0 Bewertungen0% fanden dieses Dokument nützlich (0 Abstimmungen)

4 Ansichten8 SeitenThe steel industry is developing a design-by-advanced analysis speciﬁcation for cold-formed steel construction. This eﬀort provides an opportunity to utilize the latest nonlinear structural analysis (advanced analysis) to design steel structures based on their overall system behaviour. This paper concerns the system reliability
calibrations of this design-by-analysis method, with a particular focus on cold-formed steel portal frames.

May 04, 2019

© © All Rights Reserved

PDF, TXT oder online auf Scribd lesen

The steel industry is developing a design-by-advanced analysis speciﬁcation for cold-formed steel construction. This eﬀort provides an opportunity to utilize the latest nonlinear structural analysis (advanced analysis) to design steel structures based on their overall system behaviour. This paper concerns the system reliability
calibrations of this design-by-analysis method, with a particular focus on cold-formed steel portal frames.

© All Rights Reserved

Als PDF, TXT **herunterladen** oder online auf Scribd lesen

0 Bewertungen0% fanden dieses Dokument nützlich (0 Abstimmungen)

4 Ansichten8 SeitenThe steel industry is developing a design-by-advanced analysis speciﬁcation for cold-formed steel construction. This eﬀort provides an opportunity to utilize the latest nonlinear structural analysis (advanced analysis) to design steel structures based on their overall system behaviour. This paper concerns the system reliability
calibrations of this design-by-analysis method, with a particular focus on cold-formed steel portal frames.

© All Rights Reserved

Als PDF, TXT **herunterladen** oder online auf Scribd lesen

Sie sind auf Seite 1von 8

Engineering Structures

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

advanced analysis

⁎

Francisco Sena Cardoso, Hao Zhang, Kim J.R. Rasmussen, Shen Yan

The University of Sydney, Australia

A B S T R A C T

The steel industry is developing a design-by-advanced analysis speciﬁcation for cold-formed steel construction. This eﬀort provides an opportunity to utilize the latest

nonlinear structural analysis (advanced analysis) to design steel structures based on their overall system behaviour. This paper concerns the system reliability

calibrations of this design-by-analysis method, with a particular focus on cold-formed steel portal frames. Four typical portal frames are considered. The system

reliability assessment takes into account all important random variables. A limit-state design criterion is developed which is consistent with a desired level of system

safety.

system strength given by the DDM. Moreover, the current Direct

Several steel structures design speciﬁcations permit design by ad- Strength Method is based on computational member elastic buckling

vanced nonlinear structural analysis. Advanced analysis has received stability analysis, while the DDM requires a fully nonlinear system

considerable attention from the research communities for several dec- analysis.

ades. Fully nonlinear analyses have been developed capable of accu- The Direct Design Method must satisfy certain modelling/analysis

rately capturing the true behaviour of steel structures subject to com- requirements articulated in the standards, as well as structural relia-

plex buckling modes. In professional practice, as early as 1998, the bility requirements. Appendix I of AISC360-10 [2] requires that the

Australian Standard AS4100 [1] included the design by advanced DDM must account for the natural variabilities in system, member and

analysis method. It allows internal actions to be obtained by nonlinear connection resistance, and provide a structural reliability for the frame

structural analysis. However, connection and section capacities are still no less than the current member-based design method. To fulﬁll this

required to be evaluated and checked. The scope of using advanced requirement, a resistance factor at the system level can be combined

analysis in AS4100 is limited to fully braced compact cross-sections. with the nominal frame strength predicted by the advanced analysis.

More recently, the American Steel Speciﬁcation AISC 360-10 [2] sig- The design equation is given by:

niﬁcantly broadened the scope of advanced analysis (referred to as

“inelastic analysis” in the Speciﬁcation). It permits the design of a steel

ϕs Rn ⩾ ∑i γi Qni (1)

frame to be based directly on the nonlinear structural analysis without where ϕs and Rn are the resistance factor and nominal resistance of the

the need for checking capacities of individual members. For this reason, system, respectively, and Qni and γi denote the structural load and the

the design by advanced analysis method is termed as the Direct Design corresponding load factor. The resistance Rn is the ultimate load-car-

Method (DDM) in this paper. In AISC 360-10, the advanced analysis rying limit of the frame at incipient instability of the frame. Rn is

method can be applied to non-compact cross-sections and members not computed using the nominal values of structural parameters. The

fully braced, provided the relevant limit states are captured in the structural failure risk due to the uncertainties in structural resistance is

analysis. The recently revised version of the Australian/New Zealand controlled through the use of ϕs. The current resistance factors stipu-

Standard for Cold-formed Steel Structures AS/NZS4600 [3] now also lated in the speciﬁcation, e.g. 0.9 for ﬂexural members, were developed

features provisions for the DDM by advanced analysis. for individual member safety check, as opposed to the system-level

The Direct Design Method is not to be confused with the Direct check in the design-by-analysis method. The system resistance factor

Strength Method, which is permitted in Australia/New Zealand [3], and needs to be determined by a structural reliability calibration procedure

North America [4]. The Direct Strength Method is a design method for conducted at the system level.

thin-walled members (columns and beams) [5]. The Direct Strength The system reliability implications of the DDM have been examined

⁎

Corresponding author.

E-mail address: shen.yan@sydney.edu.au (S. Yan).

https://doi.org/10.1016/j.engstruct.2018.12.054

Received 13 August 2018; Received in revised form 16 December 2018; Accepted 17 December 2018

0141-0296/ © 2018 Elsevier Ltd. All rights reserved.

F. Sena Cardoso et al. Engineering Structures 182 (2019) 164–171

in a limited number of case studies, e.g., [6–9]. Buonopane and Schafer yielding behaviour of the steel [13]. The nominal value of the yield

[6] investigated the system reliabilities of Load and Resistance Factor stress is 550 MPa, with a modulus of elasticity of 200 GPa.

