1. Introduction
Over the past 45 years there have been many studies [e.g. Veletsos and Newmark,
1960; Iwan, 1980; Miranda, 2000] investigating the response of structures with dif-
ferent hysteretic rules to earthquake records using inelastic time-history analyses.
Initially, the purpose of these studies was to determine the required force reduction
factors, R, (or behaviour factors q) to be used in force-based seismic design that
would result in a specified displacement ductility µ being achieved. Typically, it was
found that the results were rather insensitive to the hysteretic rule adopted, and that
for initial (elastic) periods greater than about 0.5 seconds the equal-displacement
approximation, which states that R = µ, provides an adequate approximation of
the behaviour. For periods shorter than about 0.5 seconds, R < µ was required,
with the value of R decreasing as the period reduced. These findings agreed with
the earlier recommendations of Newmark and Hall [1982].
More recently, a number of studies [e.g. Judi et al., 2000; Miranda and
Ruiz-Garcia, 2002] have investigated the response of inelastic systems with different
hysteretic rules with the specific intent of determining the relationship between duc-
tility and equivalent viscous damping for use in Direct Displacement-Based Design
229
December 20, 2005 13:47 WSPC/124-JEE 00236
Fu
me Fy r ki
F
ki keff
he
∆y ∆d
(a) (b)
30 5%
0.5
Elasto - Plastic
0.4
Displacement (m)
10%
Steel Frame
Damping (%)
20 15%
Concrete Frame 0.3 20%
∆d 30%
0.2
10 Unbonded
Prestressing
0.1
Te
0 0
1 2 3 4 5 6 0 1 2 3 4 5
Displacement Ductility Period (seconds)
(c) (d)
Fig. 1. Fundamentals of Direct Displacement-Based Design. (a) Simulation of MDOF structure
with SDOF, (b) effective stiffness versus ductility (c) equivalent viscous damping versus ductility,
and (d) design displacement spectrum.
An aspect of DDBD that has not received much attention is the characterisa-
tion of the initial elastic viscous damping, where “elastic” is used in this paper to
distinguish from the “equivalent viscous damping” used in the substitute structure
approach. In the past, as is common in many areas of seismic engineering, it has
been assumed that a value of 5% initial elastic damping is appropriate, and should
be added to the hysteretic damping, as indicated in Fig. 1(c). Thus, the fraction
of critical damping, ξeq , for the linear substitute structure is given as a function of
the target displacement ductility, µ, by an expression of the following form:
ξeq = ξeq,v + ξeq,h,µ , (1)
where ξeq,v and ξeq,h,µ represent energy dissipation resulting from viscous damping
and inelastic hysteresis, respectively, with ξeq,v generally assumed to be 5%, as noted
above.
However, it is apparent that there is a possible inconsistency here. In DDBD
the 5% initial elastic damping is related to the secant stiffness to maximum dis-
placement, whereas it is normal in SDOF dynamic analysis to relate the elastic
damping to either the initial (elastic) stiffness, or less commonly, a stiffness that
varies as the structural stiffness degrades with inelastic action (tangent stiffness).
Although the terminology is inexact, since the damping coefficient is proportional
to the square root of the stiffness, we will term these “initial stiffness proportional”
and “tangent stiffness proportional” damping, as they are analogous to damping
assumptions used for MDOF dynamic analysis. Since the response velocities of
the “real” and “substitute” structures are expected to be similar under seismic
response, the damping force, which is equal to the product of the damping coeffi-
cient and the velocity, will differ significantly, since the effective stiffness keff of the
substitute structure is approximately equal to keff = ki /µ (for low post-yield stiff-
ness). It would appear that some adjustment would be needed to the value of initial
elastic damping assumed in DDBD to ensure compatibility between the “real” and
“substitute” structures. Without such an adjustment, the verification of DDBD by
inelastic time-history analysis would be based on incompatible assumptions.
