness ratio of


linear segment j

where u G (ti ) = u g (ti ) + fS (ti )=m ej !2 u(ti )

u(ti ): Displacement at t = ti
u(t
i ): Velocity at t = ti
fS (ti )=m: Pseudo-acceleration at t = ti
u g (ti ): Ground acceleration at t = ti
u g : Ground acceleration increment
t: Time increment

Positive eective stiness

Negative eective stiness

Zero eective

Zero

Amn

ej 0; ej  0

ej 0; 0

stiness

eective

and
Bmn

!D = ! (ej  )
E = cos(!D t) exp(!t)
F = sin(!D t) exp(!t)

!D = ! (ej +  )
E = cosh(!D t) exp(!t)
F = sinh(!D t) exp(!t)

ej = 0 0
 = 2!
G = exp(t)

stiness
ej = 0
=0

2 1=2

2 1=2

A11
A12
A21
A22

E + (!=!D )F
F=!D
ej !2 A12
E (!=!D )F

1
(1 G)=
0
G

1
t
0
1

B11
B12
B21
B22

(A11 1)=(ej !2 )
(A12 2!B11 t)=(ej !2 t)
A12
B11 =t

(A12 t)=
(B11 =t + t=2)=
A12
B11 =t

t 2 =2

t 2 =6
A12

B11 =t

For ej = 2 : A11 = (1 + !t) exp(!t); A12 = t exp(!t); A22 = (1 !t)exp(!t).


Also applicable to 0ej 2 .

is to be made at each and every step to capture the peak displacements at the unloading points.
Although not essential, the same control may be made in the linear response as well to pinpoint
the peak displacements precisely. The displacement control is performed for the intersection
points of linear segments of the hysteresis loops dened by specied displacements, such as the
yield points and zero force crossings. The time step t is iteratively reduced in the vicinity of
those transition points in order to determine the response precisely with an acceptable relative
error margin. The bisection and=or secant methods can be eciently used in the iteration
process to pinpoint the transition at a reduced time step, t, and the remaining part of the
time step, (1 )t, can be used immediately after the transition point, followed by the use
of the regular time step t.
Copyright ? 2003 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2003; 32:871890

880


M. N. AYDINOGLU
AND Y. M. FAHJAN

The piecewise exact method has been implemented in an in-house computer software,
KOERINON, developed by the authors at the Kandilli Observatory and Earthquake Research
Institute (KOERI), Istanbul. The current version of the software includes the bilinear and
peak-oriented (modied CloughJohnston and bilinear Takeda) hysteresis models [12; 13]. It
performs the time history analysis for a specied ground motion as well as the response spectrum analysis on the basis of constant ductility and constant strength reduction concepts [14],
producing a variety of inelastic response spectra. P-delta eects can be represented by either the stability coecient or the so-called geometric frequency, however the latter is shown
below to be the representative P-delta parameter for the response spectrum analysis.

5. REPRESENTATIVE P-DELTA PARAMETER FOR RESPONSE


SPECTRUM ANALYSIS
In the existing literature on SDOF response spectrum analysis, such as those dealing
with the constant-ductility response spectra, P-delta eects are traditionally characterized
by the constant values of the stability coecient [11; 1517]. However it is shown below
that the use of a constant stability coecient in any type of a period-dependent response
spectrum is misleading, because the stability coecient itself is a period-dependent
parameter.
From Equation (5), the stability coecient may also be expressed as
=

!G2
!2

(18)

in which !G may be appropriately called as the geometric frequency [18], which is


dened as
!G2 =

g
h

(19)

