Download as pdf or txt
Download as pdf or txt
You are on page 1of 19

Engineering Structures 212 (2020) 110511

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Sensitivity analysis of double layer Diamatic dome space structure collapse T


behavior

R. Vaezi Vaznaa, M. Zarrinb,
a
Department of Civil Engineering, Sahand University of Technology, Tabriz, Iran
b
Department of Civil Engineering, K. N. Toosi University of Technology, Tehran, Iran

A R T I C LE I N FO A B S T R A C T

Keywords: Collapse behavior of double layer dome space structures depends on different factors such as load pattern,
Epistemic Uncertainty material mechanical properties, and member’s physical characteristics, which have a random nature that makes
Probabilistic Investigation it difficult to predict accurate behavior of this type of structure. In this paper, in order to comprehensively assess
Tornado Diagram Analysis the double-layer dome space structure behavior, different uncertain parameters such as the mechanical char-
FOSM method
acteristics of steel material, mass and applied gravity load, and also structure’s and members’ geometric im-
Snow Load
perfection are taken into account in nonlinear static pushover analysis. Different models with rise to span (R/S)
ratios of 0.1, 0.2, and 0.3 under incremental symmetric and asymmetric snow loads are investigated. In order to
determine the importance of these uncertain parameters and rank them according to their effects on the
structural response parameters, two different methods of Tornado Diagram Analysis (TDA) and First-Order-
Second-Moment (FOSM) are utilized. Using these methods enables the sensitivity analysis of the collapse be-
havior of the structures to the considered random variables. The results of the analyses reveal that there is a close
agreement between the TDA and FOSM analyses results in ranking the random variables according to their
importance. In addition, these methods highlight the high sensitivity of the structural response parameters to the
various sources of uncertainty, and indicate the necessity of applying these uncertainties in the analysis and
design of this type of structure.

1. Introduction is very difficult even in well-fabricated steel construction. Thus, there


will be a great potential for a space structure to be built with numerous
One of the common types of curved shape latticed structures is the imperfections. Geometrical imperfection such as initial curvature of the
dome space structure. This type of structure is successfully adopted for members, node position deviation, and also mechanical imperfection
specialized utilizations, such as the 21 distant Early Warning Line dome such as initial stress or strain due to rolling and welding, and lack of fit
built in Canada in 1956 and the 1958 Union Tank Car Company dome [2,3] can be considered as the most common types of imperfection in
near Baton Rouge [1]. Nowadays, double-layer dome space structures space structures. The effects of these kinds of imperfection have been
are extensively utilized in the construction industry. Double-Layer widely investigated by the researchers both probabilistically and de-
space trusses are efficient for covering large areas. They are usually terministically. Borri and Spinelli [4] investigated the effect of two
more redundant than the typical single-layer space structures, and may types of joint (eccentric axial force in the members) and surface geo-
remain stable once a member fails. Therefore, the main focus of this metry (or schemes) random imperfection on buckling and post-buckling
study is on the double layer dome structural behavior evaluation. behavior of single-layer reticulated shell. They concluded that the
Accurate prediction of structural response is a very problematic structure’s behavior is greatly sensitive to the structure’s out-of-scheme
issue, since there are various sources of uncertainty in the modeling and imperfection. Analysis of the ultimate limit state of a single-layer re-
loading parameters. It is necessary to have a clear understanding of how ticulated vault structure has been conducted by Tegola [5] with con-
these uncertainties affect the performance of the structures. Obtaining a sideration of the non-rectilinearity of geometric axis of the bars and also
predefined sufficient degree of accuracy during the manufacturing variation of diameter with respect to its mean value as the two types of
process of a space frame, and producing its members perfectly straight geometric imperfection. The study conducted by El Sheikh [6] on


Corresponding author. Cell: +98 9127903571.
E-mail address: mo_zarrin@yahoo.com (M. Zarrin).
URL: https://www.researchgate.net/profile/Mohamad_Zarrin/research (M. Zarrin).

https://doi.org/10.1016/j.engstruct.2020.110511
Received 20 June 2019; Received in revised form 26 January 2020; Accepted 9 March 2020
0141-0296/ © 2020 Elsevier Ltd. All rights reserved.
R.V. Vazna and M. Zarrin Engineering Structures 212 (2020) 110511

collapse capacity of 52 single-layer barrel vaults with 13 different kinds uncertainties on the structural response. In this procedure, the struc-
of configuration revealed that the 0.001 length amount of member’s tural response parameters corresponding to the upper and lower bounds
lack of fit and 100 mm joint misplacement in the central nodes of the of the random variables are estimated. The interval of these response
vaults result in notable capacity reduction. Yadollhi et al. [7] have demands, referred to as swing, is a sensitivity measurement. Sensitivity
carried out thirty analysis cases to estimate the load capacity of a analysis makes it possible to identify the most important uncertain
double-layer barrel vault by considering the initial curvature of the variable of the structure. The topic of sensitivity analysis of steel and
critical members as the imperfection in the structure. This study re- concrete structures against different types of load patterns has been
ported that the structure's capacity decreases by 41 percent due to the studied by different researchers. Hezarjaribi et al. [18] has conducted
member’s initial curvature imperfection. Rudsari and Gordini [8] have sensitivity analyses of jacket type platforms against extreme wave loads
examined the effect of initial curvature of the members on a double- to recognize the most important parameters affecting nonlinear re-
layer grid space structure’s behavior. They deduced that this structure sponse of jacket platform. In a study conducted by Haselton [19] on a 4-
with different types of support conditions is completely sensitive to the story reinforced concrete frame, the result of sensitivity analysis has
member’s initial curvature imperfection. Xin Chen and Shi-Zhao [9] been used to compute the relationship between the model random
have used random and conformable imperfection mode method to in- variables and the structural response for the derivation of the gradient
vestigate the effect of structure geometric imperfection (nodal devia- needed in the FOSM calculation. However, this method has been rarely
tion) on single-layer dome collapse behavior. They came to the con- used by other scholars on space frame structures so far. Hui Jun et al.
clusion that even a small amount of imperfection might hugely impact [20] performed sensitivity analyses on double-layer cylindrical re-
the structure’s behavior in single layer dome structure. Feng Fan et al. ticulated shells to evaluate the effect of random variable’s variation on
[10] has compared the buckling behavior of seven types of commonly structures’ response. Zhong et al. [21] performed a sensitivity analysis
used single-layer reticulated domes via applying various buckling mode on single-layer reticulated domes to examine the sensitivity of typical
shapes as the structure geometric imperfection to the perfect model. seismic responses to the uncertainties in structural modeling para-
In addition to the geometrical imperfection uncertainty, other cru- meters. They concluded that the variability in structural damping,
cial sources of uncertainty are in regard to the element’s mechanical yielding strength, steel ultimate strain, dead load, and snow load has
and material properties, and also the value of applied gravity load. The significant effects on the seismic responses.
effects of these sources of uncertainty on various types of steel struc- FOSM method is another approach to determine the structure’s
tures are widely investigated. The steel frame structures collapse be- stochastic response considering random input variables. The FOSM
havior has been early recognized to be highly sensitive to the afore- approach is derived from mathematical calculations, which uses a first-
mentioned parameters. Asgarian and Ordoubadi [11] investigated the order Taylor series along with the first and second moments of the in-
effect of structural uncertainties on the seismic performance of steel puts variables [22]. Through using this method, it would be possible to
moment frames. Jalali et al. [12] have used a number of modeling rank random variables according to their importance, and also to cal-
parameters that best reflected the structures’ variability due to the lack culate the standard deviation of the limit state functions. Hezarjaribi
of knowledge for performing the modeling uncertainty analysis on steel et al. [18] have used FOSM method in order to estimate the mean and
moment frames with a generic locally reinforced connection. Also, standard deviation of jacket structure response. Nour El-Din and Kim
Pirmoz and Marefat [13] worked on the reliability assessment of a [23] have studied the seismic sensitivity of a fixed type steel offshore
compression column in seismic EBFs by considering uncertainty in platform. They have used FOSM method to investigate the effect of each
gravity dead and live load. In another research, Zarrin et al. [14] random variable on some demand parameters. Haselton [19] examined
evaluated the variability of seismic demand and capacity of a case study the effects of modeling uncertainties on the collapse capacity of re-
jacket platform considering different sources of uncertainty. To this inforced concrete frames using the FOSM reliability method.
end, the sensitivity analysis was conducted not only in the particular The main objective and goals of this paper are: firstly, a holistic
earthquake level, but also in other ranges of intensity using Incremental assessment of the effect of some key factors such as rise to span (R/S)
Dynamic Analysis (IDA) [15]. Recently, the probabilistic performance ratio, different vertical snow load patterns, and structure geometric
of four Steel Moment Resisting Frames (SMRF) were investigated in imperfection (node deviation) on collapse behavior of double-layer
[16] using the IDA method, and the stochastic seismic performance of dome space structures. The main motivation for this assessment is that
one of them (5 story frame) was investigated considering the aleatoric the simultaneous influence of these factors has not sufficiently been
and epistemic uncertainties. Furthermore, the effects of the mass and examined in double-layer dome space structures. Secondly, the sensi-
gravity load uncertainties were pursued in [11,14,16]. These studies tivity analysis of collapse behavior of double-layer dome structure to
showed that the material’s mechanical properties and the value of ap- various sources of uncertainty via using both TDA and FOSM methods
plied gravity load intensely affect the structure’s behavior. Therefore, to evaluate the effect of each random variable on the nonlinear static
considering these sources of uncertainty in the snow-load capacity response of the structure. The main purpose for this sensitivity analysis
evaluation of the dome space structure stands at the core of this study. is that according to the previous probabilistic and deterministic studies
It can be concluded from the above-reviewed research studies that in the scope of space structures, it appears that in most of the studies,
the prediction of the response of real-life structural systems is inevitably the effect of just members’ geometric imperfection (initial curvature
associated with the aforementioned uncertain parameters. Given that and lack of fit), and also structures’ geometric imperfection (node de-
space structures are usually used to cover the places with a large viation), mostly in single-layer space structures, have been assessed. In
number of people such as stadiums, halls, etc., therefore their collapse addition, the effect of other sources of uncertainty in material me-
imposes a drastic loss of life and assets to the community. In order to chanical properties, gravity load, and also ranking of the random
prevent the occurrence of these kind of damages, it is necessary that the variables according to their importance by using TDA method has not
space structure's response sensitivity to be assessed comprehensively. been adequately addressed in previous research works. Furthermore,
Thus, for finding out the most important uncertain parameters which calculation of standard deviation of limit state function and importance
play a pivotal role in these structures’ behavior, a holistic sensitivity of uncertain parameters through the utilization of the FOSM method
analysis with the consideration of various sources of uncertainties have have not yet been performed on space structures. Finally, the results of
to be done. There are several methods to conduct sensitivity analysis the TDA and FOSM methods are compared in order to gain a good
and determine the significance of the uncertain parameters. Two of the judgment about the accuracy of TDA method for recognizing the un-
most well-known methods are called Tornado Diagram Analysis (TDA) certain parameters that play a pivotal role in double-layer dome space
[17] and First-Order-Second-Moment (FOSM) method. TDA is a tech- structures’ collapse behavior. To this end, in this paper the collapse
nique that is used to address the effect of different sources of behavior of the double-layer dome space structure with three different