The discussion in the previous paragraph raises the question of how to char-
acterise initial elastic damping in inelastic time-history analysis (ITHA). Typi-
cally research papers reporting results on SDOF ITHA state that 5% initial elastic
damping was used, without clarifying whether this has been related to the ini-
tial or tangent stiffness. This is because in SDOF analyses, no special damping
model is required, and the damping coefficient is assumed to be constant. With
MDOF analyses, the situation is often further confused by the adoption of Rayleigh
damping, which is a combination of mass proportional and stiffness proportional
damping. It is believed that many analysts consider the choice of initial elastic
damping model to be rather insignificant for either SDOF or MDOF inelastic anal-
yses, as the effects are expected to be masked by the much greater energy dissi-
pation associated with hysteretic response. This is despite evidence by others [e.g.
Otani, 1981] that the choice of initial damping model between a constant damping
December 20, 2005 13:47 WSPC/124-JEE 00236
hysteretic rules such as the modified Takeda degrading stiffness rule which has
comparatively high stiffness in post-yield cycles.
The second point considers foundation damping. Assume an essentially elastic
foundation, and a structural response with low post-yield stiffness. Then, during the
intervals of time in which the structure exhibits ductility and deforms plastically,
the foundation forces, and hence foundation deformation, remain constant. Clearly
no energy is dissipated by the foundation during these time intervals. It is thus clear
that the effects of foundation damping in SDOF analysis are best represented by
tangent stiffness related to the structural response, unless the foundation response
is separately modelled by springs and dashpots.
It is conceivable that the non-structural damping force is displacement-
dependent rather than force-dependent, and hence a constant damping coefficient
may be appropriate for the portion of “elastic” damping that is attributable to non-
structural forces. There are two possible contributions to non-structural damping
that should be considered separately:
• Energy dissipation due to hysteretic response of the nonstructural elements
• Energy dissipation due to sliding between nonstructural and structural elements
For a modern frame building, separation between structural and non-structural
elements is required, and hence they should not contribute significantly to damping.
Further, even if not separated, the lateral strength of all non-structural elements is
likely to be less than 5% of the structural lateral strength (unless the non-structural
elements are masonry infill). If we assume 10% viscous damping in these elements,
an upper bound of about 0.5% equivalent viscous damping related to the structural
response seems reasonable. Nonstructural elements are unlikely to play a significant
role in the response of bridges.
Sliding will normally relate to a frictional coefficient, and the weight of the non-
structural element. Unless the non-structural elements are masonry, the frictional
force is likely to be negligible. It should be noted that it is probable that so-called
nonstructural infill initially contributes more significantly to strength, stiffness, and
damping than is the case with (e.g.) lightweight partitions. However, it is known
that the strength degrades rapidly for drift levels > 0.5% (which is generally less
than structural yield drift). The damping force is also likely to degrade rapidly. The
effects of non-structural masonry infill should be modelled by separate structural
elements with severely degrading strength and stiffness — not by increased viscous
damping.
It is thus recommended that for modern buildings with separated or lightweight
non-structural elements, viscous energy dissipation should be modelled by a
tangent-stiffness proportional damping coefficient. Note that the specification of
tangent-stiffness proportional Rayleigh damping for MDOF, multi-storey build-
ings will not have the desired effect, since most of the elastic damping in the
critical first mode will be mass proportional, which is constant with inelastic
action. Consider the basic form for the fraction of critical damping in the Rayleigh
December 20, 2005 13:47 WSPC/124-JEE 00236
damping model:
1 α
ξ= + βωn , (2)
2 ωn
where α and β are the coefficients associated with mass proportional and stiffness
proportional damping respectively, and ωn is the circular frequency. If we specify
the same value (say 5%) for ξ at two different frequencies, where the higher one is
κ times the lower (fundamental) frequency, then:
1 α
ξ= + βω1 , (3)
2 ω1
and
1 α
ξ= + βκω1 . (4)
2 κω1
Subtracting Eq. (4) from κ times Eq. (3):
α κ2 − 1
ξ(κ − 1) = , (5)
2ω1 κ
and hence the damping α/2ω1 associated with mass proportional damping in the
first mode is ξκ/(κ + 1), while the damping attributed to stiffness proportional
damping in the first mode is ξ/(κ + 1).
Consider the case where we specify 5% damping at T1 = 1.5 sec and T2 = 0.3 sec.