Design (LRFD) and the DDM using 16 planar two-storey two-bay gravity For cold-formed steel sections, coiling, uncoiling, cold bending to

frames. The uncertainties in steel yield stress, dead and live loads were shape, and straightening of the formed member lead to a complicated

considered. The uncertainties in other parameters, e.g., Young’s mod- set of initial stresses in the section. Likewise, as a result of the manu-

ulus and sectional properties, were ignored. Zhang et al. [9] examined facturing process, the yield stress of the section corners is typically

the system reliabilities of a number of simple structures, including a enhanced, a phenomena commonly referred to as corner strength en-

continuous beam and four frames with diﬀerent levels of redundancy hancement. Regarding the modelling of residual stresses and corner

and capability of load redistribution. The study discussed the reliability strength enhancement in ﬁnite element analysis, three main approaches

implications of the LRFD and the second-order inelastic method in have been followed: (i) model the actual manufacturing process, by

AISC360-10. A similar work has been conducted by Thai et al. [7] for assuming the stress-strain curve of virgin steel, followed by the mod-

two low-rise planar frames with partially restraint connections. A elling of the actual coiling, uncoiling, roll-forming or press-braking

general reliability framework for assessing the system reliabilities of processes, e.g., in Pastor et al. [14], (ii) model residual stresses and

steel frames was developed in [10]. Using the reliability framework in corner strength enhancement independently, e.g., in Crisan et al. [15],

[10], systematic reliability calibrations were conducted for the DDM for and (iii) simply ignore simultaneously both eﬀects assuming that the

typical planar steel frames [11] and space steel frames comprising favourable eﬀect of corner strength enhancement is cancelled by the

members of locally stable hollow sections [12]. All these aforemen- unfavourable eﬀect of residual stress, e.g., in Sarawit [16], Sena Car-

tioned studies focused on structures comprising compact members with doso and Rasmussen [17]. The present study adopts the third approach,

full lateral bracing. assuming that the enhanced yield stress in the highly worked corner

The present study concerns the derivation of system resistance compensates the eﬀect of residual stress.

factors suitable for the limit state design by advanced analysis of cold- As observed from joint component tests [18], the moment-rotation

formed steel portal frames comprising members of locally unstable curves of typical bolted portal frame joints can be reasonably described

sections. Section 2 introduces the four baseline frames used for the by a multi-linear curve. Thus, a bi-linear curve in Fig. 2 is adopted to

reliability calibration, the nonlinear ﬁnite element (FE) structural model the partial rigidity of the joints. The bi-linear moment-rotation

models and the system failure modes. The four baseline frames are curve is deﬁned by four parameters, M1, M2, K1 and K2. The present

chosen to represent typical constructions of cold-formed portal frames. study does not consider joint failure. It is assumed that at the ultimate

Section 3 presents the framework of the reliability calibration, in- state of the frame, the joints have not reached their ultimate moment

cluding the probabilistic models of random variables, and the system capacities or their ductility limits.

reliability assessment tool. In Section 4, the reliability calibration re- Three types of initial geometric imperfections are relevant for a

sults are presented and discussed. System resistance factors consistent cold-formed frame, i.e., imperfections at frame, member and cross-

with a desired system reliability level are suggested. section levels. The member imperfections of a space member include

strong axis out-of-straightness (camber imperfection), weak axis out-of-

2. Baseline frames and FE models straightness (bow imperfection), and twist imperfection. Fig. 3 de-

monstrates the sectional imperfections, including the local and distor-

2.1. Descriptions of the frames tional imperfections.

It is assumed that the camber, bow and twist imperfections along

Wind design of four portal frames are considered to derive the the member length is described by a half sinusoidal (single wave) curve,

system resistance factor for the advanced analysis. The frames are de- i.e.,

noted by WF1 to WF4. Series WF1 and WF2 have a relatively short span π·z

of 8 m, comprising members made from single channels. Series WF3 αi (z ) = Ai ·sin ⎛ ⎞

⎝ L ⎠ (2)

and WF4 have a large span of 14 m, comprising members made from

back-to-back channels. Fig. 1 shows the layout of the frames. These in which αi represents the ith member imperfection (camber, bow, and

frames represent the common construction of cold-formed portal frames twist), z is the coordinate along the member, L represents the length

with typical values of span, heights and pitch angles as used in practice. between end restraint points, and Ai is the amplitude of the ith member

The geometries and cross-section sizes of WF1-WF4 are summarized in imperfection. For the cross-section distortional imperfection, the mag-

Table 1. The frames are selected to cover a variety of failure modes; this nitude (δ in Fig. 3) varies along the member by:

will be discussed in some details in Section 2.3.

π·z ⎞

The steel is considered as elastic-perfectly plastic. Sensitivity ana- δdis (z ) = Adis ·sin ⎛

⎜ ⎟

lyses suggested that the common failure modes of portal frames with ⎝ Ld ⎠ (3)

cold-formed steel sections are not particularly inﬂuenced by the post- in which δdis (z ) represents the distortional imperfection magnitude at

location z, Ld is the buckling half-wavelength for the distortional im-

perfection, and Adis is the maximum amplitude (at the middle of Ld).

Similarly, the variation of the local imperfection, δloc (z ) , along the

member is given by

π·z ⎞

δloc (z ) = Aloc ·sin ⎛

⎜ ⎟

⎝ Ll ⎠ (4)

and Aloc is the maximum amplitude.

The nominal parameters for the four frames are given in Table 2,

including the connection stiﬀness parameters and the initial geometric

imperfections. In particular, the design (nominal) value of frame out-of-

plumbness angle is 1/500 (as speciﬁed in AS4600 [3] and AISI S100-10

[4]). Design value of member out-of-straightness (camber and bow

Fig. 1. Geometry of the portal frame and wind load. imperfections) is L/1000. Note that the nominal frame models do not

165

F. Sena Cardoso et al. Engineering Structures 182 (2019) 164–171

Table 1

Dimensions and section sizes of the frames (unit: mm).

WF1 WF2 WF3 WF4

Hapex 6000 7500 9900 6900

Span 8000 8000 14,000 14,000

Column (H × W × t) C(302 × 96 × 1.5) C(203 × 76 × 2.4) 2C(203 × 76 × 2.4) 2C(254 × 76 × 1.2)

Rafter (H × W × t) C(352 × 108 × 2.4) C(352 × 108 × 3.0) 2C(352 × 108 × 3.0) 2C(352 × 108 × 3.0)

Table 2

Connection properties and initial geometric imperfections (nominal values).