Note that !G2 is a constant quantity for a pendulum with a given height and it is independent
of its initial stiness and hence its initial period. Thus it is clear from Equation (18) that
the stability coecient is a period-dependent parameter. This observation also provides a
means for estimating the possible range of stability coecients, which cannot be specied
arbitrarily for a pendulum with a given height. For example, if the height is taken as h = 3 m
for a pendulum with an initial period of T = 0:5 s, specifying a stability coecient of, say,
 = 0:05 is meaningless because the stability coecient consistent with this period cannot
exceed  = 0:02.
It is worth noting that the constant geometric frequency is not only applicable to a SDOF
system. It has been shown that it is also a constant quantity for a MDOF system responding
in its fundamental mode [18]. Thus it can be concluded that in the SDOF response spectrum
analysis with multi-linear hysteresis models, for example, in developing the constant-ductility
acceleration response spectra or the strength-based and ductility-based displacement amplication spectra (see the examples below), the geometric frequency squared, !G2 , should be
considered as the representative P-delta parameter. The corresponding stability coecient, ,
is then calculated from Equation (18) as a period-dependent variable.
Copyright ? 2003 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2003; 32:871890

881

UNIFIED FORMULATION OF THE PIECEWISE EXACT METHOD

6. EXAMPLES
The following examples are presented to demonstrate the capabilities of the piecewise exact
method and its computer implementation in KOERINON.
As a rst example, the non-linear time-history response of a SDOF system with an initial
period of T = 1:20 s and a linear damping factor of  = 0:05 is given for the classical El

0.05

Displacement, u (t)

-0.05

-0.1

-0.15

-0.2
0

10

15

(a)

20

25

30

35

40

25

30

35

40

Time, t (s)
0.05

Displacement, u (t)

-0.05

-0.1

-0.15

-0.2
0

10

(b)

15

20
Time, t (s)

Figure 4. Inelastic displacement time-history of a SDOF system (T = 1:2 s,  = 0:05,


Ry = 4) for the El Centro (1940) NS record without the P-delta eect: (a) bilinear
model, (b) modied CloughJohnston model.
Copyright ? 2003 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2003; 32:871890


M. N. AYDINOGLU
AND Y. M. FAHJAN

882
0.05

Displacement, u (t)

-0.05

-0.1

-0.15

-0.2
0

10

15

(a)

20

25

30

35

40

25

30

35

40

Time, t (s)
0.05

Displacement, u (t)

-0.05

-0.1

-0.15

-0.2
0

10

(b)

15

20
Time, t (s)

Figure 5. Inelastic displacement time-history of a SDOF system (T = 1:2 s,


 = 0:05, Ry = 4) for the El Centro (1940) NS record without the P-delta eect
( = 0:05; !G2 = 1:37 rad=s2 ): (a) bilinear model, (b) modied CloughJohnston model.

Centro (NS) record of the 1940 Imperial Valley earthquake. Bilinear and peak-oriented
(modied CloughJohnston) models are used with zero post-yield stiness. The strength reduction factor of the SDOF system is taken as Ry = 4, resulting in a yield pseudo-acceleration
of Say = fy =m = 0:80 m=s2 (0:0815 g) for the given initial period. The stability coecient is
taken as  = 0:05, which corresponds to a geometric frequency squared of !G2 = 1:37 rad=s2
in accordance with Equation (18). Figures 4 and 5show the displacement response timeCopyright ? 2003 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2003; 32:871890

883

UNIFIED FORMULATION OF THE PIECEWISE EXACT METHOD

Pseudo-acceleration, f s(t)/m (m/s )

0.5

-0.5

-1
-0.2

-0.15

-0.1

(a)

-0.05

0.05

0.05

Displacement, u (t )

Pseudo-acceleration, f s (t)/m (m/s2)

0.5

-0.5

-1
-0.2

-0.15

(b)

-0.1

-0.05
Displacement, u (t )

Figure 6. Hysteretic response of a SDOF system (T = 1:2 s;  = 0:05; Ry = 4)


for the El Centro (1940) NS record without the P-delta eect: (a) bilinear
model, (b) modied CloughJohnston model.

histories without and with P-delta eect for both hysteresis models, respectively, and the
associated hysteresis loops are given in Figures 6 and 7. It is clearly observed that the
P-delta eect amplies the displacement response excessively in the bilinear model compared
to the peak-oriented model, although the latter model exhibits comparably higher response
when the P-delta eect is excluded.
Copyright ? 2003 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2003; 32:871890