Hence κ = 5. Then, even if we specify tangent-stiffness proportional Rayleigh damp-
ing, only 0.83% is stiffness proportional, while 4.17% is mass proportional, and hence
acts in an identical manner to initial-stiffness damping when the structure responds
inelastically.
proportional damping in the program, although for SDOF analysis, the results
will be identical to the mass proportional case. As noted above, “initial-stiffness
proportional” is a misnomer for SDOF analysis, as the damping coefficient, and thus
the damping force, is actually proportional to the square root of the initial stiffness
(see Eq. (8)). The steady-state response of the pier corresponded to a displacement
ductility of about 7.7 — at the upper limit of reasonable ductile response.
Results for the stabilised loops, ignoring the transitory first three seconds
of response are plotted in Fig. 2(a) (initial-stiffness proportional damping) and
Fig. 2(b) (tangent-stiffness proportional damping). In each case the hysteretic
response associated with nonlinear structural response is plotted on the left, and the
elastic damping force-displacement response is plotted to the right. The areas inside
the loops indicate the relative energy absorption. For the case with initial-stiffness
2000 2000
1000 1000
Damping Force (kN)
Stiffness Force (kN)
0 0
-1000 -1000
-2000 -2000
-0.1 0 0.1 -0.1 0 0.1
Displacement (m) Displacement (m)
(a)
2000 2000
1000 1000
Damping Force (kN)
Stiffness Force (kN)
0 0
-1000 -1000
2
- 000 -2000
-0.1 0 0.1 -0.1 0 0.1
Displacement (m) Displacement (m)
(b)
Fig. 2. Steady-state inelastic response of a SDOF model to harmonic excitation. (a) Initial-
stiffness proportional damping, and (b) tangent-stiffness proportional damping.
December 20, 2005 13:47 WSPC/124-JEE 00236
r ki r ki r ki
ki
Fy β Fy ki
ki ku ki ki
r ki
r ki r ki r ki
(a) (b) (c)
Fig. 3. Hysteresis models included in parameter study. (a) Modified Takeda model, r = 0.002
and 0.05, (b) Bilinear model, r = 0.002 and 0.05, and (c) Flag model, r = 0.05, β = 0.35 and 0.70.
December 20, 2005 13:47 WSPC/124-JEE 00236
238
Equivalent Viscous Damping ratio M. J. N. Priestley & D. N. Grant
2
Takeda, r=0.002 r = 0.05
0.2
1.6
Flag, β =0.70
Takeda, r=0.05
Flag, β =0.35 1.2
0
0 2 4 6 8 0 2 4 6 8
Displacement Ductility, µ Displacement Ductility, µ
(a) (b)
Fig. 4. Definition of equivalent linear properties. (a) Equivalent viscous damping and (b) effective
period for different hysteresis rules.
0.8
0.6
Spectral Displacement
0.4
Solid Line: Target Spectrum
Dashed Line: Average Spectrum
Thin lines: Individual Spectra
0.2
0
0 1 2 3 4
Period (seconds)
spectra for 5% damping for the five artificial records, their average, and the design
displacement spectrum are compared in Fig. 5.
There is a tendency to use “real” earthquake records for parameter studies of
this kind, as it is more representative of expected response than will be the case
with artificial records. In such cases the real accelerograms are scaled to match the
design spectrum at the initial period of the structure analysed. However, unless the
accelerogram also matches the design spectrum at the degraded effective period at
maximum displacement response, the results will not be appropriate for the design
spectrum. The use of a large number of “real” accelerograms will not improve this,
unless the average spectrum matches the design spectrum over the full range of
December 20, 2005 13:47 WSPC/124-JEE 00236
periods. In the longer period range (say from 1.5 to 4 seconds), the “real” records
need to be chosen carefully as individual records tend to have acceleration spectra
that decrease more rapidly than given by typical uniform risk spectra. For this
reason, and to reduce the number of analyses necessary, the bulk of the study in
this section was carried out using artificial records, although the results from the
El Centro record, which showed greater scatter, but similar trends, were included
in the averaged results.
