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

net/publication/335638722

Geometric imperfection models for CFS structural members, Part I: Comparative

review of current models

Preprint · September 2019

DOI: 10.13140/RG.2.2.26255.53927

CITATIONS READS

0 118

4 authors, including:

Shafee Farzanian Arghavan Louhghalam

Johns Hopkins University University of Massachusetts Dartmouth
7 PUBLICATIONS   8 CITATIONS    35 PUBLICATIONS   205 CITATIONS   

SEE PROFILE SEE PROFILE

Benjamin Schafer
Johns Hopkins University
269 PUBLICATIONS   4,760 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

GOALI: Enabling Advanced Wind Turbine Tower Manufacturing with Reliability-Based Design View project

Steel Foam View project

All content following this page was uploaded by Shafee Farzanian on 05 September 2019.

The user has requested enhancement of the downloaded file.

 Geometric imperfection models for CFS structural members,
Part I: Comparative review of current models

S. Farzanian1 , A. Louhghalam1 , B.W. Schafer2 , M. Tootkaboni1,⇤

University of Massachusetts Dartmouth, North Dartmouth, MA
Johns Hopkins University, Baltimore, MD

Abstract

The role of unavoidable geometric imperfections complicates the load-displacement response of thin-walled cold-
formed steel (CFS) structural members. While the inclusion of geometric imperfections in shell finite element col-
lapse simulations has long been the subject of research, no consensus exits on modeling their full three-dimensional
spatial pattern. This paper, as the first part in a two-part series, provides a summary of e↵orts geared towards charac-
terizing and modeling geometric imperfections. We provide the necessary ingredients for systematic characterization
of elastic buckling mode shapes, due to their frequent use in approximating geometric imperfections, and make a con-
scious e↵ort to express the existing geometric imperfection models found in the literature through three components:
imperfection shape, magnitude, and combination coefficient. Di↵erent choices and selections of these components,
pertaining to either the intrinsic buckling behavior of the members and/or physical measurements are critically exam-
ined. The details required to develop high fidelity shell finite element nonlinear collapse analysis of CFS members
seeded with the multitude of di↵erent geometric imperfection models are presented. We then perform a comparative
study of collapse analysis results obtained from adopting di↵erent imperfection modeling strategies. The noticeable
variability in the load carrying capacity and the stability behavior highlights an urgent need for reliable probabilistic
frameworks for this class of problems. A stochastic framework to generate samples of imperfect shell FE models
from limited available data and validate the code-prescribed or practice-oriented geometric imperfection models is
presented in a companion paper (see Part II [1]).
Keywords: Geometric imperfection models, shell FE collapse analysis, CFS structural members, modal
characterization, elastic buckling modes

⇤ Corresponding author
Email addresses: shafee.farzanian@umassd.edu (S. Farzanian), alouhghalam@umassd.edu (A. Louhghalam), schafer@jhu.edu
(B.W. Schafer), mtootkaboni@umassd.edu (M. Tootkaboni)
1 University of Massachusetts Dartmouth
2 Johns Hopkins University

Preprint submitted to Journal of Thin-walled Structures January 5, 2019

1. Introduction

Material minimization, driven by the need to push the efficiency in performance and cost to its boundaries, often
leads to thin-walled structures with ultra-slender features that can undergo sudden large deformation under loading,
known as mechanical instabilities or buckling. Today, applications of thin-walled structural members encompass
a wide and growing field from structural engineering to material science. Examples can be seen in buildings and
bridges [2], ships, rockets, and aircraft industries [3, 4], liquid and bulk material storage vessels [5], pipes, domes, and
sandwich panels [6], as well as lattice materials made of micro- and nano-tubes [7, 8], and origami-based structures
and materials [9, 10]. A feature of these structures is that their response and deformation processes under load are
strongly related to their geometry. The geometric susceptibility has traditionally been considered as a negative attribute
and associated with undesirable decrease in load carrying capacity necessitating the consideration of exact physical
geometry (rather than nominal values) in collapse analysis. This has long been recognized, e.g., by von Kármán
[11] who devised the first practical strength prediction for thin-walled plates and later Winter [12] who empirically
corrected the strength prediction with implicit consideration of geometric imperfections, and persists in various forms
in design today.
An important class of thin-walled structures, is Cold Formed Steel (CFS) structural members. These are prod-
ucts made by roll-forming or press-braking thin sheets of steel into a variety of shapes at room temperature and have
many advantages such as high strength-to-weight ratio, high sti↵ness, long-term durability, high dimensional stability,
sustainability, non-combustibility, and fast and easy installation [13]. Like any other fabricated structure, however,
CFS members possess imperfections that are a result of manufacturing, shipping, storage, and construction processes
[14, 15]. These imperfections, often random in nature, are either in the form of di↵erences between the actual and
nominal material properties such as modulus of elasticity, yield stress, and residual stresses and strains (material im-
perfections) [16, 17] or di↵erence between the actual and nominal shape (geometric imperfections) [18, 19]. The
load carrying capacity of CFS members, like any other thin-walled structure, is highly influenced by geometric non-
linearity, in addition to material nonlinearity. Initial geometric imperfections, with their complex three-dimensional
spatially varying forms, therefore, play a key role in ultimate strength, sti↵ness, and post-buckling behavior of CFS
members. As a result, close replication of mechanical response observed in experiments, requires advanced measur-
ing techniques for accurate characterization of geometric imperfection, and use of proper simulation procedures that
include such imperfections.
This paper provides a comprehensive review of e↵orts geared towards characterizing and modeling geometric
imperfections in CFS members. We particularly focus on the e↵ect of adopting di↵erent imperfection models on
nonlinear response and make a conscious e↵ort to provide a unified platform (to the extent possible) where available
models are expressed in the form of an expansion comprising of three ingredients: imperfection shape, imperfection
magnitude, and combination coefficient. The organization of this paper is as follows. Section 2 provides the necessary
details for categorization of elastic buckling mode shapes as approximate forms for mimicking the three dimensional

2
shape of geometric imperfections. Section 3 examines geometric imperfections through contrasting the two key con-
cepts of “recorded” (or as-measured) imperfection and “modeled” imperfection. Section 4 reviews the most common
geometric imperfection models in the literature and, together with Section 5 which provides the details of nonlinear
shell buckling analysis, lays the groundwork for a study on the role of geometric imperfection models in nonlinear
collapse and load carrying capacity of CFS structural members in Section 6. Concluding remarks are provided in
Section 7.