2
R.V. Vazna and M. Zarrin Engineering Structures 212 (2020) 110511

R/S ratios of 0.1, 0.2, and 0.3 under both symmetric and asymmetric Another common type of geometric imperfection in space structures
snow loads, in two different geometrically perfect and imperfect states is the initial curvature of the members. In order to accurately estimate
(conditions) are examined in order to comprehensively investigate the the behavior of the structure, it is mandatory to consider this type of
effects of the aforementioned key factors. imperfection in the modeling procedure. Several investigations have
been deterministically and probabilistically carried out to assess the
2. Imperfection in space structures effects of this type of imperfection [5,7,8,10]. The Gamma distribution
function has been utilized in [5,7,8] for the initial curvature of the
The two important and influential types of imperfection in the do- members in probabilistic assessments of the single and double layer
main of construction of space structures are the nodal deviation space structure. On the other hand, for evaluating the behavior of a
(structure’s geometric imperfection) and the member’s initial curvature. single-layer dome space structure with imperfect members, Feng fan
The effect of nodal deviation on double-layer space structures has been et al. [10] has assumed that the initial curvature of the member obeys
scarcely assessed. For this type of imperfection to be allocated to the an extreme I distribution function. In that study, it was also assumed
designed dome space structure, different methods are currently avail- that the maximum amplitude of the initial curvature is 0.001 of the
able. Some scholars categorized these methods into three cases [24,25]. member’s length with the exceedance probability of 2.5 percent. From
The first method is called the random imperfection mode method, the the cited papers, it can be inferred that there is not any specific rule to
second is called the Consistent Imprecation Mode Method (CIMM), and determine the appropriate type of distribution function for probabilistic
the last one is the Eigen-mode Imperfection Method (EIM). Using the modeling of initial curvature of the members. Consequently, in order to
first aforementioned method is time-consuming and computational assess the sensitivity of the response to the probability distribution
expense, while most of the scholars reported that the EIM method is function itself, it is intended to use two different types of Gamma and
able to predict the critical load-bearing capacity of the structure accu- Extreme distribution function for modeling the members’ initial cur-
rately [24,26,27]. vature. The probability density function and probability distribution
According to the [4,9], it is noted that the maximum capacity of a function of δ in the Gamma functions are:
dome space structure depends not only on the size of the imperfection
1 δ
but also, to a greater degree, on the pattern of the imperfection dis- f (δ ) = k
x k − 1e− θ
tribution. By all means, there are two vital issues that should be paid
Γ(k ) θ (1)
more attention when EIM method is being used. First and foremost, in
order to model the initial geometric imperfections of a double layer 1 δ
F (δ ) = γ (k . )
dome space structure in nonlinear analysis, it is necessary to choose a Γ(k ) θ (2)
proper buckling mode. As will be discussed further ahead, the proper δ
where kandθ are the Gamma function’s parameters and γ (k . is the
θ
)
buckling mode could be considered as the mode resulting in the lowest
lower incomplete Gamma function.
buckling load of the structure. Secondly, an appropriate amplitude for
The probability density function and probability distribution func-
the chosen buckling mode shall be determined to apply to the original
tion of δ in the Extreme functions are:
structure configuration. Buckling mode is the deflection tendency or the
mode of deflection increment of a structure at the critical point. The f (δ ) = υ ∗eυ (δ − q) − e
υ (δ − q)
(3)
buckling mode corresponding to the first critical point is called the
lowest gradient buckling mode. The major principle of the EIM method υ (δ − q)
presented here is to model the initial imperfection of the structure with F (δ ) = 1 − e−e (4)
the lowest gradient buckling mode. Some research works on space
where the υ and q are the parameters of the Extreme function. In order
structures [24,25] have shown that in the EIM method, using the first
to determine these parameters, it is assumed that the maximum am-
eigenmode for considering the nodal deviation, in most cases, results in
plitude of the initial curvature δmax is 1/1000 of the members’ length;
lowest buckling load of the structure. Mohammadi et al. [27] deduced
the exceedance probability is 2.5 percent, and the probability of the
that the first buckling mode of a single-layer barrel vault eventuates the
δ = 0 is 1 percent [10].
lowest capacity of the structure. Zhi-Hua and Yang Li [28] have used
the lowest linear buckling mode in order to modify the nodal coordinate
of the suspended dome for making the structure imperfect. 3. Uncertain variables
The imperfection amplitude is considered as the maximum devia-
tion from the perfect geometry along the direction normal to a re- The main sources of uncertainty modeled in this paper consist of
ference surface. In the previous researches, different values have been two different categories. The first category is element-level uncertainty
chosen for the magnitude of deviation that is scaled with respect to the associated with the member’s geometric imperfection and the material’s
span length (L) of the structure. Gengbo Chen et al. [29] assessed the mechanical properties. The second source of uncertainty is associated
effect of geometric imperfection on single-layer reticulated shell with the structural level uncertainty including value of mass and ap-
structure and found that the value of L/700 for initial geometric im- plied gravity load (MaL) on the structure, and the node deviation un-
perfection appears to be more reasonable. Mohhamadi et al. [27] sug- certainty discussed in the previous section. Member’s geometric im-
gested that the difference between limit point and the bifurcation point perfection uncertainties include initial curvature of the member (δ ),
load level is a good criterion for geometric imperfection and re- thickness (t ) and diameter of member’s section (D ). Uncertainties as-
commended L/500 for the amplitude. Through holistic investigation on sociated with material include yield stress (Fy ), ultimate stress (Fu ) and
various types of dome space structures, Feng Fan et al. [24] reported Young’s modulus (Es ). Eventually, for comprehensive probabilistic as-
that applying the maximum initial geometrical imperfection of L/300, sessment of the double layer dome space structure, seven random
the value of influence coefficient would be minimum. According to the variables are taken into account. Table 1 represents quantitative in-
specification of the Chinese lattice shell code [26], the maximum geo- formation regarding the studied uncertainties, including the mean or
metric imperfection that is caused during the construction stage should median values of the input random variables, their Coefficient of Var-
be restricted within span/300 ratio. In this paper, according to the goal iation (COV), and the type of distribution along with the references
of investigating the effects of geometrical construction errors on used to quantify each of the uncertainties. It has to be noted that the
structural behavior, it is considered to follow the Chinese code for correlation coefficient of 0.75 is considered between Fy and Fu para-
obtaining the worst case of imperfection mode. Consequently, the value meters [25], and other random variables are assumed to be non-cor-
of L/300 is considered for the amplitude of geometric imperfection. related.

3
R.V. Vazna and M. Zarrin Engineering Structures 212 (2020) 110511

Table 1
The statistical characteristics of random variables.
Random variable Symbol Mean or median COV Type Reference

Vertical load and mass M aL Computed 0.10 Normal [30]


Yielding stress fy 240MPa 0.07 Lognormal [31]
Ultimate stress fu 370MPa 0.04 Lognormal [31]
Modulus of elasticity Es 2 ∗ 105MPa 0.05 Normal [32]
Member diameter D Computed 0.02 Normal [21,33,64]
Member thickness t Computed 0.064 Normal [32]
Members initial imperfection ImpG 0.001 ∗ memberlength 0.8164 Gamma [5,7,8]
Members initial imperfection ImpE 0.001 ∗ memberlength 0.24 Extreme [10]

4. Design and modeling of the dome (0.143). They have found that the collapse behavior of the structures is
directly related to the R/S ratio of the dome, and also concluded that
4.1. Configuration of the dome the generalized conformable imperfection mode method in the domes
with the R/S ratios of 1/4 and 1/5 is necessary to be considered. In
Double layer Diamatic dome can be described as a pie-shaped sector order to optimize the aerodynamic performance of single-layer re-
repeated radially around the crown of a dome [34,35]. Herein, the joint ticulated cylindrical shells, San et al. [40] evaluated the behavior of
co-ordinates and member’s connectivity data required for generating these structures with different values of R/S ratio, and deduced that the
the structure configuration is obtained through configuration proces- cylindrical shells with R/S ratio in-between 1/6.25 and 1/5.5 are cap-
sing capabilities of formex algebra [36]. Fig. 1 shows the characteristics able of keeping good aerodynamic and mechanical performance. In this
of the double layer Diamatic dome space structure, which is in- study, to evaluate the sensitivity of the double-layer dome space
vestigated in this study. In Fig. 1 (a), [T] represents top layer elements, structure response to different values of R/S ratio, models are designed
[B] represents bottom layer elements, and [W] shows web members of with R/S ratio of 0.1, 0.2, and 0.3.
the dome. Fig. 1(b) shows the span, height (rise) and thickness of the
double-layer dome space structure. 4.3. Different load patterns