WF1 WF2 WF3 WF4

Eave joint

M1 [kNm] 64 64 76 76

M2 [kNm] 76 76 92 92

K1 [kNm/rad] 6400 6400 7600 4800

K2 [kNm/rad 400 400 400 300

Apex joint

M1 [kNm] 56 56 68 68

M2 [kNm] 80 80 96 96

K1 [kNm/rad] 1400 1400 17,000 10,500

K2 [kNm/rad 1500 1500 1750 1125

Column base

M1 [kNm] 60 60 60 60

M2 [kNm] 100 100 100 100

Fig. 2. Moment-rotation curve of semi-rigid joints.

K1 [kNm/rad] 6000 6000 6000 6000

K2 [kNm/rad 450 450 450 450

Out-of-plane purlin spaced 1 m spaced 0.6 m spaced spaced 0.7 m

0.7 m

Out-of-plane girts spaced spaced spaced spaced

0.76 m 0.763 m 0.76 m 0.78 m

Frame Out-of- 1/500 1/500 1/500 1/500

plumb

Bow imp., L (m) 1/1000, 1/1000, 1/1000, 1/1000, 0.78

0.76 0.763 0.76

Camber imp., L (m) 1/1000, 4.6 1/1000, 6.1 1/1000, 7.6 1/1000, 4.6

Rafter initial geometric imperfections

Bow imp., L (m) 1/1000, 1 1/1000, 0.6 1/1000, 0.7 1/1000, 0.7

Camber imp., L (m) 1/1000, 3.6 1/1000, 3.6 1/1000, 6.8 1/1000, 4.8

while the nominal structural models do not include the twist or sec-

tional imperfections, the structural analysis models used in the relia-

bility calibrations do incorporate the local, distortional and twist im-

perfections. Therefore, possible inﬂuence of random twist and sectional

imperfections on system safety are reﬂected in the derived ϕs factor.

Structural modelling and analyses are carried out using the software

ABAQUS [19]. Rafters and columns are modelled using shell-elements

(S4R element in ABAQUS). For back-to-back cold-formed channels, the

interactions between two webs are simulated by creating a TIE con-

straint between the webs. Four elements are typically used cross the

width of each ﬂat plate. The geometric imperfections are incorporated

in the structural models directly by placing the relevant FE nodes at the

Fig. 3. Sectional initial geometric imperfections for single channel and double deﬁned imperfect geometry. The most critical combination of the di-

channels, (a) distortional imperfections, and (b) local imperfections.

rections of each type of geometric imperfections is determined and used

for the nominal structural model. For the design of a main-wind-force-

include the twist imperfection or the sectional imperfections. Sensi- resisting frame, external wind load pressure is applied to the windward

tivity analyses have shown that the sectional imperfection mainly af- wall, leeward wall and the roof, as shown in Fig. 1. It is assumed that

fects the failure modes involving sectional buckling, with a reduction of the gravity load has already included the self-weight of the structure.

system ultimate strength typically no greater than 3% [13]. For other The ultimate limit states of the frames are determined from the load-

failure modes involving no cross-sectional buckling, the inﬂuence of displacement (apex drift) response curves computed by static pushover

sectional imperfection is even smaller [13]. It should be noted that analyses. If a peak point exists in the load-displacement response, the

166

F. Sena Cardoso et al. Engineering Structures 182 (2019) 164–171

corresponding load is the ultimate load the frame can support. In the parameters (yield stress, elastic modulus and cross-section thickness)

case where the load-displacement response has no clear descending have been previously studied in the reliability calibrations of earlier

part, the point where the stiﬀness of the curve reduces to 5% of the versions of cold-formed seel design codes [21,22]. Other uncertain

initial stiﬀness is taken as the limit point [20]. In addition, a dis- parameters are discussed in some details herein.

placement-based criterion is also imposed, i.e., the system limit point

cannot exceed the point when the lateral displacement (at apex)

3.1. Semi-rigid joint properties

reaches 5% of the frame height. As will be discussed subsequently, in

most cases the ultimate strengths of the frames are deﬁned by the peak

Joints of cold-formed portal frames often use bolted connection

points in the load-displacement responses, except for a few cases in

systems. Although test data of cold-formed joints exist in the literature,

which the displacement criterion governed the frame strength limit.

e.g., [23–28], there are only very limited test results of nominally

Details of the nonlinear ﬁnite element models can be found elsewhere

identical joints to estimate the stochastic variability of joint properties.

[13].

Component tests of cold-formed portal frame joints were reported in

[24]. The connections were formed by connecting the webs and ﬂanges

2.3. Failure behaviours of the frames of back-to-back channels with gusset plates using bolts. Four nominally

identical apex joints and two eave joints were tested, and the joint

The four frames under given values of gravity loads are ﬁrst ex- moment-rotation responses were ﬁtted to a multi-linear curve. Based on

amined to investigate their failure modes. In all analyses, the given the four apex joint tests, the COV’s of M1, M2, K1 and K2 were found to

gravity loads are ﬁrst fully applied, then the wind load is applied in- be 0.087, 0.086, 0.232 and 0.042, respectively [24]. The results suggest

crementally until the failure of the system. that the initial stiﬀness K1 has a larger uncertainty than the other

Frame WF1 (under a gravity load of 1.2 kN/m): At a wind load of parameters. As the statistical data for cold-formed joints are very

3.04 kN/m, distortional buckling and yielding occur near the eave and scarce, the present paper assumes that cold-formed joints have similar

base of the right column. Frame sway instability occurs when the wind uncertainties as typical hot-rolled steel joints, for which a relatively

load increases to 4.57 kN/m. At the ultimate limit state, spatial plastic large supporting database is available, e.g. [29,30]. Typically, the COV

hinges formed near the eave and the base of the right column. of the initial stiﬀness of hot-rolled steel connections is on the order of

Frame WF2: under a constant gravity load of 2.4 kN/m, distortional 0.20–0.30. Adding engineering judgement, the present study assumes

buckling occurs near the eave of the right column when the applied that K1 has a relatively large COV of 0.30, K2 has a COV of 0.15, M1 has

wind load is about 5.36 kN/m. When the wind pressure is further in- a COV of 0.10 and M2 has a COV of 0.15. It is further assumed that the

creased to 8.74 kN/m, both columns start yielding and lateral-torsional means of these parameters are equal to their nominal values. All these

buckling occurs in the left column. The frame is at a state of incipient parameters are modelled by lognormal distributions.

collapse when the wind load reaches 10.91 kN/m.

