Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
DESIGN CONFIRMATION
“It’s what you learn after you know it all that counts.”
—Earl Weaver
Design procedures must introduce many simplifying assumptions into the process.
The validity of these simplifications must be tested analytically if they are to be
used with any degree of confidence. This testing is done through the use of elastic
and inelastic time history analyses. This chapter endeavors to analyze some of the
component and system designs developed in the preceding chapters. A confirmation
of any of the design procedures is not the objective. The focus is to demonstrate the
application of the procedures that might be used to confirm a design. The designer
who wishes to improve his or her design skills will study completed designs to
determine how the next design might reasonably be improved.
Several issues were raised in Chapter 3 that have a direct bearing on the design
process. These issues can be grouped into major categories as follows:
• Equal displacement.
• System ductility.
• Design strength.
• Modeling considerations.
The time history evaluations undertaken in this chapter are limited to frame and shear
wall designs developed in Chapter 3. The focus of each analysis is to probe these
identified focal issues. Similar types of analyses performed over the last 35 years have
led the author to accept the recommended design approaches developed in Chapter 3.
Implicit in the confirmation of a design by time history analysis is the compara-
bility of the design spectrum and the ground motions used to evaluate the design.
Two means of developing comparable representations of an earthquake are being
used today. One alters the ground motion so as to match the spectrum at every pe-
riod (matched spectrum) while the other amplifies (only) an actual ground motion
record to fit the design spectrum (scaled ground motion) in a particular period range.
763
Copyright © 2003 John Wiley & Sons Retrieved from: www.knovel.com
764 DESIGN CONFIRMATION
Both methods are used today but I believe that the latter is more appropriate provided
the selected ground motions are appropriate and a sufficient number of earthquake
records are used to analyze the building. This is because earthquakes have a domi-
nant frequency content that can be a benefit insofar as mitigating the impact of the
earthquake on the response of ductile structures, and this will be an observed char-
acteristic. The design spectra and those that were produced from the scaled ground
motion used to perform the analyses of this chapter are described in Figure 4.1.1. A
set of matched spectra are shown in Figure 4.1.2 to describe the inherent difference
in the two approaches.
The selected ground motions are identified in Table 4.1.1. An attempt has been
made in the scaling process to match the ground motion spectrum to the design
spectrum in the response range of the studied structures. For example, Earthquakes
1 and 5 match the design spectrum in the period range of the shear wall structure
(T ∼ = 0.95 second) while Earthquakes 2 and 3 match the design spectrum in the
period range of the frame braced structure. Earthquake 4 was included as a curiosity
because it contained a large single pulse (Figure 4.1.3a); it had no effect on any of
the studied structures.
The scaled ground motion for Earthquake 1 is shown in Figure 4.1.3b. Compare
this to the scaled and altered ground motion of this event used to develop the matched
spectrum of Figure 4.1.2b, which is shown in Figure 4.1.3c. Observe that the ac-
celerations used to develop the matched spectrum (Figure 4.1.3c) are significantly
increased in the post 10-second range and that the peak acceleration for the matched
ground motion exceeds that of the scaled ground motion. This in spite of the fact that
the spectral velocity defined by the matched spectrum is only 42 in./sec, as opposed
to the 51 in./sec defined by the scaled spectrum.
Comment: One caveat should be reaffirmed as it relates to the use of time history
analyses. Building designs are not in themselves academic efforts. Designs are per-
formed and buildings produced to satisfy a perceived societal need in a manner con-
sistent with the extant standard of care. The designer who performs a time history
analysis of a proposed design does so only to answer one question—Is the produced
design likely to meet the design objective? Academic interests may be explored in the
process but this is only to improve future designs; not to “perfect” a produced design.
Introducing design changes at this stage in the design process will create chaos, and
this usually results in a less effective design. Design changes should be resisted unless
the need is unquestionable.
The objective of this section is to review and analyze inelastic time history analyses
of the shear wall braced buildings designed in Section 3.1. The equal displacement
hypothesis and the impact of system strength on response and behavior are the focal
issues.
Figure 4.1.1 Design response spectrum and response spectra developed from scaled ground
motions.
Figure 4.1.3 Selected scaled ground motions used in the various time history analyses.
Figure 4.1.4 Response of 26-ft shear wall to Earthquake 1, Table 4.1.1 (building described
in Figures 2.4.9 and 2.4.10).
n = 1Sd (4.1.3)
= 1.45(7.71) (1 = 1.45)
= 11.2 in.
This is quite comparable to the peak value predicted by the elastic time history
analysis (n,max = 10.66 in. in Figure 4.1.4a).
An inelastic model of the building was also subjected to the ground motions
described in Table 4.1.1. The peak displacement was slightly less than 8 in. and this
warrants some study.