4.2. Rise to span ratio of the models 4.3.1. Symmetric loads


Symmetric loads of the dome space structure include the dead load
The R/S ratio is an important factor that influences the design of (the load derived from cladding weight) and the live load. The dead
space structures from various aspects. The intensity of applied dead load is established on the basis of the actual load which may be ex-
load, symmetric and asymmetric snow loads, and also wind load are pected to act on the structure constantly. Dead load mainly consists of
greatly sensitive to the R/S ratio variations. Previous studies have self-weight of the steel structure, supporting purlin, roof slab, ceiling
shown that increasing the R/S ratio from approximately 0.1 to 0.3, not board, hoisting equipment and berm. Live load, either snow or rain
only affects element’s design sections, but it also might changes the load, is specified by the ASCE7-10 [41] code. The larger value of the
collapse behavior of the structure. Chen [37] has used wind tunnel to snow and rain load should be used in design. Each space frame is de-
examine the effects of wind pressure on various types of the roof system signed with a uniformly distributed snow load and further allowed for
with R/S ratios from 0.0625 to 0.5. He found that the low R/S ratio roof drifting depending upon the shape and slope of the structure. Both the
building is suffered by the instantly peak pressures and this factor dead and snow loads are applied as a distributed load on the outer
changes the magnitude of design wind pressure. For evaluating buck- surface of the space structure.
ling behavior of different types of dome space structures, Feng fan et al.
[24] modeled a large numbers of space structures with various values of 4.3.2. Asymmetric snow load
R/S ratio ranging from 1/5 (0.2) to 1/8 (0.125). They concluded that In some cases, unbalanced snow load might pose a greater risk of
the buckling of space dome structures is very sensitive to the R/S ratio. failure to the roof structural systems than a uniform snow load. Some
To examine the geometric imperfection sensitivity of single-layer scholars have assessed the effect of different load patterns on space
double-dome free form space structure, Mousavi et al. [38] have used structure’s behavior. Gallego et al. [42] experimentally and theoreti-
models with different R/S ratio values between 1/4 (0.25) and 1/7 cally worked on the effect of different snow load patterns on steel silo

ss
[T] k ne
Thic
Height

[W]

Span

[B]

(a) (b)
Fig. 1. (a) Plan view of a double layer Diamatic dome, (b) Characteristics of the dome [39]

4
R.V. Vazna and M. Zarrin Engineering Structures 212 (2020) 110511

Fig. 2. (a) Description of the numerical model (b) Experimental behavior [55] (c) Numerical behavior.

roof. They compared the normal stresses of the members and the dis- represent the moderate to large deformation effects of inelastic buck-
placement produced in a conical steel silo roof structure by applying ling of the elements. The corotational formulation separates rigid body
symmetric and asymmetric load. They found that the asymmetric load modes from local deformation using, as reference, a single coordinate
pattern produces higher normal stresses and displacements than the system that continuously translates and rotates with the element as the
balanced loads. Kato and Iwamoto [43] examined the buckling beha- deformation proceeds [48]. The fiber discretization approach has been
vior of a grid single layer dome with diagonal braces under uniform and utilized for modeling the cross section of each segment. The assumed
non-uniform snow load. In order to apply the non-uniform load, they material for the fibers is considered uniaxially, and the strain in each
subjected half of the structure to the snow load. They reported that the fiber is calculated based on the strain at centroid strain and curvatures
diagonal braces impact the buckling behavior of the dome under dead at the sections, considering Bernoulli’s assumption that plane sections
and non-uniform load. Cai et al. [44] have used different load patterns remain plane and normal to element axis after bending. The assigned
including symmetric and asymmetric load to carry out nonlinear sta- stress-strain relationship for the steel material of the members is the
bility analysis of single layer hybrid grid shell. Menegotto-Pinto model [49]. One of the advantages of this type of
material is the gradual transition from linear region to nonlinear range
4.3.3. Procedure of applying asymmetric snow load of response. Furthermore, the cyclic behavior of this material is fairly
According to the ASCE7-10 [41], unbalanced snow load shall be consistent with experimental tests including Bauschinger effects. The
applied to the downwind 90° sector in plan view. At both edges of this Fatigue uniaxial material command in OpenSees has been used to
sector, the load shall decrease linearly to zero over the sector of 22.5° consider the ultimate stress and strain. The material has the ability to
each. There shall be no snow load on the remaining 22.5° upwind trigger failure based on a maximum or minimum strain. In order to
sector. There are various factors that affect the value of the snow load include global buckling in the brace response, it is modeled with eight
such as snowfall intensity, redistribution of snow load by wind, the force-based elements, and an initial camber equal to 1/1000 of element
geometry of buildings and all surrounding buildings, and etc. In this length is applied to the mid node of the element (see Fig. 2a). Utilizing
paper, symmetric and asymmetric snow loads are calculated according this technique along with using co-rotational formulation for co-
to ASCE7-10 code [41] of structural design. These symmetric and ordinate transformation [48], buckling and post-buckling behavior of
asymmetric snow loads are also used for evaluating the effect of dif- braces can be taken into consideration accurately.
ferent load patterns on structure’s behavior. More details about the The efficiency of the force-based beam column element, for the
procedure of calculating symmetric and asymmetric snow loads are tubular structural components modeling of the steel members, has been
presented in Appendix. verified in some previous studies. Asgarian et al. [50] modeled the post
buckling and also hysteretic behavior of pinned and fixed tubular struts,
4.4. Structure design which were subjected to cyclic loading, by using the forced based fiber
beam column element. In addition to this, the results were verified by a
The design of the archetype domes is carried out in compliance with comparison with experimental tests. Asgarian et al. [51] and Honarvar
the provisions in ASCE 7-10 [41] and AISC 360-10 (LRFD) [45] codes. et al. [52] used the same element in their studies to predict the seismic
As is highlighted in the previous section, the design loads for the dome- response of experimentally tested X-braced jackets. Moreover, Alanjari
shaped structure are computed based on ASCE 7 load combinations et al. [53] and Uriz et al. [54] worked on the seismic performance of
[41]. In addition to the aforementioned snow and dead loads, the nu- individual tubular strut tested by black et al. [55] upon buckling in
meric values of wind and thermal load, the earthquake in both × and y compression, as well as their capability in dissipating energy in con-
directions are also applied during the design phase. Member sections secutive cycles utilizing the same element as the case here. Previous
are selected from standard pipe sections constructed from st37 steel studies using a Fiber-type beam–column element models [56,57] have
material. The Young modulus, the Yield stress, and the Poison ratio are verified the benefits of using this type of brace model in the simulation
assumed to be 2.1 × 105 a , 240MPa and 0.3, respectively. The char- of special concentrically braced frames. The same element is recently
acteristics of all design member sections are reported in the Appendix. used by Yang et al. [58] to investigate the Seismic response and failure
mechanism of single-layer latticed domes. All the aforementioned stu-
dies confirmed the efficiency of force-based beam column element, in
4.5. Numerical modeling
case of modeling the structural components.
In order to evaluate the accuracy and validity of the finite element
A numerical model of the aforementioned dome space structure has
modeling procedure, the experimental and numerical strut model tested
been developed using OpenSees software [46]. A very well formulated
by Black et al. [55] have been assessed in this study. A fiber brace
nonlinear distributed plasticity beam-column element model within the
model was constructed in OpenSees using the fiber-based, beam-
OpenSees software has been used to model the dome structure ele-
column model to mimic this previously-tested strut [55]. The strut se-
ments. The element is based on the force formulation [47], a method
lected consisted of an extra-strong 4-in. diameter steel pipe. The strut
that introduces profound advantages over the very common displace-
was welded to physical clevises at both ends to represent pin connec-
ment beam-column element formulation. This method has been proven
tions, and the distance between the centers of the clevises was 118.4 in.,
numerically reliable, even in the case of softening behavior of buckled
giving the member a slenderness (kl/r) ratio of 80. The numerical
elements. The inelastic beam-column element with small deformation
model consists of eight force-based elements with the consideration of
theory in the present study relies on the co-rotational theory to

5
R.V. Vazna and M. Zarrin Engineering Structures 212 (2020) 110511

the non-rectilinearity of the strut member as shown in Fig. 2(a). F


n
Fig. 2(b) and (c) respectively show the experimentally and numerically 1 ∂g 1
obtained axial force–axial displacement hysteresis loops for this spe- = g (μ X1 . μ X2 . ⋯...μ Xn ) + ∑ (Xi − μ X1 ) +
1! i=1 ∂Xi 2!
cimen. From Fig. 2, it is obvious that there is an excellent agreement n n
∂ 2g
between the results of the tests and force-based brace model. ∑ ∑ (Xi − μ X1)(Xj − μ Xj )
i=1 j=1 ∂Xi ∂Xj (6)
5. Probabilistic assessment methods Considering only the first order terms of equation (6) and ignoring
higher order terms, F can be approximated as:
5.1. Procedure for probabilistic evaluation
n
1 ∂g
F ≈ g (μ X1 . μ X2 . ⋯⋯μ Xn ) +
1!
∑ (Xi − μ X1 ) ∂X
Consideration of different sources of uncertainties is accounted via i=1 i (7)
nonlinear static pushover analysis, in which both geometrically and
materially nonlinear behavior are considered, and the structure is By taking expectation from both sides, the mean of F , μF , can be
subjected to incrementally-increasing snow load. A displacement-con- expressed as:
trol analysis procedure is adopted for pushover analysis, in which the μF ≈ g (μ X1 . μ X2 . ⋯..μ Xn ) (8)
largest increment of the displacement is 0.00001 m, which resolves
some convergence problems. To obtain the response curve, the amount Utilizing the second moment of F as expressed in (7) and after some
of Load Proportionality Factor (LPF), which is increased in each step, is algebraic calculations, standard deviation of F can be derived as:
plotted versus the dome apex point vertical displacement. In order to n 2 n
∑ σX2i ⎛ ∂g (X1. X2 . ⋯..Xn ) ⎞ ∑ ∑ j≠1 ρ Xi Xj ⎛ ∂g (X1. X2 . ⋯..Xn) ⎞
n
examine the structure’s behavior more accurately, three different re- σF2 ≈ ⎜ ⎟ + ⎜ ⎟

