Proof Researchgate

See discussions, stats, and author profiles for this publication at: https://www.researchgate.
net/publication/361268334
Equivalent static loads for double-layered domes supported by multistorey

buckling-restrained braced frames
Article in International Journal of Space Structures · June 2022

DOI: 10.1177/09560599221097834
CITATION READS
1 84
3 authors:
Deepshikha Nair Yuki Terazawa

Tokyo Institute of Technology Tokyo Institute of Technology
6 PUBLICATIONS 20 CITATIONS 59 PUBLICATIONS 296 CITATIONS
SEE PROFILE SEE PROFILE
Toru Takeuchi
Tokyo Institute of Technology
274 PUBLICATIONS 2,596 CITATIONS
SEE PROFILE
All content following this page was uploaded by Deepshikha Nair on 01 August 2022.
The user has requested enhancement of the downloaded file.

Page 1 of 35 International Journal of Space Structures
1
2
3 International Journal of Space Structures
XX(X):1–35
4 Equivalent static loads for ©The Author(s) 2021
5 Reprints and permission:
6 double-layered domes supported by sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/ToBeAssigned
7
8
multistorey buckling-restrained braced www.sagepub.com/
SAGE
9 frames
10
11
12 Deepshikha Nair1 , Yuki Terazawa1 and Toru Takeuchi1
13
14
15
16 Abstract
17
Unlike multistorey frames, gridshells exhibit coupled horizontal-vertical response, closely spaced
18
19 modes and substructure-roof interaction. Previous studies have proposed elastic horizontal and
20 vertical equivalent static seismic forces determined from the input horizontal acceleration of the
21 substructure roof, an assumed acceleration distribution, nodal masses and amplification factors
Fo
22 derived from the dynamic characteristics of the dome and substructure. To extend this methodology
23 to realistic passive-controlled substructures, this paper investigates the applicability of two simple
24
ductility reduction factors (or Rµ factors) to estimate the inelastic response spectra and compute
rP
25
26 the peak horizontal acceleration of multistorey substructures with buckling-restrained braces. These
27 are then used to obtain the equivalent static loads of the roof using amplification factors which are
28 compared with those obtained from the iterative equivalent linearisation method and the accuracies
ee
29 are verified against the results from nonlinear response history analyses. It was confirmed that
30 the Rµ factors combined with the roof amplification factors provide a simple way to estimate
31 the peak roof response with sufficient accuracy for preliminary design of domes with multistorey
rR
32
33
substructures having low post-yield stiffness.
34
35
ev
Keywords
36 double-layered domes, seismic response, multistorey substructures, R-factors, equivalent static
37
38
loads, buckling-restrained braced frames
iew
39
40
41
42 1 Introduction
43
44 Equivalent static lateral force method often forms the precedent for seismic design of regular nonlinear
45 multistorey buildings as prescribed by most design codes like the ASCE-7-16 [1], NZS1170.5 [2],
46 Eurocode 8 [3] and the Building Standard Law of Japan [4]. For example, the ASCE-7-16 provides
47 response reduction factors (R-factors [1]), to estimate the design base shear and the horizontal storey
48
49
forces based on the type of buildings. However, gridshells are often treated as irregular or special
50 structures owing to the complex roof-substructure interaction and closely-spaced multiple participating
51 modes [5, 6] and therefore require advanced analyses. Modal response-spectrum analyses (RSA) is a
52 relatively simple and efficient method often used to compute the peak response of elastic gridshell
53 domes and are recommended in code specifications [6–8]. The method estimates the peak response of
54 each participating mode using the spectral acceleration of the elastic design spectrum and combines
55
56
57 1 Tokyo Institute of Technology, Japan
58
59
Corresponding author:
60 Deepshikha Nair, Department of Architecture & Building Engineering, Tokyo Institute of Technology
2-12-1-M1-29 Ookayama, Meguro-ku Tokyo, Japan
Email: nair.d.aa@m.titech.ac.jp
https://mc.manuscriptcentral.com/sps
Prepared using sagej.cls [Version: 2017/01/17 v1.20]
International Journal of Space Structures Page 2 of 35
2 International Journal of Space Structures XX(X)
1
2 the modal responses as per the appropriate combination rule. In the context of gridshell structures, its
3 application is limited to elastic structures as it can not be extended to domes with nonlinear substructures
4 as there are no prescribed scaling or response reduction factors for higher modes nor is it evident how
5 multistorey substructures response estimated using the standard reduction factors can be applied for roof
6
7
structures [5, 6].
8 Metal gridshells supported by substructures in regions of severe seismic hazard need to be designed
9 to withstand the high seismic demand. The seismic response of roofs with simple symmetric lattice
10 geometries like the arch, dome and cylinder have been extensively studied by varying the parameters
11 like the span (30m-150m), half-subtended angle (20°-40°), out-of-plane stiffness, rise-to-span ratios
12
13
and substructure stiffness [6, 7, 9–13]. It was found that increasing the out-of-plane bending stiffness
14 of the members prevented failure due to buckling that is common in single-layer roofs. The number
15 of participating modes are also reduced and the response can be simply estimated as a function of four
16 dominant modes- three anti-symmetric modes (‘O1’, ‘O2’ and ‘O2.5’) and an in-plane mode ‘I’ [5,7,14].
17 For roofs with supporting substructures, the seismic input amplifies up the substructure height and the
18 substructure roof’s horizontal acceleration acts as the input for the gridshell roof. The geometry and
19
20
dynamic characteristics of gridshell lead to its coupled horizontal and anti-symmetric vertical roof
21 response. In an attempt to develop equivalent static loads, these elastic substructure-roof interactions
Fo
22 have been investigated [13, 15] and quantified as horizontal and vertical amplification factors that are
23 functions of the substructure to roof’s period ratios. The vertical response was found to amplify by up
24 to three times the input horizontal seismic input at the substructure roof level [7]. Further increasing the
rP
25 height of the substructure leads to increased participation from the higher substructure modes and the
26
interaction between the higher substructure modes (in linear range) and dominant roof modes have also
27
28 been quantified as higher mode amplification factors [16].
ee
29 Gridshell roofs are often employed in school gymnasiums or community centers which are expected
30 to be functional as shelters even after a big earthquake in countries of high seismic hazard like Japan.
31 Consequently, it becomes important to design such structures with seismic response control strategies.
rR
32
33
One efficient solution to limit the roof seismic demand by decreasing the peak substructure response is to
34 add damping devices such as the hysteretic buckling-restrained braces (BRB) to the substructure frames
35 [17] and the response estimation of such systems have also been studied [8, 13, 18, 19] using Kasai’s
ev
36 equivalent linearisation method [20, 21]. This lowered demand enables the engineers to design the roofs
37 such that they remain elastic and the reduced roof acceleration also ensures the seismic performance
38
iew
of acceleration sensitive non-structural components such as ceilings and lighting equipment. Practical
39
applications of employing passive-control devices include the Toyota stadium [22], Shimokita dome
40
41 [23] and other seismic-retrofit projects of high-school gymnasiums [24, 25]. While there exists well-
42 established response-reduction factors for design and estimation of the inelastic design spectra of
43 conventional BRB frames [26,27], their application to substructures of curved gridshell roofs has not yet
44 been explored. Therefore, as neither reliable static load procedures nor realistic R-factors are available
45 for such structural systems, the time- and data-intensive nonlinear response history analyses (NLRHA)
46
are often the only design route. With an aim to improve the preliminary design of such double-layered
47
48 domes, this paper investigates the applicability of the ductility-based reduction factors (labelled as the
49 ‘R-factor approach’) to estimate the peak substructure response using RSA which is used to propose
50 equivalent static forces of the dome. The response of simpler single-storey substructures are first studied
51 and the response estimation methods are compared with the equivalent linearisation approach. The
52 higher mode effects of multistorey substructures are then investigated using higher mode amplification
53
factors for two-storey and six-storey analysis models and the results are verified using nonlinear response
54
55 history analyses.
56
57
58 2 Peak Substructure response
59
60 This section introduces the equivalent static load procedure and discusses two R-factor approaches
(Section 2.1) and an equivalent linearisation approach (Section 2.2) to estimate the peak substructure
acceleration of a yielding substructure.
Prepared using sagej.cls
Deepshikha Nair et al. 3
1
2 AH1=FH1 × sAHeq1
c cAH2=FH2 × sAHeq2
3
4 Equivalent
5 AV1=FV1 × sAHeq1 AV2=FV2 × sAHeq2
static loads c
c
6
7
8
9
10 SAHeq1 S AHeq2
11 MR
12
13 MS3
14 = +
15 MS2
16
17 MS1
18
19
20 Substructure T , Meq1 T ,
s 2 s Meq2
s 1 s
21 model
Fo
22
23 Figure 1. Equivalent static load procedure (Nair et al. [16])
24
rP
25
26 The peak substructure roof acceleration (AHeq ) is the seismic input for the dome and is the first
27 step in the proposed equivalent static load procedure for linear double-layered domes with elastic
28 substructures [15, 16]. The substructure is modelled as a simple lumped-mass model as shown in Figure
ee
29 1 with the roof modelled as a rigid mass (MR ) and the peak substructure acceleration is computed
30
31
using response-spectrum analysis. The substructure acceleration for each participating mode is then
rR
32 multiplied by corresponding horizontal and vertical amplification factors to obtain the peak acceleration
33 distributions of the dome using an assumed mode shape (usually the dominant mode O1) [7, 16]. The
34 amplification factors quantify the roof-substructure interaction defined by the substructure to roof period
35
ev
and mass ratios [7,13,16]. These roof horizontal and vertical accelerations (Figure 1) are then multiplied
36 by the nodal masses to obtain the equivalent inertial static forces.
37
38
iew
2.1 Inelastic accelerations: RSA & R-factor

39
40 The total force reduction factor (or the response modification factor R) for idealised bilinear systems
41 is derived as follows (Uang [28]) in Equation 1 and Figure 2(a), where Rµ is the ductility reduction
42 factor and Ω is the structural overstrength factor. Rµ accounts for the reduction in the elastic design
43 force because of the energy dissipation capacity and the reserve strength that exists between the actual
44
45
structural yield level and the first yield level is defined as Ω [28]. Rµ is therefore defined as the ratio of
46 the elastic strength demand to the inelastic strength demand (Equation 1 [28]) and is used to derive the
47 inelastic seismic acceleration spectrum.
48
49 Fy (µ = 1)
R = Rµ Ω ; Rµ = (1)
50 Fy (µ = µi )
51
52 2.1.1 Newmark-Hall methodology: The Veletsos-Newmark-Hall methodology [30] (labelled and
53 referred to as the ‘Newmark’ method hereon) is one of the earliest and commonly used methods to
54 derive the inelastic seismic design spectra (Figure 2(b)), where the 5 per cent damped spectrum for the
55
elastoplastic system is obtained by drawing a curve similar to that of the elastic system but displaced
56
57 downward by the reduction factor. The reduction factor proposed depends on the position of the period
58 on the spectrum.
59
60 • For constant-acceleration region of the spectrum where the equal energy rule applies [30], where
the area under the load-deformation diagram up to the maximum deformation is the same as that
of an elasto-perfectly plastic system :
1
2
3
4
5
6
7
8 R Rµ
9
10
11 Ω
12
13
14
15
16
17 (a) Definition of Rµ (b) Construction of inelastic spectra [29]
18
19 Figure 2. (a) Idealisation of structural response & R factors, Picture taken from Uang. [28],
20 (b) Construction of design spectra, Picture taken from Riddell et al. [29]
21
Fo
22
23 p
24 Rµ = 2µ − 1 (2)
rP
25 • For constant-velocity region of the spectrum where the equal displacement rule applies [30]:
26
27 Rµ = µ (3)
28
ee
29 2.1.2 Nassar & Krawinkler : To estimate the inelastic strength demands, strength reduction factors
30 Rµ were derived as a function of the target ductility ratio µ and the period T , using a least square fit
31
rR
regression analysis for a single degree of freedom system with post-yield stiffness ratios (p) of 0, 2 and
32
33 10 % subjected to 39000 time history analyses for T= 0.1-4.0 s, µ=1-8 for a damping of 5% [31] given
34 in Equation 4 (labelled and referred to as the ‘N&K’ method hereon)
35
ev
36
Ta b
Rµ = [c(µ − 1) + 1]1/c ; c= a
+ (4)
37 T +1 T
38
iew
39
where a = 1 for p <=2%, a = 0.8 for p =10% and b = 0.42, 0.37 and 0.29 for p =0, 2 and 10%
40 respectively.
41 2.1.3 Lee & Han: Lee & Han [32] further studied the reduction factors for different hysteretic models
42
43
and proposed Rµ as a function of the period and ductility using a two-stage regression analysis. The
44 equations for elasto-perfectly plastic models are given in Equations 5-6.
45
46 Rµ = A0 {1 − exp(−B0 × T )} (5)
47
48
49 A0 = 0.99 × µ + 0.15; B0 = 23.69 × µ−0.83 (6)
50 For comparison, the reduction factors obtained from the three methods are plotted below considering
51
52
a fundamental period in the short-period range as 0.4s (period of the benchmark model described in
53 Section 3 and elastoplastic behaviour.
54 While these ductility reduction factors were proposed to develop the inelastic response spectra to
55 estimate the design base shear for elastoplastic models, in this study, these ductility reduction factors
56 are used to estimate the approximate peak spectal acceleration of the substructure as a function of the
57 obtained period and the target roof displacement ductility. The acceleration reduction ratio Ra is defined
58
59
as the ratio of the peak inelastic acceleration to the peak elastic acceleration which can be computed
60 using the inverse of the ductility reduction ratio [30] as shown in Equation 7. For this study, from
hereon, the two most conservative methods -Newmark and N&K methods (Figure 3) are considered and
discussed for the estimation of peak substructure acceleration.
1
2
3 15
4 N&K
5 Newmark
6 10
Lee&Han,
R =
7
8
9
5
10
11
12
0
13 2 4 6 8 10 12 14
14
15
16
17 Figure 3. T =0.4s, Reduction factor as a function of target ductility
18
19
20 Aeq1 1
21 Ra = ; Ra = (7)
Fo
22 A1 Rµ
23
2.2 Inelastic accelerations: RSA & Equivalent linearisation
24
rP
25 An alternative simplified theory for estimation of the peak inelastic response of any bilinear system is
26 the equivalent linearisation method proposed by Kasai et al. [20, 21]. It uses mathematical expressions
27 as a function of the dynamic properties of the system to consider the effect of period shift as well as
28
ee
the effect of damping by evaluating the dissipated energy. The multistorey building is idealised as an
29
30 SDOF (single-degree of freedom) model and the target ductility is iterated to obtain a more accurate
31 ductility value which is a function of the achieved damping and the secant stiffness ratio which in turn
rR
32 is a function of the ductility. The proposed reduction in displacement Rd and acceleration (or force) Ra
33 depends on the stiffness of the frame, braces and the energy dissipation devices.
34 The equivalent stiffness (Keq ) and equivalent damping ratio (heq ) of the system are obtained using
35
ev
the equivalent linearisation procedure. For each of the participating modes (where i = mode number), a
36
37 starting value of ductility ratio (target ductility) µti is assumed and heqi and the equivalent stiffness ratio
38 Keqi /K1i are iterated using Equations 8-11 until µi converges, where K1i is the initial stiffness, Dh is
iew
39 the reduction factor to adjust the damping ratio from the base damping ratio hoi (5% in this study) for
40 periods less than 2s, and the subscript (j) is the jth step of the iteration [18,20]. The equivalent (secant)
41 period Teqi is then computed using Equation 12 [20].
42
43 Keqi /K1i = 1/µi + (1 − 1/µi )K2i /K1i (8)
44
45 2(K1i /K2i ) K1i /K2i + µi − 1
46 heqi = hoi + ln (9)
47 πµi (K1i /K2i )µi K2i /K1i
48 q
49 Dhi = (1 + 25hoi )/(1 + 25heqi ) (10)
50 q
51 µi(j) = µti(1) Dhi(j−1) / Keqi(j−1) /K1i (11)
52
53 q
54 Teqi = Ti K1i /Keqi (12)
55
56 For each of the participating modes, the reduction in displacement Rd and acceleration Ra can then be
57 computed as a function of the periods Teqi and Ti using Equations 13-16 [20] depending on the regions
58 of the spectrum the periods lie. This method is labelled and referred to as the ‘Kasai’ method hereon.
59 1. For Ti and Teqi in the constant acceleration region:
60
Teqi Teqi + Ti
Rdi = Dhi ( ) (13)
Ti 2Ti
1
2 2. For Ti in the constant acceleration region and Teqi in the constant velocity region, where Tc is the
3 corner period (in this study, Tc =0.52s as described later in Section 3.1)):
4 " #
5 Teqi Tc (Tc /Ti − 1)2
6 Rdi = Dhi − (14)
7
Ti Ti 2(Teqi /Ti − 1)
8
9 3. For Ti and Teqi in the constant velocity region:
10
11
Teqi
Rdi = Dhi (15)
12 Ti
13
14
The reduction in acceleration may then be computed using Equation 16 [20, 21].
15 2
16 Ti
Rai = Rdi (16)
17 Teqi
18
19 3 Analysis models: Single-storey substructures
20
21 First, the peak substructure response of simple single-storey analysis models are discussed in this section
Fo
22 to compare the response from each of the response estimation methods.