The elastic and inelastic responses of the wall system to Earthquake 1 are described
in greater detail in Figure 4.1.5. Observe how the periodicity of the inelastic response
is lengthening. The behavior pattern follows the characteristic yielding response
described in Section 1.1.1. The structure yields 2+ seconds into the earthquake
(y = 3.2 in. in Figure 3.1.6b) and continues to displace to a drift of 5 in. in the elastic
model and slightly more in the inelastic model. This corresponds to a displacement
ductility demand of about 1.56. √
The associated period shift is 1.25 ( 1.56) times. The period of the yielding
structure should be 1.2 seconds (1.25 (0.95)), and this is what is predicted by the
time history analysis (Figure 4.1.5b). Observe that the response of the inelastic model
after 3 seconds (Figure 4.1.5a) never exceeds 6 in. (t = 6 seconds) and thereafter is
limited to 4 in.—essentially elastic. Understanding this behavior requires a review
of the characteristics of the driving earthquake. Observe (Figure 4.1.1a) that the
response drops significantly in the now period of interest (T = 1.25 seconds) range.
The spectral acceleration response in the 1.3- to 1.5-second period range is only one-
third of that in the 1-second range. In essence we have a new or different structure
(T = 1.2 seconds) responding to a different earthquake ground motion (Sv ∼ = 30
in./sec). Spectral projections based on the new structure and the response spectrum
Figure 4.1.5 Enlarged time history responses of the 26-ft shear wall to Earthquake 1.
for Earthquake 1 (u ∼ = 8.6 in.) are not likely to confirm the time history projections
(max = 7.6 in.) because a portion of the strong ground motions (t = 2 seconds) have
passed (Figure 4.1.3b)—and it was these larger ground motions (0.75g) that drove
the initial structure (T = 0.95 second).
The drift experienced in the next cycle (t = 3.5 seconds) is 9.5 in. for the elastic
model but only 7.6 in. for the inelastic model. Now the ductility demand is on the
order of 2.4 (7.6/3.2) and another period shift might reasonably be anticipated. The
√
period shift should amount to µs , or 1.55 times the elastic period. This magnitude of
period shift is not confirmed in the response because the subsequent ground motions
appear to excite higher modes in the structure. Figure 4.1.6 describes behavior in
the plastic hinge region in the pre 8-second range (Figure 4.1.6a) and in the 12- to
14-second response range (Figure 4.1.6b). Observe how the stiffness in the 12- to
14-second range is consistent with the observed period change.
k12 ∼
= 1.3 (see Figure 4.1.6)
k8
T12 = 1.3T8
T12 ∼
= 1.3(0.95)
∼
= 1.2 seconds
Figure 4.1.6 Hysteretic response in the plastic hinge region of the 26-ft shear wall (inelastic
time history of Figure 4.1.3a).
it virtually impossible for the inelastic system to resonate with the ground motion.
This will logically make it unlikely that the inelastic system will ever reach the
response projections of the elastic system, which are, in their peak ranges (T = 1
second for Earthquake 1), largely attributable to resonant effects.
Observe that in Earthquakes 1, 3, and 5 (Figure 4.1.1) a significant reduction
of response is apparent in the period range of 1.2 to 1.5 seconds. Obviously care
must be taken to select earthquakes that reasonably describe potential ground motion
characteristics. This argues for the adoption of an artificial ground motion similar to
the ones described in Figure 4.1.2.
The matched design spectra described in Figure 4.1.2 are of a lesser intensity than
the design spectrum described in Figure 4.1.1. The spectral velocity in the 1-second
range is
TSa
Sv = (Eq. 4.1.1)
2π
(1.0)(0.68)(386.4)
=
6.28
= 42 in./sec
The ultimate displacement (u ) of the 26-ft shear wall braced system (T = 0.95
second) would be
TSv
Sd = (Eq. 4.1.2)
2π
0.95(42)
=
6.28
= 6.35 in.
u = 1Sd (Eq. 4.1.3)
= 1.45(6.35)
= 9.2 in.
The elastic time history analyses, using the ground motions that produced the spectra
of Figure 4.1.2, were consistent in their prediction of a peak displacement response
of slightly more than 8 in. Of particular interest is the response of the system to the
altered El Centro 90° ground motion described in Figure 4.1.3c. The elastic response
is slightly less than the spectral prediction of 9.2 in., and the early peaks (t = 5
seconds) developed for the scaled El Centro 90° ground motion (Figure 4.1.4a) have
been suppressed. The displacement peaks now occur at t = 12 seconds (Figure
4.1.7a). The inelastic response (Figure 4.1.7b) describes a similar shift in the peak
response (to t = 12 seconds), but in this case the displacement response of the
inelastic response is essentially the same as the elastic response. This tends to confirm
Figure 4.1.7 Response of the 26-ft shear wall to the (matched) ground motion described in
Figure 4.1.3c.
the fact that the large reductions observed in the response to the scaled ground motion
(Figure 4.1.4) are entirely attributable to the characteristics of the ground motion.
Clearly, this analysis supports the equal displacement proposition.