The failure mode of frame WF3 (under a gravity load of 1.8 kN/m)

3.2. Initial geometric imperfections

is excessive drift. The 5% drift criterion is reached when the wind load

increases to 7.63 kN/m.

Survey data of out-of-plumbness of cold-formed frames do not exist

For frame WF4 (under a gravity load of 2.4 kN/m), yielding starts

in the archival literature. Instead, the present study uses the database of

near the base of the right column when the wind load is about 5.79 kN/

the out-of-plumbness of general steel structures [31,32]. A lognormal

m. At a wind load of 8.40 kN/m, the eave of the right column develops

random variable can describe the out-of-plumbness angle, with a zero

distortional buckling. Frame sway instability occurs when the wind

mean and a standard deviation of 1/610 [13].

load increases to 9.81 kN/m. At that time spatial plastic hinges develop

Table 4 summarizes the statistics of the amplitudes of member im-

near the base-plate and eave of the right column.

perfections reported in the literature. The results are quite scattered,

e.g., the mean major axis out-of-straightness varies between 1/1340

3. System reliability calibration and 1/4794. The weighted average values (weighting by the number of

data in each source) of the mean major and minor axes member out-of-

Table 3 presents the statistics of the uncertain variables, including straightness are 1/2490 and 1/2249, respectively. These two values are

elastic modulus, yield stress, cross-section thickness, geometric im- signiﬁcantly smaller than the typical design value of member out-of-

perfections, and joint properties. For comparison, the design (nominal) straightness in main stream speciﬁcations (1/1000 in AS4600 [3] and

values are also listed. The nominal values are used in the design stage, AISI S100-10 [4]). The weighted average values in the last row of

while the statistics are used for the probabilistic study and reliability Table 4 are used as the statistics of the member imperfection

calibration. The stochastic characteristics of some of the uncertain

Table 4

Table 3 Statistics of the amplitudes of member imperfections.

Nominal values and statistics of the uncertain parameters.

Ref. # of Camber (×L)* Bow (×L)* Twist ratio

Nominal mean Std Dev Dist. type data

Mean Std Dev Mean Std Dev Mean Std Dev

Frame out-of-plumb 1/500 0 1/610 Normal [o/m] [o/m]

Section thickness t tn tn 0.05tn LogNormal

Bow imp. L/1000 L/2250 L/3500 Normal [43] 32 1/4794 1/11470 1/1912 1/4646 – –

Camber imp. L/1000 L/2500 L/3300 Normal [44] 12 – – 1/2359 1/1842 – –

Twist (degree/m) 0 0.26 0.2 Normal [45] 31 1/3673 1/3984 1/2715 1/4475 0.1312 0.0984

Distortional imp 0 0.81t 0.62t Normal [46] 4 1/2510 1/2829 1/2462 1/6655 0.1083 0.0492

Local imp 0 0.375t 0.34t Normal [47] 89 1/1564 1/1937 1/2685 1/4475 – –

E En En 0.06En LogNormal [48] 23 1/4578 1/8403 1/4444 1/14190 – –

Fy Fyn 1.1Fyn 0.11Fyn LogNormal [49] 29 1/1340 1/3797 1/1148 1/4342 0.3609 0.1969

M1 M1n M1n 0.1M1n LogNormal [50] 210 1/3477 1/5643 1/2242 1/3054 0.3740 0.2428

M2 M2n M2n 0.15M2n LogNormal [51] 24 1/1427 1/2422 1/2457 1/5596 – –

K1 K1n K1n 0.3K1n LogNormal Weighted 1/2490 1/3829 1/2249 1/3716 0.2590 0.1968

K2 K2n K2n 0.15K2n LogNormal Average

Note: the subscript n represents the nominal value. * L = length between restraint points.

167

F. Sena Cardoso et al. Engineering Structures 182 (2019) 164–171

Table 5 95 were studied in [33]. It was found that Wmax has a mean of 0.90W50

Survey of local imperfection results. and a COV of 0.35, in which W50 represents the design wind load with a

Ref. Shape of cross-section # of data Local Distortional 50-year return period wind speed. An Extreme Type I distribution can

ﬁt Wmax. This wind load statistic has been used extensively in the de-

Mean StDev Mean StDev velopment of LRFD criteria in the United States. Note that the wind

directionality factor was estimated based on a simple geometric ana-

[43] Channels and hats 41 0.420t 0.720t – –

[44] Channels 12 0.400t 0.330t – – lysis at a time when there were no data on directionality eﬀects. Thus in

[44] – 45 – – 2.140t 1.290t [33], it was assumed that the directionality factor had a mean of 0.85

[52] Channels 45 – – 0.790t 0.270t (which is equivalent to its nominal value). Recent experimental re-

[53] Channels 11 0.550t 0.220t – – search has revealed that this assumption was conservative [34]. The

[53] Channels 33 – – 0.680t 0.300t

mean value of the directionality factor is closer to 0.72 (which is 0.835

[45] Channels 31 0.480t 0.150t 0.470t 0.170t

[46] Channels 4 0.224t 0.119t 0.750t 0.220t of its nominal value) for non-hurricane regions. Using the updated

[54] Channels 24 – – 1.360t 0.350t statistics of the directionality factor, the mean of Wmax in ASCE7-16

[47] Channels 88 0.280 t 0.070 t 0.260 t 0.490 t decreases from 0.9W50 to 0.75W50, with the COV unchanged [34], as

[51] Channels 24 0.306t 0.282t 0.350t 0.850t

shown in Table 6.