23
24 3.1 Design Acceleration spectrum
rP
25
A site with high seismic hazard from the southern California region was chosen to develop a 5% damped
26
27 DBE (design basis earthquake) acceleration spectrum (Figure 6) using the ASCE 7-16 procedure [1])
28 with SDS = 1.4g (short period spectral acceleration), SD1 = 0.73g (spectral response acceleration at a
ee
29 period of 1s) and TL = 8s (long-period transition period obtained from Figure 22-14 of ASCE 7-16 [1]).
30
31 3.2 BRBF substructure model
rR
32
33
The 60m spanned substructure has one storey 5m high and the roof dead load was assumed to be 2kPa
34 (considering a structural weight of about 1kg/m2 for every meter of span [33], a 15% allowance for the
35 connections, and 1.3kPa nonstructural dead load). It was assumed that all the lateral forces are resisted
ev
36 by the buckling-restrained braces (BRBs [34]) placed along the perimeter in a diagonal configuration,
37 and so in the 3-d ETABS [35] analysis substructure model shown in Figure 4(a), all the beams and
38
iew
columns were modelled using elastic beam and column elements with the section sizes listed in Table 2.
39
The roof is modelled as a rigid mass in the substructure model. The BRBs for the substructure (Figure
40
41 4) were designed for the vertically distributed forces obtained as per ASCE 7-16 [1], where hx is the
42 height from base to level x and k is taken as 2 (Equation 17), Sa is the design acceleration obtained
43 for the code-defined approximate fundamental period Ta = Ct × hxn (Ct = 0.073 and x = 0.75 for steel
44 BRBFs as per Table 12.8.1 of ASCE 7-16 [1]), R was taken as 8 (as recommended for BRBFs as per
45 Table 12.2.1 of ASCE 7-16 [1]), Ie was assumed as 1, giving a base shear ratio Cb = 0.2, Vx is the storey
46
shear, BRB Vx is the storey shear resisted by the BRBs (in this study 100%) and BRB Vix is the storey
47
48 shear resisted by each BRB as given in Table 1.
49
50 Sa W wx hx k
V = ; Fx = Cvx V ; Cvx = P (17)
51 R/Ie n
52 wi hi k
i=1
53
54 Three substructure models were designed to include substructures with different BRB design yield
55 displacements. Model-B (benchmark model) as shown in Figure 4(a) was designed with a standard
56 BRB yield drift ratio of 1/750 [34]. A second model (Model-A) was designed for the same design base
57 shear with a shorter BRB yield displacement of 0.06% such that the fundamental period is shorter and
58
59
lies on the constant-acceleration region of the spectrum (Figure 6) and a third model (Model-C) was
60 designed with a longer BRB yield displacement of 0.2% with the fundamental period in the constant-
velocity region as given in Table 1. The axial force-displacement of BRBs were determined following
the design guidelines by Takeuchi and Wada [34] and the BRB parameters are given in Table 1. In the
1
2 analysis FE model, the BRBs have been modelled as links [35,36] with bilinear hysteretic characteristics
3 and a post-yield stiffness ratio (p) of 2% typical of BRBFs [34]. To investigate the effect of post-
4 yield stiffness, additional substructure models with p=10% and 25% were created keeping the yield
5 displacements the same across all models of the same series. The models with higher post-yield stiffness
6
7
like the p=25% models represent substructures with moment-frames that contribute to the lateral force
8 resistance and remain elastic after the BRBs yield. These are typically seen in structures in Japan [34]
9 which adopt rigid moment connections in the supporting frames resulting in higher post-yield stiffness
10 ratios. Additional models with p = 0% were also constructed to validate and confirm the approaches
11 proposed for elastoplastic SDOF systems in Section 2.1.
12
13
14 3.3 Roof model
15 The double-layered dome was designed for a span of 60m and a half subtended angle (θ) of 30°. The
16 lattice member sizes and vertical offsets (d) between the section centerlines are listed in Table 3. For
17
18
simplicity, the double-layer lattice (Figure 4 (b)) was modelled using equivalent beams with out-of-
19 plane stiffness modification factors [16] with moment connections and pinning the roof lattice perimeter
20 nodes. This dome model was used to identify the roof modes and is denoted as the ‘roof model’. The
21 fundamental period of the roof (TR ) was found to be 0.22s with the roof exhibiting the anti-symmetric
Fo
22 O1 mode shape [7, 13].

23
The combined analysis models were then constructed combining the roof model and substructure
24
model. The design of roof members were checked by performing NLRHA analysis on the combined
rP
25
26 model using the load combination of 1.2DL+EQ (as per Section 2.3.6 of ASCE-7-16 [1]) and ground
27 motions specified in Section 3.4 to check the member forces against their design strength as per the
28 guidelines in AISC 360-16 [37]. The roof geometry was modelled in Grasshopper [38] and imported to
ee
29 ETABS [35] for analysis. The roof member beams in ETABS were modelled as ‘frame’ objects which
30
are general three-dimensional beam elements [36].
31
rR
32
Table 1. Equivalent Lateral Forces: Proportioning BRBs using R = 8
33
34 Model-A (T1 = 0.3s)
35
ev
h (m) kN kN kN kN mm kN/m (m) MPa mm2

Storey
36 Wx Vx BRB Vx BRB Pi δy BRB Keqi Lt σy Ac
37 RFL 5 6064 1061 1061 54 1.85 29160 6.4 90 600
38
iew
Model-B (T1 = 0.4s)

39
40 Storey
Wx Vx BRB Vx BRB Pi δy BRB Keqi Lt σy Ac
41
RFL 5 6064 1061 1061 54 4.12 13125 6.4 235 230
42
43 Model-C (T1 = 0.5s)
44 Storey
45 Wx Vx BRB Vx BRB Pi δy BRB Keqi Lt σy Ac
46 RFL 5 6064 1061 1061 54 6.17 8748 6.4 235 230
47 where Pi =Yield axial force, δy =Yield axial deformation,Keqi =Equivalent axial stiffness,
48 Lt =total BRB length and Ac =Area of core in plastic zone [34]
49
50
51 Table 2. L60 : Substructure model data
52 (a) Storey heights and seismic weights (b) Section sizes (σy =235MPa)
53 Storey Height (m) Weight (kN) Member Section Shape Section Size (mm)
54 RFL 5 6064 Column SHS 450×450×25
55
56
57
Table 3. Roof models: Member section sizes (σy =235MPa)
58
59 Model Dead Load (DL) Double-layer
offset Single-layer beam m Tension beam
60 (kPa) beam (mm)d (cm) (roof member) (mm) (mm)
L60 2 2×ϕ165.2 t7.1∗
150 ϕ307.5 t7.5∗ 50.1 ϕ816 t16
∗
ϕ =Outer Diameter, t =Thickness of circular hollow section, m = Out-of-plane stiffness modification factor
1
2 The first two modes contributing to 99% of the mass participation of the combined model (Model-B)
3 are shown in Figure 5. The first mode is the substructure T1 mode oscillating in-phase with the roof’s
4 dominant O1 mode [7]. The second mode exhibits the two modes oscillating out-of-phase. The effects of
5 substructure stiffness and period ratios on the dominant modes have been discussed in detail in previous
6
7
studies [7, 16].
8 Roof as rigid mass View: Perpective View: Plan
9 Seismic Input
10
11
MRFL
12
A A’
13
K1 ,ho A O A’
14
15
16 SDOF model
17
Tension roof
18
beams member beams
19
20 BRB link Force-deformation (kN-mm)
BRB as link elements L=60m
21 100 View: Perpective
Fo
80 p=2%
22 60
23 40
20
24 0
-25 -20 -15 -10 -5 0 5 10 15 20 25
Rigid
rP
25 -40
connections
-60
26 -80
Model-B
δy =0.13%
27 -100
Pinned base
28
ee
(a) Substructure model (b) Roof model

29
30 Figure 4. Single-storey substructure model and roof model
31
rR
32
33
34
35
ev
36
37
38
iew
39
40
41
42 Mode-1: O1+T1 (0.47s) Mode-2: O1-T1 (0.22s)
43
44 Figure 5. Model-B: Combined model mode shapes
45
46
47
48 3.4 Input ground motions
49 11 natural ground motions were selected from the Pacific Earthquake Engineering Center strong motion
50 database [39]. An initial suite was obtained using a standard search criteria of minimum magnitude of
51
6.5, a maximum magnitude of 7.5, and maximum shear velocity of 400m/s. 11 of the selected ground
52
53 motions (Horizontal-1 direction) listed in Table 4 were then spectrally matched to the design spectrum
54 (in accordance with Section 16.2 of ASCE 7-16 [1]) such that the average of the spectra for the suite
55 equals or exceeds 110% of the 5% target design spectrum over the period range of 0.2T1-1.5T1 where
56 T1 was considered as 0.4s (the fundamental period of benchmark Model-B). The response spectra of
57 the matched ground motions are shown in Figure 6.
58
59
For NLRHA performed on the combined models using these spectrally matched waves, Rayleigh
60 damping of 5% was assigned to the first and second mode and the analysis was performed using the
integration approach using the Hilber-Hughes-Taylor (HHT) method [35, 36]. Two additional levels of
ground motion intensity were considered to investigate the response at maximum-considered earthquake
1
2 (MCE) level (1.5 × DBE level as prescribed by the ASCE-7 [1]) and the serviceability level (SLE = 0.2
3 × the DBE level). As ASCE does not have a defined SLE design level, for this study, the SLE spectrum
4 spectrum was defined by retaining the shape of the DBE spectra as is common practice in countries like
5 Japan and New Zealand, and then scaling it by 0.2 (following the average ratio between the SLE and
6
7
DBE design spectra of typical seismic codes as reviewed by Chandler et. al. [40]).
8 Table 4. Input ground motions
9
10 Ground motion no. Earthquake* Year Station name M Rjb (km) PGA(g)
11 1 Trinidad 1980 Rio Dell Overpass E-Ground 7.20 76.06 0.16
12 2 Spitak Armenia 1988 Gukasian 6.77 23.99 0.20
3 Loma Prieta 1989 Emeryville; Pacific Park 6.93 76.87 0.25
13
4 Loma Prieta 1989 Fremont - Emerson Court 6.93 39.66 0.19
14 5 Loma Prieta 1989 Hayward - BART Sta 6.93 54.01 0.16
15 6 Loma Prieta 1989 Sunnyvale - Colton Ave. 6.93 23.92 0.21
16 7 Northridge-01 1994 LA - 116th St School 6.69 36.39 0.21
17 8 Northridge-01 1994 LA - Baldwin Hills 6.69 23.50 0.24
18 9 Northridge-01 1994 LA - Cypress Ave 6.69 28.98 0.22
19 10 Northridge-01 1994 LA - N Faring Rd 6.69 12.42 0.28
11 Northridge-01 1994 LA - Obregon Park 6.69 35.43 0.36
20
Rjb =closest horizontal distance between the vertical projection of the rupture plane and recording station
21
Fo
M= moment magnitude, PGA= unscaled peak ground acceleration

22 *These records were obtained from the NGA-West2 online ground motion database tool
23
24
rP
25
26 Roof O1
Model-A Model-B
27
Spectral Acceleration (cm/s2 )
2000 Model-C
28 GM1 GM9
ee
29 GM2 GM10
1500 GM3 GM11
30 GM4 Mean
31 1000
GM5 Target DBE (5%)
rR
GM6 SLE (5%)

32 GM7 MCE (5%)
33 500
GM8
34
35
ev
0
36 10 -2 10 -1 10 0 10 1
37 Period (s)
38
iew
Figure 6. Target design spectrum and earthquake levels