Comment: Clearly, the selection of ground motions used in the analysis of this
system raises several issues. Is an artificial earthquake that creates a constant spectral
intensity reasonably assumed? If not, the designer must study the proposed ground
motions prior to performing a time history analysis to insure that anomalies do not
exist in the period range of interest. This latter approach seems more reasonable. In
either case the structural engineer and the geotechnical consultant must work together
to insure that the proposed system is reasonably tested.
The elastic and inelastic response of this structure (Figure 2.4.9) to the ground motion
of Earthquake 2 (Figure 4.1.1b) is interesting. The elastic response (Figure 4.1.8a)
Figure 4.1.8 Roof displacement response of the 26-ft shear wall to Earthquake 2, Table 4.1.1.
suggests two ground pulses, one in the 3- to 7-second range and another in the 12-
second range. Figure 4.1.8b shows the inelastic response and identifies a period shift
quite similar to that caused by Earthquake 1 (see Figure 4.1.5b). Observe that the
maximum inelastic and elastic displacement responses are the same. The behavior
patterns, however, are quite different. The first inelastic excursion in the positive
sense provokes a slightly larger inelastic response, while the negative sense inelastic
response is significantly greater (50%) than the elastic response and the periodicity
of the response has increased.
The response spectra for Earthquake 2 (in Figure 4.1.1b) is quite different from the
spectra produced by ground motions 1, 3, and 5. Instead of suggesting a reduction in
response, the response actually increases. When the second pulse arrives (t = 12
seconds), the spectral acceleration imposed on the inelastic structure (T = 1.2
seconds) is about 0.67g, while that imposed on the elastic model (T = 0.93 second)
is on the order of 0.6g.
The displacement response for these models suggested by a spectral analysis is
developed as follows. First combine Eq. 4.1.1, 4.1.2, and 4.1.3 to create
1T 2 Sa
= (4.1.4a)
4π 2
1.45T 2 (386.4)Sag
=
39.4
and, given the steepness of the response spectrum in this period range, this confirms
the elastic time history prediction of 8 in.
The response of the inelastic structure, using the same procedures, should be
= 13.7 in.
and this is significantly greater than the 8 in. predicted by the inelastic time history
analysis (Figure 4.1.8b).
The reduced displacement suggests that energy dissipation may indeed have re-
duced the response. Consider the curvature ductility suggested by Figure 4.1.6a,
which reflects plastic hinge behavior at inelastic displacements in the 8-in. range.
0.000035
µφ =
0.00001
= 3.5
and this is quite consistent with the reported peak inelastic displacement of slightly
less than 8 in. It seems that structural damping is impacting the inelastic system
response, at least for this single-degree-of-freedom system whose system ductility
and dissipated energy are reasonably quantified (Figure 4.1.6a).
The curvature response described in Figure 4.1.6a is acceptable from the stand-
point of suggested strain states. The maximum curvature is slightly less than 0.00004
rad/in. and Table 2.4.3 suggests that induced material strains will be
εc = 0.0016 in./in.
εs = 0.01094 in./in.
From a design acceptance perspective, this meets our objectives. From a design
procedures perspective, it merits some analysis.
Computer programs adopt one of two possible element models. One models the
end of an element as a rotational spring; the other creates a fiber model. The spring
model is consistent with the model used in the design; specifically one that assumes
a constant curvature in the prescribed plastic hinge region. The fiber model creates
a linearly changing curvature from the face of the support to the inception of elastic
behavior, provided this point coincides with a node. In other words, it develops the
hinge length for the user. The distinction is important because our design basis was
developed from testing. Test results were reduced to strain states using the plastic
hinge model (Figure 2.1.8). A plastic hinge length of 13 ft (0.5-w ) was used in the
analysis of the example 26-ft long wall.
The time history analysis was performed on a fiber model. Nodes were established
at each floor level (Figure 2.4.9), and the computer adopted a plastic hinge length of
10.5 ft or one floor. The conclusions reported in Figure 4.1.4 are for an average curva-
ture in the plastic hinge region and, as a consequence, are consistent with the curvature
model described in Figure 2.1.10. The curvatures developed in the design process
follow the curvature model described in Figure 2.1.11. The two are easily compared.
The reported curvature is clearly consistent with that developed for conceptual design
purposes. The computer model adopted a plastic hinge length of 126 in. This is more
conservative than the hinge length adopted in the conceptual design. The impact on
the comparative analysis would not change our confirmation of the behavior predicted
by the computer analysis for
4.4
θp =
1197
= 0.00367 radian
0.00367
φp =
126
= 0.000029 rad/in.
φu = φy + φp
= 0.000039 rad/in.
• Time history based conclusions must be developed with care and rationally
explained.
• Ground motions used to describe inelastic behavior should be compatible with
anticipated levels of spectral intensity in the pre- and postyield behavior ranges.
• The “equal displacement” hypothesis is reasonably used to design structures of
this type.