[55] Channels 30 0.480t 0.170t 0.890t 0.210t

Weighted average 0.375t 0.34t 0.81t 0.62t The wind load statistics for AS/NZS1170.2 were examined in

[35–38], and are summarized in Table 6. It is estimated that Wmax has a

Note: t = section thickness. mean of 0.68W50 and a COV of 0.39. These values are not dissimilar to

the updated American wind load statistics.

amplitudes, assuming normal distributions. In the probabilistic studies In the aforementioned studies, the means of Wmax were expressed

(simulations), the amplitudes of the member imperfections are sampled using W50 since the loading standards traditionally deﬁned the design

randomly, and then randomly assigned a direction. Eq. (2) is then used (nominal) wind loads using the 50-year-return period wind speeds.

to determine the member imperfections at any location along the However, in current American and Australian wind load standards, the

member length. nominal wind speeds are deﬁned using a high-return-period (i.e.,

The statistical data of the local imperfections from the literature are 500–700 years) as opposed to the traditional 50-year return period to

summarized in Table 5. Results from diﬀerent sources are combined to eliminate the need for the “cyclone factor” in earlier loading standards.

obtain the weighted average values. It was found that the local im- The wind load for T-year return period, denoted by WT, is approxi-

perfection has a mean value of 0.375t, and a standard deviation of mately related to W50 as [34]:

0.34t; the mean of the distortional imperfection is 0.81t with a standard

WT = W50 [0.36 + 0.1 ln(12T )]2 (6)

deviation of 0.62t. These statistics are adopted in the present study.

In ASCE7-16 [39], the nominal wind load Wn (for Risk Category II

structures) corresponds to a return period of 700 years, thus W700/

3.3. Wind load statistics

W50 = 1.6. Using W700 as the nominal wind load Wn, the mean of Wmax

is expressed as 0.75W50 = 0.75W700/1.6 = 0.47Wn. In AS/NZS1170.2

Wind loading is the major inﬂuence in cold-formed structural de-

[40], the design wind speed Wn is for a return period of 500 years (for

sign. It is therefore important to evaluate wind load statistics carefully.

common structures with Importance Level 2 and a 50-year service life).

In a general form, the wind load, W, on a structure may be written as

Therefore, W500/W50 = 1.5. Accordingly, the mean of Wmax is rewritten

W = c·Cp·E ·G·D·V 2 (5) as 0.68W50 = 0.45W500 = 0.45 Wn. It can be seen that the wind loads

in ASCE7-16 and AS/NZS1170.2 have similar statistics, i.e., the mean of

in which c is a constant, V = wind speed, Cp = pressure coeﬃcient, wind load Wmax is about 0.45Wn ∼ 0.47Wn, with a COV approximately

E = exposure factor, G = gust factor, and D = directionality factor. In 0.35–0.40. In the present paper, the wind load Wmax is modelled as an

design, the wind load of interest is the maximum wind load, denoted by Extreme Type 1 largest distribution. The mean and COV of Wmax are

Wmax, corresponding to the maximum wind speed, Vmax, to occur in taken as 0.47Wn and 0.40, respectively. This wind load probabilistic

50 years (typical service lifetime). All the parameters in Eq. (5) (except model covers both the wind loads in ASCE7-16 and AS/NZS1170.2.

for the constant c) are random. The estimation of the statistics of these

random parameters is a very diﬃcult task requiring extensive sup- 3.4. System reliability analysis method

porting data. In certain cases, the Delphi questionnaire method (expert

judgement) has been used to estimate the wind load statistics [33]. As In general, wind loads govern the design of light gauge steel frames.

more data becomes available, the wind load probabilistic models are Therefore, the ϕs for cold-formed portal frames is calibrated considering

updated [34]. Table 6 summarises the statistics for the wind pressure wind load combination. The reliability calibration of gravity loads only

parameters in the American loading standards ASCE7-95 and ASCE7-16 is not presented in this paper, but can be found elsewhere [13].

and the Australian standard AS/NZS1170.2-11. The safety of a structure can be quantiﬁed by its failure probability,

The wind load statistics for the American loading standard ASCE7- Pf. Direct Monte Carlo simulation is often used for evaluating structural

reliability [41]. However, this method is computationally prohibitive

Table 6 for the present study, as capturing the rare event of structural failure

Wind load statistics for ASCE7 and AS/NZS1170.2. requires a considerable amount of trails (on the order of 105 or larger),

ASCE7-95 ASCE7-16 AS/NZS 1170.2 each trail requiring a nonlinear shell element-based analysis. Thus, the

simpliﬁed reliability analysis method introduced in [10] is used in the

Mean/ COV Mean/ COV Mean/ COV present study. In this method, the stochastic characteristics of the lat-

nominal nominal nominal

eral resistance of the frame is ﬁrst estimated using random sampling

Cp 0.91 0.15 0.91 0.15 0.95 0.15 techniques, and then compared with the wind loads to approximate the

E 0.97 0.15 0.97 0.15 0.90 0.20 failure probability.

G 0.96 0.12 0.96 0.12 0.95 0.10 Wind-resistance design must satisfy Eq. (7), i.e.,

D 1.0 0.12 0.835 0.12 0.81 0.1

Vmax/V50 1.0 0.12 1.0 0.12 1.0 0.12 ϕs Rn = 1.2Dn + Wn (7)

Wmax/W50 0.90 0.35 0.75 0.35 0.68 0.39

where Dn = dead load, and Wn = wind load. To evaluate the system

168

F. Sena Cardoso et al. Engineering Structures 182 (2019) 164–171

reliability, the limit state function corresponding to Eq. (7) can be ex- Table 7

pressed as: Statistics of lateral resistance and system reliability index of WF1.

in which Rw represents the lateral capacity of the frame (simultaneously 1.2 0.9 3.927 4.575 0.127 2.43

subjected to the dead load Dn). The system fails if Rw is smaller than 0.8 3.414 2.79

0.7 2.902 3.21

Wmax.

The core of the reliability analysis is to obtain the stochastic model 0.8 0.9 4.169 4.687 0.147 2.29

for Rw. This can be achieved using the random sampling technique 0.8 3.654 2.63

0.7 3.139 3.20

through the following steps:

0.45 0.9 4.377 4.915 0.145 2.29

0.8 3.857 2.64

1. Generate a frame with randomly sampled geometric, material and

0.7 3.335 2.99

stiﬀness properties, and loaded with the dead load Dn. Note that the

dead load is treated deterministically herein since its uncertainty is R̄w = mean of Rw, VR = COV of Rw.

signiﬁcantly smaller than that of wind load.