39
40
41
3.5 Comparison of peak substructure response
42
43 For each of the three substructure models, the elastic design spectral acceleration (Sa1 ) was used
44 to obtain the elastic base shear (V1 ) and then extrapolate the maximum elastic substructure roof
45 displacement (D1 ) at the substructure roof level (point A of Figure 4(a)) from the initial stiffness (K1 )
46
as shown in Equation 18. The maximum roof level acceleration (point A of Figure 4(a)) A1 of the
47
48 substructure is obtained from the corresponding roof displacement and the fundamental period and
49 the response spectrum (Sa1 ) as shown in Equation 18. The reduction factors (Rµ and Ra ) were then
50 computed following Equations 2-4 from the target ductility µt using Equation 19 and Equation 7. For
51 the equivalent linearisation approach, Equations 8-16 were used. The mean response obtained from the
52 NLRHA performed on the combined models were also used to obtain the mean displacement ductility
53
and reduction factor using Equations 20-21.
54
55 V1 = M1 Sa1 D1 = V1 /K1 A 1 = D1 ω 2 (18)
56
57 µt = D1 /Dy (19)
58
59 µN LRHA = Dmean /Dy (20)
60 Rµ−N LRHA = V1 /Vmean (21)
1
Post-yield stiffness ratio (p) increasing
2
p=0%, Kasai p=2%, Kasai p=10%, Kasai p=25%, Kasai
3 p=0%, Newmark p=2%, Newmark p=10%, Newmark p=25%, Newmark
4 p=0%, N&K p=2%, N&K p=10%, N&K p=25%, N&K
p=0%, NLRHA p=2%, NLRHA p=10%, NLRHA p=25%, NLRHA
5 20 20 20 20
Reduction factor Rµ
6
15 Rμ =μ
Reduction factor R
Reduction factor R
Reduction factor R
15 15 15
7 MCE
8 10 10 Rμ =√2μ-1 10 10
9 5 5 DBE 5 5
SLE
10
11 5 10 15 20 5 10 15 20 5 10 15 20 5 10 15 20
NLRHA
12 1 1 NLRHA 1 NLRHA 1 NLRHA
Acceleration reduction Ra
13
Reduction factor Ra
Reduction factor Ra
Reduction factor Ra
0.8 0.8 0.8 0.8
14 0.6 0.6 0.6 0.6
15 0.4 0.4 0.4 0.4
16 0.2 0.2 0.2 0.2
17 0
5 10 15 20
0
5 10 15 20
0
5 10 15 20
0
5 10 15 20
18 NLRHA NLRHA NLRHA NLRHA
19 (a) p=0% (b) p=2% (c) p=10% (d) p=25%

20
21 Figure 7. Substructure Model-B: Reduction factors as functions of achieved displacement ductility
Fo
22
23
24 To compare the peak substructure responses from all the methods against the NLRHA results, the
reduction in acceleration Ra = Aeq /A1 and Rµ is plotted against the displacement ductility µN LRHA
rP
25
26 for Model-B in Figure 7. It was observed that for models with low post-yield stiffness, p < 10%,
27 all the estimation methods provide good accuracy in estimating the acceleration and the acceleration
28
ee
reduction values are nearly equal to those obtained from the NLRHA. When the structure is subjected
29 to SLE level earthquakes, the obtained ductility values are close to 2 with small reduction in response.
30
31
When the intensity of earthquake is increased to DBE level, the mean ductility values increase to a
rR
32 range of 8-14 depending on the post-yield stiffness. For MCE levels, the values increase further to
33 15-20 but the reduction in response (Ra ) is not linearly proportional to the increase in ductility. The
34 results in Figure 7 exhibit a nonlinear relationship between Rµ and µ that is close to the Rµ and µ
35
ev
curve proposed by Newmark for short-period structures in Figure 3. Newmark’s method and N&K’s
36 methods provide good accuracy with Newmark’s method being more conservative than the NLRHA
37
38
response when the post-yield stiffness is low (p <= 2%). This suggests that for short period structures
iew
39 with low post-yield stiffness ratio (p <= 2%), the actual ductility reduction factors (from NLRHA) lie
40 between the Newmark’s factors obtained from the equal energy (Equation 2) and the equal displacement
41 rule (Equation 3) as can be seen from the dotted lines in Figures 7(a) and (b). As the post-yield
42 stiffness increases to p = 25% (Figure 7(d)), the errors in the response become larger. This is because
43 the equations by Newmark have been proposed considering elastoplastic systems. As the post-yield
44
45
stiffness increased, the reduction ratio Ra increased but the R-factor methods estimated lower peak
46 accelerations assuming an elastoplastic response. However, as the Newmark method is conservative for
47 short-period structures, the errors in response estimation was less than 20% for models with p <= 10%
48 (Figure 7(a),(b) and (c)). Among the estimation methods, Kasai’s approach (Table 5) provides the most
49 accurate estimate of the response to that obtained from NLRHA for all cases as this approach is more
50 mathematical and less empirical making it applicable to any bilinear system. The actual ductility values
51
and damping ratios are estimated leading to better accuracies in response estimation.
52
53 Table 5. Substructure models: Kasai method’s equivalent linearisation results, p=2%
54
55 ho1 heq1 Keq1 /K1 Dh Rd Ra µt1 T1 Teq1 TR
56 Model-B-SLE 0.05 0.12 0.58 0.75 1.11 0.64 1.77 0.45 0.59 0.23
57 Model-B-DBE 0.05 0.44 0.10 0.43 1.55 0.16 8.87 0.45 1.41 0.23
58 Model-B-MCE 0.05 0.47 0.07 0.42 1.90 0.12 13.31 0.45 1.77 0.23
59
60
1
2 4 Peak Roof Response
3
4
4.1 Substructure T1 mode amplification factors
5 The substructure-roof interactions for single-storey structures are defined by the amplification factors
6 FH1 and FV 1 [7] that depend on the mass (RM ) and period (RT ) ratios [7] given in Equation 22.
7
These are characterised as ratios of the effective modal mass (Meq ) and fundamental period (T1 ) of
8
9 the substructure model, which includes the roof mass, relative to the total roof mass (MR ) and period
10 of the (dominant) fundamental roof O1 mode (TR ) [7, 16]. The mass and period ratios increase as the
11 substructure gets heavier (such as in the case of an RC substructure) or as the substructure stiffness
12 increases [16]. Heavy substructures with RT 1 < 1.5 and RM > 2 produce higher amplification between
13 the T1 and O1 modes due to resonance, as noted by Takeuchi et al. [7]. To account for this resonance
14 ′ ′
induced amplification, the amplification factors have been modified to FH and FV (when RT 1 < 1.5 and
15
16 RM 1 > 2), as given by Equations 25-26. [7].
17 The proposed amplification factors FH1 and FV 1 are defined by Equations 23-26 [7]. A vertical
18 calibration factor Cv θ (Equation 24) of 1.85θ was adopted as previously proposed based on a numerical
19 study investigating the influence of the half-subtended angle on the peak vertical acceleration [7]. After
20 obtaining the peak substructure response s AHeq , the peak roof accelerations of the combined model c AH
21
Fo
and c AV are estimated using the amplification factors and the dominant roof O1 mode shape adopted for
22
23 the envelope, which has the horizontal and vertical distributions given by Equations 27-28 [7], where x
24 and y are the coordinates of roof nodes, the roof center is located at {x,y} = {0,0} and L is the span of
rP
25 the dome. Note that in this paper, the preceding subscripts r, s and c refer to the roof, substructure and
26 combined models, respectively.
27
s Meq s T1
28 RM = , RT 1 = (22)
ee
29 r MR r TR
30
31
 
p 3 (0 < RT 1 ≤ 5/36) 
rR
32
33
FH1 = 5/(4RT 1 ) (5/36 < RT 1 ≤ 5/4) (23)
34

1 (5/4 < RT 1 ) 
35
ev
 
36  p 3C V θ (0 < RT 1 ≤ 5/16)
37 FV 1 = ( 5/RT 1 − 1)C V θ (5/16 < RT 1 ≤ 5) (24)
38
iew

0 (5 < RT 1 ) 
39
40 s
41 ′ 1
42 FH1 = FH1 2 + 2 2
(25)
43
(1 − RT 1 ) + (1/RM 1 )θ
44 s
45 ′ 1
46 FV 1 = FV 1 2 + 2 2
(26)
47 (1 − RT 1 ) + (1/RM 1 )
48
49
p
x2 + y 2
π
50 c AHi (x, y) = s AHeqi {1 + (FHi − 1) cos } (27)
51 L
52 p
53 x 2π x2 + y 2
c AV i (x, y) = s AV eqi FV i p sin (28)
54 x2 + y 2 L
55
56 In previous studies [13, 18], for the Kasai’s equivalent linearisation approach, the period ratios were
57 computed using the substructure’s inelastic periods RT eq1 [13] and the rationale behind this assumption
58
59
has not been confirmed yet. Therefore, to investigate the relationship between the amplification factors
60 and the inelastic period ratios, the vertical amplification factors computed from the mean NLRHA
response divided by the peak substructure acceleration FV = c AV max /s AHeq have been plotted against
the elastic (RT 1 ) and inelastic (RT eq1 ) period ratios (Equation 29).
1
2 3.5 3.5
3 p=25%
4 3 3 p=10%
p=2%
5 2.5 2.5
6
2 2
7
8 1.5 1.5
9
1 1
10
11 0.5 0.5
12 0 0
13 0 2 4 6 8 0 2 4 6 8 10 12
14
15 (a) Period ratio calculated using T1 (b) Period ratio calculated using Teq1
16
17 Figure 8. Vertical amplification factors w.r.t period ratios
18
19
20
3.5 3.5
21 p=25%
Fo
22 3 3 p=10%
23 p=2%
2.5 2.5
24
2
rP
25 2
26 1.5 1.5
27
28 1 1
ee
29 0.5 0.5
30
0 0
31 0 1 2 3 4 5 6 0 5 10 15
rR
32
33
(a) Period ratio calculated using T1 (b) Period ratio calculated using Teq1
34
35
ev
Figure 9. Horizontal amplification factors w.r.t period ratios

36
37
38
iew
s T1 s Teq1
39 RT 1 = ; RT eq1 = (29)
40 r TR r TR
41 It was observed that the obtained amplification factors plotted against the elastic period ratios (Figures
42 8(a) and 9(a)) are in better agreement as opposed to the plots against the inelastic period ratios (Figures
43 8(b) and 9(b)). The largest errors between the proposed curves and the obtained factors were observed
44
45
for the p = 2% and p = 10% models as can be seen in Figure 8(b). As the post-yield stiffness increases,
46 the more conservative the proposed factors becomes. The horizontal amplification factors have also
47 been plotted for comparison as FH = c AHmax /s AHeq . In case of horizontal amplification, the errors in
48 the calculation remain the same as the horizontal amplification factor becomes constant at 1 (FH1 = 1)
49 for all period ratios larger than 1.25 (RT 1 > 1.25). As a result, the proposed horizontal amplification
50 factor is constant at 1 (FH = 1) for both the cases. Therefore, in this study, the more accurate elastic
51
52
period ratios (Equation 22) are adopted for computation of the amplification factors.
53
4.2 Direct-R approach
54
55 The ‘Modal response spectrum analysis’ method in ASCE-7-16 [1] recommends calculation of design
56 level force-related parameters by dividing the elastic response directly by R/Ie (=8 in this study).
57 Therefore, another straightforward method of estimating DBE-level response is to divide the peak
58
59
elastic roof response directly by the R-factor. Employing a single reduction factor for all modes to
60 estimate the peak accelerations may underestimate the overall response as the actual reduction in
acceleration depends on the achieved ductility and additional modal damping which may not correspond
to the R-factor as assumed in this approach. Furthermore, the applicability of this method to models
1
2 with fundamental periods in the constant-acceleration region (or short-period range) where the ‘equal-
3 displacement rule’ is not valid [30] has not been proved yet. Nevertheless, this method is also used
4 for computing DBE level response using a reduction factor of 8 for comparison and is labelled as the
5 ‘Direct-R’ method hereon.
6
7
4.3 Comparison of peak roof response
8
9 The peak roof accelerations for the Model-A, Model-B and Model-C were calculated using the first
10 mode amplification factors in Equations 23-28. For the equivalent linearisation approach, s AHeqi was
11 computed considering the effect of both the period shift and the damping [13] using Equations 8-16.
12 For the R-factor approach methods, the peak substructure acceleration was obtained using Equation 7.
13
14
The proposed ridgeline A-O-A’ (Figure 4) accelerations are compared with the mean NLRHA results in
15 Figures 10-11.
16 When the post-yield stiffness ratio (p) is low (Figure 10), the peak substructure accelerations for
17 Model-C in the constant velocity region obtained from the Newmark and N&K method were nearly
18 equal to the peak accelerations obtained from Kasai’s method leading to similar accuracy in the overall
19 roof response. Model-A exhibited the largest amplifications in both the horizontal and vertical directions
20
21
as the substructure periods were closest to the roof’s period (smallest period ratios RT ) among the
Fo
22 three models. In small ductility regions (Figure 10 SLE results), all methods give reasonably accurate
23 results. In large ductility regimes like the DBE and MCE levels, the Newmark method presents the most
24 conservative response estimates for models in the constant acceleration region (Model-A and Model-B)
rP
25 while the Kasai method is the most conservative for structures in the constant velocity regions (Model-
26 C). For structures with fundamental periods in the constant velocity region (Model-C), the simpler
27
28
Direct-R factor and Krawlinkler’s approach also estimate the DBE response with reasonable accuracy.
ee
29 When the post-yield stiffness ratio (p) is large (Figure 11), the peak substructure accelerations for
30 all models were underestimated by the Newmark and the N&K method (Figure 7 in Section 3.5). For
31 the DBE and MCE level response, the overall roof response is severely underestimated by the R-factor
rR
32 methods. These errors increase as the level of the intensity increases from SLE to MCE. Nevertheless,
33 all the methods give reasonably accurate results in small ductility regions (Figure 11 SLE results) as
34
the structure is largely in the elastic range. As observed for the low post-yield ratio models, Model-
35
ev
36 A exhibited the largest amplifications in both the horizontal and vertical directions. Kasai’s relatively
37 rigorous equivalent linearisation method accurately estimates the substructure response (Figure 7) and
38 therefore presents a conservative roof response envelope (Figure 11) for models in both the constant
iew
39 acceleration and velocity regions. It is therefore recommended to use the equivalent linearisation
40 approach when p is higher than 2%.
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
1
2 SLE DBE MCE
3
(a) Model-A T1 =0.30s
4
Horizontal Acceleration (m/s2 )