2. Elastic buckling modes: the link between nonlinear behavior and geometric imperfections

Buckling mode shapes are inherent attributes of a structural member and depend on several factors including its
cross-sectional profile, length, loading and boundary conditions. The nonlinear behaviour of CFS members, similar
to other thin-walled structures, is sensitive to the presence of geometric imperfection shapes that are affine to these
buckling modes. Selected suites of these modes are therefore utilized in most current geometric imperfection models.
Various analytical and numerical methods, with di↵erent limitations and capabilities [20], exist for the elastic
buckling analysis of CFS members. Analytical methods are simple, but inherently limited in their applicability [21–
anl
23]. They provide the buckling mode shapes i for either individual plates or plate assemblages for some particular
boundary conditions. More general computational tools such as the plate/shell Finite Element Method (FEM), Finite
Strip Method (FSM), constrained Finite Strip Method (cFSM), Generalized Beam Theory (GBT), and constrained
Finite Element Method (cFEM), on the other hand, can handle a wide variety of loading, geometry and boundary
conditions.
Shell FEM, for example, can handle arbitrary geometry, loading, and boundary conditions. The challenge in using
FEM, however, is that it provides no mechanical organization of the many possible buckling modes. In addition, the
FE
number of independent FEM buckling modes i can be very large (equal to the number of nodes multiplied by the
number of degrees of freedom (DOF); typically six). For the aim of geometric imperfection modeling, this requires the
inclusion of many buckling modes to capture the full stability solution, which makes the selection of modes subjective
and less straightforward.
Semi-analytical methods such as FSM o↵er a powerful alternative for stability analysis of prismatic members.
FSM was proposed by Cheung [24], popularized by Hancock [25], and made accessible by the open source program
CUFSM [26]. FSM uses special shape functions for longitudinal interpolation of the unknown displacement field
and as such achieves reduced computational cost, when compared to FEM in which full longitudinal discretization is
FS
needed. The number of independent FSM buckling modes i is modest (equal to the cross-sectional node numbers
multiplied by four–DOF per node–and by the number of longitudinal terms retained in the series expansion for the
displacement field). For geometric imperfection modeling, FSM provides a more convenient and less subjective
procedure (when compared with FEM), in particular for cases where the solution can be fully represented with only
a few terms in the expansion. However, the modes provided by FSM are “quasi-pure”; without specialized treatment

3
the buckling modes extracted from an FSM solution may possess features that pertain to the interaction/coupling
between local-distortional modes and/or distortional-global modes (see [27]). While researchers continue to develop
specialized variants of FSM such as those based on spline or Chebyshev interpolation or reduced kernel particle
method [28–30], other solution techniques such as cFSM, GBT, and cFEM (see [31–37]) have been developed which
dec
provide decomposed buckling modes i through enforcing a series of mechanics-based criteria to arrive at the
G D L S T/O
individual pure modes, i.e. global ( i ), distortional ( i ), local ( i ), and other ( i ) buckling modes.

2.1. Efficient decryption of stability behavior via FSM and cFSM

The key feature of the semi-analytical techniques is that the stability solution can be captured, efficiently, with min-
imum number of buckling modes. This is of particular importance when geometric imperfections are to be simulated
through a limited number of eigen modes. While in practical situations buckling modes rarely appear in isolation, and
typically some modal coupling is present, utilizing pure buckling mode shapes, provided by FSM and its constrained
variant (cFSM), GBT or cFEM, has found its place as a more efficient and objective procedure for geometric imper-
fection modeling. In what follows we provide a brief overview on how FSM and cFSM allow for efficient exploration
of a stability solution and the identification of pure eigen modes. This overview is preceded with a description of the
modes of instability in CFS members to ensure a smooth transition for the reader less familiar with the mechanics of
this class of structural members.

2.1.1. Global modes

From an analytical standpoint, the global buckling of thin-walled CFS members has been studied in earlier works
G
[21, 23, 38–40]. Generally speaking, a global buckling mode ( i ) involves two orthogonal rigid-body translations of
the cross-section and one rigid-body rotation of the cross-section and is completely defined by the DOF associated
with main nodes only, i.e. those nodes at the junction or end of the flat plates comprising the cross-section [41]. A set
of three coupled di↵erential equations governing the flexural and torsional buckling results in a characteristic cubic
equation which yields the critical global buckling modes and loads. CFS members can buckle by flexure about one of
G G
the principle axes (flexural buckling), including pure minor-axis ( Fn ) or major-axis ( F j) flexural buckling, or twist
G
about shear center (pure torsional buckling T ), or a combination thereof (flexural-torsional buckling).
G
Depending on the type of cross-section (see Table 1), the set of global eigenmodes i contains: (i) three pure
G G G
global buckling modes ( Fn , F j, T ), for doubly symmetric sections (I-section, square or rectangular tube, cylinder,
etc.) or point-symmetric sections (Z-section, cruciform, etc.), (ii) one pure flexural buckling mode about the axis
G
of non-symmetry ( Fn ) and two flexural-torsional buckling modes about the axis of symmetry and the longitudinal
G G
axis ( F jT 1 and F jT 2 ), for singly-symmetric sections like many CFS members with open cross sections (plain/lipped
channel, hat- or T-section, etc.), (iii) three coupled flexural-torsional buckling modes, minor-axis flexural-torsional
G G
buckling mode ( FnT ), major-axis flexural-torsional buckling mode ( F jT ), and minor- major-axis flexural-torsional
G
buckling mode ( FnF jT ), for non-symmetric sections. In all these cases the order of the three global buckling modes

4
and that of the associated critical loads depend on section properties and e↵ective member length. As an example, for
G
a lipped channel section, if the torsional rigidity of the section is low, the flexural-torsional buckling failure ( F jT 1 )
G
occurs first or, if the member is too long, the minor-axis flexural buckling FnT happens first.