2. Carry out an advanced analysis to compute the frame’s lateral re- Step 4. Repeat Step 2 and Step 3 for each combination of Dn and ϕs.

sistance Rw of the frame. For each Dn, a ϕs versus βs curve is plotted. The average curve for the

3. Repeat Steps 1 and 2 to obtain samples of the lateral resistance Rw, three values of Dn is used to represent the relationship between ϕs

and compute its empirical statistics (sample mean, standard devia- and βs for the frame.

tion, and distribution type).

This procedure is demonstrated using the example of WF1. Table 7

Step 1 can be performed using the standard random sampling presents the values of Wn associated with the three values of Dn and

technique, or more eﬃcient techniques such as Latin Hypercube three values of ϕs, i.e. a total of nine combinations of Dn and ϕs. For

Sampling (LHS). This study typically used 200 LHS simulations to cover each pair of (Dn, Wn, ϕs), the frame is at its ultimate limit state ac-

the probability space of Rw. Once the statistics of Rw is obtained, the cording to the design Eq. (7).

system failure probability is equivalent to evaluating the probability Next, a sample distribution of the lateral resistance of WF1 under

that Rw is smaller than Wmax. This represents the simplest form of re- the (deterministic) nominal dead load Dn is generated using 200 LHS

liability evaluation, and can be computed easily [41]. simulations. Since the system lateral resistance is generally inﬂuenced

Limit state structural design has traditionally used the reliability by the gravity loads acting on the frame, the probabilistic analyses of

index, β, to quantify the structural reliability. The probability of failure the system lateral resistance need to be conducted for all three values of

Pf is related to β by β = Φ−1(1 − Pf), and Φ denotes the cumulative Dn. Fig. 4 plots the histograms of the lateral resistance (Rw) of WF1

distribution of a unit normal [22,41]. In this paper, the system relia- under three values of Dn (1.2 kN/m, 0.8 kN/m and 0.45 kN/m). The

bility index is denoted by βs, with the subscript “s” to emphasise it is a mean values and COVs of Rw for WF1 are listed in Table 7. It suggests

“system” reliability. that the COVs of Rw under diﬀerent Dn are quite similar, in a narrow

range of 0.13 and 0.15. Rw can be modelled as a lognormal.

4. Reliability calibration results With the probabilistic information of Rw and Wmax, the reliability

indices of WF1 for each case of ϕs and Dn are computed using the direct

Using the reliability analysis method presented in Section 3, the Monte Carlo technique and listed in the last column of Table 7. For a

system reliability index βs of a given system can be computed, and the prescribed target βs, the required ϕs can be directly obtained from

relation between βs and ϕs can be established. Obviously, the required Table 7. For example, to achieve a target βs of 2.5, ϕs needs to be 0.855

value of ϕs for achieving a given target reliability index may not be the for WF1.

same for diﬀerent frames. Even for the same frame, the required re- Tables 8–10 summarize the loading information, lateral resistance

sistance factor depends on the loading scenario, notably the wind-to- models, and reliability analysis results of the other three baseline

gravity load ratio. Since wind loads are more uncertain than the dead frames. It can be seen that the uncertainty of WF3’s ultimate strength is

load, the required resistance factor would decrease as the wind load relatively low, about 0.05–0.06. Similar observation is also made in the

becomes more dominant. However, for design purposes, a single re- case of WF2 under the particular dead load Dn of 3.6 kN/m. This is

sistance factor is desirable. To achieve this goal, the reliability cali- because the failures of WF3 and WF2 (with Dn of 3.6 kN/m) are con-

bration considers three diﬀerent levels of design dead load for each of trolled by the deformation criterion. This failure mode is less inﬂuenced

the four baseline frames. The three values of dead load represent (re- by the uncertainty of material strength (Fy). Also, the randomness in

latively) heavy, medium and light dead loads. Accordingly, the design material stiﬀness (Young’s modulus) is signiﬁcantly less than the

wind load that the frame can support becomes more dominant as the variability associated with steel yield stress. On the other hand, the

applied dead load is reduced. Under each design dead load, three values failure of other frames are aﬀected more by the randomness in both

of ϕs (0.7, 0.8 and 0.9) are considered to establish the link between ϕs material strength and stiﬀness. The COVs for the ultimate strengths of

and system reliability index βs. these frames vary between 0.11 and 0.18, with a typical value of 0.15.

For a structural system designed according to Eq. (7), the linkage Table 11 summarizes the average values of βs (at three levels of dead

relating βs and ϕs is determined using the following four steps. load) for the four frames considering three possible system resistance

factors (ϕs = 0.7, 0.8 and 0.9). For a given value of ϕs, the reliability

Step 1. Establish a “nominal” FE structural model using the nominal indices for the four frames are comparable. For instance, for a ϕs value

values of frame properties. For a prescribed design dead load Dn and of 0.90, the reliability indices of the four frames vary between 2.34 and

a given system resistance factor ϕs, choose the nominal wind load 2.56, with an average value of 2.41. Alternatively, the required ϕs for

Wn such that the frame satisﬁes Eq. (7). achieving a certain reliability index can be readily obtained from

Step 2. Perform a probabilistic analysis (simulations) to estimate the Table 11.

statistics of the lateral resistance Rw of the frame designed in Step 1 Hot-rolled steel members subject to dead and live loads have a

(see Section 3.4). target reliability index approximately 2.6–2.8, and about 2.5 under

Step 3. Evaluate the reliability associated with the limit state: g wind loads in the existing AISC LRFD [21,22,42]. Historically, the

() = Rw − Wmax.

169

F. Sena Cardoso et al. Engineering Structures 182 (2019) 164–171

Table 9

Statistics of lateral resistance and system reliability index of WF3.

Dn ϕs Wn R̄w VR β

0.8 6.002 2.71

0.7 5.212 3.10

0.8 6.107 2.69

0.7 5.317 3.13

0.8 6.175 2.7

0.7 5.387 3.05

Table 10

Statistics of lateral resistance and system reliability index of WF4.