• It seems unlikely that the ultimate displacement of this system will exceed 11
in. (0.9%). Given this and the strains predicted by the pushover analysis (Figure
3.1.6b), there seems no reason to believe that performance objectives will not
be met.
11
y = 7.5 in. φy h2w
40
u = 22.86 in. (Response spectra)
µ = 4 (Objective)
µ = 3 (Attained)
The issue now addressed is the advantage that might result from an increase in flexural
strength. Accordingly, the response of the proposed design, six #6 bars (ρs = 0.25%),
is compared with an otherwise identical wall reinforced with six #9 bars (ρs =
0.57%).
These walls were subjected to the ground motion of Earthquake 3 (Table 4.1.1)
whose response spectrum is described in Figure 4.1.1c. Observe that the ground mo-
tion based spectrum is less than the design spectrum at the probable elastic building
period of 1.92 seconds, but exceeds the design spectrum in the 2.1- to 2.7-second
period range. Accordingly, any period shift is likely to produce an increase in dis-
placement.
The results of the time history analyses are presented in Figure 4.1.9. The peak
elastic displacement occurs at 9 seconds and corresponds to almost 18 in. This is
consistent with the response predicted using the design spectrum (22.86 in.). The peak
inelastic displacement is 15 in. (Figure 4.1.9b) for the proposed design (six #6 bars)
but almost 18 in. (Figure 4.1.9d) for the stronger wall (six #9 bars). This is explained
following the concepts discussed in Section 1.1.1. Observe that the stronger wall (six
#9 bars) did not yield at the 6-second excursion while the design wall (six #6 bars)
did. This resulted in a reduced displacement demand on the design wall in subsequent
cycles (Figure 4.1.9b) while the stronger wall had a deferred first plastic excursion
that coincided with a larger pulse. Accordingly, there may often be an advantage
associated with early yielding. This is not a conclusion that can be applied to the
design of a wall, but it is not an uncommon occurrence.
Several observations are made:
• The proposed 16-ft wall design (ρ = 0.25%) meets our performance objectives.
• The experienced deflection should be in the 15-in. range (1.2%), and this is twice
that expected of the 26-ft wall.
• Concrete strain states will also be higher in the 16-ft long wall (εc = 0.0023
in./in. see Figure 3.1.7c) than those in the 26-ft long wall (εc = 0.0016 in./in.—
see Figure 3.1.7a). The deflection predicted for the stronger wall was 18 in. This
suggests a concrete strain of 0.0028 in./in. (Figure 3.1.7c). Structural damage
should not be expected in either case.
Figure 4.1.9 Response of various strength shear wall braced structures to Earthquake 3.
• Concrete strain states are not directly related to wall strength (Figure 3.1.7c)
but rather to displacement demand and, as a consequence, in shear wall braced
buildings performance will not be improved by increasing strength.
• In near fault locations wall designs should be studied by subjecting them to pulse
type excitations.
• Design procedures developed in Chapters 2 and 3 reasonably predict behavior
and seem to produce systems that will survive the criterion earthquake.
The predicted level of spectral acceleration at the roof will increase significantly as
the building is stiffened and strengthened.
1Sd 4π 2
1Sa = (see Eq. 4.1.1 and 4.1.2) (4.1.5)
T 2 µ
Wall A
u
µ =
y
11.2
=
3.2
= 3.5
Wall B
18
µ =
7.5
= 2.4
Wall C
15
µ =
7.5
= 2.0
u 4π 2
An = (see Eq. 4.1.5) (4.1.6)
T 2 µ
Wall A
11.2(4)(3.14)2
An =
(1)2 (3.5)
= 126 in./sec 2 (0.33g)
Wall B
18(4)π 2
An =
(2)2 (2.4)
= 74 in./sec 2 (0.19g)
Wall C
15(4)π 2
An =
(2)2 (2.0)
= 74 in./sec 2 (0.19g)
The objective of this section is to study the behavior of the frames designed in Section
3.2. Analytical models start with the simplest form. Models are then modified so as to
describe the peculiarities associated with precast systems. The impact model changes
could have on predicted response is then discussed.
Photo 4.1 Precast clad concrete frame, Mariners Island, San Mateo, CA. (Courtesy of Engle-
kirk Partners, Inc.)
in the shear wall examples reviewed. A similar conclusion is reached for frame braced
structures. The intuitive argument suggests that an increase in strength will cause a
more violent response, leading to larger displacements and induced levels of strain.
The frames designed in Section 3.2 were subjected to inelastic time history anal-
yses. Component behavior was modeled using a bilinear elastic perfectly plastic rep-
resentation of behavior—alternative behavior models will be considered in Section
4.2.2. Two frame types, each with three different levels of strength, are considered:
the three-bay frame of Figure 3.2.2a and the four-bay alternative. Frame characteris-
tics are summarized in Table 4.2.1 and identified as to number of bays and level of
strength (for example, 3-1 refers to three bays—strength level 1).