0.15
Tangent Stiffness
0.1
Initial Stiffness
Displacement (m)
0.05
-0.05
-0.1
0 4 8 12 16 20
Time (seconds)
Fig. 6. Displacement response of a SDOF model, T = 0.5 sec, to 1.5 × El Centro 1940
NS component, using tangent-stiffness and initial-stiffness viscous damping. (Takeda hysteresis,
R = 4, r = 0.05).
high, have been omitted when greater than about 2.7, so that the trends at larger
periods are more apparent.
From these plots, and Table 1, the following conclusions can be drawn:
TS
2 2 2
IS
0.8 0.8 0.8
0 0.4 0.8 1.2 1.6 2 0 0.4 0.8 1.2 1.6 2 0 0.4 0.8 1.2 1.6 2
Period (seconds) Period (seconds) Period (seconds)
(a)
TS
2 2 2 TS
0 0.4 0.8 1.2 1.6 2 0 0.4 0.8 1.2 1.6 2 0 0.4 0.8 1.2 1.6 2
Period (seconds) Period (seconds) Period (seconds)
(b)
Fig. 7. Average ratio of peak inelastic displacement to elastic displacement for modified Takeda
hysteresis. (TS = Tangent-stiffness proportional damping, IS = Initial-stiffness proportional
damping, R = force-reduction factor). (a) r = 0.002 and (b) r = 0.05.
As noted above, Figs. 7 to 9 show the results of the tangent-stiffness and initial-
stiffness analyses as separate lines. In order to facilitate a quantitative examination
of the effect of the choice of damping model, the T/I results of Table 1 have been
plotted for the three types of hysteresis model in Fig. 10. For Takeda hystere-
sis, the tangent-stiffness displacements are about 10%, 20%, and 35% higher than
initial-stiffness displacements, on average, for force-reduction factors of 2, 4, and 6
December 20, 2005 13:47 WSPC/124-JEE 00236
2 2 2
TS
TS IS
1.2 1.2 1.2
IS
0 0.4 0.8 1.2 1.6 2 0 0.4 0.8 1.2 1.6 2 0 0.4 0.8 1.2 1.6 2
Period (seconds) Period (seconds) Period (seconds)
(a)
2 2 2 TS
0 0.4 0.8 1.2 1.6 2 0 0.4 0.8 1.2 1.6 2 0 0.4 0.8 1.2 1.6 2
Period (seconds) Period (seconds) Period (seconds)
(b)
Fig. 8. Average ratio of peak inelastic displacement to elastic displacement for Bilinear hysteresis.
(TS = Tangent-stiffness proportional damping, IS = Initial-stiffness proportional damping, R =
force-reduction factor). (a) r = 0.002 and (b) r = 0.05.
respectively; for Bilinear hysteresis, the ratios are approximately 4%, 8%, and 15%,
while for Flag hysteresis the average ratios are about 20%, 30% and 50%. Generally
the ratios decrease as the period increases. It is felt that, except for Bilinear with
low force-reduction factors, these ratios are too large to be ignored in ITHA.
2 2 TS 2 TS
IS
0.8 0.8 0.8
0 0.4 0.8 1.2 1.6 2 0 0.4 0.8 1.2 1.6 2 0 0.4 0.8 1.2 1.6 2
Period (seconds) Period (seconds) Period (seconds)
(a)
TS
2 2 2 TS
IS
0.8 IS 0.8 0.8
0 0.4 0.8 1.2 1.6 2 0 0.4 0.8 1.2 1.6 2 0 0.4 0.8 1.2 1.6 2
Period (seconds) Period (seconds) Period (seconds)
(b)
Fig. 9. Average ratio of peak inelastic displacement to elastic displacement for flag hysteresis.
(TS = Tangent − stiffness proportional damping, IS = Initial − stiffness proportional damping,
R = Force − reduction factor). (a) β = 0.35, and (b) β = 0.70.
is multiplied by the velocity to obtain the viscous damping force, and therefore
the viscous energy dissipation. If the energy dissipation is to be the same in the
inelastic model and the substitute structure, then the fraction of critical damping
assigned in design should depend on the assumed relationship between c and ξ (see
Secs. 4.1.1–4.1.3, below), and will, in general, not be identical to ξ. Furthermore,
for damping models in which c changes with inelasticity, the appropriate value of
ξeq,v will depend on the ductility.