i=1 ⎝ ∂Xi ⎠ i=1 ⎝ ∂Xi ⎠


sponse quantities of the structure have been considered to be in-
vestigated via sensitivity analyses. These quantities are maximum value ⎛ ∂g (X1 . X2 . ⋯..Xn ) ⎞
⎜ ⎟
of LPF, vertical displacement of the dome apex point corresponding to ∂Xj (9)
⎝ ⎠
the max LPF, and the LPF value when apex point reaches to 0.05 m (in
the 0.2 and 0.3 R/S ratio structures) and 0.1 m (in the 0.1 R/S ratio where ρ Xi Xj is the correlation coefficient between two random variables
structure). The values of 0.05 m and 0.1 m correspond to the point Xi and Xj , and ∂g (X ) is the gradient of the F with respect to the random
∂xi
where the mean parameter model (base case structural model) ap- variablei , which is approximately calculated by using a finite difference
proximately loses 50 percent of its maximum capacity (referred here- approach. Herein, for perturbing the g function, two extreme values
after to “residual capacity”). corresponding to 5th and 95th percentiles have been considered [14]. It
should be noted that in order to calculate the numeric value of ∂g (X ) ,
∂xi
5.2. Tornado diagram analysis (TDA) the subtraction of the upper and lower percentiles of the Xi random
variable is calculated and denoted as h. Moreover, the subtraction of the
In the present study, the TDA method is adopted to identify the calculated value of g function corresponding to these percentiles of Xi
relative importance of the uncertain model parameters by means of random variable is supposed to be Δg . Therefore, the numeric value of
incremental nonlinear static analyses. As stated before, maximum value ∂g (X )
is obtained using the following equation:
∂xi
of the LPF, vertical displacement of the dome apex point, and the value
of LPF corresponding to 0.05 m (0.1 m in 0.1 R/S ratio model) dome ∂g (X ) Δg
= lim
apex point displacement are utilized as three different types of response ∂x i h→0 h (10)
parameters (limit states) in sensitivity analysis. TDA is a common tool
used to depict the sensitivity of a response parameter to changes in the
6. Results of the analyses
selected model random variables. In the TDA method, in order to de-
termine the prominent uncertain parameters affecting these limit states,
6.1. The effect of models’ R/S ratio on the value of design load
each random variable is perturbed individually, while all the others are
kept at their mean (best estimate) value. For perturbing a specific
The total magnitude of different types of load depends on the dome's
random variable, two extreme values corresponding to 5th and 95th
configuration characteristics such as the value of R/S ratio. The total
percentiles of its probability distribution are assigned to that random
dead load excluding the self-weight is applied to the outer surface of the
variable in the model, and the structure’s response is extracted each
dome as a gravitational load, thereupon, the more outer surface area
time. This procedure is repeated for all the random variables. The re-
increases, the greater the total applied dead load will be. On the other
sults are then displayed as a special type of bar graph, with bar for each
hand, as highlighted in the Appendix, the entire value of symmetric and
input random variable displaying the variation of the response para-
asymmetric snow load decreases by increasing the R/S ratio. Therefore,
meter from the nominal value. Finally, a tornado diagram can be uti-
it can be concluded that the designed dome space structure’s self-weight
lized to demonstrate the sensitivity of demand parameters to each
is utterly tied to its configuration, and the model with lower self-weight
random variable [17].
can be considered as an optimum designed model. It is noteworthy that
from the economical point of view, the final construction cost of the
5.3. First order second moment method (FOSM) dome space structure is highly dependent to the self-weight of the
structure, which is the total weight of the steel used in the dome space
In this method, the target function, g, which relates the different structure. Therefore, a lower self-weight value is preferable for an op-
random variables in demand estimation is written as: timum design.
F = g (X1 . X2 . ⋯...Xn ) Tabulated in Table 2 are the total values of self-weight, dead load,
(5)
symmetric and asymmetric snow loads of the models. According to this
where X1 . X2 . ⋯..Xn are random variables and F denotes the demand table, the dome with the R/S ratio of 0.1 has the lowest dead load value.
parameters. Tylor series expansion of the response function g around On the other hand, the 0.3 R/S ratio dome has the lowest symmetric
random variables mean value need to be derived in order to determine and asymmetric snow loads. However, this table illustrates that the
the mean and standard deviation of response function. Therefore, the domes with the R/S ratios of 0.2 and 0.1 have an almost equal dead
following can be employed: load value. The variability of the total asymmetric snow load values in

6
R.V. Vazna and M. Zarrin Engineering Structures 212 (2020) 110511

Table 2
Design outputs of the models.
R/S ratio Models’ self- Self-weight Design loads
weight (KN) difference
(percent) Dead load Dead loaddifference Symmetric snow Symmetric snow Asymmetric snow Asymmetric snow
(KN) (percent) (KN) difference (percent) (KN) difference (percent)

0.1 78.49 37.11 214.20 −3.77 439.02 52.2 248.53 61.97


0.2 57.24 ––––– 222.60 ––––– 288.46 ––––– 153.44 –––––
0.3 63.70 11.27 262.65 17.99 264.52 −8.29 139.60 −9.01

accordance to the R/S ratios is more than the symmetric snow load under asymmetric snow load. In addition, the values of LPF in different
values in this table. As stated in the appendix, the calculation of models are very close to each other under the symmetric snow load.
asymmetric snow load totally depends on the slope at eaves of the However, the value of LPF slightly increases by rising the R/S ratio. The
domes, and this is directly dependent on the configuration (R/S ratio) of reason behind this lies in the optimum [59] design procedure of these
the dome. Moreover, the values of symmetric and asymmetric snow structures. The models were designed in such a way to return the same
loads are approximately close to each other in the domes with the R/S elements’ stress ratio in different parts of the dome such as the top,
ratios of 0.2 and 0.3. These facts along with the geometric configuration bottom and web layers in the three different R/S ratio domes
of the dome with the R/S ratio of 0.2 make it to has a lower self-weight For investigating the effect of structure geometric imperfection
than the other two domes. The relative difference between self-weight, (nodal deviation) of the double-layer dome space structure, the re-
dead load, symmetric and asymmetric snow loads of the domes with the sponse behavior of the perfect and imperfect structure should be
R/S ratio of 0.1 and 0.3 with respect to the dome with R/S ration of 0.2 compared. Previous researches have shown that the geometric im-
are also tabulated in Table 2. The symmetric and asymmetric snow perfection could has remarkable effect on maximum capacity of the
loads of the model with 0.3 R/S ratio are approximately 9 percent lower single-layer space structures. Mousavi et al. [38] have shown that im-
than the model with 0.2 R/S ratio, while its dead load is 18 percent plementing the buckling mode of the single layer double-dome free
more than the dome with R/S ratio equal to 0.2. This fact also leads to form space structure as the geometrical imperfection can strikingly
the lower self-weight of the 0.2 R/S ratio model compared to the model reduce structure capacity. Cai et al. [44] imposed imperfection in the
with R/S ratio of 0.3. Furthermore, the symmetric and asymmetric positive and negative vertical direction (+Z and –Z direction) on the
snow loads of the dome with the R/S ratio of 0.1 are significantly higher perfect hybrid single-layer space structure, and deduced that applying a
than the other two models, which leads to a much more self-weight of small value of imperfection in these structures may lead to a remarkable
this model. All in all, regarding the structures’ self-weight, the 0.2 R/S change of the failure load.
ratio dome is the most optimum configuration. To evaluate the maximum capacity of the domes, the total absolute
value of snow load presented in Table 3 is pursued. This table shows
6.2. Result of nonlinear static analyses that although the designed domes under asymmetric snow load shows a
larger value of maximum LPF than the domes under symmetric snow
In order to evaluate the nonlinear collapse behavior of the double- load, the absolute capacity value of the models decreases when they are
layer dome space trusses and having an estimation of different response subjected to the asymmetric snow loads. The differences in perfect
parameters such as maximum capacity, collapse displacement, residual models' absolute value of capacity under symmetric and asymmetric
capacity, and initial stiffness of the models, the nonlinear static analyses snow loads are 27, 35, and 43 percent in the dome with R/S ratio of 0.1,
have been performed subjecting the models to a monotonically in- 0.2, and 0.3, respectively. This means that, as an example in the dome
creasing pattern of uniform and non-uniform snow loads. Fig. 3(a, b) with the R/S ratio of 0.1, the asymmetric loading pattern decreases 27
shows the load proportionality factor versus apex point displacement percent of the dome's capacity under the snow load. Initial stiffness of a
curve of the perfect structure’s behavior when all the random variables dome is another key factor that can influence the structures’ response.
have been set to their mean values. The numeric values of the afore- As it turns out from Fig. 3(d, e), both of the R/S ratio and load pattern
mentioned response parameters are presented in Table 3. In this table, affect the initial stiffness of the double-layer dome space structure.
the absolute value of maximum capacity of the models is also presented Despite using stronger design sections in the dome with the R/S ratio of
as a multiplication of LPF by the designed symmetric and asymmetric 0.1, Fig. 3(e) indicates that this dome is more flexible than the two
snow loads. Assessing Fig. 3 and Table 3, the effect of some basic key other domes, because of its lower initial stiffness which is attributed to
factors such as R/S ratio, symmetric and asymmetric loading pattern on its low R/S ratio configuration. In addition to this, the initial stiffness of
the behavior of double-layer dome space structures can accurately be the domes under asymmetric snow load is higher than its value under
examined. symmetric snow load. According to Table 3 and Fig. 3(d), the domes
As an example description of the characteristics of the response under asymmetric snow load show a lower value of collapse displace-
curves, according to Fig. 3 (a) for the dome with the R/S ratio of 0.2, ment than the domes under symmetric snow load, which results in
the structure behaves linearly up to 0.012 m displacement of the dome larger initial stiffness for the domes under asymmetric snow load.
apex point. However, after buckling of a set of compressional members As mentioned before, the first buckling mode is used in this paper to
in 2.524 value of LPF (max capacity of the structure) and correspond- apply geometric imperfection both positively and negatively. Fig. 4 il-
ingly 0.013 m of apex point displacement, the structure experiences a lustrates the response curve of the perfect and imperfect models under
sudden drop in its capacity. Thereafter, it regains its ability to withstand symmetric and asymmetric snow loads. According to Fig. 4(a)–(c), the
the applied forces up to 0.02 m displacement. Finally, due to buckling effect of geometric nodal imperfection on the maximum capacity are
of other members, the structure would not be able to keep its stability more than its effect on the post-buckling behavior of the considered
state in new deformed shape, and would accelerate toward collapse. domes (compare the response near 0.1 m displacement). Further, these
From post buckling area of response curve, it is apparent that the figures reveals that in contrast to single-layer space structure [9,25,27],
structure loses approximately half of its capacity in 0.05m displacement double-layer dome space structures are relatively less-sensitive to the
of apex point. geometric nodal imperfection. For instance, Zugasti et al. [60] in-
According to Fig. 3(a, b, c), it is obvious that the value of LPF is vestigated the influence of geometrical imperfections on the behavior of
almost similar in all the domes with three different R/S ratio values squared plan-form single-layer dome space structures, and reported that

7
R.V. Vazna and M. Zarrin Engineering Structures 212 (2020) 110511

Fig. 3. Collapse behavior of the domes and comparison of response parameters (The “Sym.” represents symmetric and the “Asym.” represents asymmetric).