M. N. AYDINOGLU
AND Y. M. FAHJAN

884

Pseudo-acceleration, f s (t)/m (m/s )

0.5

-0.5

-1
-0.2

-0.15

(a)

-0.1

-0.05

0.05

0.05

Displacement, u (t )

Pseudo-acceleration, f s (t)/m (m/s )

0.5

-0.5

-1
-0.2

-0.15

(b)

-0.1

-0.05
Displacement, u (t )

Figure 7. Hysteretic response of a SDOF system (T = 1:2 s;  = 0:05; Ry = 4) for the


El Centro (1940) NS record with the P-delta eect ( = 0:05; !G2 = 1:37 rad=s2 ): (a)
bilinear model, (b) modied CloughJohnston model.

In the second example, constant-ductility response spectra are demonstrated for the same
earthquake record in terms of strength reduction factor spectra (Ry spectra) without and with
the P-delta eect, as shown in Figures 8 and 9 for the same hysteretic models, respectively.
The target ductility factors are taken as
= 1:5; 2; 3; 4; 5; 6 and the linear damping factor is
taken as  = 0:05. The P-delta eect is represented by a constant value of the geometric
Copyright ? 2003 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2003; 32:871890

885

UNIFIED FORMULATION OF THE PIECEWISE EXACT METHOD

12

Bottom to top
= 1.5, 2, 3, 4, 5, 6

Strength Reduction Factor, Ry

10

0
0

0.5

(a)

1.5

2.5

2.5

Period, T (s)
12

Bottom to top
= 1.5, 2, 3, 4, 5, 6

Strength Reduction Factor, Ry

10

0
0

0.5

(b)

1.5
Period, T (s)

Figure 8. Strength reduction factor spectra of the El Centro 1940 NS record for specied
target ductility factors
= 1:5; 2; 3; 4; 5; 6 and  = 0:05 without the P-delta eect: (a) bilinear
model, (b) modied CloughJohnston model.

frequency squared of !G2 = 1:00 rad=s2 , which yields period-dependent stability coecients, ,
as depicted on the right-hand side vertical axes in Figure 9. It is seen that the P-delta eect
reduces the strength reduction factor, in other words it increases the strength demand, in both
hysteretic models particularly in the long-period range with increasing stability coecients.
However strength demand is relatively less in the peak-oriented model compared to the bilinear
model. In the short-period range the P-delta eect is negligible due to decreasing stability
coecients with the decreasing initial period in accordance with Equation (18).
Copyright ? 2003 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2003; 32:871890


M. N. AYDINOGLU
AND Y. M. FAHJAN

886
12

0.60

Bottom to top
= 1.5, 2, 3, 4, 5, 6

0.40

0.30

0.20

0.10

0.00

0.5

(a)

1.5

2.5

Period, T (s)
12

0.60

Bottom to top
= 1.5, 2, 3, 4, 5, 6

10
Strength Reduction Factor, Ry

Stability coefficient,(T )

0.50

0.50

0.40

0.30

0.20

0.10

Stability coefficient,(T )

Strength Reduction Factor, Ry

10

0.00

0.5

(b)

1.5

2.5

Period, T (s)

Figure 9. Strength reduction factor spectra of the El Centro (1940) NS record for
specied target ductility factors
= 1:5; 2; 3; 4; 5; 6 and  = 0:05 with the P-delta eect
(!G2 = 1:00 rad=s2 ): (a) bilinear model, (b) modied CloughJohnston model (curve with
dashed lines represents the relationship between stability coecient and period).