Earthquakes 2 and 3 were selected because their spectral response was essentially
the same as that of the design spectrum in the 2- to 3-second period range (see
Figures 4.1.1b and 4.1.1c). The responses of the various models to the selected ground
motions are presented in Table 4.2.2.
It seems clear that a generalization of behavior is not possible; however, intuitive
logic tends to be supported by the responses described. Consider the following propo-
sitions:
• An increase in the strength of the frame will not guarantee a reduction in the
level of experienced drift.
• Response is a function of the characteristics of the ground motion.
The impact of ductility on response was discussed in Section 1.1.1, and it should be
obvious that the nature of the earthquake can affect the peak displacement more than
the strength of the system.
The response spectrum for Earthquakes 2 and 3 identify essentially identical
spectral velocities in the period range of interest, yet the range of predicted peak
displacement response is significantly higher on average for Earthquake 2 than for
Earthquake 3.
Figures 4.2.1a, through 4.2.1c describe the responses of the various three-bay
frames to the ground motion of Earthquake 2. For Frame 3-1 the first inelastic excur-
sion is the largest drift and the impact of a ground displacement pulse at t ∼
= 4 seconds
is significantly reduced. Compare this with the response of Frame 3-3, where the first
positive pulse produces a response in the elastic range. Now the pulse at t ∼
= 4 seconds
produces the first inelastic excursion, the consequences of which are a series of elastic
excursions about a new baseline created by a permanent or residual drift of 16.5 in.
These anomalies aside, it seems reasonable to imagine that peak displacements of a
ductile frame would become larger as the strength of the frame is increased. In Section
Figure 4.2.1 Response of the various three-bay frames to the ground motion of Earthquake 2.
1.1.1 it was pointed out that it was the strength of the spring that connects the mass to
the ground that drives the mass and, as a consequence, a weakening of the spring tends
to reduce the displacement response on all but the first inelastic excursion. This is the
essence of base isolation so it is reasonable to assume that the displacement response
would increase as the linkage between the ground and the mass is increased.
The addition of one bay does not seem to materially impact the level of experienced
roof drift. Figure 4.2.2 describes the response of the four-bay frame (4-2) to Earth-
quake 2 and, as can be seen, it is quite similar in character to the response of Frames
3-2 and 3-3, which bracket it in terms of strength. The encouraging aspect from a
design perspective is that peak drifts seem to be consistent with spectral projections.
Residual drifts should logically be reduced by an increase in frame strength, but
this is not a reliable consequence. Observe (Table 4.2.2) that the residual drift of
the various three-bay frames to Earthquake 3 and the various four-bay frames to
Earthquake 2 exhibited the logical reduction in residual drift, while the other residual
drift patterns were very irregular. Design objectives were discussed in Chapter 3
regarding residual displacements, but it seems clear that an increase in strength will
not guarantee reduced residual drifts.
From a performance perspective, damage and failure potential can only be related
to induced concrete strains. If it is assumed that the drift of Frames 3-2 and 3-3 will be
essentially the same, then Frame 3-3 will experience much more damage than Frame
3-2 because, absent a change in member size, an increase in strength will reduce
the available ductility in its components. Table 4.2.3 describes critical strain states
and 4.2.2 because the stiffness of the member is presumed to be constant (Figure
4.2.3a).
Stiffness degradation should generally serve to reduce the displacement response
of a structure. In order to appreciate this fact, consider the response spectrum for
Earthquake 3 (Figure 4.1.1c). A structure whose fundamental period is in the 0.5- to
1-second range will be tuned to this earthquake, and resonance would be a logical con-
cern. When, however, the structure is forced into the inelastic range, the periodicity
of the response will also change and the consequences of resonance will be avoided.
Logically, this suggests that a reasonable level of ductility should be included in the
cast-in-place hybrid
Stiffness degradation (HC) 1.0 0.5
Slip control (HS) 1.0 0.2
Before proceeding with the time history analyses, both the cast-in-place and hybrid
building models were subjected to a static displacement controlled response and
member behavior was checked to insure that plastic hinge behavior objectives had
been attained. Building displacement responses are shown in Figure 4.2.5. Observe
that building responses reflect the characteristics of the components. Component be-
haviors are described in Figure 4.2.6. Compare the experimental response described
in Figure 2.1.2 for the cast-in-place beam with the analytical model described in
Figure 4.2.6a. Figures 4.2.6b and c describe how several experimental cycles compare
with analytical cycles. Similarly, compare the experimentally determined hysteretic
behavior of the hybrid system (Figures 2.1.47a and 2.1.47b) with the developed ana-
lytical model (Figure 4.2.6d). Observe that the “slip” feature is reasonably captured.
The hybrid beam dissipates about 60% of the energy dissipated by the cast-in-place
beam.
Both models were analyzed using the ground motion for Earthquake 3 (Figure
4.1.1c), which predicts a fairly constant response in the period range of interest (0.55
second). Building responses are shown in Figure 4.2.7. The response of the hybrid
system (Figure 4.2.7b) is greater than that predicted for the cast-in-place system
(Figure 4.2.7a), as might be expected.