5 8
Kasai N&K
8
Kasai Direct-R
8
Kasai N&K
6 6
Newmark Mean NLRHA
6
Newmark
N&K
Mean NLRHA
6
Newmark Mean NLRHA
7
8 4 4 4
9 2 2 2
10
11 0
-40 -20 0 20 40
0
-40 -20 0 20 40
0
-40 -20 0 20 40
12 Ridge coordinates (m) Ridge coordinates (m) Ridge coordinates (m)
13
Vertical Acceleration (m/s2 )

6 6 6
14
15
4 4 4
16
17
18 2 2 2
19
20 0
-40 -20 0 20 40
0
-40 -20 0 20 40
0
-40 -20 0 20 40
Fo
22 (b) Model-B T1 =0.45s

23

8 8 8
24 Kasai N&K Kasai Direct-R Kasai N&K
rP
25 6
Newmark Mean NLRHA
6
Newmark
N&K
Mean NLRHA
6
Newmark Mean NLRHA
26
27 4 4 4
28
ee
2 2 2
29
30 0
-40 -20 0 20 40
0
-40 -20 0 20 40
0
-40 -20 0 20 40
rR
32
6 6 6
33
34 4 4 4
35
ev
36
2 2 2
37
38
iew
39
0 0 0
-40 -20 0 20 40 -40 -20 0 20 40 -40 -20 0 20 40
41 (c) Model-C T1 =0.55s
42
8 8 8
44
Newmark Mean NLRHA Newmark Mean NLRHA Newmark Mean NLRHA
6 6 N&K 6
45
4 4 4
46
47 2 2 2
48
49 0
-40 -20 0 20 40
0
-40 -20 0 20 40
0
-40 -20 0 20 40
51
6 6 6
52
53 4 4 4
54
55 2 2 2
56
57
0 0 0
58 -40 -20 0 20 40 -40 -20 0 20 40 -40 -20 0 20 40
60
Figure 10. Roof ridgeline accelerations (post-yield stiffness ratio p = 2%)
1
2 SLE DBE MCE
3
(a) Model-A T1 =0.30s
4

5 10
Kasai N&K
10
Kasai Direct-R
10
Kasai N&K
6 8 Newmark Mean NLRHA 8 Newmark
N&K
Mean NLRHA 8 Newmark Mean NLRHA
7 6 6 6
8 4 4 4
9
10 2 2 2
11 0
-40 -20 0 20 40
0
-40 -20 0 20 40
0
-40 -20 0 20 40
13

8 8 8
14
15 6 6 6
16
17
4 4 4
18 2 2 2
19
20 0
-40 -20 0 20 40
0
-40 -20 0 20 40
0
-40 -20 0 20 40
Fo
22 (b) Model-B T1 =0.45s

23

10 10 10
rP
25 8 Newmark Mean NLRHA 8 Newmark

N&K
Mean NLRHA 8 Newmark Mean NLRHA
26 6 6 6
27 4 4 4
28
ee
2 2 2
29
30 0
-40 -20 0 20 40
0
-40 -20 0 20 40
0
-40 -20 0 20 40
rR
32
8 8 8
33
34 6 6 6
35
ev
4 4 4
36
37 2 2 2
38
iew
39
0 0 0
-40 -20 0 20 40 -40 -20 0 20 40 -40 -20 0 20 40
41 (c) Model-C T1 =0.55s
42
10 10 10
44 8 Newmark Mean NLRHA 8 Newmark Mean NLRHA 8 Newmark Mean NLRHA

N&K
45 6 6 6
46 4 4 4
47 2 2 2
48
49 0
-40 -20 0 20 40
0
-40 -20 0 20 40
0
-40 -20 0 20 40
51
8 8 8
52
53 6 6 6
54 4 4 4
55
56 2 2 2
57
0 0 0
58 -40 -20 0 20 40 -40 -20 0 20 40 -40 -20 0 20 40
60
Figure 11. Roof ridgeline accelerations (post-yield stiffness ratio p = 10%)
1
2 5 Effect of higher substructure modes
3
5.1 2-storey analysis models
4
5 To investigate the applicability of the approach of using R-factors for substructures with higher modes,
6 the single-storey substructure was replaced with a two-storey substructure. As in the single-storey
7 models, the first storey and the roof level are each 5m high and the first storey has a floor weight of
8
7kPa and the roof dead load is assumed to be 2kPa (Section 3.1). The substructure model is shown
9
10 in Figure 12 and the gravity frame section sizes are also adopted from the single-storey model given in
11 Table 6). The BRBs for the 2-storey substructure models were designed using the same design procedure
12 as outlined in Section 3.1 and the properties of the BRBs and design storey shears are given in Table 7.
13 The combined analysis models A, B and C were then constructed combining the roof model (Section
14 3.3) and the substructure models.
15
16 Roof BRB link Force-deformation (kN-mm)
17 as rigid mass 200
150 p=2%
18 100
50
19 0
-25 -20 -15 -10 -5 0
-50
5 10 15 20 25
20 -100
-150
Model-A
21 -200
δy =0.06%
A’
Fo
A
200
22 150
100
23 50
0
24
-25 -20 -15 -10 -5 0 5 10 15 20 25
-50
-100
Model-B
rP
25 -150
-200
δy =0.13%
26 200
150
27 100
50
28 0
ee
-25 -20 -15 -10 -5 0 5 10 15 20 25
Model-C
-50
29 -100
δy =0.20%
-150
30 Pin-ended
-200
BRB
31 beams
(a) Perspective view
rR
32 A A’ RFL
33 MRFL
5m 1FL
34 M1FL
35 5m
ev
36 Lumped mass model

5m (b) A-A’ Elevation view
37
38
iew
Figure 12. 60m Substructure model and modelling assumptions

39
40 The substructure periods and roof periods are mapped on the design spectrum in Figure 13. Model-
41
B and Model-C have fundamental periods in the constant velocity region and Model-A with a stiffer
42
43 substructure has a relatively shorter fundamental period in the constant acceleration region. The
44 fundamental periods contribute to a mass participation ratio of about 90% and most of the remaining
45 10% came from the higher T2 mode which lies in the constant acceleration region for all the models. The
46 substructure T2 modes are also in close proximity to roof’s dominant O1 and O2 modes giving rise to
47 second mode period ratios RT 2 = T2 /TR in the range of 1 (where T2 is the substructure T2 mode period
48
and TR is the roof’s O1 mode as defined previously by the authors [16]). Consequently, in these models,
49
50 the vertical amplification factors quantifying the T2-roof interactions are larger and more significant
51 than those from the T1-roof interactions with the longer RT 1 period ratios [16].
52
Table 6. 2-storey substructure model data
53 (a) Storey heights and seismic weights (b) Section sizes
54
Storey Height (m) Weight (kN) Member Section Shape Section Size (mm)
55
RFL 10 6064 Column SHS 450×450×25
56 1FL 5 6032 Beam I beam 340×250×9×14
57
58
59 5.2 Inter-storey drifts
60
The mean peak substructure inter-storey drifts and mean residual inter-storey drifts obtained from
NLRHA on the combined models are given in Figure 14. For structures having four stories or less,
1
2 Table 7. Equivalent Lateral Forces: Proportioning BRBs (σy = 235, 205MPa)
3 Model-A (T1 = 0.38s, T2 = 0.15s,)
4 Storey
h (m) kN kN kN kN kN mm kN/m mm2
5 Wx Vx BRB Vx BRB Vix BRB Pi δy BRB Keqi Ac
6 RFL 10 6064 1157 1157 40 70 1.85 37770 341
7 1FL 5 6032 2117 2117 60 105 1.85 56558 511
8 Model-B (T1 = 0.56s, T2 = 0.23s,)
9 h (m) kN kN kN kN kN mm kN/m mm2
10 Storey
Wx Vx BRB Vx BRB Vix BRB Pi δy BRB Keqi Ac
11 RFL 10 6064 1157 1157 40 70 4.12 16997 297
12 1FL 5 6032 2117 2117 60 105 4.12 25451 446
13 Model-C (T1 = 0.69s, T2 = 0.28s,)
14
h (m) kN kN kN kN kN mm kN/m mm2
15 Storey
Wx Vx BRB Vx BRB Vix BRB Pi δy BRB Keqi Ac
16
RFL 10 6064 1157 1157 40 70 6.17 11331 298
17
1FL 5 6032 2117 2117 60 105 6.17 16967 456
18
where Pi =Yield axial force in each BRB, δy =Yield axial deformation
19 Keqi =BRB equivalent axial stiffness, Ac =Area of core as defined in Takeuchi [34]
20
21
Fo
Target DBE (5%) Target DBE (5%) Roof O1=0.22s

22
Spectral Acceleration (cm/s2 )
2000 Roof O2=0.11s

Roof O1
23 Roof O1
Model-A-T2 Model-B-T2
Roof O2 Model-A-T1 Model-A T1=0.38s
24 1500 Model-B-T1 Model-C-T2 Model-A T2=0.15s
rP
Roof O2
25 Model-C-T1
Model-B T1=0.56s
26 1000 Model-B T2=0.23s
27 Model-C T1=0.69s
Model-C T2=0.28s
28 500
ee
29
30 0
10 -2 10 -1 10 0 10 1 10 -2 10 -1 10 0 10 1
31 Period (s) Period (s)
rR
32
33 Figure 13. Substructure and roof periods on the design acceleration spectrum
34
35
ev
36 ASCE-7 [1] prescribes an allowable storey drift of 2.5% for DBE level. The obtained mean storey drifts
37 are within the permissible ASCE limits for all the models. Typical values of mean residual drift are
38 about 0.3% for p=2% models under the DBE level. These values increase to 0.6% under the MCE level.
iew
39 If 0.5% residual drift is assumed to be the limit beyond which the structures are no longer practically
40
usable [41], then the models with less than or equal to 2% post-yield stiffness ratio may not meet this
41
42 immediate occupancy performance level under MCE level excitation.
43
44
45 SLE SLE
10 10
46
DBE DBE
MCE MCE
47
48 8 8
49
Elevation (m)
Elevation (m)
50 6 6
51
52
4 4
53
54
55 2 2
56
57 0 0
58 0 0.01 0.02 0.03 0.04 0.05 0 0.01 0.02 0.03
Storey drift ratio Storey drift ratio
59
60 (a) Storey drift ratio (b) Residual drift ratio
Figure 14. Model-B p = 2%: Mean of peak storey drift ratios of substructure
1
2 5.3 Peak Substructure response
3 For each of the three substructure models, modal pushover analyses corresponding to the first and
4
5
second translational modes were performed to obtain the initial base shear to roof displacement stiffness
6 (K1i ) as shown in Equation 18. The elastic design spectral acceleration of the first and second modes
7 (Sa1 and Sa2 ) were used to obtain the elastic base shears (Vi ) and then extrapolate the maximum
8 elastic substructure roof displacement (Di ) at point A of Figure 12(a) from the initial stiffness (K1i )
9 as shown in Equation 18. The peak elastic RFL level acceleration Ai of the substructure is also obtained
10 from the corresponding roof displacement and the fundamental period and the response spectrum as
11
12
shown in Equation 18. The reduction factors (Rµi and Rai ) were then computed for both T1 and T2
13 modes following Equations 2-4 from the target ductility µti using Equation 7 and Equation 19. For the
14 equivalent linearisation approach, Equations 8-16 were used. The peak inelastic base shear and roof
15 acceleration of the substructure model were then obtained by combining the modal base shears using
16 the SRSS modal combination rule using Equations 30 and 31.
17
18
19 5000 3000
(Idealised) Model-A (Idealised) Model-A
20 (Obtained) Model-A (Obtained) Model-A
4000 (Idealised) Model-B (Idealised) Model-B
21
Base Shear (kN)
Base Shear (kN)
(Obtained) Model-B (Obtained) Model-B

Fo
(Idealised) Model-C
22
(Idealised) Model-C
(Obtained) Model-C
2000 (Obtained) Model-C
3000
23
24 2000
1000
rP
25
26 1000
27
28 0 0
ee
0 0.005 0.01 0.015 0.02 0 1 2 3 4

29
Roof drift ratio Roof drift ratio 10 -3
30
31 (a) First mode (b) Second mode
rR
32
33 Figure 15. Modal pushover curves and idealisation of structural response (p = 2% models)
34
35
ev
36
s
37 V1 2 V2 2
sV = ( ) +( ) (30)
38 Rµ1 Rµ2
iew
39
40
s
A1 2 A2 2
41 s Aeq = ( ) +( ) (31)
42 Rµ1 Rµ2
43
44
45 1.2 1.2 1.2
Base shear coefficient Cb
46 1 p =2% 1 p =10% 1 p =25%

47 0.8 0.8 0.8
48 0.6 0.6 0.6

49 0.4 0.4 0.4
Model-A, Kasai
50 0.2 0.2 0.2
Model-A, Newmark
Model-A, N&K
51 0
Model-A, NLRHA
0 0 Model-B, Kasai
52 SLE DBE MCE SLE DBE MCE SLE DBE MCE
Model-B, Newmark
Model-B, N&K
53 15 15 15 Model-B, NLRHA
Sub acceleration Aeq (m/s2)

Model-C, Kasai
54 p =2% p =10% p =25% Model-C, Newmark
Model-C, N&K
55 10 10 10 Model-C, NLRHA
56
57 5 5 5
58
59
0 0 0
Figure 16. Base shear ratios and peak substructure (RFL) acceleration
1
2 1.2 1.2 1.2
p =2% p =10% p =25%
3 1 1 1
4 0.8 0.8 0.8 Model-A T1

Model-A T2
5 0.6 0.6 0.6 Model-B T1
Model-B T2
6 0.4 0.4 0.4
Model-C T1
Model-C T2
7 0.2 0.2 0.2