Finally the third example demonstrates the constant strength reduction spectra for the same
earthquake record in terms of the strength-based displacement amplication spectra [14; 18]
as shown in Figures 10 and 11. In these spectra, a typical ordinate represents the ratio of
the inelastic spectral displacement to the elastic spectral displacement for a given strength
reduction factor. When the P-delta eect is excluded, the strength-based spectral displacement
amplication, which is denoted by SdaR in Figure 10, corresponds to the C1 coecient of the
Copyright ? 2003 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2003; 32:871890

887

UNIFIED FORMULATION OF THE PIECEWISE EXACT METHOD

Strength-based spectral displacement amplification, SdaR

10

Bottom to top
Ry = 1.5, 2, 3, 4, 5, 6

9
8
7
6
5
4
3
2
1
0
0

0.5

(a)

1.5

2.5

2.5

Period, T (s)

Strength-based spectral displacement amplification, SdaR

10

Bottom to top
Ry = 1.5, 2, 3, 4, 5, 6

9
8
7
6
5
4
3
2
1
0

(b)

0.5

1.5
Period, T (s)

Figure 10. Strength-based displacement amplication spectra of the El Centro (1940) NS record for
specied strength reduction factors Ry = 1:5; 2; 3; 4; 5; 6 and  = 0:05 without the P-delta eect: (a) bilinear model, (b) modied CloughJohnston model.

displacement coecient method given in FEMA 356 [19] for the performance-based seismic
evaluation [14; 18]. On the other hand when the P-delta eect is included, the strength-based
spectral displacement amplication, which is denoted by SdaS in Figure 11, corresponds to
the product of C1 and C3 coecients dened in the same method [18; 19]. In the strengthbased displacement amplication spectra, the constant strength reduction factors are taken as
Ry = 1:5; 2; 3; 4; 5; 6 and the linear damping factor is taken as  = 0:05. The P-delta eect is
again represented by a constant value of the geometric frequency squared of !G2 = 1:00 rad=s2
Copyright ? 2003 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2003; 32:871890


M. N. AYDINOGLU
AND Y. M. FAHJAN

888

0.50

Bottom to top
R y = 1.5, 2, 3, 4, 5, 6

Ry = 3

7
0.30

5
4

0.20

3
2

0.10

1
0

0.00
0

0.5

(a)

1.5

2.5

Period, T (s)
0.50

10

Bottom to top
R y = 1.5, 2, 3, 4, 5, 6

Ry = 5

7
0.30

5
4

0.20

3
2

Stability coefficient, (T)

0.40

Ry = 6

Strength-based spectral displacement amplification,


S daS

Stability coefficient, (T )

0.40

Ry = 6

Strength-based spectral displacement amplification,


S daS

10

0.10

1
0

0.00
0

0.5

(b)

1.5

2.5

Period, T (s)

Figure 11. Strength-based displacement amplication spectra of the El Centro (1940) NS record
for specied strength reduction factors Ry = 1:5; 2; 3; 4; 5; 6 and  = 0:05 with the P-delta eect
(!G2 = 1:00 rad=s2 ): (a) bilinear model, (b) modied CloughJohnston model (curve with dashed lines
represents the relationship between stability coecient and period).

and the period-dependent stability coecients, , are depicted on the right-hand side vertical
axes in Figure 11. It is seen that the P-delta eect may lead to the dynamic instability of
the SDOF system at certain periods where the displacement response becomes excessive for
a given strength reduction factor. In the bilinear model the dynamic instability occurs in the
period range between 1.0 and 1.5 s for systems with strength reduction factors Ry = 3; 4; 5; 6.
However in the more stable peak-oriented model it occurs for only Ry = 5; 6 at much longer
Copyright ? 2003 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2003; 32:871890

UNIFIED FORMULATION OF THE PIECEWISE EXACT METHOD

889

periods. The dynamic instability phenomenon has recently been investigated in detail by Aydinoglu [18].

