INCREMENTAL DYNAMIC ANALYSIS TO ASSESS THE PERFORMANCE, CAPACITY AND DAMAGE
4.3 METHODOLOGY EMPLOYED IN THE IDA STUDY
4.4.1 S TRUCTURAL M AXIMUM D EMANDS
4.4.1.1 Radial Interstorey Drifts
Maximum values of global radial interstorey angular drifts for the IDA of the two sets of
the five prototype structures were computed and are presented in Figures 4-2 and 4-3. Tables 4-
6 and 4-7 list their overall medians for both sets of structures where they show strong
resemblance between both sets of results. Figure 4-2 is for the structures with elastically
behaving columns, Figure 4-3 presents results for the structures with columns allowed to behave
inelastically. The presented results had been further manipulated by setting a cut-off when
values of interstorey drifts exceed 10%, effectively implying structural collapse; see FEMA 350
(2000) Table 4-10. Figures 4-2 (a) and 4-3 (a) exhibit the median values of radial interstorey
drifts and Figures 4-2 (b) and 4-3 (b) display the 90th percentile values. Figures 4-2 (c) and 4-3
(c) show the median values of radial interstorey drifts normalised by their intensity measures
The rationale for normalising the maximum values of global radial interstorey angular
drifts by their relevant IM values, S (Ta 1=2s), can be found in Cornell et al (2002) and Jalayer and
Cornell (2003). Jalayer and Cornell (2003) provided the following expression for interstorey
drift demand as a power function of spectral acceleration and lognormal random variable ε:
bε
(4.2)
θ = a Sa
where a and b are least-squares regression coefficients (in log-log space) for displacement-based
demands, such as interstorey drifts (θ). For medium- to long-period structures the well-known principle of equal displacement (Veletsos and Newmark, 1960), renders b = 1. The lognormal
random variable ε is a random variable with unit median that encompasses both epistemic and aleatory uncertainties, which at this stage can be taken as one for this deterministic study.
Consequently, the expression is simplified to:
(4.3)
a = θ/ Sa
Therefore without randomness, the term θ/Sa, or θ/IM, reduces to a simple single
constant, a. However, variation in a can be expected due to the randomness of the input and the
output (response). Therefore, θ/IM in this case represents the magnitude of interstorey drifts at the same universal value of IM = Sa(T1=2s) = 1g, thus facilitating the comparison between
different levels of IM and for all structures utilised in this study.
On inspecting figures 4-2 (a to d) and 4-3 (a to d) along with their numerical data, both
results were identical up to MCE values of Sa(T1=2s) = 0.25g. This verifies that structural
performance followed the inherent strong-column/weak-beam principle with no plastic hinges
occurring in columns except at ground and roof levels. Beyond MCE, overall structural
instability started to occur where it can be seen strongly evident at 3DBE values at Sa(T1=2s) ≥
0.45g, incipient collapse is evident. At this intensity level, interstorey drifts in excess of 8% may
be expected which is well beyond their stable capacities (4%). Such instabilities and erratic
behaviours were marked in the analyses with columns allowed inelastic behaviour due to the
On further inspection of Figures 4-2 (c) and (d) together with 4-3 (c) and (d), one can realise a
noticeable increase in the values θr/IM for the bisymmetrical structure (Tx = Ty) followed by that of
Tx/Ty = 1.5 when compared with other structures of Tx/Ty = {2.0, 2.5, 3.0}, specifically marked in the
90th percentile figures. Table 4-3 lists the increases in both median and 90th percentile values for Tx/Ty
= 1.0 and 1.5 when compared with their overall medians. Such increased response can be attributed to
the rather more flexible nature of bisymmetrical structures and those with close bidirectional
fundamental periods in comparison with those with Tx/Ty≥ 2, as well as the apparent synergic coupling
of the two principal bidirectional responses. This finding emphasises the importance of considering the
bidirectional fundamental period ratio (Tx/Ty) when studying displacement-based demands. In essence
one can postulate that for structures with close bidirectional fundamental periods, the expected drift
values would be higher by 3-15% than those with periods well apart with Tx/Ty > 1.5.