2.1.2. Distortional modes

Unlike global or local buckling modes, the definition of distortional buckling modes (also traditionally known as
D
sti↵ener buckling or local-torsional buckling modes [42]) is more challenging. Distortional buckling modes i involve
changes in the cross-sectional shape characterized by rotation of the flange at the flange/web junction in members with
edge sti↵ened elements (e.g., lipped channels or Z sections) or the displacement of an intermediate sti↵ener normal to
its plane. Distortional buckling emerges at considerably longer wavelengths than the local buckling modes but often
shorter wavelengths than the global buckling mode.
To determine the critical load and mode shape for distortional buckling (exact, approximate, or simplified) ana-
lytical models with di↵erent assumptions and treatments of the flange and web have been suggested by researchers.
Lau and Hancock [43], for example, proposed an analytical model that is based primarily on the assumption that the
flange acts as an isolated column undergoing flexural-torsional buckling, while the web provides elastic restraint to
the flange. Schafer and Pekoz [44] developed an analytical model to allow for the impact of applied stresses on the
web’s rotational sti↵ness, thus allowing for the case when distortional buckling is triggered by instability of the web
as opposed to the flange; see [42] for more details. These analytical methods, while viable for a number of scenarios
have limited applicability. Numerical methods such as cFSM that enforce a set of mechanics-based criteria are there-
fore often used for obtaining the distortional buckling modes. The number of distortional modes, which similar to
global buckling modes are defined by the degrees of freedom associated with main nodes only [41], depends on the
cross-section topology (see Table 1) and can be calculated as [45]:

ND = nm 4 (1)

where nm is the number of main nodes; those nodes at the junction or end of the flat plates comprising the section. For
example, an AISI S200 lipped channel stud [46] with four internal fold lines (five flats and nm = 6) has two distortional
buckling modes, a Sigma stud with eight fold lines (nine flats and nm = 10) has six distortional buckling modes and
so on. Similar rules can be derived for other specific cross-section topologies, e.g. hollow sections. However, for a
general topology, no simple and general rule exists.

2.1.3. Local modes

L
Local buckling modes i involve plate-like buckling/deformations of cross-sectional elements confined only to
within the section. That means the line junctions between elements remain straight and angles between elements
do not change. The associated buckling length for the local mode is the smallest among the three modes (global,
distortional, local), and typically less than the width of plate elements comprising the cross-section. The analytical

5
solutions available for the local mode are mainly for single plate elements overlooking the fact that adjacent elements
of a cross-section interact with each other in reality. While closed-form approximations of the buckling stress for the
local mode exist in the literature [47], numerical methods (e.g., cFSM) are often used to extract these modes and their
associated critical loads. Unlike the global and distortional modes, local modes are defined by the degrees of freedom
associated with both main nodes (see Section 2.1.1) and sub-nodes, i.e., those nodes within a flat plate discretized into
multiple strips. The number of independent local buckling modes for CFS members depends on both the cross-section
topology and mesh density, and can be calculated as [45]:

NL = 2n s + 2nm e + nm i (2)

where n s is the number of subnodes which are nodes located at any internal location where the angle between two
adjoining strips is 0 , nm i is the number of internal main nodes located at any internal fold line, and nm e is the
number of external main node located at the termination of any open segments. For instance, for a lipped channel stud
section with “x1” mesh density consisting of only a single strip in the flats, with the exception of the web flat with
two strips, n s = 1, nm e = 2, nm i = 4, and NL = 10. Also, in a lipped channel stud with “x2” mesh density in which
the number of strips in the web, flange, and lips are doubled, the result is NL = 22.

2.1.4. Shear/transverse extension or other modes

S T/O
The shear and transverse extension modes or the other buckling modes i have negligible participation in the
stability solution (see Figure 3) and as such are not considered in CFS geometric imperfection modeling; see [41, 48]
for further details. The number of shear/transverse extension or other modes for CFS members can be calculated as
[45]:

NS T/O = 2(nm + n s 1) (3)

with nm and n s defined before. For example, for a lipped channel stud section with “x1” and “x2” mesh densities
(defined above), NS T/O = 12, and NS T/O = 24, respectively.

6
 Table 1: Categorization of buckling mode shapes for CFS members; The entries under FS and dec correspond to one longitudinal term in the expansion
bi i i

bi

FE FS dec
i i i

G D L S T/O
i i i i

Mode Number
Sectional Meshinga Sectional Symmetryb Sectional Foldsc Sectional Meshing Sectional Meshing

C-x1 C-x2 C-xn DS/PS SS NS 4 8 n C-x1 C-x2 C-xn C-x1 C-x2 C-xn
FE FS FS FS G G G D D D L L L S T/O S T/O S T/O
1 1 1 1 1 Fn Fn FnT 1 1 1 1 1 1 1 1 1
FE FS FS FS G G G D D D L L L S T/O S T/O S T/O
2 2 2 2 2 Fj F jT 1 F jT 2 2 2 2 2 2 2 2 2
FE FS FS FS G G G D D L L L S T/O S T/O S T/O
3 3 3 3 3 T F jT 2 FnF jT 3 3 3 3 3 3 3 3
: : : : : : : : : : : : :
FS D L S T/O
: : 28 : : 6 : 10 : : 12 : :
FS L S T/O

7
: : 52 : : 22 : 24 :
FE FS D L S T/O
N NFE NFS ND NL NS T/O

a
The entries correspond to lipped channel studs with 7-node (C-x1) and 13-node (C-x2) discretizations of the cross section
b
DS: Doubly Symmetric, PS: Point Symmetric, SS: Singly Symmetric, and NS: Non-Symmetric; Mathematically, there is one additional
global buckling mode for each of these types, the axial compression buckling mode where only unit warping (longitudinal) displacements
occur [45].
c
The entries correspond to a lipped channel stud with 4 folds and a Sigma stud with 8 folds
2.1.5. Illustrative examples: stability solutions and bi

Table 1 is a condensed illustration of the contrast between FEM, FSM, and cFSM in how each of these methods
examine the stability solution and how di↵erent factors such as sectional symmetry, sectional folds and discretization
a↵ect the space of extracted eigenmodes. FEM extracted modes, albeit providing a powerful basis for exhaustive
exploration of a stability solution, are not ideal for geometric imperfection modeling. In this section we use a set of
carefully selected profiles from CFS lipped channel sections, with di↵erent dominant buckling modes (see Figure 1),
to demonstrate how FSM and cFSM can be used to obtain the critical eigenmode shapes. These eignemode shapes, as
mentioned earlier, are frequently used to approximate the shape of geometric imperfections in FE collapse modeling.
The ingredients of the modal characterization (modal decomposition and identification [49]) process and the intrica-
cies that may arise as a result of mode coupling are also laid out. The loading is assumed axial and simple-simple
boundary condition is adopted for all cases. The material is assumed to be linear elastic, isotropic, and homogeneous
with Young’s modulus E = 210 GPa and Poisson’s ratio ⌫ = 0.3.