It is interesting to note that the magnitude of the increase in response is consistent
with that predicted by Eq. 3.1.45. Consider the following:
Cast-in-Place System
max
µ =
y
7.0
= (see Figure 4.2.7a)
1.5
= 4.7
√
µ−1
ζeq = √ (Eq. 1.1.9d)
π µ
= 0.17
ζ̂eq = 17 + 5 (Eq. 1.1.9a)
= 22
3.38 − 0.67 ln ζ̂eq
R̂ = (Eq. 3.1.45)
2.3
= 0.57
Figure 4.2.6 (Continued) Beam behavior models static displacement controlled behavior.
Hybrid System
max
µ =
y
8.2
= (see Figure 4.2.7b)
1.5
= 5.5
√
µ−1
ζeq = √ (Eq. 1.1.9d)
π µ
= 0.18
Now, adjust the equivalent damping to account for the reduction in energy absorbed
(Figure 1.1.8). Presuming that the hybrid system is 60% effective,
ζeq = 0.6ζeq
= 0.11
ζ̂eq = 11 + 5 (Eq. 1.1.9a)
= 16
3.38 − 0.67 ln ζ̂eq
R̂ = (Eq. 3.1.45)
2.3
= 0.66
The resultant increase in displacement should be on the order of
R̂hybrid 0.66
=
R̂cast-in-place 0.57
= 1.16
The recorded moment curvatures for the cast-in-place beam and the hybrid beam are
presented in Figure 4.2.8. Observe that the curvature demand imposed on the hybrid
beam is almost 40% greater than that imposed on the cast-in-place beam.
Conclusions
Figure 4.2.8 Recorded hysteretic behavior of frame beam (see Figure 4.2.7).
Photo 4.2 Precast concrete clad poured-in-place concrete ductile frame braced condo-
minium, The Remington, Los Angeles, CA, 2002. (Courtesy of Magee Architects, Inc.)
Figure 4.2.9 Plastic hinge distribution in Frame 3-2, = 17 in., t = 2.5 seconds (see Figure
4.2.1b).
t = 2.5 seconds. As a consequence, Figure 4.2.9 describes how a response in the first
mode excursion might distribute inelastic actions.
First, note that the rotational ductility demand is concentrated in the lowermost
part of the frame, about six to seven floors extending over a height that corresponds
to the depth of the frame. Accordingly, these lower floors might be viewed as an
effective plastic hinge region for the frame. The contrast is even more pronounced
when described in terms of the postyield rotation demand. The postyield rotation
demand, θp , in the lower floors (2 and 3) is 6.5 times that imposed on level 11.
The real concern is with the probable level of strain induced in the concrete. This
can be estimated fairly easily.
The range of postyield rotation (θp ) is between 0.014 radian ( = 17 in.) and
0.031 radian at a building drift of 30 in. This corresponds to a range of postyield
curvature demands of between 0.00078 and 0.0017 rad/in. One can either resort to
computer developments similar to those whose conclusions are presented in Table
2.2.1 or use shorthand methodologies to identify the range of strain states.
λo Mn = 1000 ft-kips
h = 36 in.
a ∼
d− = 30 in.
2
λ o Mn
Cc =
d − a/2
1000
=
2.5
= 400 kips
Cc
a=
0.85fc b
400
=
0.85(5)(20)
= 4.7 in.
a
c=
β1
4.7
=
0.8
= 5.9 in.
This would be a conservative estimate because it does not account for the compression
steel. The depth to the neutral axis would be reasonably assumed to be on the order of
5 in. The induced level of postyield concrete strains would be estimated as follows:
λo Mn
φy = (Computer model)
EIe
1000(12)
=
4000(27,200)
= 0.00011 rad/in.
φy -c
θy = (Conjugate beam)
6
0.00011(276)
=
6
= 0.005 radian
θp
+ 1 = 4.4 (see Figure 4.2.9)
θy
θp = 3.4θy
= 3.4(0.005)
= 0.017 radian
θp
φp = (Eq. 2.1.9)
-p
0.017
=
18
= 0.00096 rad/in.
εcp = φp c
= 0.00096(5)
= 0.0048 in./in.
εcu = εcp + εcy
= 0.0048 + 0.003
= 0.0078 in./in.
Based on this, one might conclude that spalling was not likely to occur in the lower
level frame beams.
In summary, it seems reasonable to conclude that
drifts of 3.75 to 4.04 in. were predicted for Frame 3-2. The predicted yield base shear
was 435 kips. In Section 3.2.6 a sequential yield analysis was performed on Frame 3-2
(Figure 3.2.8a). Postyield behavior did not start until the building drift reached almost
7 in., (DR = 0.39%) and the base shear associated with first yield (λo Mn = 1000
ft-kips) was over 600 kips. This apparent contradiction is a direct consequence of
modeling assumptions used to describe the behavior of the beam (see Figure 2.1.3).