establishment of span/300 nodal imperfection can reduce the ultimate the collapse-initiation location from support area to the top layer of the
load factor (the collapse load for the structure given as a factor of the dome (red color members in Fig. 5). The effect of this change of failure
applied load) of the imperfect dome to 0.4 of its perfect counterpart. initiation location is also discernable in the pushover curve in Fig. 4(b).
Quiet similar results was observed by [61] for suspended dome systems. This finding implies that the nodal geometric imperfection could
It was reported that at the value of maximum nodal imperfection, the change the response of the dome in some cases. Consequently, it may
buckling capacity of the system reduces by almost 50 percent. not be recommended to ignore this issue in the double layer dome space
As shown in Fig. 5, the total collapse of the models occurs through structure design. Comparing the effect of positive and negative im-
buckling of the black color members in bottom layer, which play a key perfection on the model’s behavior, it appears that the positive im-
role in transferring the applied loads to the supports. As an exception, in perfection increases the value of LPF, while negative imperfection de-
the dome with the R/S ratio of 0.2, the geometric imperfection changes creases the LPF value. Thereupon, for examining the model's sensitivity

Table 3
The results of nonlinear static analyses of the models.
No. model Model’s type Maximum LPF Collapse displacement (m) Maximum snow load (KN)

1 R/S = 0.1, sym. snow, perfect 2.3298 0.0246 1022.833


2 R/S = 0.1, sym. snow, negative imperfection 2.2622 0.0256 993.155
3 R/S = 0.1, sym. snow, positive imperfection 2.3836 0.0234 1046.453
4 R/S = 0.1, asym. snow, perfect 3.0057 0.0136 747.023
5 R/S = 0.1, asym. snow, negative imperfection 2.8880 0.0141 717.771
6 R/S = 0.1, asym. snow, positive imperfection 3.1330 0.0133 778.662
7 R/S = 0.2, sym. snow, perfect 2.5243 0.0123 728.165
8 R/S = 0.2, sym. snow, negative imperfection 2.3932 0.0130 690.348
9 R/S = 0.2, sym. snow, positive imperfection 2.5745 0.0112 742.646
10 R/S = 0.2, asym. snow, perfect 3.0730 0.0057 471.537
11 R/S = 0.2, asym. snow, negative imperfection 2.9311 0.0059 449.764
12 R/S = 0.2, asym. snow, positive imperfection 3.3262 0.0054 510.390
13 R/S = 0.3, sym. snow, perfect 2.7966 0.0121 739.762
14 R/S = 0.3, sym. snow, negative imperfection 2.7477 0.0131 726.827
15 R/S = 0.3, sym. snow, positive imperfection 2.8579 0.0113 755.977
16 R/S = 0.3, asym. snow, perfect 3.0189 0.0052 421.463
17 R/S = 0.3, asym. snow, negative imperfection 2.9261 0.0053 408.508
18 R/S = 0.3, asym. snow, positive imperfection 3.1807 0.0049 444.052

8
R.V. Vazna and M. Zarrin Engineering Structures 212 (2020) 110511

(a) R/S ratio of 0.1 (b) R/S ratio of 0.2

(c) R/S ratio of 0.3


Fig. 4. The comparison of collapse behavior under symmetric and asymmetric snow loading and the collapse initiation area (The “NegImp” and “PosImp” in the
legends represent imperfection in negative (inward) and positive (outward) vertical direction, respectively).

6.3. Results of sensitivity analyses

6.3.1. Evaluating pushover sensitivity analyses


Due to space limitation, only the pushover sensitivity analysis re-
sults of the perfect dome with the R/S ratio of 0.2, as the optimal
configuration, under the symmetric snow load have been chosen to be
presented in more detail in this section. At first, the random variables
are set to their mean values in order to extract the mean behavior of the
models, and then, they are set to their upper and lower bound quantiles
for obtaining the perturb behavior of the models.
As seen in Fig. 6(a), the yield stress random variable indicates a
greater effect on models’ response than the modulus of elasticity. The
upper and lower bound quantile of the yield stress random variable
change the whole response symmetrically around the mean model re-
sponse, while the variation of the modulus of elasticity just perturbs the
response at the collapsing point, and its curve tends to the mean model
curve in post-failure region. In other words, the variation in modulus of
elasticity does not have any effect on the residual response. In contrast,
the initial stiffness of the dome seems insensitive to the yield stress
Fig. 5. The location of buckled members. variation, but the alternation of elastic modulus changes the initial
stiffness remarkably. According to the Fig. 6(b), considering the initial
to the variation of random variables in the following sections, it is stiffness, the maximum and residual capacity, it can be concluded that
preferred to use negative imperfection in order to capture the most in all these response parameters, the thickness uncertain parameter
critical response. changes the response behavior more than the diameter random vari-
able. Both variables alter the whole response symmetrically from the
pre to post-failure region. Weighing their influence on the model’s re-
sponse against the influence of the other random variables, it can be

9
R.V. Vazna and M. Zarrin Engineering Structures 212 (2020) 110511

(a) The yield stress and elastic modulus variation’ (b) The diameter and thickness variation’ effect
effect

(c) The Gamma and Extreme imperfection variation’ (d) The MaL and ultimate stress variation’ effect
effect
Fig. 6. Comparison of random variable’ influence on structures’ pushover response behavior (The “Up” and “Low” in the legends represent the upper and lower
extremes of the uncertain parameters, respectively).

deduced that these random variables can be counted as the influential of this parameter are obtained from a normal distribution function, it is
uncertain parameters in the dome with R/S ratio of 0.2. Fig. 6(c) seen in this figure that the structure shows asymmetric behavior with
compares the response behaviors in which the Gamma and Extreme respect to the mean response. On the other hand, the ultimate stress
imperfection random variables are set to their upper and lower bound uncertain parameter has a negligible effect on the response curve. This
quantiles. This figure reveals that applying the upper bound quantile of figure illustrates that, in the whole response curve, the models with
Gamm imperfection strikingly changes the maximum capacity of the perturbed ultimate stress values show approximately the same response
model, and the potential for changing the collapse mechanism can be as the model with the mean values of random variables.
seen, while implementing the Extreme imperfection’s upper bound
quantile do not change the response behavior at all. The reason lies in 6.3.2. Evaluating TDA results
the shape of the Extreme probability density function (PDF), where the The investigation of the response statistics sensitivity to the con-
upper bound quintile of the Extreme imperfection is only 11 percent sidered uncertain parameters is summarized below. First, all para-
larger than its mean value in this paper case study. The results also meters of the finite element model were set equal to their respective
show that increasing the value of imperfection decreases the initial mean values, and the dome response were extracted using incremental
stiffness of the dome. The lower bound quantile of both types of the nonlinear static analysis. Next, this procedure was repeated one-para-
imperfection random variables changes the structure’s behavior simi- meter-at-a-time, setting each parameter to its 5th and 95th percentile
larly, and it causes the structure to show more post-failure strength, while holding the remaining parameters at their mean values. The re-
which results in the more residual capacity of the model. sponse variation can be displayed through a tornado diagram, and the
Based on Fig. 6(d), having the greatest impact on the double-layer absolute difference between the two extremes is called the swing [62].
dome space structure’s response, the mass and load random variable As is seen in Fig. 7, in the tornado diagram, the swings due to the
can be considered as the most influential uncertain parameter. The various random variables are displayed in the descending order of the
upper bound quantile of this random variable decreases the initial swing size from top to bottom. A larger swing size implies a larger effect
stiffness and maximum capacity of the model, while its lower bound of the corresponding random variable on the dome maximum capacity.
quantile increases them. Although the upper and lower bound quantiles By counting the total number of times each random parameter has been

10
R.V. Vazna and M. Zarrin Engineering Structures 212 (2020) 110511

Fig. 7. Tornado diagram for the maximum capacity of domes with three different R/S ratios.

ranked in the top first among all models, it can be found that in all the maximum capacity of the domes to these random variables is ap-
models, the mass and load (MaL) random variable is the most important proximately close to each other. Moreover, the Gamma imperfection
uncertain parameter. Fig. 7 shows that in all cases, in spite of con- with the largest value of COV among the random variables is considered
sidering symmetric normal distribution function for MaL random vari- within the second important group of uncertain parameters. Therefore,
able, the model structures show a somewhat asymmetric behavior it might be concluded that the COV value does not necessarily have
around the mean models at the collapsing (buckling) point. The direct impact on the significance of the variables. The yield stress, Ex-
thickness, diameter, and Gamma imperfection are considered as the treme imperfection, and modulus of elasticity uncertain parameters
second important group of random variables. According to Fig. 7, im- approximately have the same effect on the maximum capacity of the
plementing the upper and lower bound quintiles of the thickness, dia- double-layer dome space structure, and are supposed to be considered
meter, yield stress and modulus of elasticity random variables sym- as the third important group of random variables. Furthermore, it can
metrically change the maximum capacity with respect to the mean be seen that the ultimate stress has virtually no impact on the struc-
model maximum capacity at the point of failure. However, the Gamma ture’s response. Consequently, the uncertainty in these four parameters
imperfection has an asymmetrical effect on the maximum capacity of could be neglected without leading to a significant loss of accuracy.
the models. This might be attributed to this fact that the ratio of the The random variables that their upper bound quantile decrease
upper bound quintile of the Gamma imperfection to its mean value is structure’s capacity are called demand parameters. On the other hand,
2.6, while the ratio of the lower bound quintile of this variable to its the random variables that their upper bound quantile increase the
mean value is 0.11. Furthermore, as will be explained further ahead, models’ capacity are called capacity random variable [22]. Therefore,
the change of failure initiation position in the structure (as a result of in this study, the MaL and member’s imperfection are demand para-
the asymmetrical PDF of this random variable) might have caused this meters while the other random variables are capacity parameters. It is
asymmetrical behavior. In addition, in the models with the R/S ratio of noteworthy that the identification of the demand or capacity random
0.2 under asymmetric snow load, the diameter random variable shows a variables is done based on inspecting the pushover curves of the models
somewhat asymmetrical effect. with the upper and lower bound quantiles of the random parameters.
As seen in Fig. 7, in all the models, the importance of the thickness For instance, according to Fig. 6(d), the MaL random variable is a de-
random variable is greater than the diameter parameter. However, the mand parameter for the model with the R/S ratio of 0.2. The same
key issue that should be noticed here is that although the thickness conclusions can be made in regard to the pushover curves of the other
random variable has approximately four times greater COV than the models, which are not presented herein due to the space limitation.
diameter uncertain parameter, it appears that the sensitivity of the Total failure in double-layer space structure occurs when the structure