Figure 1: Geometry of three CFS lipped channel cross-sections (a) Case 1 (b) Case 2 (c) Case 3

Figure 2 provides FSM stability solutions along with isolated cFSM solutions and the corresponding categorized
modes bi which are obtained using CUFSM employing the modal basis option [45]. Each panel in this figure
contains the signature curve (m = 1 only, in red), the first 15 single-term stability solutions (in green), and the many-
term solution (black dashed line). Each single-term solution corresponds to a case where only one term is retained in
the series expansion for the deformation field while the many-term solution corresponds to retaining 15 terms in the
expansion. It can be seen that, for simple-simple boundary condition, the many-term stability solution can be obtained
by stitching together a sufficient number of first single-term solutions. A close look at the stability solutions depicted
here, however, shows that the number of terms required to produce the exact solution is problem dependent. While
in Figure 2(a) the many-term solution with 15 terms fails to match the exact solution at lengths around 3000 mm, the
inclusion of 3 and 6 terms in Figure 2(b) and (c) respectively provides stability solutions that match the exact solution

8
for a wide range of lengths.

Figure 2: FSM stability solutions along with pure mode cFSM solutions (modal decomposition) for (a) Case 1 (b) Case 2 (c) Case 3

The signature curves (red), which assume one half-sine wave for the variation of displacement field along the
member, often provide the essential information for local, distortional, and global buckling modes even in the presence
of some coupling between these modes. For example, in Figure 2(a) (Case 1), the two unique minima (see [50] for
more details) at half-wavelengths of 116 and 570 mm correspond to the critical lengths for local and distortional
buckling modes, respectively, and the descending branch at longer lengths (⇠> 1500 mm) corresponds to the global
mode. However, for particular cross-sections with non-unique minima, as in Case 3 depicted in Figure 2(c), the

9
signature curve may fail to identify the pure buckling modes. In such cases, to warrant the extraction of pure buckling
modes (modal decomposition) a technique equipped with constraints (e.g. cFSM) that forces the deformation field
to conform to a certain class of eigenmodes needs to be used. The identification of such isolated mode shapes is,
in fact, an essential step when imperfection magnitudes obtained from measurements need to be tied to a particular
imperfection shape (see Section 3.2). Figure 2 also illustrates the pure local (blue), distortional (black), and global
(orange) buckling modes for the cases investigated here. The lowest mode (the sold lines) and a few higher eigenmodes
(dotted lines of same color with the lowest mode) are included. The pure mode stability solutions are also overlaid
with their associated cross sectional deformations. The first five, four, and two local modes (out of ten possible local
D D
modes for “x1” mesh density, see Table 1), the two possible distortional modes ( 1, 2) and the three possible global
G G G
modes ( Fn , F jT 1 , F jT 2 ), for Cases 1-3 respectively, are illustrated. It is noted that, in contrast to local and distortional
modes, the sequence of global modes is not unique and depends on sectional properties and the length of member.

Figure 3: Modal identification for (a) Case 1 (b) Case 2 (c) Case 3

Figure 3 provides modal identification plots for the three examples examined in Figure 2. The plots are scaled by
the load factor so that the identification may be visualized along with the signature curve. It is seen in Figure 3(a) that
for Case 1 local buckling is the dominant mode of instability for a wide range of lengths (lower than ⇠ 3000 mm). This
can also be deduced from Figure 2(a) where it is observed that the many-term stability solution can be obtained by
10
stitching the first local minima regions of multiple single-term solutions. Similar to Case 1, modal identification plots
for Cases 2 and 3 are depicted in Figure 3(b) and (c). It can be seen that Case 2 represents a predominantly distortional
buckling behavior at least for a wide range of lengths (the length interval associated with the predominantly red part
in Figure 3(b)). Case 3, however, is more complicated as the dominant buckling mode is not identifiable. Figure 3(c)
in fact indicates (i) there exists a measurable local-distortional coupling for a range of lengths between 100 to 1000
mm (ii) the amount of local-distortional coupling changes as a function of buckling half-wavelength. The information
contained within the plots of Figure 3, along with the encapsulation of stability solutions in any of the panels in
Figure 2, convey an important message: while higher eigenmode solutions separate significantly from the full stability
L D G G
solution, the lowest (first) decomposed buckling modes (e.g. 1, 1, Fn / F jT 1 ) can be used to efficiently capture
the full solution for a wide range of lengths. This ability of a few first pure mode shapes to generate the full
stability solution makes them the most suitable choices for mimicking the geometric imperfection shapes (see
Section 3.2.1).

3. Geometric imperfections: recorded vs modeled

A close look at the way geometric imperfections G(x, y, z) are introduced into a shell FE analysis of CFS mem-
bers indicates two distinct categories: recorded imperfections, Gr (x, y, z), and modeled imperfections, Gm (x, y, z).
Recorded imperfections are grounded in measured data and tie the shell FE models to physical reality. Modeled im-
perfections, on the other hand, are inspired by the intrinsic behavior of the member and approximate the buckling
mode shapes for modeling convenience.

3.1. Recorded imperfections: Gr (x, y, z)

The quality and resolution of recorded or as-measured geometric imperfections Gr (x, y, z) is hinged to the mea-
surement technologies and platforms. Most imperfection measurements of CFS members, in the past, were made by
contact devices such as displacement sensors, calipers, rulers, etc. This leads to a sparse representation of the actual
imperfection profile, known as a partial-field recorded imperfection Grp (x, y, z). Such sparse data [51–54] often re-
ports maximum imperfection magnitudes, or some imperfection statistics (e.g. mean, variance, di↵erent quantiles) for
given points across the section, and is hard to use in the creation of as-measured shell FE models–see Section 3.2 for
how statistics from partially recorded imperfections is used to tie model imperfections Gm (x, y, z) to physical reality.
One may approximate the full imperfection data at points which weren’t measured (across the section and/or along
the member) using mechanical interpolation [52], or data processing techniques (see [1, 55]).
Recent advancements in measurement technologies have o↵ered versatile platforms for recording imperfections in
CFS members. These include, but are not limited to, photogrammetric [56, 57] and laser-based [18, 19, 58] methods
which o↵er highly attractive capabilities, such as non-contact and fast-speed measurement and accurate capturing of
di↵erent geometric features while being quite inexpensive. The gathered data at the end of the process is a dense point

11
cloud, known as full-field recorded imperfection Gr f (x, y, z) which can directly be used to develop imperfect shell
FE models as close to physical reality as possible. An important application of recorded geometric imperfections,
whether measured sparsely and extended to a denser grid or measured at a dense grid from the beginning, is that they
can be used to construct data-informed stochastic models of geometric imperfection profile. Such models are used
to generate statistically similar samples of imperfect CFS members which are then used, often within an uncertainty
quantification framework, for the purpose of validating di↵erent geometric imperfection models, which is the focus
of our companion e↵ort [1].