The analysis process used to generate Figure 3.2.8a adopted a frame behavior model
that is quite different from that used to predict postyield beam strain states in Section
3.2.3.1. The difficulty encountered using the sequential yield model (Figure 3.2.8a)
lies in the fact that concrete strain states are hard to predict, for behavior at idealized
yield has been in the postyield domain for some time (see Section 2.1.1). This diffi-
culty is overcome when we adopt steel first yield (εs = εy ) as a point of departure.
The following yield baseline was developed in Section 3.2.3.1.
εsy
φy = (Eq. 2.1.7a)
0.67d
0.002
=
0.67(33)
= 0.00009 rad/in.
φy -c
θy = (Conjugate beam)
6
0.00009(276)
=
6
= 0.00414 radian
y ∼
= 4 in. (Section 3.2.3.1, Step 8)
θpb = φp -p
= 0.0018(18)
= 0.0324 radian
δp = θj hx
= 0.027(120)
= 3.24 in.
For the eight stories of assumed postyield response, this corresponds to a postyield
component of building drift of
p = 8(3.24)
= 26 in.
u = y + p
= 4 + 26
= 30 in.
u
µ=
y
30
=
4
= 7.5
Observe that this is reasonably consistent with the strain states conclusions developed
from the time history analysis (Section 4.2.3), given the fact that no postyield rotation
was presumed above level 8.
Conclusion: We should be convinced that the selected system will meet our perfor-
mance objectives. Further, that the use of a system ductility/overstrength factor of 8
was appropriate.
Conclusions
• Inelastic time history analyses are most effectively used to confirm the existence
of a well-distributed hinging pattern, one that avoids soft regions and strength
degradation.
• The temptation to fine-tune beam strengths at this point in the design process
should be resisted unless the need is clearly identified.
Frame 3-1
fc = 5 ksi
P = 1820 kips
M = 2097 ft-kips
θp = 0.01 radian
εc = 0.007 in./in.
P
= 0.28
Ag fc
Frame 3-2
fc = 5 ksi
P = 2141 kips
M = 2025 ft-kips
θp = 0.0135 radian
εc = 0.0135 in./in.
P
= 0.33
Ag fc
Frame 3-3
fc = 5 ksi
P = 3100 kips
M = 2330 ft-kips
θp = 0.02 radian
εc = 0.024 in./in.
P
= 0.48
Ag fc
Frame 4-1
fc = 5 ksi
P = 1820 kips
M = 1903 ft-kips
θp = 0.012 radian
εc = 0.0095 in./in.
P
= 0.28
Ag fc
Frame 4-2
fc = 5 ksi
P = 2430 kips
M = 2300 ft-kips
θp = 0.0165 radian
εc = 0.0155 in./in.
P
= 0.375
Ag fc
Clearly, the concrete strengths used in Frames 3-2, 3-3, and 4-2 need to be increased
so as to reduce the level of concrete strain and improve the level of available ductility,
but this should have been discovered much earlier in the design process.
Conclusion: Frame strength must not exceed the reasonable postyield deformation
capability of the compression side column.
My
φy = (Reasonably elastic behavior)
EIe
My -
θy =
EIe 6
θy = C1 -
Figure 4.2.10 (a) Time history response of irregular Frame 3-2I to Earthquake 2. (b) Hinge
propagation—irregular frame of part a.
Hence the rotation at yield for members of equal strength but unequal spans becomes
-1
θy1 = θy2
-2
θcp
µ1 =
θby1
θcp
µ2 =
θby2
The relationship is further exacerbated when the centroid of beam postyield rotation
moves away from the face of the column.
µθ 1 -c1 − -p
=
µθ 2 -c2 − -p
23 − 1.5
=
13 − 1.5
= 1.87
An obvious mitigation would be to increase the strength of the interior beam, and this
can be done if it does not increase the reinforcement ratio to undesirable levels. Recall
that an identical reinforcement program in adjoining beams was viewed as being
desirable from a constructibility perspective, so a clear need should be perceived
before adopting a dissimilar reinforcing program.
Consider the strain states imposed on the plastic hinge region on either side of the
interior column if the postyield component of story drift is 1.8%.
0.022
=
18
= 0.0012 rad/in.
εcp = φb2p c
= 0.0012(5)
= 0.006 in./in.
εcu = εcy + εcp
= 0.003 + 0.006
= 0.009 in./in.
Observe that postyield concrete strain states are essentially the same on both sides of
the interior column. Hence, the rotation ductility demands described in Figure 4.2.10b
tend to deceive, and an increase in the strength of the interior beam will only tend to
promote a more uniform yielding of the system, but not significantly alter induced
postyield strain states.