Dn ϕs Wn R̄w VR β

0.8 7.495 2.98

0.7 6.430 3.37

0.8 7.813 2.88

0.7 6.745 3.24

0.8 8.567 2.72

0.7 7.283 3.11

Table 11

Reliability indices βs of four frames (average of three levels of Dn).

ϕs WF1 WF2 WF3 WF4 Average

0.8 2.69 2.65 2.70 2.86 2.73

0.7 3.14 2.94 3.09 3.24 3.10

those of hot-rolled steel members, partly due to the higher live-to-dead

load ratio in cold-formed steel members. Under gravity loads only, cold-

formed steel structural members are calibrated to fulﬁl a target member

reliability index of approximately 2.5 [3,4]. For the Direct Design

Method, if a target reliability index for frame systems can be estab-

lished, the ϕs versus βs relationship in Table 11 can assist the code

writers to arrive at a suitable value of ϕs. For instance, to achieve a βs

value of 2.5 under wind loads, the required ϕs for WF1 to WF4 would be

0.856, 0.856, 0.861 and 0.910, respectively. The average value of ϕs is

0.87. Considering that resistance factors in design standards are cus-

Fig. 4. Histograms of lateral resistance (Rw) of WF1. tomarily rounded to the nearest 0.05, a ϕs of 0.85 may be adopted. The

ﬁnal choices of target system reliability indices for the advanced ana-

Table 8 lysis method ultimately rest with the speciﬁcation committee.

Statistics of lateral resistance and system reliability index of WF2.

Dn ϕs Wn R̄w VR β 5. Conclusion

3.6 0.9 12.223 13.328 0.071 2.36

0.8 11.135 2.63 A reliability analysis has been presented for designing cold-formed

0.7 9.918 2.91 steel portal frames by the Direct Design Method using advanced non-

1.8 0.9 9.245 10.602 0.105 2.44 linear structural analysis. The system reliability calibration is based on

0.8 8.400 2.70 four frames representing typical constructions of cold-formed steel

0.7 7.568 2.97 portal frames, with various failure modes. Through simulation techni-

1.0 0.9 8.580 9.546 0.107 2.34 ques, the probabilistic characteristics of the ultimate lateral strengths of

0.8 7.717 2.63 the baseline frames are determined. It is found that the ultimate

0.7 6.853 2.95 strength of frame WF3 has a relatively small uncertainty, with a COV of

R̄w = mean of Rw, VR = COV of Rw.

about 0.05. This is because the failure of frame WF3 is controlled by the

deformation criterion, thus less inﬂuenced by the uncertainty of

170

F. Sena Cardoso et al. Engineering Structures 182 (2019) 164–171

material strength. The limit states of other frames are deﬁned by the [22] Ellingwood BR. LRFD: implementing structural reliability in professional practice.

peak points of the load-displacement responses, and thus aﬀected by Eng Struct 2000;22:106–15.

[23] Chung KF, Ho HC, Wang AJ, Yu WK. Advances in analysis and design of cold-

the randomness in both material strength and stiﬀness. The ultimate formed steel structures. Adv Struct Eng 2016;11(6):615–32.

strengths of these frames have coeﬃcients of variation in the range [24] Zhang X. Steel portal frames with locally unstable members PhD thesis The

between 0.11 and 0.18, with a typical value of 0.15. University of Sydney; 2014.

[25] Dubina D. Structural analysis and design assisted by testing of cold-formed steel

The derived values of ϕs for the four frames are comparable. If as- structures. Thin-Walled Struct 2008;46(7–9):741–64.

suming a system reliability index of 2.5 for wind design, the values of ϕs [26] Chung KF, Lau L. Experimental investigation on bolted moment connections among

of the four frames vary between 0.856 and 0.910; the averaged ϕs is cold formed steel members. Eng Struct 1999;21(10):898–911.

[27] Wong MF, Chung KF. Structural behaviour of bolted moment connections in cold-

0.87. This system resistance factor may be used for cold-formed portal formed steel beam-column sub-frames. J Constr Steel Res 2002;58(2):253–74.

frames with similar failure modes to those examined in the present [28] Blum HB. Long-span cold-formed steel double channel portal frames PhD thesis The

work. The variation of system reliability level (βs) as a function of the University of Sydney; 2017.

[29] Rauscher TR, Gerstle KH. Reliability of rotational behavior of framing connections.

frame resistance factor ϕs developed in this paper can help the speci-

Eng J Am Inst Steel Construct 1992;29(1):12–9.

ﬁcation committees to arrive at the suitable ϕs for the Direct Design [30] Bjorhovde R, Brozzetti J, Colson A. Connections in steel structures: behaviour,

Method. strength and design. CRC Press; 1988.

[31] Beaulieu D, Adams PF. The results of a survey on structural out-of-plumbs. Can J

Civ Eng 1978;5(4):462–70.

Acknowledgments [32] Lindner J, Gietzelt R. Assumptions for imperfections for out-of-plumb of columns.

Stahlbau 1984;53:97–102.

This research was supported by the Australian Research Council [33] Ellingwood BR, Tekie PB. Wind load statistics for probability-based structural de-

sign. J Struct Eng 1999;125(4):453–63.

under Discovery Grant DP110104263. Francisco Sena Cardoso was [34] McAllister TP, Wang N, Ellingwood BR. Risk-informed mean recurrence intervals

provided with a supplementary scholarship by the Centre for Advanced for updated wind maps in ASCE 7–16. J Struct Eng 2018;144(5):06018001.

Engineering at the University of Sydney. This support is gratefully ac- [35] Pham L, Holmes JD, Leicester RH. Safety indices for wind loading in Australia. J

Wind Eng Ind Aerodyn 1983;14(1–3):3–14.

knowledged. [36] Holmes JD. Wind loads and limit states design. Civ Eng Trans, IE Aust

1985;27:21–6.

References [37] Henderson DJ, Ginger JD. Vulnerability model of an Australian high-set house

subjected to cyclonic wind loading. Wind Struct 2007;10(3):269–85.