11
R.V. Vazna and M. Zarrin Engineering Structures 212 (2020) 110511

(a) R/S ratio 0.2 (b) R/S ratio 0.3


Fig. 8. The collapse behavior of perfect models under symmetric snow load (The “Up” and “Low” in the legends represent the upper and lower extremes of the
uncertain parameter, respectively).

could not carry the re-distributed load due to buckling of the first set of the residual capacity is totally sensitive to the MaL uncertain para-
compression members and the structure undergoes a progressive col- meter. The thickness, yield stress, and diameter variables could be
lapse. In this study, buckling of support-connected members in almost considered as the second influential group of uncertain parameters. As
all the models causes total failure in double-layer dome space structure. it turns out from this figure, more or less the same random variables,
But in the domes with the R/S ratio of 0.2 and 0.3 with perfect geo- which were important corresponding to the sensitivity of maximum
metry (under symmetric snow load), as is shown in Fig. 8, im- capacity, are also dominant when the residual capacity of the models is
plementing the lower bound quantile of imperfection parameter causes examined. The only exception is the yield stress parameter. Fig. 10 il-
the domes to show a different type of behavior, which results in a lustrates that, in most cases, the thickness parameter is more important
gentler slope in the post-failure area of response curve. This observation than the yield stress except in the model with the R/S ratio of 0.2 under
is attributed to the change of the initial buckling members’ location asymmetric snow load and the model with the R/S ratio of 0.3 under
from support area in the bottom layer to the members in the top layer. symmetric snow load. The remaining random variables have negligible
This modified post-failure behavior leads to a one-sided swing in the effect on residual capacity of the case study double-layer dome space
TDA diagram of the residual capacity (see Fig. 10) in these two models. structures.
Additionally, when the upper bound quantile of imperfection is im- Although the Gamma imperfection uncertain parameter was con-
plemented to the members as an initial curvature, in spite of a re- sidered within the second vital group of random variables in the ca-
markable decrease in capacity, the double-layer dome shows a notice- pacity TDA, according to Fig. 10, it is obvious that this parameter does
able potential for changing the collapse mechanism from total collapse not have a striking effect on the residual capacity of the models except
to local collapse with a dynamic snap through. Comparing the swing of in the perfect dome with the R/S ratio of 0.2 under symmetric snow
the Gamma and Extreme imperfection in Fig. 7, a major finding re- load. In this model, the Gamma and Extreme imperfection parameters
garding the importance of member’s imperfection on double-layer play a key role in changing the residual strength of the dome because,
dome space structures can be observed that the distribution function of as previously mentioned, applying upper bound quantile of the im-
imperfection random variable plays a pivotal role on their importance. perfection alters the collapse initiation location from the support area
It is interesting to note that the snow load pattern, R/S ratio and to the top layer of the dome.
being geometrically perfect or imperfect are not influential factors in By evaluating both the maximum and residual capacity TDA dia-
ranking the random variables according to their importance. But, ir- grams, such a strong influence of mass and load, thickness, and dia-
respective to these factors, the mass and load, Gamma imperfection, meter uncertain parameters can be understood. Moreover, through si-
thickness, and diameter are the random variables that can be con- multaneous examination of Figs. 7 and 10 and assessing the variation of
sidered as the key uncertain parameters which have the potential to these three uncertain parameters in these figures, it can be inferred that
remarkably change the structure capacity. variation of these uncertain parameters changes the whole response
According to Fig. 9, the sensitivity of dome apex point collapse curve in pre and post-failure area. In contrast, the remaining random
displacement to uncertain parameters is investigated too. Based on this variables do not show a remarkable effect on the residual capacity of
figure, it appears that the Fy and Es are the most important random double-layer dome space structures.
variables which cause a great deal of variation on collapse displace-
ment. In addition, this figure shows that except for the aforementioned
random variables, the importance ranking of other random variables 6.4. Result of the FOSM analysis
differs in the different models, and finding a specific trend for the im-
portance of uncertain parameters is very difficult. Fig. 9 shows that, in Table 4 summarizes the results of FOSM approach providing the
most cases, the Gamma and Extreme imperfection random variables do coefficient of variation of the capacity LPF and the LPF value in 0.05 m
not have a noticeable effect on collapse displacement of double-layer (0.1 m in 0.1 R/S ratio model) apex point displacement (residual ca-
dome space structures and can be considered as a random variable with pacity) in all the models. Tabulated in this table suggest that different
moderate importance. Similar to the capacity TDA, the ultimate stress types of load pattern and some specific characteristics of the dome such
random variable has much less impact on the collapse displacement of as R/S ratio and geometric imperfection do not have a remarkable ef-
double-layer dome space structures. fect on the COV of the maximum capacity of double-layer dome space
It can be well seen in the Tornado Diagram presented in Fig. 10 that structures. Based on the results presented in this table, an average COV
of 0.2 is proposed for double-layer dome space structure capacity under

12
R.V. Vazna and M. Zarrin Engineering Structures 212 (2020) 110511

Fig. 9. Tornado diagram for the collapse displacement of domes with three different R/S ratios.

snow loading. On the other hand, the residual capacity of the domes lines show the importance factor of the random variables in the TDA
with the R/S ratio of 0.3 have a greater value of COV than the two other and FOSM methods, respectively.
domes. On average, the COV value of 0.25 is proposed for the residual As seen in Fig. 11, there is a close similarity between the results of
capacity of this type of structure. the TDA and FOSM methods, and it can be concluded that alike to the
The first term in Eq. (9) shows the direct contribution of random results of the TDA analysis, the MaL random variable with the im-
variable variations on the computed variance of the limit state function portance factor of 0.7 is considered as the most effective uncertain
(response parameters), and also shows which random variable mostly parameter. Likewise, the thickness, Gamma imperfection, and diameter
affect it [63]. According to this part of the equation, the random vari- of the member’s section are the other three high-rank random variables.
∂g These three random variables have importance factor of 0.35 to 0.45 in
able with the largest value of ∂X σXi has the greater effect on the limit
i
state function’s variance. Consequently, according to Eq. (11), im- the domes with the R/S ratios of 0.2 and 0.3, while their importance
∂g factor is around 0.3 to 0.45 in the dome with the R/S ratio of 0.1.
portance vector (ω ) which is a normalized form of σ
∂Xi Xi
may rank the
The importance factor of the random variables for residual capacity
random variables according to their importance: using both the TDA and FOSM methods are demonstrated in Fig. 12.
∂g This figure indicates that the uncertain parameters in both the TDA and
σ
∂Xi Xi
ωi = FOSM methods almost have the same importance. As highlighted in the
∂g
‖ ∂X σXi ‖ TDA sensitivity section, the parameter which influences the residual
i (11)
capacity most considerably is the MaL parameter with a higher value of
The importance factor of the random variables in the TDA method is importance factor than the other random variables. The yield stress,
calculated by dividing the swing value of the individual random vari- thickness, and diameter of members’ section are nearly the most
able to the norm of all swing values. The importance factor of each dominant parameters in this response limit state parameter. As men-
random variable is a value between 0 and 1. According to this value, the tioned in the previous section, when the lower bound quantile of
most influential uncertain parameter is a random variable with a Gamma imperfection random variable is implemented to the member’s
greater value of importance factor. These importance factors for the as the initial curvature, the collapse-beginning area in the dome
maximum capacity of the models in both the FOSM method (ωi ) and the changes, and as a result, the residual capacity increases in the perfect
TDA method are compared in Fig. 11. In this figure, the red and blue domes with the R/S ratios of 0.2 and 0.3 under symmetric snow load.

13
R.V. Vazna and M. Zarrin Engineering Structures 212 (2020) 110511

Fig. 10. Tornado diagram for the residual strength of domes with three different R/S ratios.

Table 4 7. Conclusion
Results of the FOSM analyses for the double layer dome space structures.
R/S ratio Load pattern COV
The present work studies the influence of some key factors such as
the rise to span ratio, snow load pattern, and structure’s geometric
Maximum capacity Residual capacity imperfection on design and collapse behavior of double-layer dome
space structure. The effect of different sources of uncertainty in mate-
0.1 Symmetric snow, Perfect 0.195 0.208
Symmetric snow, Imperfect 0.198 0.210
rial mechanical behavior, member’s geometric imperfection, and the
Asymmetric snow, Perfect 0.179 0.216 applied gravity dead load on the nonlinear static collapse behavior of
Asymmetric snow, 0.181 0.217 the dome space structure are investigated. Moreover, the most influ-
Imperfect ential uncertain parameters on model’s response parameters are re-
0.2 Symmetric snow, Perfect 0.194 0.326 cognized by using the TDA and FOSM analyses. The most notable
Symmetric snow, Imperfect 0.187 0.207 findings can be summarized as follows:
Asymmetric snow, Perfect 0.215 0.248

• The value of R/S ratio plays a pivotal role in determining the total
Asymmetric snow, 0.222 0.248
Imperfect
value of the applied dead, symmetric and asymmetric snow loads.
0.3 Symmetric snow, Perfect 0.201 0.301
The R/S ratio of 0.2 is recommended as the optimum configuration
Symmetric snow, Imperfect 0.197 0.405
Asymmetric snow, Perfect 0.204 0.300 for designing double-layer dome space structures, since the dome
Asymmetric snow, 0.209 0.296 with R/S ratio of 0.2 has the minimum value of self-weight com-
Imperfect pared to the domes with R/S ratios of 0.1 and 0.3.
• In case of designing the domes with different R/S ratios in such a
way to return the same elements’ stress ratio in different parts of the
The two later figures reveal that the ultimate stress random variable dome structure, the collapse capacity LPF values of the domes with
with the lowest value of importance factor is an insignificant uncertain different R/S ratios under both symmetric and asymmetric snow
parameter. load patterns are approximately close to each other. On the other
hand, asymmetric snow load decreases the absolute value of the

14
R.V. Vazna and M. Zarrin Engineering Structures 212 (2020) 110511

Fig. 11. The importance factor of random variables in TDA and FOSM methods (maximum capacity).