8 0 0 0
10
Figure 17. Kasai method: Reduction in substructure modal acceleration
11
12
13 The peak substructure accelerations s Aeq and base shear ratio Cb (=s V /W where W is the seismic
14 weight) for all the models are compared with the mean NLRHA response in Figure 16. The typical
15 DBE level base shear ratios are low at around 0.3 and increase to 0.35 for MCE level for p = 2%
16 models. These values further increase with increasing post-yield stiffness with Model-A with the lowest
17
reduction and the highest base shear ratio at about 0.9. As observed for the single-storey models,
18
19 the R-factor methods estimate the base shear and accelerations with good accuracy for p = 2%, with
20 Newmark’s method being the most conservative of all the estimation methods in the short-period range.
21 As the post-yield stiffness ratio increases (p = 25%), Kasai method is the most accurate if compared
Fo
22 with the mean NLRHA response. For SLE level response and low-ductility regimes, all methods are
23 equally accurate.
24
The achieved damping, reduction ratios and the periods corresponding to Kasai’s evaluation method
rP
25
26 (Section 2.2) are also shown in Tables 8-11. To compare the reduction in acceleration for each
27 mode across models with different post-yield stiffness, the reduction in peak substructure acceleration
28 Rai = Aeqi /Ai obtained from the Kasai method is plotted for modes T1 and T2 for Model-A, Model-B
ee
29 and Model-C in Figure 17. Among the three models, the differences in the reduction ratios for the same
30 post-yield stiffness and earthquake intensity were less than 10%. This suggests that variation in BRB
31
design yield displacements (within the reasonable range of 0.05%-0.2% [34]) have relatively minor
rR
32
33 effects on the overall peak substructure responses. For models with p = 2%, the reduction in T1 mode
34 is the lowest for the SLE level with typical Ra1 values at around 0.7. These values reduced further to 0.2
35 for the DBE level and 0.1 for MCE level. As the post-yield stiffness increases to p = 10%, the reduction
ev
36 ratio increases to 0.3 for the DBE level and 0.2 for the MCE level. The models with p = 25% therefore
37 exhibit the lowest reductions in acceleration with typical Ra1 values at around 0.8 for the SLE level,
38
iew
0.4 for the DBE level and 0.3 for the MCE level. However, for the T2 mode response, the increase
39
40
in post-yield stiffness has a relatively minor effect on the reduction ratios as the additional damping
41 is lesser and the ductility values are smaller in the higher mode (Tables 9 and 11). Thus, substructures
42 designed with a high post-yield stiffness can achieve reduced residual inter-storey drifts (Figure 22b) but
43 the corresponding gridshell roofs may have to be designed for higher design forces due to the increased
44 seismic input.
45
46 Table 8. p = 2% models: Peak substructure First-mode (T1) parameters
47 heq1 Keq1 /K1 Dh Rd Ra µt1 T1 Teq1 TR
48
Model-A-SLE 0.10 0.62 0.79 1.14 0.71 1.63 0.38 0.48 0.23
49 Model-A-DBE 0.43 0.11 0.44 1.74 0.20 8.14 0.38 1.12 0.23
50 Model-A-MCE 0.47 0.07 0.42 2.15 0.15 12.21 0.38 1.41 0.23
51
Model-B-SLE 0.09 0.65 0.82 1.01 0.66 1.53 0.56 0.69 0.23
52 Model-B-DBE 0.42 0.12 0.44 1.25 0.16 7.66 0.56 1.59 0.23
53 Model-B-MCE 0.46 0.08 0.43 1.52 0.12 11.48 0.56 2.00 0.23
54 Model-C-SLE 0.07 0.78 0.92 1.04 0.81 1.24 0.69 0.77 0.23
55 Model-C-DBE 0.38 0.16 0.46 1.14 0.19 6.18 0.69 1.70 0.23
56 Model-C-MCE 0.44 0.10 0.43 1.36 0.14 9.27 0.69 2.15 0.23
57
58
59 5.4 Peak Roof response
60
The multistorey structures have been previously found to exhibit significantly higher amplification,
especially in the vertical direction for RT 1 > 2 [16] which may be attributed to the substructure T2
1
2 Table 9. p = 2% models: Peak substructure Second-mode (T2) parameters
4 Model-A-SLE 0.05 1.00 1.00 1.00 1.00 1.00 0.15 0.15 0.23
5 Model-A-DBE 0.08 0.73 0.88 1.12 0.82 1.37 0.15 0.18 0.23
6 Model-A-MCE 0.15 0.49 0.69 1.19 0.58 2.06 0.15 0.22 0.23
7 Model-B-SLE 0.05 1.00 1.00 1.00 1.00 1.00 0.23 0.23 0.23
8 Model-B-DBE 0.08 0.72 0.87 1.12 0.81 1.40 0.23 0.27 0.23
9 Model-B-MCE 0.16 0.49 0.68 1.18 0.57 2.10 0.23 0.33 0.23
10 Model-C-SLE 0.05 1.00 1.00 1.00 1.00 1.00 0.28 0.28 0.23
11 Model-C-DBE 0.08 0.71 0.86 1.13 0.80 1.42 0.28 0.33 0.23
12 Model-C-MCE 0.16 0.47 0.67 1.20 0.57 2.13 0.28 0.41 0.23
13
16 Model-A-SLE 0.09 0.71 0.84 1.09 0.78 1.63 0.38 0.45 0.23
17 Model-A-DBE 0.21 0.34 0.60 1.33 0.45 8.14 0.38 0.65 0.23
18 Model-A-MCE 0.19 0.31 0.62 1.45 0.44 12.20 0.38 0.68 0.23
19 Model-B-SLE 0.08 0.74 0.86 1.01 0.74 1.53 0.56 0.65 0.23
20 Model-B-DBE 0.21 0.35 0.61 1.03 0.36 7.65 0.56 0.95 0.23
21 Model-B-MCE 0.20 0.31 0.62 1.11 0.34 11.48 0.56 1.01 0.23
Fo
22 Model-C-SLE 0.06 0.84 0.94 1.03 0.86 1.24 0.69 0.75 0.23
23 Model-C-DBE 0.20 0.37 0.61 0.99 0.37 6.18 0.69 1.12 0.23
24 Model-C-MCE 0.20 0.33 0.61 1.06 0.35 9.27 0.69 1.20 0.23
rP
25
27
heq2 Keq2 /K2 Dh Rd Ra µt2 T2 Teq2 TR
28
ee
Model-A-SLE 0.05 1.00 1.00 1.00 1.00 1.00 0.15 0.15 0.23
29 Model-A-DBE 0.07 0.79 0.91 1.09 0.86 1.37 0.15 0.17 0.23
30 Model-A-MCE 0.12 0.60 0.76 1.11 0.67 2.06 0.15 0.20 0.23
31
rR
Model-B-SLE 0.05 1.00 1.00 1.00 1.00 1.00 0.23 0.23 0.23
32 Model-B-DBE 0.07 0.79 0.90 1.08 0.85 1.40 0.23 0.26 0.23
33 Model-B-MCE 0.13 0.61 0.74 1.08 0.66 2.10 0.23 0.29 0.23
34 Model-C-SLE 0.05 1.00 1.00 1.00 1.00 1.00 0.28 0.28 0.23
35
ev
Model-C-DBE 0.07 0.76 0.90 1.11 0.84 1.42 0.28 0.32 0.23
36 Model-C-MCE 0.12 0.56 0.75 1.16 0.65 2.13 0.28 0.37 0.23
37
38
iew
39 mode and its interaction with the roof’s predominant modes in its close proximity. To account for these
40 higher-mode effects, the T2-roof horizontal and vertical amplification factors were proposed [16] as a
41 function of the period ratio RT 2 as defined in Equations 32-33. After obtaining the peak substructure
42
43
response, the peak roof accelerations for the combined models were calculated for each of the modes
44 from the inelastic response s AHeqi and the corresponding first and second mode amplification factors
45 in Equations 23-28. For the equivalent linearisation approach, the substructure’s peak modal response
46 s AHeqi was computed considering the effect of both the period shift and the damping [13]. For the R-
47 factor approach methods, the peak inelastic substructure acceleration for each mode was obtained using
48 Equation 7. For comparison, the DBE level response was also computed using the Direct-R method
49
(Section 4.2) and a reduction factor R = 8 for both the modes.
50
51 The modal accelerations at each node were then combined using Equations 34-35 to obtain the
52 combined response envelope. This study uses an absolute summation rule [13, 16]. The equivalent static
53 seismic forces for each node may then be computed from the nodal mass mk and acceleration c AH (x, y)
54 or c AV (x, y) at position (x, y) using Equation 36.
55
56
57 FH2 = 1 (32)
58  
0 RT 2 ≤ 1/5
59

 

60
 6CV θ(RT 2 − 1/5) 1/5 < RT 2 < 7/10

FV 2 = (33)
 3CV θ 7/10 ≤ RT 2 ≤ 21/16 
 p

RT 2 > 21/16

5/(RT 2 − 1) − 1)CV θ

1
2
2
3 X
4 c AH (x, y) = |c AHi (x, y)| (34)
5 i=1
6 2
7
X
c AV (x, y) = |c AV i (x, y)| (35)
8
i=1
9
10 fH (x, y) = mk c AH (x, y), fV (x, y) = mk c AV (x, y) (36)
11
12 The proposed ridgeline A-O-A’ (Figure 4) accelerations are compared with the mean NLRHA results
13 in Figures 18-20. For comparison, results accounting only for the substructure T1 mode (s AHeq1 ) with
14 the corresponding roof amplification factors (FH1 and FV 1 ) are also shown and labelled ‘T1’ and for
15
the substructure T2 mode (s AHeq2 ) response obtained with the corresponding roof amplification factors
16
17 (FH2 and FV 2 ) are also shown and labelled ‘T2’.
18 The ductility ratios from the first mode T1 are large resulting in longer secant periods (Tables 8 and
19 10) and thereby smaller substructure accelerations from the first mode as depicted by the dotted T1
20 lines in Figures 18-20. The second mode accelerations achieve smaller ductility ratios (Tables 9 and
21 11) resulting in near-elastic peak substructure (s AHeq2 ) response. Furthermore, the T2 mode exhibits
Fo
22
shorter period ratios (RT 2 ) leading to larger vertical roof amplification factors and therefore, the higher
23
24 mode contribution towards the overall vertical roof response is much larger than the first mode, even
rP
25 though the mass participation of the higher mode is small. For p < 10%, all the estimation methods
26 present near identical responses for Model-B and Model-C (with T1 modes in the constant velocity
27 region) accurately capturing the mean NLRHA response. For Model-A (with T1 mode in the constant
28
ee
acceleration region), the Newmark method presents the most conservative response among the proposed
29
R-factor approaches. For p < 10%, the R-factor approaches results are conservative if compared to the
30
31 mean NLRHA response. For DBE level response results using direct-R factor method, the accuracy
rR
32 is relatively low as it assumes much larger force reduction factors in both the modes and may not be
33 realistic for multistorey structures with significant second mode contributions. For structures with higher
34 post-yield stiffness like the p = 25% models, the peak substructure response is underestimated by the R-
35
ev
factor methods leading to unconservative overall peak roof responses (Figure 20). These errors increase
36 as the level of the intensity increases from SLE to MCE with increasing ductility values. The more
37
38
accurate Kasai method is therefore recommended for estimating responses of structures with higher
iew
39 post-yield stiffness ratios.

40
41 5.5 6-storey analysis models
42 To further investigate the effects of substructure higher-modes, a 6-storey substructure with a height of
43
30m was also constructed as shown in the Figure 21. Commensurate with the geometry of the previous
44
45 substructure models, the storey height for the 6-storey substructure is kept constant at 5m, each storey
46 has a floor weight of 7kPa and the roof dead load is assumed to be 2kPa (Section 3.1). The gravity frame
47 section sizes are also adopted from the 2-storey model given in Table 6. The BRBs were designed using
48 the same design procedure and a standard post-yield stiffness ratio p=2% as outlined in Section 3.1 and
49 the properties of the BRB links and design storey shears are given in Table 12. The combined analysis
50
model (Model-B) was then constructed combining the roof model (Section 3.3) and the substructure
51
52
model. To investigate the effect of post-yield stiffness, an additional substructure model with links
53 modelled with a post-yield stiffness p=25% were created keeping the yield displacements the same.
54 For brevity, only the Model-B is discussed in this paper for 6-storey models.
55
56
57
58
59
60
1
2 SLE DBE MCE
3 (a) Model-A (T1 =0.38s, T2 =0.15s)
4
25 25 25
5 Kasai-T1 N&K-T2 Kasai-T1 Direct-R-T2 Kasai-T1 N&K-T2
Newmark-T1 Kasai-T1+T2 Newmark-T1 Kasai-T1+T2 Newmark-T1 Kasai-T1+T2
6 20 N&K-T1
Kasai-T2
Newmark-T1+T2
N&K-T1+T2
20 N&K-T1
Direct-R-T1
Newmark-T1+T2
N&K-T1+T2
20 N&K-T1
Kasai-T2
Newmark-T1+T2
N&K-T1+T2
7 15
Newmark-T2 Mean NLRHA
15
Kasai-T2
Newmark-T2
Direct-R-T1+T2
Mean NLRHA 15
8
N&K-T2
10 10 10
9
5 5 5
10
11 0
-40 -20 0 20 40
0
-40 -20 0 20 40
0
-40 -20 0 20 40
13 15 15 15
14
15 10 10 10
16
17
5 5 5
18
19
20
0 0 0
-40 -20 0 20 40 -40 -20 0 20 40 -40 -20 0 20 40
Fo
22 (b) Model-B (T1 =0.56s, T2 =0.23s)