In keeping with Demand and Capacity Factor Design (DCFD), Jalayer and Cornell
(2003), and on inspecting Figures 4-2 (c) and (d) together with 4-3 (c) and (d) as well as their
data up to MCE one can substantiate that interstorey drifts are candidates of high level of
confidence as the reported standard deviation and coefficient of variations (COV) are markedly
low. Experience, and many authors such as Shome and Cornell (1999), suggests that interstorey
drifts are distributed lognormally about their medians with their standard deviations of their
natural logarithms, βθ| Sa. Note that, βθ| Sa = σlnθ|Sa is the standard deviation of the natural
logarithms of θ, herein referred to as the dispersion and given by the symbol β. According to Benjamin and Cornell (1970), COV values are approximately numerically equal to the lognormal
standard deviation (σlnx) for moderate levels of dispersion less than 0.3. Consequently these
COV values are only considered as a measure of confidence in values of the considered results
when their values COV < 0.3 showing low variability in results. On studying COVθrp|Sa of
median-to-median of θr median responses (p=50%) and 90th percentile demands (p=50%), the median of COV values was 0.065 for medians and 0.07 for 90th percentiles, Table 4-3. Further,
structural design capacity, i.e., Sa(T1=2s) ≤ Sa(MCE). This finding strongly suggests a high level
of confidence in the probabilistic demand representation of maximum radial interstorey drifts
due to their substantial independence of structural configuration, albeit noticeably dependant on
the biaxial interaction for T /Tx y ≤ 1.5. Furthermore, θr/IM substantial independence of levels of
IM can be realised from Figures 4-2 (c) and (d) along with 4-3 (c) and (d), where the values of
θr/IM largely maintain same values (almost a vertical line in the figures) for different structural configurations and IM values up to MCE. Consequently, θr is linearly proportional to IM, Sa(T1=2s) ≤ Sa(MCE), validating the equal displacement principle, thus b≈ 1.
th
Regression analysis was conducted on the results of median and 90 percentile values of
θr to evaluate the coefficients of power equation θ = a Sab. Table 4-8 lists the median values of a
and b for both the median response and 90th percentile code-maximum demands at Sa(T1=2s) ≤
Sa(MCE) and > Sa(MCE). Median values of b for median response and code-maximum demands
were respectively 1.127 and 1.055 for Sa(T =2s) 1 ≤ Sa(MCE), very close to one (equal
displacement). While for Sa(T =2s) > S1 a(MCE), values of b were respectively 1.4 and 1.96 (i.e.
close to parabolic). Table 4-9 shows COV values of a and b, where COV(b) were 0.057 << 0.3
for Sa(T =2s) 1 ≤ Sa(MCE) and 0.24 < 0.3 for Sa(T =2s) > S1 a(MCE), confirming their credible
values. Value of a code-maximum demands at Sa(T =2s) 1 ≤ Sa(MCE) was 0.086 with COV(a) =
0.144 < 0.3, renders it credible for implementation. However, its value of 0.264 for ≤ Sa(MCE)
has COV(a) = 0.632 >> 0.3, indicating high variability due to erratic behaviour for motions
larger than MCE. In essence, one can postulate that θr are predominately dependent on Sa(T1)
and differential interstorey capacities and marginally on differential biaxial capacities.
Accordingly, one can implement the power equation of:
bγ
θ = a Sa biaxial,θε (4.4)
where values of a and b are obtained from Table 4-8 and γbiaxial,θ is the θr biaxial factor equals
{1.03–1.15} Tx/Ty≤ 1.5 for 90th percentile maxima and one for Tx/Ty > 1.5. The value of γbiaxial,θ
γbiaxial,θ= 1.15 – 0.24 (Tx/Ty – 1) (4.5)
Equation (4.4) is suitable for medium to long period seismic-code designed structures that
follows the equal displacement principle. In addition, it can be credibly implemented in rapid
calculation of expected maximum values of θr and in the DCFD framework. 4.4.1.2 Structural Maximum Response of Lateral Storey Drifts
As in section 4.4.1.1, Figures 4-4 and 4-5 exhibit the demands of global maxima of lateral
radial storey drifts (δr) per unit IM and Tables 4-6 and 4-7 list their overall medians values. Both
figures and tables show strong resemblance as with the case of θr and thus drawing the same above- mentioned conclusion that the considered structures with columns allowed to behave inelastically
acted similar to those with elastic columns up to MCE. Consequently, the structures with realistic
(inelastic) columns followed the inherent Capacity Design principles and their columns behaved
essentially elastically up to MCE. Beyond MCE, the 90th percentile maximum values of δr
erratically increase exhibiting apparent collapse behaviour. Note the maximum lateral storey drifts
in all cases occur at the roof level, as later discussed in the following section 4.4.2.
Figures 4-4 (c) and (d) together with 4-5 (c) and (d) for Sa(T1=2s) ≤ 0.10g (elastic
behaviour) show that the values of δr/IM were almost identical, i.e., b ≈ 1. However, δr/IM values for Sa(T1=2s) ≥ 0.15g (DBE) and ≤ 0.25g (MCE) show reducing magnitudes of lateral
sways due to structural softening. As the structures underwent inelastic deformation their
fundamental periods were extended reducing their structural Sa intake, and consequently a
decrease in the values of δr resulted. This indicates that the coefficient b is less than one. A similar change in behaviour is reported by Ramamoorthy et al (2006) that shows discontinuities
at the onset of first yield. Nonetheless, the values of δr/IM maintain the same ratios with their
overall medians, particularly up to MCE.