maximum capacity of the domes for carrying the snow load. Indeed, approximately four times greater than the diameter’s COV. In ad-
comparing the dome's maximum capacity value when they are dition, in spite of having the largest value of COV among the random
subjected to the symmetric and asymmetric snow loads, it is obvious variables, the Gamma imperfection is not the most influential
that the domes are able to carry less asymmetric snow load than the random variable. Therefore, it might be concluded that the COV
symmetric one. value does not necessarily have a direct impact on the significance
• Double-layer dome space structures are far less sensitive to the of the variables.
structure's geometric imperfection (node deviation) than single- • The yield stress, Extreme imperfection, and modulus of elasticity
layer one. Nevertheless, in some cases (for instance the imperfect uncertain parameters can be considered as the third influential
dome with R/S ratio of 0.2 under symmetric snow load), the node group of random variables with far less impact on structural capa-
deviation changes the collapse initiation area, and shows the po- city than the other parameters. Moreover, the ultimate stress para-
tential for changing the collapse mechanism of the dome. Therefore, meter almost has no influence on the response parameters.
it is not recommended to ignore the node deviation in double layer Therefore, in probabilistic assessments of the capacity of this type of
space structures at all. structure, the uncertainty in these four parameters could be ne-
• The results of the TDA method for the maximum capacity show that, glected without leading to a significant loss of accuracy.
in all cases, the mass and load random variable is the most influ- • The apex point collapse displacement value shows a remarkable
ential uncertain parameter. This is in general agreement with the sensitivity to the variation of the yield stress and modulus of elas-
reported results of some other studies for other types of structures ticity parameters. Therefore, these uncertain parameters are sug-
[14,23]. The thickness, diameter, and Gamma imperfection are gested as the most influential random variables in determining the
considered as the second important group of random variables. apex point collapse displacement of the dome space structure. On
Consequently, the variation in these four random variables is re- the other hand, the residual capacity of the models are mostly
commended to be considered in static response evaluation of the sensitive to the mass and load uncertain parameter, and the thick-
double-layer dome space structure. ness and diameter uncertain parameters are considered as the
• The sensitivity of the maximum capacity to the variation of the second crutial group of random variables.
members’ thickness and diameter in all the models are approxi- • The results of the FOSM analyses show that, in all the models, the
mately equal, while the thickness random variable COV value is COV value of the maximum capacity is approximately equal to 0.2,

15
R.V. Vazna and M. Zarrin Engineering Structures 212 (2020) 110511

Fig. 12. The importance factor of random variables in TDA and FOSM methods (residual capacity).

while some differences between the COV values of the residual ca- assessments.
pacity of the domes with R/S ratio of 0.1, 0.2, and 0.3 are observed. • Ranking of the random variables does not depend on some specific
However, on average, COV values of 0.2 and 0.25 are proposed for key factors such as the symmetric and asymmetric snow load pat-
the maximum and residual capacity of the domes, respectively. terns, rise to span ratio of the dome, and being imperfect while these
• Based on the results of the FOSM method for ranking the random parameters are very important in designing the double-layer dome
variables according to their contribution in calculating the COV space structure. Actually, almost the same trend of the random
value of the response parameters, it is observed that there is a close variables’ ranking for each response parameter can be seen in all the
similarity between the results of the TDA and FOSM methods. models.
Therefore, it is inferred that the TDA method is able to accurately
recognize the most influential random variables.
• Since the results of the TDA and FOSM methods are almost the same, Declaration of Competing Interest
considering the simplicity and ease of implementation, the TDA
method is recommended for sensitivity analysis of this type of The authors declare that they have no known competing financial
structure. On the other side, when the computation of the variation interests or personal relationships that could have appeared to influ-
(standard deviation) of the response parameters is of interest, the ence the work reported in this paper.
FOSM method is proposed as the simplest method of probabilistic

Appendix

Design and details of the models

Based on the ASCE7-10 [41], the ground snow load determination for a site is based on an extreme value statistical analysis of data available in
the vicinity of the site using a value with a 2 percent annual probability of being exceeded. There are two different types of snow load patterns that
the dome structures have to be designed for i.e. balanced and unbalanced snow loads. For obtaining the numeric values of snow load on the sloping
roof, it is mandatory to calculate snow load on a flat roof at first. The Eq. (A.1) shows the way of calculating flat roof snow load.
Pf = 0.7 ∗ Ce ∗ Ct ∗ Is ∗ Pg (A.1)

16
R.V. Vazna and M. Zarrin Engineering Structures 212 (2020) 110511

Fig. A1. The top layer slope.

Table A1
applied snow loads on the models.
R/S ratio 0.1 R/S ratio 0.2 R/S ratio 0.3

a (Degree) Cs Snow Load a (Degree) Cs Snow Load a (Degree) Cs Snow Load


N / m2 N / m2 N / m2

Sym. Asym. Sym. Asym. Sym. Asym.

Ring 1 2.8 1 1132.0 565.5 3.6 1 1132.0 565.5 4.6 1 1132.0 565.5
Ring 2 4.3 1 1132.0 833.0 10.9 0.9 1029.1 838.4 13.7 0.86 980.1 830.1
Ring 3 7.1 0.96 1094.8 1099.7 18.2 0.79 901.7 1110.0 22.9 0.72 819.4 1093.8
Ring 4 9.9 0.92 1045.8 1367.2 25.4 0.68 776.2 1382.0 32.1 0.58 660.0 1358.4
Ring 5 12.8 0.88 995.8 1633.8 32.7 0.57 648.8 1382.0 41.2 0.44 500.8 1020.3
Ring 6 15.6 0.83 946.8 1901.4 40 0.46 522.4 1044.8 50.4 0.3 341.0 682.1

Fig. A2. Asymmetric snow load pattern (this figure is taken from ASCE 7–10, (a) R/S ratio 0.1, (b) R/S ratio 0.2 and 0.3.

Table A2
Design section of the members.
Element color R/S ratio of 0.1 R/S ratio of 0.2 R/S ratio of 0.3

Section type D.out (mm) Thick. (mm) Section type D.out t Section type D.out t
(mm) (mm) (mm) (mm)

D33.7T3.2 33.7 3.2 D26.9T2.6 26.9 2.6 D26.3T2.6 26.9 2.6

D33.7T3.2 33.7 3.2 D33.7D2.6 33.7 2.6 D33.7D3.2 33.7 3.2

D42.4T3.2 42.4 3.2 D33.7T2.6 33.7 2.6 D33.7D3.2 33.7 3.2

D42.4T3.2 42.4 3.2 D33.7T2.6 33.7 2.6 D33.7T2.6 33.7 2.6

D48.3T3.2 48.3 3.2 D42.4T2.6 42.4 2.6 D42.4T2.6 42.4 2.6

D52T3.2 52 3.2 D42.4T3.2 42.4 3.2 D42.4T3.2 42.4 3.2

D52T3.2 52 3.2 D42.4T2.6 42.4 2.6 D42.4T3.2 42.4 3.2

D52T3.2 52 3.2 D42.4T3.2 42.4 3.2 D42.4T3.2 42.4 3.2

In which:
Ce is exposure factor that is determined from Table 7-2 of ASCE7-10. In this study, based on terrain category (C) and being partially exposed, the
amount of Ce is considered to be equal to 1.
Ct is thermal factor that is determined from Table 7-3 of ASCE7-10 and is assumed to be equal 1 in this paper.
Is is importance factor that is extracted from Table1.5-2 of ASCE7-10. Since the structure’s failure poses substantial risk of human life, the value of
importance factor is supposed to be equal 1.10.

17
R.V. Vazna and M. Zarrin Engineering Structures 212 (2020) 110511

Fig. A3. (a) Top layer of the dome, (b) Bottom layer of the dome, (c) Web layer of the dome.

Pg is ground snow load which is considered to be 1.5 KN2


m
The sloped roof balanced snow load, Ps , is obtained by multiplying the flat roof snow load, Pf , by the roof slope factor, Cs :
Ps = Cs ∗ Pf (A.2)
In which: Cs is slope factor determined from Fig. 7-2 of ASCE7-10. In this study, as is shown in Fig. A.1, the domes are formed with six different
sloped rings. Each ring has its own Cs factor according to its slope with respect to the horizon. The a1 to a6 in this figure represents the degree
between the specific sloped ring and the horizon line. Tabulated in Table A.1 are the slope values of each ring and its corresponding Cs parameter in
the domes with different R/S ratios.
Based on ASCE7, the calculation of asymmetric snow load totally depends on the slope at eaves of the domes. According to the Fig. A.2, two
different methods are used in order to calculate the asymmetric snow load of the models. The dome with the R/S ratio of 0.1 has eaves slope of fewer
than 30°, therefore its asymmetric snow load pattern follows the Fig. A.2(a) method, while having the eaves slope greater than 30°entails the
asymmetric snow load of the domes with the R/S ratio of 0.2 and 0.3 follow the Fig. A.2(b) method. The calculated numeric values of symmetric and
asymmetric snow load of the models have shown in Table A.1.
The AISC 360–10 code has been used for designing models under different aforementioned types of loads. The elements’ design section of the
models and their position in the structure has shown in table A.2 and Fig. A.3, respectively.
In Table A.2, sections’ characteristics are tabulated according to the elements’ color in the Fig. A.3.