23 25 25 25
24 Kasai-T1
Newmark-T1
N&K-T2
Kasai-T1+T2
Kasai-T1
Newmark-T1
Direct-R-T2
Kasai-T1+T2
Kasai-T1
Newmark-T1
N&K-T2
Kasai-T1+T2
rP
20 20 20
25 N&K-T1
Kasai-T2
Newmark-T1+T2
N&K-T1+T2
N&K-T1
Direct-R-T1
Newmark-T1+T2
N&K-T1+T2
N&K-T1
Kasai-T2
Newmark-T1+T2
N&K-T1+T2
26
Newmark-T2 Mean NLRHA Kasai-T2 Direct-R-T1+T2 Newmark-T2 Mean NLRHA
15 15 Newmark-T2 Mean NLRHA 15
N&K-T2
27 10 10 10
28
ee
5 5 5
29
0 0 0
30 -40 -20 0 20 40 -40 -20 0 20 40 -40 -20 0 20 40
rR
32 15 15 15
33
34 10 10 10
35
ev
36 5 5 5
37
38
iew
0 0 0
39 -40 -20 0 20 40 -40 -20 0 20 40 -40 -20 0 20 40
Ridge coordinates (m) Ridge coordinates (m) Ridge coordinates (m)
40
41 (c) Model-C (T1 =0.69s, T2 =0.28s)
42 25 25 25
43
Kasai-T1 N&K-T2 Kasai-T1 Direct-R-T2 Kasai-T1 N&K-T2
20 N&K-T1 Newmark-T1+T2 20 N&K-T1 Newmark-T1+T2 20 N&K-T1 Newmark-T1+T2
44 Kasai-T2
Newmark-T2
N&K-T1+T2
Mean NLRHA
Direct-R-T1
Kasai-T2
N&K-T1+T2
Direct-R-T1+T2
Kasai-T2
Newmark-T2
N&K-T1+T2
Mean NLRHA
15 15 15
45 Newmark-T2
N&K-T2
Mean NLRHA
46 10 10 10
47 5 5 5
48 0 0 0
49 -40 -20 0 20 40 -40 -20 0 20 40 -40 -20 0 20 40
50
51 15 15 15
52
53 10 10 10
54
55 5 5 5
56
57 0 0 0
-40 -20 0 20 40 -40 -20 0 20 40 -40 -20 0 20 40
59
60 Figure 18. p = 2% 2-storey models: Ridgeline accelerations
1
2 SLE DBE MCE
3 (a) Model-A (T1 =0.38s, T2 =0.15s)
4
25 25 25
6 20 N&K-T1
Kasai-T2
Newmark-T1+T2
N&K-T1+T2
20 N&K-T1
Direct-R-T1
Newmark-T1+T2
N&K-T1+T2
20 N&K-T1
Kasai-T2
Newmark-T1+T2
N&K-T1+T2
7 15
15
Kasai-T2
Newmark-T2
Direct-R-T1+T2
Mean NLRHA 15
8
N&K-T2
10 10 10
9
5 5 5
10
11 0
-40 -20 0 20 40
0
-40 -20 0 20 40
0
-40 -20 0 20 40
13 20 20 20
14
15 15 15 15
16 10 10 10
17
18 5 5 5
19
20
0 0 0
-40 -20 0 20 40 -40 -20 0 20 40 -40 -20 0 20 40
Fo
22 (b) Model-B (T1 =0.56s, T2 =0.23s)

23 25 25 25
24 Kasai-T1
Newmark-T1
N&K-T2
Kasai-T1+T2
Kasai-T1
Newmark-T1
Direct-R-T2
Kasai-T1+T2
Kasai-T1
Newmark-T1
N&K-T2
Kasai-T1+T2
rP
20 20 20
25 N&K-T1
Kasai-T2
Newmark-T1+T2
N&K-T1+T2
N&K-T1
Direct-R-T1
Newmark-T1+T2
N&K-T1+T2
N&K-T1
Kasai-T2
Newmark-T1+T2
N&K-T1+T2
26
N&K-T2
27 10 10 10
28
ee
5 5 5
29
0 0 0
30 -40 -20 0 20 40 -40 -20 0 20 40 -40 -20 0 20 40
rR
32 20 20 20
33
34
15 15 15
35
ev
10 10 10
36
37 5 5 5
38
iew
0 0 0
39 -40 -20 0 20 40 -40 -20 0 20 40 -40 -20 0 20 40
40
41 (c) Model-C (T1 =0.69s, T2 =0.28s)
42 25 25 25
43
20 N&K-T1 Newmark-T1+T2 20 N&K-T1 Newmark-T1+T2 20 N&K-T1 Newmark-T1+T2
44 Kasai-T2
Newmark-T2
N&K-T1+T2
Mean NLRHA
Direct-R-T1
Kasai-T2
N&K-T1+T2
Direct-R-T1+T2
Kasai-T2
Newmark-T2
N&K-T1+T2
Mean NLRHA
15 15 15
45 Newmark-T2
N&K-T2
Mean NLRHA
46 10 10 10
47 5 5 5
48 0 0 0
49 -40 -20 0 20 40 -40 -20 0 20 40 -40 -20 0 20 40
50
51 20 20 20
52 15 15 15
53
54 10 10 10
55
5 5 5
56
57 0 0 0
-40 -20 0 20 40 -40 -20 0 20 40 -40 -20 0 20 40
59
1
2 SLE DBE MCE
3 (a) Model-A (T1 =0.38s, T2 =0.15s)
4
25
5 25
Kasai-T1 N&K-T2 Kasai-T1
Newmark-T1
Direct-R-T2
Kasai-T1+T2
25
Kasai-T1 N&K-T2
Newmark-T1 Kasai-T1+T2 Newmark-T1 Kasai-T1+T2
6 20 N&K-T1
Kasai-T2
Newmark-T1+T2
N&K-T1+T2
20 Krawlinkler-T1
Direct-R-T1
Newmark-T1+T2
Krawlinkler-T1+T2
20 N&K-T1
Kasai-T2
Newmark-T1+T2
N&K-T1+T2
7
Kasai-T2 Direct-R-T1+T2
Newmark-T2 Mean NLRHA Newmark-T2 Mean NLRHA
15 15 Newmark-T2 Mean NLRHA
15
Krawlinkler-T2
8 10 10 10
9
5
10
5 5
11 0
-40 -20 0 20 40
0
-40 -20 0 20 40
0
-40 -20 0 20 40
13 20 20 20
14
15 15 15 15
16 10 10 10
17
18 5 5 5
19
20
0 0 0
-40 -20 0 20 40 -40 -20 0 20 40 -40 -20 0 20 40
Fo
22 (b) Model-B (T1 =0.56s, T2 =0.23s)

23
Horizontal Acceleration (m/s2)
25 25 25
24 Kasai-T1
Newmark-T1
N&K-T2
Kasai-T1+T2
Kasai-T1
Newmark-T1
Direct-R-T2
Kasai-T1+T2
Kasai-T1
Newmark-T1
N&K-T2
Kasai-T1+T2
rP
20
25 20 N&K-T1
Kasai-T2
Newmark-T1+T2
N&K-T1+T2
N&K-T1
Direct-R-T1
Newmark-T1+T2
N&K-T1+T2
20 N&K-T1
Kasai-T2
Newmark-T1+T2
N&K-T1+T2
26
N&K-T2
27 10 10 10
28
ee
5 5 5
29
0 0 0
30 -40 -20 0 20 40 -40 -20 0 20 40 -40 -20 0 20 40
rR
32 20 20 20
Vertical Acceleration (m/s2)
33
34
15 15 15
35
ev
10 10 10
36
37 5 5 5
38
iew
0 0 0
39 -40 -20 0 20 40 -40 -20 0 20 40 -40 -20 0 20 40
40
41 (c) Model-C (T1 =0.69s, T2 =0.28s)
42 25 25 25
43
20 N&K-T1 Newmark-T1+T2 20 Krawlinkler-T1 Newmark-T1+T2
20 N&K-T1 Newmark-T1+T2
44 Kasai-T2
Newmark-T2
N&K-T1+T2
Mean NLRHA
Direct-R-T1
Kasai-T2
Krawlinkler-T1+T2
Direct-R-T1+T2
Kasai-T2
Newmark-T2
N&K-T1+T2
Mean NLRHA
15 Newmark-T2 Mean NLRHA
45 15 Krawlinkler-T2
15
46 10 10 10
47 5 5 5
48 0 0 0
49 -40 -20 0 20 40 -40 -20 0 20 40 -40 -20 0 20 40
50
51 20 20 20
52 15 15 15
53
54 10 10 10
55
5 5 5
56
57 0 0 0
-40 -20 0 20 40 -40 -20 0 20 40 -40 -20 0 20 40
59
1
2 Table 12. Equivalent Lateral Forces: Proportioning BRBs (σy = 235MPa)
3 Model-B (T1 = 0.93s, T2 = 0.39s,)
4 h (m) kN kN kN kN kN mm kN/m mm2
Storey
5 Wx Vx BRB Vx BRB Vix BRB Pi δy BRB Keqi Ac
6 RFL 30 6064 1483 1483 43 75 4.12 18343 321
7 6FL 25 6032 2713 2713 77 138 4.12 33550 588
8 5FL 20 6032 3696 3696 106 188 4.12 45715 801
9 4FL 15 10968 5038 5038 144 256 4.12 62305 1091
10 3FL 10 10968 5932 5932 169 302 4.12 73364 1285
2FL 5 10968 6379 6379 182 325 4.12 78894 1382
11
12
13
14
15 Roof members
16
17
18
19
20
21
Fo
22
23
24 BRB link Force-deformation
rP
25 (kN-mm)
26
p=2%
600
27 400
28 200
ee
29 -25 -20 -15 -10 -5 0

-200
5 10 15 20 25
Model-B
30 -400
-600 δy =0.13%
31
rR
32
33 Pinned beams BRB as link elements
34
35
ev
O
36 Pinned Rigid
z
37 connection connection
38 A A’
iew
x RFL
39 6FL
40 5FL
MF Column
41 4FL Tension
42
MF Beam floor weight 3FL beam
43 Pinned
θ = 7kPa 2FL
44 storey height = 5m connection detail
θ=30° Base
45 L/2=30m 5m 5m 5m
46
47 Figure 21. 6-storey Model-B: Perspective and elevation view
48
49
50
51 5.6 Inter-storey drifts
52
53 The mean peak inter-storey drifts and mean residual inter-storey drifts for the substructure of the
54 combined models (Model-B) are given in Figure 22. ASCE-7 [1] prescribes an allowable storey drift of
55 2.0% for DBE level. The obtained mean storey drifts are quite uniform along the substructure height
56 and are within the permissible ASCE limits for all the models with post-yield stiffness. Typical values
57 of mean residual drift are about 0.3% for p=2% models under the DBE level. These values increase
58
59
to 0.4% under the MCE level within the 0.5% limit for immediate occupancy criteria for most stories
60 except the roof floor. While increasing the post-yield stiffness to p=25% reduces the peak inter-storey
drifts, the improvement in residual drifts is much more significant with typical residual drifts for all the
three levels within 0.2%.
1
35 35
2 p=2%-SLE p=2%-SLE
3 30
p=2%-DBE
p=2%-MCE 30
p=2%-DBE
p=2%-MCE
4 p=25%-SLE
p=25%-DBE
p=25%-SLE
p=25%-DBE
5 25 p=25%-MCE
25 p=25%-MCE
Elevation (m)
Elevation (m)
7 20 20
8
9 15 15
10
10 10
11
12
5 5
13
14 0 0
15 0 0.02 0.04 0 0.005 0.01 0.02
16 Storey drift ratio Storey drift ratio
17 (a) Storey drift ratio (b) Residual drift ratio
18
19 Figure 22. Model-B: Mean of peak storey drift ratios of substructure
20
21
Fo
22 5.7 Peak Substructure acceleration

23 The first substructure mode T1 (fundamental translational mode) has a period of 0.93s and a mass
24
participation of 77% and the second translational mode T2 has a period of 0.39s and contributes to
rP
25
26 another 15%. The periods are longer than those of the 2-storey models resulting in longer period ratios.
27 These longer period ratios imply that the roof-substructure interaction is weaker and corresponds to
28 smaller vertical amplification factors which have an inverse relationship with the period ratios in this
ee
29 range. The mass participation of second mode is larger if compared to the 2-storey substructure models
30 leading to higher contribution of the T2 mode to the peak substructure response.
31
rR
32
For each of the models, modal pushover analyses was conducted to obtain the first and second modal
33 pushover curves and initial base shear to roof displacement stiffness (K1i ) was obtained as shown in
34 Figure 23. For p = 2% model, the first mode pushover curve has a post-yield stiffness of 2% and the
35
ev
second mode pushover curve was found to have a higher post-yield stiffness of 11% as the BRBs of all
36 the storeys do not yield at the same time and many of the braces remain elastic when the substructure is
37 pushed using a force distribution based on the second mode shape. Similarly, for the p = 25% model,
38
iew
the first mode pushover curve has a post-yield stiffness of 25% and the second mode pushover curve
39
40 was found to have a post-yield stiffness ratio of 36%.
41
42 10 4 10 4
3 3
43 (Idealised) p=2% (Idealised) p=2%
(Obtained) p=2% (Obtained) p=2%
44 (Idealised) p=25% (Idealised) p=25%
Base Shear (kN)
Base Shear (kN)
(Obtained) p=25%
45 2 2
(Obtained) p=25%
46
47
48 1 1
49
50
51 0 0
52 0 0.005 0.01 0.015 0.02 0 1 2 3 4 5 6
53 Roof drift ratio Roof drift ratio 10 -3
54
(a) First mode (T1) (b) Second mode (T2)
55
56 Figure 23. Model-B: Modal pushover curves and idealisation of structural response
57
58
59
As for the 2-storey and single-storey models, the peak response of each model was computed for three
60 levels of earthquake intensity-SLE, DBE and MCE. The reduction in acceleration (Rai ) and achieved
damping parameters for each substructure mode from the Kasai method are given in Tables 13-16.
For the DBE level, as both the seismic demand and reduction factors are large, the roof displacement
1
2 ductility factors obtained from the first mode µ1 were about 6. The second mode had a significant
3 contribution to damping as well as can be seen from the higher heq2 values and lower Dh2 values in
4 Tables 14 and 16 if compared to the 2-storey models in Tables 9 and 11. The shift in periods and
5 reduction in acceleration was thus larger than those observed for the second mode in 2-storey models.
6
8
9
10 Model-B-SLE 0.06 0.87 0.97 1.04 0.91 1.11 0.94 1.00 0.23
Model-B-DBE 0.37 0.18 0.47 1.10 0.20 5.53 0.94 2.19 0.23
11
Model-B-MCE 0.43 0.11 0.44 1.30 0.15 8.30 0.94 2.78 0.23
12
13
14
Table 14. p = 2% models: Peak substructure Second-mode (T2) parameters
15
17 Model-B-SLE 0.05 1.00 1.00 1.00 1.00 1.00 0.39 0.39 0.23
18 Model-B-DBE 0.23 0.43 0.58 1.08 0.46 2.96 0.39 0.60 0.23
19 Model-B-MCE 0.31 0.33 0.51 1.10 0.36 4.44 0.39 0.69 0.23
20
21
Fo

24 Model-B-SLE 0.05 0.91 0.98 1.03 0.94 1.11 0.94 0.98 0.23
rP
25 Model-B-DBE 0.20 0.39 0.61 0.98 0.38 5.53 0.94 1.49 0.23
26 Model-B-MCE 0.20 0.34 0.61 1.04 0.35 8.29 0.94 1.60 0.23
27
28
ee