The relationship between damping force, Fd , and displacement, for steady-state
harmonic response under Bilinear hysteresis is illustrated in Fig. 11 for three damp-
ing models: A constant damping coefficient, tangent-stiffness proportional damping,
December 20, 2005 13:47 WSPC/124-JEE 00236
1.5 1.5 2
Displacement Ratio ∆TS/∆IS
R=6
1.4 1.4 1.8 R=4
R=4 R=6
1.3 1.3 1.6 R=6
R=6
R=4
R=6
1.2 1.2 1.4
R=4
R=2
1 1 1
0 0.4 0.8 1.2 1.6 2 0 0.4 0.8 1.2 1.6 2 0 0.4 0.8 1.2 1.6 2
Period (seconds) Period (seconds) Period (seconds)
Fd
Fs c = constant
Y
c = α m + β kt
U
r ki
c = β kt
Y ki ki
U U ∆
Substitute ∆
structure, Y
k = keff Substitute Y
Bilinear structure,
U
hysteresis ξeq,v = ξ
(a) (b)
Fig. 11. Steady-state harmonic response of different hysteretic and damping models. (a) Bilinear
hysteresis, and (b) corresponding damping force versus displacement for various damping models
and substitute structure.
and Rayleigh damping with a 50% mass and 50% tangent stiffness contribution at
the system period of vibration. As discussed below, Rayleigh damping is not com-
monly used in SDOF analyses, as it requires at least two modes of vibration to
fully determine the coefficients of the model. Points labelled “Y” and “U” rep-
resent changes in stiffness due to yielding and unloading, respectively, and the
subsequent changes in the damping coefficient for the latter two models. For each
model, the specified fraction of critical damping, ξ, is identical, but clearly the
energy dissipation, represented by the area contained in the “viscous hysteresis”
loops, is model-dependent. In addition, the substitute structure model for which
ξeq,v = ξ has been specified is shown in the same figure. In this case, the damping
December 20, 2005 13:47 WSPC/124-JEE 00236
coefficient and the stiffness of the equivalent linear hysteresis model are constant,
so the shape of the damping force-displacement loop is elliptical for steady-state
harmonic response. It is clear from Fig. 11 that ξeq,v = ξ will not result in the
same viscous energy dissipation as the ITHA, regardless of hysteresis model and
damping model.
For DDBD applications, it will be important to obtain appropriate values of ξeq,v
for hysteresis models and damping models representative of real structural response.
This is of particular importance for verifying DDBD with ITHA, as in this case the
models are known exactly, and this source of error can be effectively removed from
the comparison. For actual design applications, there is more uncertainty in the
modelling, particularly for the viscous damping. In this case, we believe that it is
appropriate to use values corresponding to tangent stiffness proportional damping,
as discussed in Sec. 3.2.4.
of critical damping, ξ:
c = 2mωi ξ = 2ξ mki . (8)
In the substitute structure model, the damping coefficient is determined with
respect to the secant stiffness, keff , and will therefore be given by:
c = 2ξeq,ν mkeff . (9)
For steady-state response of the inelastic model and the substitute structure at
the same displacement amplitude, the velocity time histories will be nearly identical.
Although this assumption will be less accurate under earthquake loading, it is
consistent with the approach commonly used for the hysteretic component of the
equivalent viscous damping [Jacobsen, 1960], and should provide adequate results.
Identical velocity over time implies that the viscous energy dissipation will also be
equal if:
ξeq,ν = ξ · λ1 , (10)
where
v s
u
u ki µ
λ1 = t = , (11)
keff 1 + rµ − r
and µ and r are the displacement ductility and post-yield stiffness ratio, respectively.