3.2. Modeled imperfections: Gm (x, y, z)

There exists a plethora of di↵erent strategies for modeling initial geometric imperfections in shell FE collapse
analysis of CFS members. A close look at the literature reveals that the geometric imperfection “models” that lie at
the core of these strategies, while seemingly di↵erent and colored by the needs of the analyst (see Section 4), have a
lot in common, in that they all assume an imperfection profile in the following form:
n
X
Gm (x, y, z) = ci ↵i di (x, y, z) (4)
i=1

comprising of three ingredients, imperfection shapes di (x, y, z), magnitudes ↵i , and combination coefficients ci .

3.2.1. Imperfection shapes

The imperfection, or deformed shapes di (x, y, z) in the expression of Gm (x, y, z) are chosen to be in close asso-
ciation with the buckling modes bi (x, y, z) as discussed in Section 2. Note, di (x, y, z) are not necessarily the same
as bi (x, y, z), and often are an approximation or simplification of the buckling mode shapes. In the widely used “tra-
B
ditional modal approach” to CFS imperfection modeling, for example, five imperfection shapes, including bow d,
C T
camber d, and twist d imperfections representing global or overall out-of-straightness reflecting sweeps about the
L D
weak, strong, and longitudinal axes of the member, plus local d and distortional d imperfections approximated by
cross-sectional elements’ out-of-flatness and out-of-straightness (type I and type II cross-section imperfection data
respectively; see [51]) are considered.
The process of utilizing a selected set of eigen modes bi to approximate the set of imperfection shapes di involves
some subtleties and nuances that deserve attention. For example, di↵erent characteristics of the CFS member under
consideration, such as its cross-sectional profile, length, boundary conditions, and loading, a↵ect the set of buckling
modes bi and consequently the set of imperfection shapes di . While a “pure” major-axis flexural buckling mode
G
Fj does not happen in isolation–e.g. singly symmetric members such as channels always couple major-axis flexural
C
and torsional buckling–its imperfection shape counterpart known as camber d may be considered in a geometric
imperfection model [52]. However, properly mimicking the geometric imperfection shape affine to the flexural-
torsional buckling mode requires that correct contributions of pure major-axis flexural buckling and pure torsional
buckling (see Section 2.1.1) be considered. In addition, pure deformed shapes or any combination thereof with
12
arbitrarily degrees of contribution could be considered in approximating the geometric imperfection profile despite
the fact that some coupling between local-distortional modes and/or distortional-global modes typically appears in
intrinsic buckling modes. Furthermore, although there exits a unique critical half-wave length for each of the local
L D
b or distortional b buckling modes, di↵erent half-wave lengths or longitudinal distributions could be considered
in approximating di . Moreover, warping deformations, which are enforced in global buckling modes extracted from
CUFSM, are often neglected when simple half-sine wave distributions for rigid-body transverse translation or torsion
of the cross-section are used to approximate di . Finally, dents and other localized imperfections that are common in
practice are generally ignored when bi are used to mimic di .

3.2.2. Magnitudes
The magnitude, ↵i , corresponding to particular di (x, y, z), tie the modeled imperfection Gm (x, y, z) to physical
reality. These magnitudes can be obtained from statistics of recorded geometric imperfections [51–54], from simple
empirical relations (see Section 4), or based on quality control tolerances provided in standards such as AISI S200
[46]. Note, however, that the definition of imperfection magnitudes ↵i is specific to the measurement platform used to
record the geometric imperfections; see the traditional modal approach discussed in Section 3.2.1.

3.2.3. Combination coefficients

The combination coefficients ci pertain to how to combine the imperfection shapes already multiplied by magni-
tude values to find the lowest load carrying capacity, or the load capacity that is close to experimental observations.
The coefficients are often obtained in such a way that a suitable norm of the coefficient vector is equal to 1, that is:
v
t n
X
kck2 = c2i = 1 (5a)
i=1

kck1 = max(|ci |) = 1 (5b)

with c the vector of combination coefficients.

4. Existing geometric imperfection models

The literature on nonlinear collapse analysis of CFS members includes many geometric imperfection models that
have been proposed by researchers and di↵erent structural design codes. In what follows, we review some of the
most common models that relate directly to the discussion in the preceding section (see also [59]). The consistent
nomenclature of Equation (4) is adopted for presenting di↵erent choices of the three components in a geometric
imperfection model Gm (x, y, z), that is the imperfection shapes di (x, y, z), the magnitudes ↵i , and the combination
coefficients ci .

13
4.1. Dawson and Walker (1972)

Dawson and Walker [60] presented closed form solutions for elastic local buckling of simply supported plates
seeded with generalized geometric imperfections. Unlike an assembly of plates comprising a thin-walled member,
such single plates have a simpler analytical stability solution. Explicit expressions for load carrying capacity, end
shortening, and sti↵ness of these members were derived and the shape of the initial geometric imperfection was
assumed to be the same as the plate buckling mode:

1
X 1
X i⇡x j⇡y
Gm (x, y) = 1.0 ai j t cos(
|{z} ) cos( ) (6)
i=1,3,... j=1,3,... ci j
|{z} | a {z b}
↵
ij
di j (x, y)

where a, b, and t are length, width, and thickness of plate, respectively and ai j are prescribed coefficients such that

1
X 1
X
↵= ai j t (7)
i=1,3,... j=1,3,...

is the amplitude of the initial geometric imperfection at x = 0 and y = 0. Three di↵erent expressions proposed by
other researchers [61–64] were considered to estimate the magnitude of the local imperfections:

↵= t (8a)
0.5
↵= ( y / cr ) t (8b)

↵= ( y / cr )t (8c)

where , , are general constants to be determined experimentally, y is the yield stress, and cr is the critical
buckling stress. Although, assuming = = = 0.2 in the above equations provided adequately conservative
results, they recommended Equation (8c) with = 0.2 as the most suitable generalized imperfection amplitude for the
analysis of CFS members.