30
31
rR
32 Model-B-SLE 0.05 1.00 1.00 1.00 1.00 1.00 0.39 0.39 0.23
Model-B-DBE 0.19 0.51 0.63 1.05 0.54 4.13 0.39 0.55 0.23
33
Model-B-MCE 0.20 0.46 0.61 1.08 0.50 6.20 0.39 0.58 0.23
34
35
ev
36 5.8 Peak roof response

37
38
After obtaining the peak substructure response, the peak roof accelerations for the combined models
iew
39 were calculated for each of the modes from the inelastic response s AHeqi and the corresponding first
40 and second mode amplification factors in Equations 23-28 as outlined in Section 5.4. The peak roof
41 accelerations for the combined models listed in Table 13-16 were calculated using the first and second
42 mode amplification factors in Equations 23-28. The proposed envelope of ridgeline A-O-A’ (Figure 21)
43 accelerations are compared with the mean NLRHA results in Figures 24-25. The second mode achieves
44
45
smaller ductility ratios (Tables 14 and 16) and the reduction in acceleration is lesser if compared to
46 the T1 mode response. Therefore, in large ductility regimes for the DBE and MCE levels, s AHeq2 (T2
47 dashed lines) is higher than s AHeq1 (T1 dotted lines). Furthermore, the T2 mode exhibits relatively
48 shorter period ratios (RT 2 < RT 1 ) leading to larger vertical roof amplification factors (FV 2 > FV 1 ) and
49 therefore, the higher mode contribution towards the overall vertical roof response is much larger than
50 the first mode, even though the mass participation of the higher mode is small. For p = 2%, the response
51
52
from Newmark’s method (T1+T2 response) is nearly equal to the Kasai method (T1+T2 response) in
53 Figure 24, conservatively capturing the mean NLRHA response. For the DBE level response results
54 using direct-R factor method, the accuracy is relatively low as it assumes much larger force reduction
55 factors than actually achieved in both the modes and may not be realistic for multistorey structures with
56 significant second mode contributions. As observed in the previous 2-storey models (Figures 16-17), for
57 structures with higher post-yield stiffness like the p = 25% models, the peak substructure response is
58
59
underestimated by the R-factor methods leading to unconservative overall peak roof responses (Figure
60 25). These errors increase as the level of the intensity increases from SLE to MCE with increasing
ductility values. Therefore, the iterative but accurate Kasai method is recommended for estimating
responses of structures with higher post-yield stiffness ratios.
1
2 SLE DBE MCE
3 Horizontal Acceleration (m/s2)

30 30 30
4 Kasai-T1
Newmark-T1
N&K-T2
Kasai-T1+T2
Kasai-T1
Newmark-T1
Direct-R-T2
Kasai-T1+T2
Kasai-T1
Newmark-T1
N&K-T2
Kasai-T1+T2
5 20
N&K-T1
Kasai-T2
Newmark-T1+T2
N&K-T1+T2
20
N&K-T1
Direct-R-T1
Newmark-T1+T2
N&K-T1+T2
20
N&K-T1
Kasai-T2
Newmark-T1+T2
N&K-T1+T2
6
N&K-T2
7
10 10 10
8
9
10 0
-40 -20 0 20 40
0
-40 -20 0 20 40
0
-40 -20 0 20 40
12 20 20 20

13
15
14 15 15
15 10 10 10
16 max
min
17 5 5 5
18 0 0 0
19 -40 -20 0 20 40 -40 -20 0 20 40 -40 -20 0 20 40
Ridge coordinates (m)
20 Ridge coordinates (m) Ridge coordinates (m)
21
Fo
22 Figure 24. 6-storey (Model-B: T1 =0.94s, T2 =0.39s) p = 2%: Ridgeline accelerations

23
24
rP
25 SLE DBE MCE

26

30 30 30
28
ee
N&K-T1 Newmark-T1+T2 N&K-T1 Newmark-T1+T2 N&K-T1 Newmark-T1+T2

Kasai-T2 N&K-T1+T2 Direct-R-T1 N&K-T1+T2 Kasai-T2 N&K-T1+T2
20 20 20
29 Newmark-T2 Mean NLRHA Kasai-T2
Newmark-T2
Direct-R-T1+T2
Mean NLRHA
30
N&K-T2
31 10 10 10
rR
32
33 0
-40 -20 0 20 40
0
-40 -20 0 20 40
0
-40 -20 0 20 40
35
ev
20 20 20
36
37 15 15 15
38
iew
10 10 10 max
39
min
40 5 5 5
41
0 0 0
42 -40 -20 0 20 40 -40 -20 0 20 40 -40 -20 0 20 40
44
45 Figure 25. 6-storey (Model-B: T1 =0.94s, T2 =0.39s), p = 25%: Ridgeline accelerations
46
47
48 6 Equivalent static loads
49 The proposed peak roof accelerations were then used to compute the corresponding equivalent static
50 loads using Equation 36. Static analyses were performed by applying the obtained vertical and horizontal
51
52
loads simultaneously to the roof nodes in each of the patterns (corresponding to the roof’s dominant
53 anti-symmetric mode shape O1 expressed by Equations 27 and 28) as shown in Figure 26 [13]. The
54 maximum response (axial force and bending moment) in each member from the four static load cases
55 may be considered as the preliminary seismic demand. The detailed results for the 2-storey p = 2%
56 and p = 25% models are compared with the mean NLRHA response in Figures 27-28. The proposed fit
57 lines are also shown using dashed lines. The axial forces from the R-factor approaches are close to the
58
59
NLRHA response except for the direct-R method. Furthermore, the R-factor methods’ and the Kasai
60 method’s member responses are nearly equal for models with p = 2%. For p < 10%, all the estimation
methods present near identical responses for Model-B and Model-C (with T1 modes in the constant
velocity region) accurately capturing the mean NLRHA response. For Model-A (with T1 mode in the
1
2 fH fH
(1) (3)
3
4
5 fV fV
6
7
8
9
10 (2) fH (4) fH
11
12
13 fV fV
14
15
16
17 Figure 26. Static Load Patterns
18
19
20 constant acceleration region), the Newmark method followed by the Kasai method present the most
21 conservative response among the investigated estimation methods. The errors in the R-factor methods
Fo
22 are more evident in the models with higher post-yield stiffness (Figure 28) where the Kasai method
23 (shown in black markers) exhibits a much more conservative DBE and MCE level axial forces and
24 bending moments when compared to the NLRHA method. As discussed in Section 5.4, this is due to the
rP
25 underlying assumptions of R-factor approaches which were proposed considering EPP or low post-yield
26
27
stiffness behaviour and thus are recommended for models with relatively low post-yield stiffness.
28 The member responses of the 6-storey models are also compared in detail in Figures 29-30. While
ee
29 the second mode possesses higher post-yield stiffness in the 6-storey cases, similar to the 2-storey
30 models’ responses, the accuracy of the Newmark method and the Kasai method remained nearly equal
31 for models with p = 2%. As observed for the peak ridge accelerations, the differences in proposed
rR
32 member forces are more evident in the models with higher post-yield stiffness where the Kasai method
33
34
exhibits a more conservative response with the errors in the R-factor method estimates larger in the
35 higher ductility ranges (the DBE and MCE level). As the equivalent static load results from the proposed
ev
36 RSA and amplification factor approach provide an enveloped response, the obtained axial forces and
37 bending moments do not necessarily correspond to the critical combination of the forces and moments
38
iew
obtained from the response-history analyses [5, 7]. Therefore, it is recommended to use the proposed
39 equivalent static load procedure in the preliminary design stages as a baseline method for initial sizing
40
41
and approximate estimation of the member forces. For the final design member check, it is recommended
42 to use the more rigorous NLRHA method.
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
1
2 Model SLE DBE MCE
3 1000 1000 1000
4 800 800 800
5
Proposed (kN)
Proposed (kN)
Proposed (kN)
600 600 600
6
7 400 400 400
8 200 200 200
9 0 0 0
0 200 400 600 800 1000 0 200 400 600 800 1000 0 200 400 600 800 1000
10 A NLRHA (kN) NLRHA (kN) NLRHA (kN)
11 (a) Axial forces (kN)
12 1000 1000 1000
13 Kasai Kasai Kasai
Kasai-fit Kasai-fit Kasai-fit
14
Proposed (kN-m)
Proposed (kN-m)
Proposed (kN-m)
800 800 800
Newmark Newmark Newmark
Newmark-fit Newmark-fit Newmark-fit
15 600 N&K
N&K-fit
600 N&K
N&K-fit
600 N&K
N&K-fit
16 400 400 Direct-R
Direct-R-fit
400
17 200 200 200

18 0 0 0
19 0 200 400
NLRHA (kN-m)
600 800 1000 0 200 400
NLRHA (kN-m)
600 800 1000 0 200 400
NLRHA (kN-m)
600 800 1000
20
(b) Bending moments (kN-m)
21
Fo
22 1000 1000 1000
23 800 800 800

Proposed (kN)
Proposed (kN)
Proposed (kN)
24 600 600 600
rP
25 400 400 400

26 200 200 200
27
28
0 0 0
ee
0 200 400 600 800 1000 0 200 400 600 800 1000 0 200 400 600 800 1000
29 B NLRHA (kN) NLRHA (kN) NLRHA (kN)

31 1000 1000 1000
rR
32
Kasai Kasai Kasai
Proposed (kN-m)
Proposed (kN-m)
Proposed (kN-m)
800 800 800
33 600
Newmark-fit
N&K 600
Newmark-fit
N&K 600
Newmark-fit
N&K
34 N&K-fit N&K-fit
Direct-R
N&K-fit
400 400 400
35
ev
Direct-R-fit
36 200 200 200
37 0
0 200 400 600 800 1000
0
0 200 400 600 800 1000
0
0 200 400 600 800 1000
38
iew
NLRHA (kN-m) NLRHA (kN-m) NLRHA (kN-m)

39 (b) Bending moments (kN-m)
40 1000 1000 1000
41 800 800 800
42
Proposed (kN)
Proposed (kN)
Proposed (kN)
43
600 600 600
44 400 400 400
45 200 200 200
46 0 0 0
47
0 200 400 600 800 1000 0 200 400 600 800 1000 0 200 400 600 800 1000
C NLRHA (kN) NLRHA (kN) NLRHA (kN)
49
1000 1000 1000
51
Proposed (kN-m)
Proposed (kN-m)
Proposed (kN-m)
800 800 800

52 600 N&K
N&K-fit
600 N&K
N&K-fit
600 N&K
N&K-fit
53 400 400 Direct-R
Direct-R-fit
400
54 200 200 200

55 0 0 0
56 0 200 400 600 800 1000 0 200 400 600 800 1000 0 200 400 600 800 1000
57
59
60 Figure 27. 2-storey p = 2% models: Comparison of member axial forces and bending moments
1
2 Model SLE DBE MCE
3 1000 1000 1000
4 800 800 800
5
Proposed (kN)
Proposed (kN)
Proposed (kN)
600 600 600
6
7 400 400 400
8 200 200 200
9 0 0 0
0 200 400 600 800 1000 0 200 400 600 800 1000 0 200 400 600 800 1000
10 A NLRHA (kN) NLRHA (kN) NLRHA (kN)
12 1000 1000 1000
14
Proposed (kN-m)
Proposed (kN-m)
Proposed (kN-m)
800 800 800
15 600 N&K
N&K-fit
600 N&K
N&K-fit
600 N&K
N&K-fit
16 400 400 Direct-R
Direct-R-fit
400
17 200 200 200

18 0 0 0
19 0 200 400
NLRHA (kN-m)
600 800 1000 0 200 400
NLRHA (kN-m)
600 800 1000 0 200 400
NLRHA (kN-m)
600 800 1000
20
21
Fo
22 1000 1000 1000
23 800 800 800

Proposed (kN)
Proposed (kN)
Proposed (kN)
24 600 600 600
rP
25 400 400 400

26 200 200 200
27
28
0 0 0
ee
0 200 400 600 800 1000 0 200 400 600 800 1000 0 200 400 600 800 1000

31 1000 1000 1000
rR
32
Kasai Kasai Kasai
Proposed (kN-m)
Proposed (kN-m)
Proposed (kN-m)
800 800 800
33 600
Newmark-fit
N&K 600
Newmark-fit
N&K 600
Newmark-fit
N&K
34 N&K-fit N&K-fit
Direct-R
N&K-fit
400 400 400
35
ev
Direct-R-fit
36 200 200 200
37 0
0 200 400 600 800 1000
0
0 200 400 600 800 1000
0
0 200 400 600 800 1000
38
iew

40 1000 1000 1000
41 800 800 800
42
Proposed (kN)
Proposed (kN)
Proposed (kN)
43
600 600 600
44 400 400 400
45 200 200 200
46 0 0 0
47
0 200 400 600 800 1000 0 200 400 600 800 1000 0 200 400 600 800 1000
C NLRHA (kN) NLRHA (kN) NLRHA (kN)
49
1000 1000 1000
51
Proposed (kN-m)
Proposed (kN-m)
Proposed (kN-m)
800 800 800

52 600 N&K
N&K-fit
600 N&K
N&K-fit
600 N&K
N&K-fit
53 400 400 Direct-R
Direct-R-fit
400
54 200 200 200

55 0 0 0
56 0 200 400 600 800 1000 0 200 400 600 800 1000 0 200 400 600 800 1000
57
59
1
2 Model SLE DBE MCE
3 1000 1000 1000
4 800 800 800
5
Proposed (kN)
Proposed (kN)
Proposed (kN)
600 600 600
6
7 400 400 400
8 200 200 200
9 0 0 0
0 200 400 600 800 1000 0 200 400 600 800 1000 0 200 400 600 800 1000
12 1000 1000 1000
14
Proposed (kN-m)
Proposed (kN-m)
Proposed (kN-m)
800 800 800
15 600 N&K
N&K-fit
600 N&K
N&K-fit
600 N&K
N&K-fit
16 400 400 Direct-R
Direct-R-fit
400
17 200 200 200

18 0 0 0
19 0 200 400
NLRHA (kN-m)
600 800 1000 0 200 400
NLRHA (kN-m)
600 800 1000 0 200 400
NLRHA (kN-m)
600 800 1000
20
21
Fo
23
24
rP
25
26 Model SLE DBE MCE
27 1000 1000 1000
28
ee
800 800 800