7. CONCLUSIONS
The so-called piecewise exact method based on the recursive solution of equation of motion
of SDOF systems is re-formulated in a unied format for the non-linear response including the
P-delta eect, in which the post-yield stiness of the hysteresis loops could become negative.
The complete solution of equation of motion is formulated to handle the positive, negative
or zero stiness of any linear segment of the multi-linear hysteresis loops, and the resulting
simple recurrence equations provide the exact solution including the P-delta eect.
The accuracy of the solution is easily controlled with a suciently small time step size.
In order to capture the transition points at the intersection of linear segments, simple iterative procedures based on the secant and=or bisection methods can be used. Thus, the general
form of the piecewise exact method developed in this paper provides an extremely ecient
tool for the non-linear time-history and response spectrum analysis of SDOF systems excluding or including the P-delta eect. A number of examples are presented to demonstrate the
implementation of the method.
On the other hand, it is shown in the paper that the stability coecient should not be used
as the controlling P-delta parameter in the development of the response spectra, because of
the fact that it is a period-dependent quantity. Instead, the so-called geometric frequency is
shown to be period-independent and constant quantity for a SDOF system as well as for a
MDOF system responding in its fundamental mode. Thus it is recommended that the geometric
frequency be used in lieu of the stability coecient as the representative P-delta parameter
for the inelastic response spectra.

ACKNOWLEDGEMENTS

The authors are grateful to two anonymous reviewers for their valuable comments towards the enhancement of the paper.
REFERENCES
1. Otani S. Hysteresis models for earthquake response analysis. In Earthquake Resistance of Reinforced Concrete
Structures, ed. T. Okada. University of Tokyo Press: Tokyo, 1993; 387398.
2. Clough RW, Penzien J. Dynamics of Structures. McGraw-Hill; New York, 1993.
3. Chopra AK. Dynamics of Structures. Prentice-Hall: Englewood Clis, NJ, 1995.
4. Nigam NC, Jennings PC. Digital calculations of response spectra from strong-motion earthquake records.
Earthquake Engineering Research Laboratory, California Institute of Technology, June 1968.
5. Nigam NC, Jennings PC. Calculation of response spectra from strong motion earthquake records. Bulletin of
the Seismological Society of America 1969; 59:909 922.
6. Gates NC. The earthquake response of deteriorating systems. Earthquake Engineering Research Laboratory,
California Institute of Technology, EERL 77-03, March 1977.
7. Iwan WD, Gates NC. The eective period and damping of a class of hysteretic structures. Earthquake
Engineering and Structural Dynamics 1979; 7: 199 211.
8. Nau JM. Computation of inelastic response spectra. Journal of Engineering Mechanics (ASCE) 1983; 109:
279 288.
Copyright ? 2003 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2003; 32:871890

890


M. N. AYDINOGLU
AND Y. M. FAHJAN

9. Nau JM. Response spectra for bilinear hysteretic systems. Journal of Structural Engineering (ASCE) 1986;
112:17271731.
10. Gupta AK. Response Spectrum Method in Seismic Analysis and Design of Structures. CRC Press: Boston,
1990.
11. Mazzolani FM, Piluso V. Theory and design of seismic resistant steel frames. E&FN Spon, 1996.
12. Mahin SA, Bertero VV. Nonlinear seismic response of a coupled wall system. Journal of the Structural Division
(ASCE) 1976; 102:1759 1780.
13. Takeda T, Sozen MA, Nielsen NN. Reinforced concrete response to simulated earthquakes. Journal of the
Structural Division (ASCE) 1970; 96:25572573.
14. Aydinoglu MN, Kacmaz U. Consistent inelastic response spectra for displacement-based seismic evaluation.
Earthquake Engineering and Structural Dynamics, 2002, submitted for publication.
15. Bernal D. Amplication factors for inelastic dynamic P- eects in earthquake analysis. Earthquake Engineering
and Structural Dynamics 1987; 15:635 651.
16. Rahmana M. Eects of soft soil and hysteresis model on seismic demands. PhD Dissertation, Stanford
University, CA, 1993.
17. MacRae GA. P- eects on single-degree-of-freedom structures in earthquakes. Earthquake Spectra 1994;
10:539 568.
18. Aydinoglu MN. Inelastic structural capacity and seismic demand including P-delta eects, Kandilli Observatory
and Earthquake Research Institute, Department of Earthquake Engineering, Report No. 2002/1, Istanbul, Turkey,
2002.
19. Federal Emergency Management Agency. Prestandard and commentary for the seismic rehabilitation of
buildings (FEMA 356) Washington, DC, 2000.

Copyright ? 2003 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2003; 32:871890

Vous aimerez peut-être aussi