[38] Holmes JD, Kwok KCS, Ginger JD. Wind loading handbook for Australia and New

[1] AS4100. AS4100-1998 Steel structures. Sydney, Australia: Standards Australia; Zealand-background to AS/NZS 1170.2 wind actions. Report No. AWES-HB-001-

1998. 2012, Australian Wind Engineering Society; 2012.

[2] AISC. AISC 360-10 Speciﬁcation for structural steel buildings. Chicago, Illinois: [39] ASCE7-16. Minimum design loads and associated criteria for buildings and other

American Institute of Steel Construction (AISC); 2010. structures. Reston, VA: American Society of Civil Engineers; 2017.

[3] AS4600. Cold-formed steel structures. Sydney, Australia: Standards Australia; 2018. [40] AS/NZS1170.2. Structural design actions, Part 2: wind actions, Standards Australia/

[4] AISI-S100. North American speciﬁcation for the design of cold-formed steel struc- Standards New Zealand; 2011.

tural members. American Iron and Steel Institute (AISI); 2012. [41] Melchers RE. Structural reliability analysis and prediction. John Wiley & Sons;

[5] Schafer BW. Review: the direct strength method of cold-formed steel member de- 1999.

sign. J Constr Steel Res 2008;64(7–8):766–78. [42] Ellingwood BR. Probability-based codiﬁed design: past accomplishments and future

[6] Buonopane SG, Schafer BW. Reliability of steel frames designed with advanced challenges. Struct Saf 1994;13:159–76.

analysis. J Struct Eng 2006;132(2):267–76. [43] Dat DT, Peköz T. The strength of cold-formed steel columns. Technical Report No.

[7] Thai HT, Uy B, Kang WH, Hicks S. System reliability evaluation of steel frames with 80-4. Cornell University; 1980.

semi-rigid connections. J Constr Steel Res 2016;121:29–39. [44] Mulligan GP. The inﬂuence of local buckling on the structural behavior of singly-

[8] Zhang H, Ellingwood BR, Rasmussen KJR. System reliabilities in steel structural symmetric cold-formed steel columns. Technical Report No. 83-1. Cornell

frame design by inelastic analysis. Eng Struct 2014;81:341–8. University; 1983.

[9] Zhang H, Liu H, Ellingwood BR, Rasmussen KJR. System reliabilities of planar [45] Young B. The behaviour and design of cold-formed channel columns PhD thesis

gravity steel frames designed by the inelastic method in AISC 360–10. J Struct Eng Sydney, Australia: The University of Sydney; 1997.

2018;144(3):1–8. [46] Dubina D, Ungureanu V. Eﬀect of imperfections on numerical simulation of in-

[10] Zhang H, Shayan S, Rasmussen KJR, Ellingwood BR. System-based design of planar stability behaviour of cold-formed steel members. Thin-Walled Struct

steel frames, I: reliability framework. J Constr Steel Res 2016;123:135–43. 2002;40(3):239–62.

[11] Zhang H, Shayan S, Rasmussen KJR, Ellingwood BR. System-based design of planar [47] Shiﬀeraw Y, Vieira Jr LCM, Schafer BW. Compression testing of cold-formed steel

steel frames, II: reliability results and design recommendations. J Constr Steel Res columns with diﬀerent sheathing conﬁgurations. Annual stability conference,

2016;123:154–61. structural stability research council. 2010.

[12] Liu W, Zhang H, Rasmussen K. System reliability-based Direct Design Method for [48] Rossi B, Jaspart J, Rasmussen K. Combined distortional and overall ﬂexural-tor-

space frames with cold–formed steel hollow sections. Eng Struct 2018;166:79–92. sional buckling of cold-formed stainless steel sections: experimental investigations.

[13] Sena Cardoso F. System reliability-based criteria for designing cold-formed steel J Struct Eng 2010;136(4):354–60.

structures by advanced analysis PhD thesis The University of Sydney; 2016. [49] Schafer BW. Sheating braced design of wall studs. Research Report RP13-1.

[14] Pastor JBMM, Roure F, Casafont M. Residual stresses and initial imperfections in Washington, D. C.: American Iron and Steel Institute; 2013.

non-linear analysis. Eng Struct 2013;46:493–507. [50] Zeinoddini V. Geometric imperfections in cold-formed steel members PhD thesis

[15] Andrei Crisan VU, Dubina Dan. Behaviour of cold-formed steel perforated sections Johns Hopkins University; 2011.

in compression: Part 2—numerical investigations and design considerations. Thin- [51] Vieira LCM, Schafer BW. Behavior and design of sheathed cold-formed steel stud

Walled Struct 2012;61:97–105. walls under compression. J Struct Eng 2013;139(5):772–86.

[16] Sarawit AT. Cold-formed steel frame and beam-column design PhD thesis Ithaca, [52] Lau SCW, Hancock GJ. Inelastic buckling of channel columns in the distortional

NY: Cornell University; 2003. mode. Thin-Walled Struct 1990;10(1):59–84.

[17] Sena Cardoso F, Rasmussen KJR. Finite element (FE) modelling of storage rack [53] Kwon Y, Hancock G. Tests of cold-formed channels with local and distortional

frames. J Constr Steel Res 2016;126:1–14. buckling. J Struct Eng 1992;118(7):1786–803.

[18] Zhang X, Rasmussen KJR, Zhang H. Experimental investigation of locally and dis- [54] Wang H, Zhang Y. Experimental and numerical investigation on cold-formed steel

tortionally buckled portal frames. J Constr Steel Res 2016;122:571–83. C-section ﬂexural members. J Constr Steel Res 2009;65(5):1225–35.

[19] ABAQUS. version 6.10. Simulia, Dassault Systèmes, France; 2010. [55] Peterman KD. Experiments on the stability of sheathed cold-formed steel studs

[20] Ziemian RD, McGuire W, Deierlein GG. Inelastic limit states design. Part I: planar under axial loading and bending Master thesis Baltimore, Maryland: Johns Hopkins

frame studies. J Struct Eng 1992;118(9):2532–49. University; 2012.

[21] Ellingwood BR, MacGregor JG, Galambos TV, Cornell CA. Probability based load

criteria: load factors and load combinations. J Struct Division ASCE 1982;108(5).

171