The first equality in Eq. (11) applies for any hysteresis model, whereas the second
equality only applies for models with a bilinear backbone, such as the Bilinear,
modified Takeda, and Flag models considered in previous sections. For small r,
√
we have λ1 ≈ µ. Equation (11), which is plotted in Fig. 12(a), implies that if a
constant damping coefficient equivalent to 5% of critical damping is assumed to be
appropriate, then for an expected displacement ductility of µ = 4, and post-yield
stiffness ratio r = 0.05, an equivalent damping of ξeq,v = 9.33% should be assumed
in DDBD.
r=0
λ1 3
r = 0.1
2 r = 0.2
0
0 1 2 3 4 5 6 7 8 9 10
µ
(a)
1.0 1.0
λ2 λ2
0.5 0.5
r = 0.05
r = 0.2 r = 0.02
r = 0.1
r=0 r=0
0 0
0 1 2 3 4 5 6 7 8 9 10
µ
0 1 2 3 4 5 6 7 8 9 10
µ
(b) (c)
Fig. 12. Correction factors for viscous damping in DDBD. (a) λ1 for Bilinear and modified
Takeda hysteresis, (b) λ2 for Bilinear hysteresis, and (c) λ2 for modified Takeda hysteresis.
for Bilinear hysteresis, and a more complicated expression for the modified Takeda
model, given in full in [Grant et al., 2005]. Note that the r term in Eq. (14) represents
the dissipated energy for a linear viscoelastic system with stiffness rki . In Fig. 11,
this corresponds to the area of the smaller ellipse partly traced out by the damping
force between points Y and U; the extra spikes in the damping force from U to Y
are given by the second addend in Eq. (14). The inverse cosine is a result of the
assumption of harmonic steady-state response in the nonlinear model.
Equation (14), for Bilinear hysteresis, and the corresponding expression for mod-
ified Takeda hysteresis, are plotted in Figs. 12(b) and 12(c), respectively.
this is the same approach taken for the hysteretic component of the equivalent
viscous damping in Jacobsen’s approach [Jacobsen, 1930; 1960].
The factor λ1 derived in Sec. 4.1.1 is an exact correlation between the damping
coefficients of the nonlinear model and the substitute structure, as the viscous
damping remains linear in both cases. The λ1 factor adjusts the fraction of critical
damping specified in terms of initial stiffness to one appropriate for secant stiffness,
which, provided the velocity history is the same in both cases, results in exactly
the same damping force for the duration of loading. For this reason, Eq. (11) does
not require numerical validation.
The derivation of factor λ2 for tangent-stiffness proportional damping and
Rayleigh damping, however, clearly involves some degree of approximation. It is
not immediately obvious what effect replacing the nonlinear viscous damping for
Bilinear hysteresis (the solid lines in Fig. 11(b)) with an equivalent linear viscous
damping (dashed line, scaled by λ2 ) will have on the response of a SDOF system.
For the modified Takeda hysteresis model, the viscous damping force is even more
complicated, particularly when the loading is non-harmonic, as different branches of
the hysteresis model exhibit significantly different stiffnesses. At any instant in time,
the actual damping force will be different in the nonlinear and substitute structure
models, and what is important is the effect of the viscous energy dissipation on the
maximum displacement response.
It is difficult to use a direct comparison of the response of a nonlinear system
with that of the corresponding substitute structure to determine the correction
factor, λ2 . For the former system, the effect of damping due to hysteresis cannot
be separated from that due to viscous dissipation, and the substitute structure
model will require an approximation of the equivalent viscous damping due to both
components. Conversely, existing studies of the hysteretic component of equivalent
viscous damping [e.g. Judi et al., 2000; Miranda and Ruiz-Garcia, 2002] that do not
consider the model- and ductility-dependence of the elastic viscous component do
not properly separate the effects of the two halves of Eq. (1). In that case, adopting
an elastic viscous damping equal to zero in both DDBD and ITHA, (as in Blandon
et al. [2005]), could isolate the effects of hysteresis. We cannot achieve the converse
by setting the hysteretic dissipation to zero as we would no longer have a nonlinear
system.