4.2. Schafer and Pekoz (1998)

Schafer and Pekoz [51] suggested two approaches for geometric imperfection modeling that consider both im-
perfection shape and magnitude. To perform a limited study, they suggested that at least two fundamentally di↵erent
eigenmode shapes be summed together for the imperfection shape. Such modes should not be the same fundamental
modes with a di↵erent wavelength. It was shown as an example that seeding a geometric imperfection pattern with
only local or only distortional eigenmode is unconservative when compared to seeding both modes which implies the
importance of properly combining eigenmodes to form the geometric imperfection pattern. As for the magnitude,
maximum values for local and distortional geometric imperfections were proposed using empirical expressions (sim-
ple rules of thumb) and probabilistic analysis (see Table 2). Two values at representative points were introduced: ↵1 ,

14
the maximum local geometric imperfection at the middle of the web in the form of out-of-flatness of the web plate,
and ↵2 , the maximum deviation at flange-lip intersection for distortional geometric imperfection. Based on collected
data, it was suggested that local imperfection be in the form of either Equation (9a) (in terms of plate width w) or
Equation (9b) (in terms of plate thickness t) assuming w/t < 200:

↵1 ⇡ 0.006w (9a)
2t
↵1 ⇡ 6te (↵1 and t in mm) (9b)

A similar expression for distortional imperfection magnitude (valid for w/t < 100) was proposed:

↵2 ⇡ t (10)

where the thickness should be less than 3 mm for both Equations (9) and (10) to be valid. Noticing a large variation in
measured maxima, Schafer and Pekoz presented the cumulative distribution function (CDF) values of ↵1 /t and ↵2 /t
for the measured data (see Table 2) which are in good agreement with additional measurements. These percentile
values for ↵/t which are given irrespective of y and cr –that is in the same form as Equation (8a)–enable making the
connection between a given confidence level and a particular imperfection magnitude.

Table 2: CDF values for maximum imperfection [51].

type I type II
P( < )
↵1 /t ↵2 /t

0.25 0.14 0.64

0.50 0.34 0.94
0.75 0.66 1.55
0.95 1.35 3.44
0.95 3.87 4.47
Mean 0.50 1.29
St. dev. 0.66 1.07

4.3. Sivakumaran and Abdel-Rahman (1998)

Sivakumaran and Abdel-Rahman [65] proposed geometric imperfection patterns and magnitudes for specific CFS
members and validated FE results through a test program. A set of perforated and non-perforated lipped channel stub-
columns with di↵erent lengths under axial compression was examined. The columns were short enough to eliminate
the global buckling mode but long enough to accommodate various size of perforations as well as to allow longitudinal
propagation of local buckling mode. Inspired by local buckling shape of an individual plate and modifying Hancock’s
[66] suggestion for distribution of plate imperfection in CFS columns, geometric imperfection was seeded only within
15
the web (and not within the flange or the lip) as a double sine-wave distribution:
⇡x ⇡y
Gm (x, y) = 1.0 ↵ sin( ) sin( ) (11)
w w̄

with w the web width and w̄ the average of the web and flange widths resulting in one half-sine wave in transverse
direction of web and multiple half-sine waves with a wave length of w̄ in its longitudinal direction. The imperfection
magnitude at the central point of the web plate, ↵, was selected to be one-half of the upper limit recommended by BS
5950, i.e.
8 r 9
1> >
< w y >
>
=
↵= > 0.145( ) t > (12)
2> : t E > ;

where t is thickness, w/t is slenderness ratio, and E and y are material properties of plate. Note, Equation (12) is of
the same form as Equation (8b).

4.4. Sun and Butterworth (1998)

Sun and Butterworth [67] conducted FE analysis on steel single angle members under compression and eccentric
loading applied to one leg. The nonlinear stress-strain curve with an elastic part, a perfectly plastic part, and a
multilinear strain hardening, based on test data was used. For the initial geometric imperfection pattern, d, they
adopted several half-sine waves in the longitudinal direction and linear variation in the transverse direction. They
used initial imperfection magnitudes of ↵ = {0.167t, 0.333t, 0.5t, 0.667t} in their analyses. The results were then
compared with experimental tests to calibrate the amplitude for the prediction of load carrying capacity and buckling
behavior. Equal leg struts (EA 90x90x6) with four lengths were chosen for tests. The high width/thickness ratio of the
section gave rise to local buckling and the varying slenderness ratios from 50 to 150 led to both elastic and inelastic
buckling behavior. The ↵ = 0.333t amplitude was the best average match for ultimate loads although the post buckling
behavior was not predicted properly. This value is nearly identical to 50th percentile for type I imperfection in Table 2.
Sun and Butterworth also investigated the e↵ect of combination coefficient ci (first half-sine wave twisted about the
shear center in clockwise and counterclockwise directions). Except for the most slender member, the impact on both
the load carrying capacity and post-buckling behavior was tangible.

4.5. Chou et. al. (2000)

Chou et. al. [68] examined the post-buckling behavior of cold formed steel lipped channel and hat-section stub-
columns under axial compression. They compared the numerical results with experimental data by Zaras and Rhodes
[69] as well as BS 5950 predictions. For initial geometric imperfection pattern they used the first linear buckling
mode shape in contrast to their previous work [70] where they used small perturbation loads. For ↵ values, they used
0.5
0.10t, 0.50t, 1.00t and Walker’s Equation (8b) for imperfection magnitude [64], that is ↵ = ( y / cr ) t, with t the
member thickness, y the yield stress, and cr the critical buckling stress but with = 0.3. They recommended that
Walker’s expression needs to be used in the design process.
16
4.6. Gardner (2002)
Gardner [71] performed experiments and numerical simulations on cold-formed stainless steel members and com-
pared the results with the predictions of Eurocode (EC 3: Part 1.4) to propose a consistent approach for the modeling
of stainless steel structures. The investigated members comprised Square, Rectangular, and Circular Hollow Sections
(SHS, RHS, and CHS respectively) under variety of loading and boundary conditions Proper material modeling as-
sumptions including appropriate stress-strain behaviors and residual stresses were considered in the FE simulations
to enable better replication of test results. The eigenmodes obtained from elastic buckling analysis were used to de-
fine the shape of initial geometric imperfection. For stub-columns with SHS, RHS, and CHS sections, only the local
buckling mode shape (the lowest buckling mode) was seeded. However, for longer pin-ended columns (with SHS and
RHS sections) that typically buckle into global mode shape, the superposition of the first two eigenmodes, the global
and local modes, was proposed as the geometric imperfection pattern. In beam specimens, the lowest eigenmode in
pure compression (similar to those selected for the corresponding stub-columns) was used and the in-plane bending
failure mode was in the form of local buckling of the compression flange and the top portion of the web. For imper-
fection amplitude, the measured values of local and global imperfections as well as empirical relations (Equation (8)
for cold-formed carbon steel members) were used in FE simulations and the results were compared with experiments.
A regression analysis on measured local imperfection magnitudes for SHS and RHS stainless steel members showed
6
the best fit for = 7.3 ⇥ 10 in Equation (8b) and = 0.023 in Equation (8c). The results obtained from numerical
simulations with imperfection magnitudes defined by Equation (8c) then demonstrated accurate agreement with test
results and therefore Equation (13) was suggested for local imperfection magnitude of SHS and RHS members:

↵L = 0.023( 0.2 / cr )t (13)

For CHS members, however, a di↵erent equation similar to Equation (8a) was recommended:

↵L = 0.2t (14)

where 0.2 is the material 0.2% proof stress, ↵L is the local imperfection magnitude, and t is the thickness. For beam
specimens, the local imperfection magnitude was selected similar to those of stub columns. Finally, based on the
parametric study and comparison with test results for SHS and RHS pin-ended columns, Equation (15) for global
imperfection magnitude was proposed:

↵G = L/2000 (15)

where ↵G is the global imperfection magnitude in buckling direction and L is the specimen length.

4.7. Dinis and Camotim (2008, 2009, 2010)

Dinis and Camotim [72–74] employed di↵erent geometric imperfection models in shell FE collapse analysis of
CFS lipped channel columns/beams to study the e↵ect of interaction of critical buckling modes on the strength and
17
post-buckling behavior. The cross-section and length of the members were carefully selected so as to ensure the
coincident (or nearly so) happening of critical modes to maximize mode interaction e↵ects. For the imperfection
shapes di , they combined the pure buckling mode shapes (two or three modes) obtained from the elastic buckling
analyses of the column (beam) under axial load (major axis bending). As for the magnitudes, ↵ = 0.1t was considered
for the local and distortional modes, and ↵ = L/1000 for the global mode. Each of modes was normalized to exhibit
unit displacement at a selected degree of freedom, and then scaled down with the magnitude values. For the global
mode, the vertical displacement of flange-lip corner at mid-span (the selected degree of freedom), was converted to its
equivalent value as a web chord rotation at mid-span. To linearly combine the scaled imperfection shapes, they used
the combination coefficients satisfying Equation (5a). Since the imperfection shapes are indeed orthogonal buckling
modes, using this approach may result in unconservative estimates of load carrying capacity.

4.8. Zeinoddini (2011)

Zeinoddini [52] adopted di↵erent approaches for geometric imperfection modeling in nonlinear collapse analysis
to examine failure mechanisms for CFS members. Di↵erent bracing configurations were used to invoke local, dis-
tortional, and flexural-torsional failure in a 362S162-68 profile (AISI S200 nomenclature [46]). As-measured imper-
fections reported by Peterman and Schafer [54] were seeded in the form of perturbations in the perfect geometry and
mechanical interpolation was used to extend the measured imperfections to the points with no measurement. Also, a
combination of the buckling mode shapes (the traditional modal approach) was used to seed geometric imperfections.
The mode shapes, obtained via CUFSM [75], included local, distortional, and (mixed) global mode shapes with the
latter encapsulating bow, camber, and twist imperfections which can be seeded separately as half-sine waves as well.
For magnitudes ↵i , a statistical summary of type I and II imperfections (↵1 and ↵2 ) based on measured data, collected
from other researchers, was updated (see Table 2). The CDF values for the bow, camber, and twist imperfections (↵3
to ↵5 ) were provided through conducting a measurement program at Johns Hopkins University (Table 4-4 in [52]) as
well as through processing the measured data reported by others (Table 4-2 in [52]). For combination coefficients ci ,
three values in the set { 1, 0, +1}, with ci = 1 (ci = 1) representing the existence of a mode with positive (negative)
sign and ci = 0 representing the absence of a mode were considered to show the importance of the combination ap-
proach on the load carrying capacity of imperfect models. A total of (33 = 27) combinations with ↵i values equal to
50%ile of all measured data (Table 6-1 in [52]) were considered. Both 25% and 75% exceedance values with ci = 1
were also examined to explore the e↵ect of magnitude on response. Finally, in an e↵ort to go beyond the available
or recorded measurements, stochastic simulation of of geometric imperfections based on spectral representation was
also explored by Zeinoddini and Schafer; see [76], see also [77] and Part II [1] for further discussion.

4.9. Torabian et. al. (2016)

Torabian et. al. [78, 79] performed experiments and numerical simulations to develop a new direct strength design
method for CFS beam-columns under combined actions (axial load and bending moment). They provided validated

18
modeling protocols for the shell FE collapse analysis of CFS Cee- and Zee-section beam-columns for accurate predic-
tion of strength capacity. Such FE models were then used to construct the strength surface for tested specimens and
to expand the results to other CFS beam-column members. The key parameters in FE modeling including geometric
imperfections, mesh density, and residual stresses were examined to provide a uniform protocol. Di↵erent imperfec-
tion shapes, magnitudes, and combination coefficients were attempted in an e↵ort to identify those that best reproduce
the test (in terms of the statistics of the absolute strength and sti↵ness as well as the entire nonlinear response). For
L D B
the imperfection shapes, three buckling modes, the local d, the distortional d, and the mixed global (bow d and
C
camber d) were chosen. As for the magnitude, the 25th, 50th, and 75th percentile values from [76] were initially
used for the validation purpose and the 50%ile value was finally chosen for all models in the parametric study. The
combination coefficients for the three imperfection shapes were selected in such a way that it brings about the worst
L D
case scenario or the lower-bound for the ultimate strength; i.e., ci = 1 for both d and d and an appropriate sign for
the bow and camber resulting in maximum eccentricities at the mid-height of the specimens.