Conclusions
Stability and torsion are a major concern of the structural engineer. Both topics are
reasonably well understood for systems whose behavior is elastic but the extrapola-
tion to the inelastic behavior range is conjectural. A pragmatic overview is the goal
of this section.
restoring force, and this complicates any reduction of the problem to a simple design
rule or procedure.
The form adopted for the stability index is the ratio of the P shear to the mech-
anism shear:
Ptr
Q= (4.3.1)
V M hx
where Ptr is the sum of the tributary mass above the level being considered and VM
is the mechanism shear which, for our purposes, is
λo Mn
VM = (4.3.2)
hx
Through the years the stability index (Q) has been related to the elastic displacement
associated with a static lateral load criterion (y ) and Q∗ to the inelastic displace-
ment u or µy . Paulay and Priestley [4.1] adopt the inelastic approach and define a
consideration limit state for Q∗ of 8.5%.
In other words, a design should mitigate P effect when the basic design strength
VM (Eq. 4.3.2) is not enough to cause Q∗ to be less that 8.5%. In this case VM should
be increased by (1 + Q∗ ). Hence
where VM∗ is the objective design mechanism strength based on the mechanism
strength developed absent P considerations.
The identified limit state (8.5%) and mediation were developed from analytical
work done after the San Fernando Earthquake (1971), which suggested that P
effects substantially increased building drifts and residual drifts. This is not consistent
with the analytical efforts reported in this chapter or the author’s experience. Further,
inelastic drifts were, in the stability index approach, typically assumed to be a function
of the system ductility factor amplified to account for the distribution of inelastic
demand in the lower stories, a subject discussed in Section 4.2.
The compensating strength (VM∗ − VM ) has typically been based on an energy
balancing approach where the actual increase in strength is half of that required to
compensate for P effects at the anticipated displacement or ductility demand. In
the example building, this would amount to doubling the design ductility factor to
account for a concentration of inelastic displacement in the lower part of the structure
and then halving it to balance the energy dissipated.
Now apply these concepts to the frame designs of Chapter 3.
Working loosely so as to identify the potential level of additional or P compen-
sating shear on the three-bay frame (3-2), assume that we have concluded that the
following appropriately describes the action anticipated in the lower levels:
4.3.2 Torsion
Few analytical procedures are more futile than an elastic torsional analysis to demon-
strate compliance with prescriptive drift objectives. To confirm this consider the dis-
cussion regarding the design of unequal shear walls or irregular frames. A time history
treatment of torsion must consider inelastic behavior if it is to have meaning, and this
is not possible now nor is it likely to be in the near future. The intent of this section is
to describe how the impact of torsion in the inelastic range can rationally be mitigated.
Paulay [4.2] presented an excellent discussion of inelastic torsion, and the essence of
his solution was to provide an orthogonal bracing program that would remain elastic
and thereby mitigate the consequence of unbalanced torsion in the postyield behavior
range. This is certainly one solution, but it is not usually available to the design team.
I see no need to repeat Paulay’s arguments, and instead approach the problem from a
pragmatic position and offer mitigating measures that can be easily developed in the
design process.
Why does inelastic behavior impact torsion adversely? To understand the con-
cern consider the spring supported beam described in Figure 4.3.3a and the spring
between Systems 3 and 4 is significant, the center of rigidity will move to a corner
of the building and the behavior will revert to the planar condition developed for the
beam described in Figure 4.3.3a.
Eventually someone will develop an inelastic three-dimensional computer pro-
gram that will consider the dynamic impact of inelastic behavior, but in the mean
time and even given a computational assist, a practical design solution is required.
My design approach has been to balance the strength of the bracing systems so as to
match the inelastic behavior of each system to the extent possible. This is most simply
done by providing identical bracing elements. Unfortunately, this is more often than
not an unacceptable solution from an aesthetic or functional perspective (see Fig-
ure 4.3.5). Given a condition where bracing systems must be different, they must be
strength balanced to the extent possible, even if it means that strength or prescriptive
mandates be violated.
A sequential yield analysis is most effectively used to balance bracing systems.
The behavior of two significantly different wall systems is described in Figure 4.3.3.
The strength of shear wall 1 must be significantly reduced in order to match that
of wall 2. The plan aspect ratio (length to width) of the building described in Figure
4.3.4 is nearly 3, and this would make the impact of any orthogonal bracing system on
torsional response unlikely. Observe that this balancing of bracing system strengths
should minimize torsional response even if the stiffness of the two systems is in-
correctly assumed. This is fortunate because, as should be clear from the material
Clearly, one cannot expect to entirely eliminate the impact of torsion on the response
of a building, but a strength-based approach will produce much better results than
any elastic stiffness-based alternative.
SELECTED REFERENCES
[4.1] T. Paulay and M. J. N. Priestley, Seismic Design of Reinforced Concrete and Masonry
Buildings, John Wiley & Sons, New York, 1992.
[4.2] T. Paulay, “Are Existing Seismic Torsion Provisions Achieving the Design Aims?”
Earthquake Spectra, Vol. 13, No. 2, p. 249.