As with θr/IM, on further inspecting Figures 4-4 (c) and (d) together with 4-5 (c) and (d), one can realise a noticeable increase in values of δr/IM for the bisymmetrical structure (Tx = Ty) followed
by that of Tx/Ty = 1.5 when compared with other structures of Tx/Ty = {2.0, 2.5, 3.0}. This increase
is specifically marked in the 90th percentile Figures 4-4 (d) and 4-5 (d). Table 4-4 lists the increases
in both median and 90th percentile values for Tx/Ty = 1.0 and 1.5 when compared with their overall
medians. As with θr/IM, the same conclusion was drawn which emphasises the importance of considering the bidirectional fundamental period ratio (Tx/Ty) when studying displacement-based
demands. In essence for structures with close bidirectional fundamental periods, the expected values
would be higher by 1-12% than those with periods well apart with Tx/Ty≥ 2.
Finally, on inspecting Figures 4-4 (c) and (d) together with 4-5 (c) and (d) as well as their
data up to MCE one can substantiate that maximum radial storey sways are candidates of high
level of confidence as the reported standard deviation and coefficient of variations (COV) are
markedly low. On studying COVδp|Sa median-to-median of the median δr responses (p=50%)
and 90th percentile δr demands (p=90%) in Table 4-4, the median of COV values was 0.056 for medians and 0.042 for 90th percentiles. This finding strongly suggests a high level of confidence
in the probabilistic demand representation of storey lateral drifts due to their substantial
independence structural configuration, albeit noticeably dependant on the biaxial interaction for
T /Tx y ≤ 1.5. The substantial independence of δr can be realised from Figures 4-4 (c) and (d)
along with 4-5 (c) and (d) where the values of δr/IM largely maintain the same ratios with their overall medians for different structural configurations up to MCE. Values of δr/IM are, linearly proportional to IM, Sa(T =2s), for S1 a(T1) ≥ DBE.
th
Regression analysis was conducted on the results of median and 90 percentile values of
δ b
r to evaluate the coefficients of power equation δ = a Sa , to test its applicability. Table 4-8 lists
the median values of a and b for both the median response and 90th percentile code-maximum
demands at Sa(T =2s) 1 ≤ Sa(MCE) and > Sa(MCE). Median value of b for median response and
code-maximum demands were respectively 0.845 and 0.825 < 1, due to structural softening, for
Sa(T =2s) 1 ≤ Sa(MCE). While for Sa(T =2s) 1 ≤ Sa(MCE), b values were respectively 1.314 and
Sa(T =2s) 1 ≤ Sa(MCE) and 0.169 < 0.3 for Sa(T =2s) > S1 a(MCE), assuring their credible values.
Value of a code-maximum demands at Sa(T =2s) 1 ≤ Sa(MCE) was 1.196 with COV(a) = 0.123 <
0.3, renders it credible for implementation. However, its value of 3.044 for ≤ Sa(MCE) has
COV(a) = 0.489 > 0.3, indicating its variability due to erratic behaviour for motion larger than
MCE. In essence, one can postulate that δr are predominately dependent on Sa(T1) and storey
capacities and marginally on differential biaxial capacities. Accordingly, one can implement the
power equation of:
bγ ε
(4.6)
δ = a Sa biaxial,δ
where values of a and bare obtained from Table 4-8 and γbiaxial,δ is the δr biaxial factor equals
{1.02–1.11} Tx/Ty≤ 1.5 for 90th percentile maxima and one for Tx/Ty > 1.5. The value of γbiaxial,δ
can be interpolated for 1.0 ≤ T /Tx y≤ 1.5 by the following expression:
γbiaxial δ= 1.11 – 0.18 (Tx/Ty – 1) (4.7)
Equation (4.6) is suitable for medium to long period seismic-code designed structures that
follows the equal displacement principle. Again, it can be credibly implemented in rapid
calculation of expected maximum values of δr and in DCFD framework. 4.4.1.3 Structural Maximum Response of Storey Accelerations
This section follows the same framework implemented in studying both θr and δr, in order
to assess the applicability of the power function A = a Sab and its credibility in representing
global maximum maxima of radial storey accelerations (Ar). Figures 4-6 and 4-7 display the
demands of Ar per unit IM and Tables 4-6 and 4-7 list their overall medians that show strong
resemblance indicating that the considered structures with columns allowed inelastic behaviour
acted as the ones with elastic columns up to MCE. Consequently, the structures with realistic
columns followed the inherent Capacity Design principles and their columns behaved elastically
up to MCE. Beyond MCE; the 90th percentile (code-maximum) values of Ar erratically increase
Figures 4-6 (c) and (d) together with 4-7 (c) and (d) for Sa(T1=2s) ≤ 0.10g (elastic
behaviour) show that the values of Ar/IM were almost identical, i.e., b ≈ 1. However, Ar/IM
values for Sa(T1=2s) ≥ 0.15g (DBE) and ≤ 0.25g (MCE) show slight reduction in magnitudes of
radial storey accelerations due to structural softening. As the structures underwent inelastic
deformation their fundamental periods were extended reducing their structural Sa intake, and
consequently; decreasing the values of Ar. This indicates that the coefficient b is less than one.