Proposed (kN)
Proposed (kN)
Proposed (kN)
29 600 600 600
30
31
400 400 400
rR
32 200 200 200
33 0
0 200 400 600 800 1000
0
0 200 400 600 800 1000
0
0 200 400 600 800 1000
35
ev
(a) Axial forces (kN)

36 1000 1000 1000
37 Kasai
Kasai-fit
Kasai
Kasai-fit
Kasai
Kasai-fit
Proposed (kN-m)
Proposed (kN-m)
Proposed (kN-m)
800 800 800

38
iew

39 600 N&K
N&K-fit
600 N&K
N&K-fit
600 N&K
N&K-fit
40 400 400 Direct-R

Direct-R-fit
400
41 200 200 200
42 0 0 0
43
0 200 400 600 800 1000 0 200 400 600 800 1000 0 200 400 600 800 1000
44
45
47
48
49 7 Conclusions
50
51 This paper investigated the applicability of the ductility-based reduction factors to estimate the peak
52 substructure response (R-factor approaches) using RSA which forms the basis for the equivalent static
53 forces of the dome. The response of simpler single-storey substructures and the higher mode effects
54 of multistorey substructures were studied and the results from the proposed estimation methods were
55 verified using nonlinear response history analyses (NLRHA). The following conclusions were drawn
56
57
from this investigation:
58
59
1. The RSA based amplification factor approach proposed for elastic structures was extended for
60 practical yielding substructures designed with an R-factor of 8. The accuracy in the peak response
envelope was found to vary depending on the post-yield stiffness of the substructure and the method
adopted to obtain the peak substructure response.
1
2 2. Kasai’s equivalent linearisation method estimated the substructure response with accuracies very
3 similar to the NLRHA for all the studied models. This method used in combination with the
4 proposed amplification factor approach is an efficient alternative to the NLRHA method to estimate
5 a conservative envelope of the peak accelerations and the equivalent static loads for the preliminary
6
7
design of the dome.
8 3. For single-storey and first-mode dominated structures, both the R-factor methods- ‘Newmark’
9 and ‘Nassar&Krawlinkler’ apply simple ductility reduction factor formulations to estimate the
10 peak substructure responses and the accuracies are comparable to those from the NLRHA for
11 structures with low post-yield stiffness ratios (p <= 2%). The Newmark method formulated based
12 on the equal energy rule was found to have the best accuracy among the R-factor approaches for
13
short-period structures with fundamental periods in the constant acceleration region of the design
14
15 acceleration spectrum.
16 4. The accuracies of the R-factor methods in estimating the peak substructure response was low for
17 single and 2-storey structures with high post-yield stiffness ratios (p = 25%) due to the underlying
18 assumptions behind the formulations. Therefore, for structures with high post-yield stiffness ratios
19 (p >= 2%) typically found in countries like Japan with moment connections, it is recommended
20
to adopt the more accurate Kasai method in combination with the amplification factors to obtain
21
Fo
22 the peak envelope roof response. These may then be used to obtain the equivalent static loads on
23 the roof for preliminary member sizing.
24 5. While multistorey structures exhibit significant higher mode response, for substructures with low
rP
25 (first-mode) post-yield stiffness ratios, Newmark’s method in combination with the first and second
26 mode amplification factors provides a conservative roof response with accuracies nearly equal to
27
those of the iterative Kasai method. Therefore, for all structures with low post-yield stiffness ratios
28
ee
29 (p < 10%), it is recommended to adopt the simpler Newmark method in combination with the
30 amplification factors to obtain the envelope of the peak roof response.
31
rR
32 References
33
34 [1] Minimum Design Loads for Buildings and Other Structures. ASCE Standard–ASCE/SEI 7–16:
35
ev
American Society of Civil Engineers (ASCE), 2016.

36
37 [2] NZS 1170.5: 2004, Structural Design Actions Part 5: Earthquake actions-New Zealand.
38
iew
Wellington, New Zealand: Standards New Zealand, 2004.

39
40 [3] Design of Structures for Earthquake Resistance. British Standard–Eurocode 8: Part-1, 2005.
41
42 [4] The Building Standard Law of Japan on CD-ROM. The Building Center of Japan, 2016.
43
44 [5] Guide to Earthquake Response Evaluation of Metal Roof Spatial Structures. Working Group
45 8,International Association for Shell and Spatial Structures (IASS), 2019.
46
47 [6] Nakazawa S, Kato S, Takeuchi T et al. State-of-the-art of seismic response evaluation methods for
48 metal roof spatial structures. Journal of the IASS 2012; 53: 117–130.
49
50 [7] Takeuchi T, Ogawa T and Kumagai T. Seismic response evaluation of lattice shell roofs using
51
amplification factors. Journal of the IASS 2007; 48: 197–210.
52
53
[8] Kato S, Nakazawa S and Saito K. Two-mode based estimation of equivalent seismic loads and static
54
55 estimation of dynamic response of reticular domes supported by ductile substructures. Journal of
56 the IASS 2006; 47: 35–52.
57
58 [9] Kawaguchi M and Kato S. Metal space structures. Journal of the IASS 2001; 42: 21–26.
59
60 [10] Kato S, Ueki T and Mukaiyama Y. Study of dynamic collapse of single layer reticular
domes subjected to earthquake motion and the estimation of statically equivalent seismic forces.
International Journal of Space Structures 1997; 12(3-4): 191–203.
1
2 [11] Yamada S. Vibration behaviour of single-layer latticed cylindrical roofs. International Journal of
3 Space Structures 1997; 12(3-4): 181–190.
4
5 [12] Takeuchi T, Kumagai T, Okayama S et al. Seismic response evaluation of low–to–high rise lattice
6 domes. Proceedings of APCS 2009 2009; : 71–72.
7
8 [13] Nair D, Terazawa Y, Sitler B et al. Seismic response of long-span domes supported by multi-storey
9 substructures. Journal of the IASS 2020; 61: 140–157.
10
11 [14] Takeuchi T, Ogawa T, Nakagawa M et al. Response evaluation of medium-span lattice domes with
12 substructures using response spectrum analysis. In IASS2004 Symposium, Montpellier, 9.
13
14 [15] Takeuchi T, Kumagai T, Shirabe H et al. Seismic response evaluation of lattice roofs supported
15 by multistory substructures. In Shell and Spatial Structures: Structural Architecture - Towards the
16 future looking to the past. IASS Symposium. Venice, Italy, p. 362.
17
18 [16] Nair D, Ichihashi K, Terazawa Y et al. Higher mode effects of multistorey substructures on the
19 seismic response of double-layered steel gridshell domes. Engineering Structures 2021; 243:
20 112677.
21
Fo
22 [17] Takeuchi T, Xue SD, Nakazawa S et al. Recent applications of response control techniques to
23 metal spatial structures. Journal of the IASS 2012; 53(2): 99–110.
24
rP
25 [18] Kumagai T, Takeuchi T, Ogawa T et al. Seismic response evaluation of latticed domes with
26 elasto-plastic substructures using amplification factors. In Proceedings of IASS 2005. Bucharest,
27
Romania, pp. 383–390.
28
ee
29 [19] Kato S and Nakazawa S. Seismic design method to reduce the responses of single layer reticular
30
domes by means of yielding of substructure under severe earthquake motions. In Proceedings of
31
rR
32 IASS 2001. Nagoya, Japan, p. TP077.

33
[20] Kasai K, Ito H and Watanabe A. Peak response prediction rule for a SDOF elasto-plastic system
34
35 based on equivalent linearization technique. Journal of Structural and Construction Engineering
ev
36 (Transactions of AIJ) 2003; 68(571): 53–62.

37
38
[21] Kasai K and Ito H. JSSI manual for building passive control technology. Part. 8 Peak response
iew
39 evaluation and design for elastoplastically damped system. In 13th World Conference on
40 Earthquake Engineering, Vancouver, Canada. Paper. 5060.
41
42 [22] Kaneda K. Endeavors to control the vibration on long span structure. In Proc. of International
43 Symposium on Theory, Design and Realization of Shell and Spatial Structures, Nagoya, Oct., 2001.
44
45 [23] Hosozawa O and Mizutani T. Structural design of Simokita Dome. In Proceedings of IASS 2005.
46 Bucharest, p. 707.
47
48 [24] Takeuchi T, Ogawa T, Suzuki T et al. A basic study on damage-controlled design concept for truss
49 frame structures. Journal of Structural and Construction Engineering (Transactions of AIJ) 2005;
50 51: 31–37.
51
52 [25] Takeuchi T, Tsutsumi T, Ogawa T et al. Seismic retrofit of steel-structure school gymnasia with
53 energy dissipation braces. In Proceedings of IASS 2010. Shanghai.
54
55 [26] Fahnestock LA, Sause R and Ricles JM. Seismic response and performance of buckling-restrained
56 braced frames. Journal of Structural Engineering 2007; 133(9): 1195–1204.
57
58 [27] Güner T and Topkaya C. Performance comparison of BRBFs designed using different response
59 modification factors. Engineering Structures 2020; 225: 111281.
60
[28] Uang CM. Establishing R (or Rw ) and Cd factors for building seismic provisions. Journal of
Structural Engineering 1991; 117(1): 19–28.
1
2 [29] Riddell R, Hidalgo P and Cruz E. Response modification factors for earthquake resistant design of
3 short period buildings. Earthquake Spectra 1989; 5(3): 571–590.
4
5 [30] Riddell R. Inelastic response spectrum: Early history. Earthquake Engineering & Structural
6 Dynamics 2008; 37(8): 1175–1183.
7
8 [31] Nassar A, Osteraas J and Krawinkler H. Seismic design based on strength and ductility demands.
9 In 10th World Conference on Earthquake Engineering, volume 10. pp. 5861–5866.
10
11 [32] Lee LH, Han SW and Oh YH. Determination of ductility factor considering different hysteretic
12 models. Earthquake Engineering & Structural Dynamics 1999; 28(9): 957–977.
13
14 [33] King M, Whitby W and Hanshaw G. Design of the Singapore Sports Hub Roof with high strength
15 niobium steel. In IABSE Symposium Report, volume 101. International Association for Bridge and
16 Structural Engineering, pp. 1–8.
17
18 [34] Takeuchi T and Wada A. Buckling-restrained braces and applications. Japan Society of Seismic
19 Isolation, 2017.
20
21 [35] Analysis reference manual for SAP2000, ETABS, SAFE and CSIBridge. Computers and Structures,
Fo
22 Inc., Berkeley, CA, USA, 2015.

23
24 [36] Wilson EL. Three-dimensional static and dynamic analysis of structures. Computers and
rP
25 Structures, 2002.
26
27 [37] Specification for Structural Steel Buildings (ANSI/AISC 360–16). American Institute of Steel
28 Construction (AISC), 2016.
ee
29
30 [38] Rutten D. Grasshopper3d, 2015.
31
rR
32 [39] Pacific Earthquake Engineering Research Center. PEER ground motion database, 2020. URL
33 https://ngawest2.berkeley.edu/.
34
35 [40] Chandler AM and Duan X. Performance of asymmetric code-designed buildings for serviceability
ev
36 and ultimate limit states. Earthquake Engineering & Structural Dynamics 1997; 26(7): 717–735.
37
38 [41] Erochko J, Christopoulos C, Tremblay R et al. Residual drift response of SMRFs and BRB frames
iew
39 in steel buildings designed according to ASCE 7-05. Journal of Structural Engineering 2011;
40 137(5): 589–599.
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
View publication stats

Proof Researchgate

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Proof Researchgate

Uploaded by

Copyright:

Available Formats

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

Equivalent static loads for double-layered domes supported by multistorey

Article in International Journal of Space Structures · June 2022

Deepshikha Nair Yuki Terazawa

SEE PROFILE SEE PROFILE

The user has requested enhancement of the downloaded file.

2.1 Inelastic accelerations: RSA & R-factor

22 to compare the response from each of the response estimation methods.

22 O1 mode shape [7, 13].

h (m) kN kN kN kN mm kN/m (m) MPa mm2

Model-B (T1 = 0.4s)

(a) Substructure model (b) Roof model

M= moment magnitude, PGA= unscaled peak ground acceleration

GM6 SLE (5%)

Figure 6. Target design spectrum and earthquake levels

14 0.6 0.6 0.6 0.6

15 0.4 0.4 0.4 0.4

16 0.2 0.2 0.2 0.2

19 (a) p=0% (b) p=2% (c) p=10% (d) p=25%

Figure 9. Horizontal amplification factors w.r.t period ratios

Horizontal Acceleration (m/s2 )

Horizontal Acceleration (m/s2 )

Vertical Acceleration (m/s2 )

Vertical Acceleration (m/s2 )

22 (b) Model-B T1 =0.45s

Horizontal Acceleration (m/s2 )

Horizontal Acceleration (m/s2 )

Vertical Acceleration (m/s2 )

Vertical Acceleration (m/s2 )

Horizontal Acceleration (m/s2 )

Horizontal Acceleration (m/s2 )

Vertical Acceleration (m/s2 )

Vertical Acceleration (m/s2 )

59 Ridge coordinates (m) Ridge coordinates (m) Ridge coordinates (m)

Horizontal Acceleration (m/s2 )

Horizontal Acceleration (m/s2 )

Vertical Acceleration (m/s2 )

Vertical Acceleration (m/s2 )

22 (b) Model-B T1 =0.45s

Horizontal Acceleration (m/s2 )

Horizontal Acceleration (m/s2 )

25 8 Newmark Mean NLRHA 8 Newmark

Vertical Acceleration (m/s2 )

Vertical Acceleration (m/s2 )

Horizontal Acceleration (m/s2 )

Horizontal Acceleration (m/s2 )

44 8 Newmark Mean NLRHA 8 Newmark Mean NLRHA 8 Newmark Mean NLRHA

Vertical Acceleration (m/s2 )

Vertical Acceleration (m/s2 )

59 Ridge coordinates (m) Ridge coordinates (m) Ridge coordinates (m)

-25 -20 -15 -10 -5 0 5 10 15 20 25

36 Lumped mass model

Figure 12. 60m Substructure model and modelling assumptions

Target DBE (5%) Target DBE (5%) Roof O1=0.22s

2000 Roof O2=0.11s

(Obtained) Model-B (Obtained) Model-B

0 0.005 0.01 0.015 0.02 0 1 2 3 4

Base shear coefficient Cb

46 1 p =2% 1 p =10% 1 p =25%

48 0.6 0.6 0.6

Sub acceleration Aeq (m/s2)

4 0.8 0.8 0.8 Model-A T1