For this reason, the system described in Fig. 13 has been used to study the
effects of tangent-stiffness proportional and Rayleigh damping on the response of a
SDOF structure. In this system, the restoring force, Fs , is obtained from a linear
elastic model with stiffness equal to the effective stiffness of the substitute struc-
ture model (Fig. 13(a)). The calculation of the damping force (Fig. 13(b)), however,
assumes that the stiffness is actually equal to the tangent stiffness of a simulated
hysteretic model, kt,sim (simulated Bilinear hysteresis is shown in the dotted line
in Fig. 13(a)). The damping level of the simulated nonlinear system, ξ, is adjusted
until the peak displacement response under a given input ground motion matches
that of a substitute structure with specified damping, ξeq,v . The ratio ξeq,v /λ1 ξ is
December 20, 2005 13:47 WSPC/124-JEE 00236
Fs
Fd Simulated tangent
stiffness proportional
damping,
2ξ
c = ⋅ kt ,sim
∆ ω
Linear ∆
hysteresis,
k= keff Simulated
Bilinear Substitute
hysteresis for structure,
c calculation c = 2ξeq ,v mk eff
(a) (b)
Fig. 13. Numerical procedure for determining λ2 . (a) State determination and definition of sim-
ulated bilinear stiffness, and (b) damping force versus displacement for simulated tangent-stiffness
proportional damping model and substitute structure.
then equal to λ2 , from Eq. (13). Note that hysteretic damping (or its linearised
equivalent) is not included in either the substitute structure or the simulated non-
linear structure.
The procedure described above was carried out for target ductility values
between 1.0 and 10.0, over a suite of 40 ground motions records from the
FEMA/SAC database [Woodward-Clyde Federal Services, 1997]. The records, from
the Los Angeles (LA) and near fault (NF) subsets of the database, were scaled
to obtain the target displacement in the substitute structure, and therefore the
actual hazard level to which the ground motions correspond is irrelevant, provided
that the duration of each record is somewhat consistent with its intensity. Although
the analytical relationships derived in Sec. 4.1 do not predict period-dependence,
the simulations were carried out for SDOF initial stiffness corresponding to effec-
tive periods of 1.0 sec and 2.0 sec at a ductility of 4. Finally, the post-yield stiffness
ratio and, for the modified Takeda model, the unloading parameter, γ, (see Eq. (6))
were varied, although only the results for γ = 0.5 are reported here.
The median results across the ground motion suite for converged values of λ1 λ2
are shown in Fig. 14. The total multiplier is plotted, rather than just the unknown
value λ2 , as it is the product that is required for DDBD applications, and this
allows simplified design equations to be obtained directly from Fig. 14 (see below).
Figures 14(a) and 14(b), for Bilinear and modified Takeda hysteresis, respectively,
show that the analytical expressions developed in Sec. 4.1 generally underestimate
the correction factors obtained from the numerical simulations. For both models, it
may be observed that the period-dependence is also negligible. Increasing the post-
yield stiffness also tends to increase the factor λ1 λ2 towards unity, although for
the range of r values appropriate for modelling reinforced concrete elements with
the modified Takeda model, the dependence is small. Obviously, for the Bilinear
model, as r tends to one, the response tends to linear, and λ1 λ2 → 1. For the
modified Takeda model, this limit does not hold, as the applicability of the model
December 20, 2005 13:47 WSPC/124-JEE 00236
1 1
λ1λ2 λ1λ2
r=0 r=0
r = 0.1 r = 0.05
r = 0.2 T = 1 sec
T = 1 sec eff,4
eff,4 Teff,4 = 2 sec
T = 2 sec
eff,4
Analytical Analytical
0.5 0.5
0 0
0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10
µ µ
(a) (b)
Fig. 14. Results of numerical simulations for λ1 λ2 . (a) Bilinear model and (b) modified
Takeda model.
Equations (18a) and (18b) should only be used for the range of r values used in the
numerical analyses — 0 to 0.2 for the Bilinear model, and 0 to 0.05 for the modified
Takeda model — although at least Eq. (18a) exhibits the correct limit as r tends to
1.0. Using the same example as considered in Sec. 4.1.1, the required damping to
December 20, 2005 13:47 WSPC/124-JEE 00236
5. Conclusions
It is reasoned that elastic viscous damping in structures is more realistically mod-
elled in inelastic time-history analyses by a tangent-stiffness proportional viscous
damping coefficient than by a constant damping coefficient. This is in contradiction
to common analytical practice.