Nonetheless, the values of Ar/IM maintain the same ratios with their overall medians,
particularly up to MCE.
As with θr/IM and δr/IM, on further inspecting Figures 4-6 (c) and (d) together with 4-7 (c) and (d), one can realise a noticeable reduction in values of Ar/IM for the bisymmetrical
structure (T = Tx y) followed by that of T /Tx y = 1.5 when compared with other structures of Tx/Ty
= {2.0, 2.5, 3.0}. This reduction is specifically marked in the 90th percentile figures. Table 4-5
lists the decreases in both median and 90th percentile values for T /Tx y = 1.0 and 1.5 when
compared with their overall medians. For the 90th percentile demands, the reduction was about
15% for T /Tx y = 1.0 and 1.5 that is attributed to their relative higher flexibility. Note that Figure
4-7 (d) and data show the median response lie about T /Tx y = 2.5. Consequently one can
postulate that for structures with close bidirectional fundamental periods, Tx/Ty ≤ 1.5, the
expected values would be lower by 15% than those with periods well apart with T /T x y≥ 2. Finally, on inspecting Figures 4-6 (c) and (d) together with 4-7 (c) and (d) as well as their
data up to MCE; one can substantiate that maximum radial storey accelerations are candidates of
high level of confidence as the reported standard deviation and coefficient of variations (COV)
are markedly low. On studying COVAp|Sa median-to-median of the median Ar responses
(p=50%) and 90th percentile Ar demands (p=90%) in Table 4-5, the median of COV values was
0.036 for medians and 0.082 for 90th percentiles. This finding strongly suggests a high level of
confidence in the probabilistic demand representation of storey accelerations due to their
interaction for T /Tx y≤ 1.5. The substantial independence of Ar can be realised from Figures 4-6
(c) and (d) along with 4-7 (c) and (d) where the values of Ar/IM largely maintain same ratios
with their overall medians for different structural configurations up to MCE. Values of Ar/IM
are linearly proportional to IM, Sa(T =2s), for S1 a(T1) ≥ DBE.
th
Regression analysis was conducted on the results of median and 90 percentile values of
Ar to evaluate the coefficients of power equation A = a Sab, to test its applicability. Table 4-8
lists the median values of a and b for both the median response and 90th percentile code-
maximum demands at Sa(T =2s) 1 ≤ Sa(MCE) and > Sa(MCE). Median values of b for median
response and code-maximum demands were respectively 0.982 ≈ 1 and 0.795 < 1 (due to structural softening) for Sa(T =2s) 1 ≤ Sa(MCE). While for Sa(T =2s) > S1 a(MCE), values of b
were respectively 0.797 and 1.332. Table 4-9 shows COV values of a and b, where COV(b)
were 0.163 < 0.3 for Sa(T =2s) 1 ≤ Sa(MCE) and 0.2 < 0.3 for Sa(T =2s) > S1 a(MCE), assuring their
credible values. Value of a code-maximum demands at Sa(T1=2s) ≤ Sa(MCE) was 7.434 with
COV(a) = 0.385 > 0.3, indicating its high variability. However, its value of 13.939 for ≤ Sa(MCE) has COV(a) = 0.215 > 0.3, renders it credible for implementation. In essence, one can
postulate that Ar is predominately dependent on Sa(T1) and storey strength capacities and
marginally on differential biaxial capacities. Accordingly, one can implement the power
equation of:
bγ
A = a Sa biaxial,Aε (4.8)
where values of a and b are obtained from Table 4-8 and γbiaxial,A is the Ar biaxial factor equals
0.85 for Tx/Ty ≤ 1.5 for 90th percentile maxima and one for T /Tx y > 1.5. Equation (4.8) is
suitable for medium to long period seismic-code designed structures that follow the equal
displacement principle. Furthermore, it can be credibly implemented in fast calculation of