References [17] Porter KA. An overview of PEER’s performance-based earthquake engineering


methodology. In Proceedings of ninth international conference on applications of
statistics and probability in civil engineering. 2003. Jul 6.
[1] Popko ES. Divided Spheres: geodesics and the orderly subdivision of the sphere. [18] Hezarjaribi M, Bahaari MR, Bagheri V, Ebrahimian H. Sensitivity analysis of jacket-
CRC Press; 2012. type offshore platforms under extreme waves. J. Constr. Steel Res. 2013;83:147–55.
[2] El-Sheikh A. Effect of member length imperfections on triple-layer space trusses. J. [19] Haselton CB. Assessing seismic collapse safety of modern reinforced concrete frame
Eng. struct. 1997;19(7):540–50. buildings, Ph.D. thesis. Department of Civil and Environmental Engineering
[3] Sheidaii MR, Gordini M. Effect of random distribution of member length im- Stanford University; 2006.
perfection on collapse behavior and reliability of flat double-layer grid space [20] Hui-Jun L, Xue F, Zhen-Xing CH, Ming-Fu L, Guang-De L. Sensitivity and correlation
structures. J. Advances in Struct. Eng. 2015;18(9):1475–85. analysis of double-layer cylindrical reticulated shell considering several perfor-
[4] Borri C, Spinelli P. Buckling and post-buckling behaviour of single layer reticulated mance functions. J. Space Struct. 2015;30(1):13–26.
shells affected by random imperfections. Comput Struct 1988;30(4):937–43. [21] Zhong J, Zhi X, Fan F. Sensitivity of seismic response and fragility to parameter
[5] Tegola AL. Ultimate limit states of space reticular structures with random behaviour uncertainty of single-layer reticulated domes. J. Steel Struct. 2018;18(5):1607–16.
elements. J. Space Struct. 1992;7(4):345–52. [22] Mahadevan S, Haldar A. Probability, reliability and statistical method in en-
[6] El-Sheikh A. Effect of geometric imperfections on single-layer barrel vaults. J. space gineering design. John Wiley & Sons; 2000.
struct. 2002;17(4):271–83. [23] El-Din MN, Kim J. Sensitivity analysis of pile-founded fixed steel jacket platforms
[7] Yadollahi MM, Gül R, Polat R, Yadollahi BM, Gül MS. The effect of initial geometric subjected to seismic loads. J. Ocean Eng. 2014;85:1–11.
imperfection on the load bearing capacity of double layer barrel vault space [24] Fan F, Cao Z, Shen S. Elasto-plastic stability of single-layer reticulated shells. J.
structures. In 3rd ECCOMAS Thematic Conference on Computational Methods in Thin-Walled Struct. 2010;48(10–11):827–36.
Structural Dynamics and Earthquake Engineering. 2011. p. 25–8. Corfu, Greece. [25] Bruno L, Sassone M, Venuti F. Effects of the equivalent geometric nodal imperfec-
[8] Roudsari MT, Gordini M. Random imperfection effect on reliability of space struc- tions on the stability of single layer grid shells. Eng. Struct. 2016;112:184–99.
tures with different supports. J. Struct. Eng. and Mech. 2015;55(3):461–72. [26] JGJ61-2003 Technical specification for reticulated shells. Beijing: China
[9] Chen X, Shen SZ. Complete load-deflection response and initial imperfection ana- Architecture Industry Press, 2003.
lysis of single-layer lattice dome. J. Space Struct. 1993;8(4):271–8. [27] Mohammadi M, Abedi K, Taghizadieh N. Stability analysis of single-layer barrel
[10] Fan F, Yan J, Cao Z. Elasto-plastic stability of single-layer reticulated domes with vault space structures. J. Space Struct. 2012;27(4):203–18.
initial curvature of members. J. Thin-Walled Struct. 2012;60:239–46. [28] Chen ZH, Li Y. Parameter analysis on stability of a suspendome. J. Space Struct.
[11] Asgarian B, Ordoubadi B. Effects of structural uncertainties on seismic performance 2005;20(2):115–24.
of steel moment resisting frames. J. Constr. Steel Res. 2016;120:132–42. [29] Chen G, Zhang H, Rasmussen KJ, Fan F. Modeling geometric imperfections for re-
[12] Jalali SA, Banazadeh M, Tafakori E, Abolmaali A. Seismic performance assessment ticulated shell structures using random field theory. Eng. Struct. 2016;126:481–9.
of steel moment frames with generic Locally Reinforced connections. J. Constr. [30] B. Ellingwood, T.V. Galambos, M. Gregor, C.A. Cornell, Development of probability-
Steel Res. 2011;67(8):1261–71. based load criterion for American national standard A58. U.S. Government printing
[13] Pirmoz A, Marefat MS. Reliability assessment of compression columns in seismic Office, 1980.
EBFs. J. Constr. Steel Res. 2015;104:274–81. [31] JCSS, Probabilistic Modal Code-Part 1: Basis of Design. (12th draft), Joint
[14] Zarrin M, Asgarian B, Abyani M. Probabilistic Seismic Collapse Analysis of Jacket Committee on Structural Safety, available from: 2001. http://www.jcss.
Offshore Platforms. J. Offshore Mech. Arctic Eng. 2018;140(3):031601. [32] ISO Fixed Steel Offshore Structures Code 19902, Component-based calibration of
[15] Vamvatsikos D, Cornell CA. Incremental dynamic analysis. Earthquake Eng. Struct. North West European annex for environmental load factors, Research Report 088,
Dyn. 2002;31(3):491–514. 2003.
[16] Zarrin M, Abyani M, Asgarian B. A statistical study on lognormal central tendency [33] Wang D. Probabilistic risk analysis of steel frame structures. Harbin Dissertation of
estimation in probabilistic seismic assessments. J. Struct. Infrastruct. Eng. 2019. Harbin Institute of Technology; 2006. p. 18–9.
https://doi.org/10.1080/15732479.2019.1668813. [34] Abedi K, Parke GA. Progressive collapse of single-layer braced domes. J. Space

18
R.V. Vazna and M. Zarrin Engineering Structures 212 (2020) 110511

Struct 1996;11(3):291–306. [51] Asgarian B, Aghakouchack AA, Bea RG. Non-linear analysis of jacket type offshore
[35] Makkar A, Abbas S, Haslin SM. Finite Element Analysis of Diamatic, Schwedler and platforms using fiber elements. ASME J. Offshore Mech. Arctic Eng.
Diamatic-Schwedler Hybrid Domes. J. Eng. Trends Technol. 2016;39:57–62. 2006;128(3):224–32.
[36] Nooshin H. Formex configuration processing in structural engineering. London: [52] Honarvar MR, Bahari MR, Asgarian B, Alanjari P. Cyclic inelastic behavior and
Elsevier Applied Science; 1984. analytical modeling of pile leg interaction in jacket type offshore platform. Appl
[37] Chen J. A Study on the selection of optimal roof type for low-rise buildings group–in Ocean Res 2008;29(4):167–79.
a view of wind pressures action. J. Procedia Eng. 2012;31:1149–54. [53] Alanjari P, Asgarian B, Bahaari MR, Honarvar MR. On the energy dissipation of
[38] Mousavi Mehdi Abbasi, Abedi Karim, Chenaghlou Mohamadreza. Imperfection jacket type offshore platforms with different pile–leg interactions. Appl Ocean Res
sensitivity analysis of double domes free form space structures. Int. J. Str. Stab. Dyn. 2009;31(2):82–9.
2015;15(04):1450067. https://doi.org/10.1142/S0219455414500679. [54] Uriz P, Filippou FC, Mahin SA. Model for cyclic inelastic buckling of steel braces. J.
[39] Nooshin H. Space structures and configuration processing. Prog Struct Mat Eng struct. eng. 2008;134(4):619–28.
1998;1(3):329–36. [55] Black GR, Wenger WA, Popov EP. Inelastic buckling of steel struts under cyclic load
[40] San B, Xu C, Qiu Y. Three-dimensional aerodynamic optimization of single-layer reversals, UCB/EERC-80/40. Berkeley, Calif: Earthquake Engineering Research
reticulated cylindrical roofs subjected to mean wind loads. Adv. Civil Eng. Center; 1980.
2019:1–13. Article ID 4156319. [56] Lai JW, Mahin SA. Steel concentrically braced frames using tubular structural
[41] American Society of Civil Engineers, Minimum Design Loads for Buildings and sections as bracing members: design, full-scale testing and numerical simulation.
Other Structures. ASCE Standard ASCE/SEI 7-10, Published by N.Y. The Society. Int. J. Steel Struct. 2014;14(1):43–58.
[42] Gallego E, Fuentes JM, Ramírez-Gómez A, Ayuga F. Effects of different snow load [57] Hsiao PC, Lehman DE, Roeder CW. A model to simulate special concentrically
arrangements on steel silo roof structures. J. Adv. Struct. Eng. braced frames beyond brace fracture. Earthquake Eng. Struct. Dyn.
2018;21(16):2507–17. 2012;42(2):183–200.
[43] Kato S, Iwamoto T. Buckling and reliability analysis of single layer grid dome with [58] Yang D, Yun C, Wu J, Yao Y. Seismic response and failure mechanism of single-layer
diagonal brace under snow load. J. IASS. 2017;58(3):207–25. latticed domes with steel columns and braces as substructures. Thin-Walled Struct.
[44] Cai J, Xu Y, Feng J, Zhang J. Nonlinear stability of a single-layer hybrid grid shell. J. 2018;124:458–67.
Civil Eng. Manage. 2012;18(5):752–60. [59] Song M, Chen D. An improved knowledge-informed NSGA-II for multi-objective
[45] American institute of Steel Construction (AISC), Specification for Structural Steel land allocation (MOLA). Geo Spat. Inf. Sci. 2018;21(4):273–87.
Buildings, Chicago, III. USA, 2010. [60] Altuna Zugasti AM, Lopez-Arancibia A, Puente I. Influence of geometrical and
[46] Mazzoni S, McKenna F, Scott M, Fenves G. Open System for Earthquake Engineering structural parameters on the behaviour of squared plan-form single-layer structures.
Simulation (OpenSEES) - OpenSEES Command Language Manual. Berkeley, Ca: J Constr Steel Res 2012;72:219–26.
Uni. of California; 2007. [61] Kitipornchai S, Kang W, Lam HF, Albermani F. Factors affecting the design and
[47] Spacone E, Filippou FC, Taucer FF. Fiber beam column model for nonlinear analysis construction of Lamella suspen-dome systems. J Constr Steel Res 2005;61:764–85.
of RC frames. I: formulation. J Earthq. Eng. Struct. Dyn. 1996;25(7):711–25. [62] Porter KA, Beck JL, Shaikhutdinov RV. Sensitivity of building loss estimates to
[48] De Souza R. Forced-based finite element for large displacement inelastic analysis of major uncertain variables. Earthq. Spectra 2002;18(4):719–43.
frames Ph.D Thesis. California: University of California, Berkeley; 2000. [63] Heidary-Torkamani H, Bargi K, Amirabadi R, McCllough NJ. Fragility estimation
[49] Menegotto M, Pinto P. Method of analysis for cyclically loaded reinforced concrete and sensitivity analysis of an idealized pile-supported wharf with batter piles. J. soil
plane frames including changes in geometry and nonelastic behavior of elements dyn. and earthq. eng. 2014;61:92–106.
under combined normal force and bending. Lisbon: IABSE Symposium; 1973. [64] Bai Yong, Hui-bin Yan Yu, Cao Younghoon Kim, Yang Yong-yan, Jiang Hua. Time-
[50] Asgarian B, Aghakouchak AA, Bea RG. Inelastic post buckling and cyclic behavior of dependent reliability assessment of offshore jacket platforms. Ships and Offshore
tubular braces. ASME J. Offshore Mech. Arct. Eng. 2005;127(3):256–62. Structures 2015;11(6):1–12.

19

You might also like