Comparative inelastic analyses of single-degree-of-freedom systems with differ-
ent hysteretic rules, and either a constant damping coefficient or tangent-stiffness
proportional elastic damping showed that the choice of damping model can give a
very significant influence on peak response displacements, with the tangent-stiffness
assumption invariably resulting in larger displacements. The difference increases as
displacement ductility increases, and is dependent on initial period and hysteresis
rule. Although the analyses were carried out only on SDOF systems, it is expected
that similar results would be obtained for MDOF systems. This, however, needs
verification by additional studies. It is noted that care needs to be exercised when
December 20, 2005 13:47 WSPC/124-JEE 00236
References
Blandon, C. A. and Priestley, M. J. N. [2005] “Equivalent viscous damping equations for
direct displacement-based design,” J. Earthquake Eng. 9(Special Issue 2), 257–278.
Carr, A. J. [2001] Ruaumoko — A Program for Inelastic Time-History Analysis, Dept. C.
Eng. Univ. Canterbury, New Zealand.
Caughey, T. K. [1960] “Classical normal modes in damped linear systems,” J. Appl. Mech.
27, 269–271.
Grant, D. N., Blandon, C. A. and Priestley, M. J. N. [2005] “Modelling inelastic response
in direct displacement-based design,” Rept. No. ROSE 2005/03, European School for
Advanced Studies in Reduction of Seismic Risk, Pavia.
Jacobsen, L. S. [1930] “Steady forced vibrations as influenced by damping,” Transactions
ASME 52, 169–181.
Jacobsen, L. S. [1960] “Damping in composite structures,” in Proceedings of the 2nd
World Conference on Earthquake Engineering, Volume 2, Tokyo and Kyoto, Japan,
pp. 1029–1044.
Judi, H. J., Davidson, B. J. and Fenwick, R. C. [2000] “The direct displacement based
design — A damping perspective,” in Proceedings of the 12th World Conference on
Earthquake Engineering, Auckland, New Zealand, Paper No. 0330.
Iwan, W. D. [1980] “Estimating inelastic response spectra from elastic spectra,” Earth-
quake Eng. and Struct. Dynamics 8, 375–388.
December 20, 2005 13:47 WSPC/124-JEE 00236
Miranda, E. [2000] “Inelastic displacement ratios for structures on firm sites,” J. Struct.
Eng. 126, 1150–1159.
Miranda, E. and Ruiz-Garcia, J. [2002] “Evaluation of approximate methods to estimate
maximum inelastic displacement demands,” Earthquake Eng. and Struct. Dynamics
31, 539–560.
Newmark, N. M. and Hall, W. J. [1982] Earthquake Spectra and Design, EERC, Berkeley.
Otani, S. [1981] “Hysteretic models of reinforced concrete for earthquake response analy-
sis,” J. Faculty of Eng. Univ. Tokyo XXXVI(2), 24 pp.
Otani, S. and Sozen, M. A. [1972] “Behavior of multistory reinforced concrete frames
during earthquakes,” Structural Research Series No. 392, Civil Engineering Studies,
University of Illinois, Urbana.
Priestley, M. J. N. [2000] “Performance-based seismic design,” Keynote Address, 12-
WCEE, Auckland, 22 pp.
Nakaki, S. D., Stanton, J. F. and Sritharan, S. [1999] “An overview of the PRESSS five-
story precast test building,” PCI Journal 44(2), 26–39.
Veletsos, A. S. and Newmark, N. M. [1960] “Effect of inelastic behavior on the response
of simple systems to earthquake motions,” Proc. World Conference on Earthquake
Engineering, Japan, V.2, 895–912.
Wilson, E. L. and Penzien, J. [1972] “Evaluation of orthogonal damping matrices,” Int.
J. Numerical Methods in Eng. 4, 5–10.
Woodward-Clyde Federal Services [1997] “Suites of ground motion for analysis of steel
moment frame structures,” Report No. SAC/BD-97/03, SAC Steel Project.