4.10. Australian Standard (2012)

The Steel Storage Racking part of Australian Standard (AS 4084) [80] allows shell FE collapse simulation of CFS
members to replace traditional strength calculations. When performing collapse simulations, geometric imperfections
in the form of the local and distortional buckling modes need to be considered in the structural model explicitly.
The shape of the geometric imperfection is mimicked by buckling modes which are obtained from linear buckling
analysis via finite element or finite strip method. The magnitudes for local and distortional imperfections are chosen
p p
as ↵L = 0.3t y / crl and ↵D = 0.3t y / crd , respectively, with y the nominal yield stress, and crl and crd the

local and distortional buckling stresses, respectively.

4.11. Eurocode 3 - Design of Steel Structures (2007)

Another standard that allows direct shell FE collapse simulation of thin-walled structures is part 1-6 of Eurocode
3 (Strength and Stability of Shell Structures) [81]. EC3 defines dimples (both in-weld and in-plate) or other forms
of shell imperfections such as accidental eccentricities and out-of-roundness that need to be seeded as geometric
imperfections. It is recommended that the pattern of the equivalent geometric imperfections be chosen in such a form
that it has the most unfavorable e↵ect on the strength of the shell. Eigenmode-affine patterns are recommended unless
di↵erent more unfavorable patterns can be justified. The pattern of the equivalent geometric imperfections should also
reflect the constructional detailing and the boundary conditions in an unfavorable manner. The sign of imperfections
is chosen in such a manner that the maximum initial shape deviations are unfavorably oriented towards the center of
the shell curvature and the amplitude depends on the fabrication tolerance quality class.

19
5. Nonlinear shell buckling analysis: model setup and implementation details

Owing to the complexity of nonlinear stability behavior of CFS members, numerical solution techniques such
as FEM are often used to study their load carrying capacity and post-buckling response. This section describes the
details required to setup FE models for nonlinear collapse analysis of CFS members using the general purpose com-
mercial software package, ABAQUS, including the analysis types, material properties, element type and size, loading
and boundary conditions, and most importantly the seeding of geometric imperfections onto the perfect model. An
in-house set of Python, MATLAB, and Unix shell scripts coupled with ABAQUS and CUFSM software packages
for automated high performance FE computing [82, 83] is developed to facilitate the repetitive tasks pertaining to
model setup and implementation details, including those related to generating a multitude of geometric imperfec-
tion models, submission of massive FE analyses on supercomputers, and post-processing of the results. Figure 4
depicts a schematic of a 3D computational model (a lipped channel CFS member) with exaggerated magnitudes of
geometric imperfection encompassing five imperfection shapes di as defined in the traditional modal approach (see
Section 3.2.1). Further details on computational modeling and importance of di↵erent parameters and settings have
been discussed in [84].

Figure 4: Computational model with geometric imperfection; (a) combination of di (b) bow (c) camber (d) twist (e) local (f) distortional

5.1. Analysis types

The open source software package CUFSM based on FSM and cFSM is used for eigenvalue buckling analysis.
The cFSM option provides an efficient means for modal characterization and facilitates the process of seeding geo-
metric imperfections in shell FE model of the CFS member. Fully nonlinear collapse analysis aimed at estimating load
carrying capacity and capturing the post-buckling behavior is performed using the commercial FE software package

20
ABAQUS that allows proper modeling of material nonlinearity, geometric nonlinearity prior to buckling, and unstable
post-buckling response. Several approaches in ABAQUS allow for modeling the complex behavior in post-buckling
regime where the load-displacement response shows a negative sti↵ness and the structure releases strain energy to
remain in equilibrium. One alternative is to make recourse to dynamic analysis taking into account the inertia e↵ects
as the structure snaps. Another approach is to use artificial damping to stabilize the structure during a static analysis.
ABAQUS o↵ers an automated version of this stabilization approach for the static analysis procedures which is the
one used for this study. Alternatively, static equilibrium path during the unstable phase of the response can be traced
by using the modified Riks method [85] which works well for problems with post-buckling unstable behavior such as
snap-through problems.

5.2. Material properties

Accurate modeling of the basic material properties is of crucial importance if the modeling of CFS member is to
resemble the behavior observed in the lab. Although, the focus of this study is to provide a comprehensive review of
geometric imperfection modeling and its interplay with load carrying capacity, appropriate material models are used
in all FE simulations. The material is assumed to be homogeneous and isotropic and modeled as elastic-perfectly
plastic (von Mises yield criteria with isotropic hardening) with Youngs modulus E = 210 GPa, Poissons ratio ⌫ = 0.3,
and a yield stress of y = 345 MPa.

5.3. Element type and size

CFS members are usually modeled with shell elements as through-thickness deformation and tri-axial stresses are
not expected to be significant. ABAQUS o↵ers a wide variety of elements for shell modeling (e.g., S4, S4R, S4R5,
S8R, etc.). Element type S9R5 is used in our study. Di↵erent mesh densities were examined to achieve convergence.
The convergence study and the seeding of geometric imperfection models were facilitated with our in-house Python
code. Finally, to avoid element distortion in large deformations, the aspect ratio of elements is kept between 0.5 and
2 and orthogonal meshing aligned with the orientation of member’s longitudinal axis is used. The reader interested in
more details is referred to [84].

5.4. Loading and boundary conditions

As mentioned in Section 5.1, for certain classes of geometric imperfection models, the imperfection shapes are
borrowed from the set of buckling modes of the perfect member. This implies that the boundary conditions used
for nonlinear analysis and those used for buckling analysis need to match. While such compatibility is naturally
guaranteed when FE analysis is used for both buckling eigenmode analysis and nonlinear collapse analysis, care
is needed when a numerical technique other than FEM is used for buckling analysis. For example in FSM, the
semi-analytical technique that is the basis of CUFSM, the end constraints are enforced by adopting particular shape
functions in the longitudinal direction that meet the end conditions a priori.

21
 Table 3: Equivalent FE boundary conditions matching FSM boundary conditions

Fixed degree of freedom

Boundary Condition x=0 x=L
Ux Uy Uz ✓x ✓y ✓z Ux Uy Uz ✓x ✓y ✓z
Simple-Simple (S-S) X X X X X X
Clamped-Clamped (C-C) X X X X X X X X X X X X
Simple-Clamped (S-C) X X X X X X X X X
Clamped-Free (C-F) X X X X X X
Clamped-Guided (C-G) X X X X X X X X X

While in this work, CFS members with simple-simple boundary condition are considered, for the sake of complete-
ness, we are reporting equivalent FE boundary conditions matching FSM boundary conditions in Table 3. Replicating
FSM boundary conditions in FE models requires warping free ends (Uy , 0 @ x = 0 and x = L). A constraint
(fixed warping) is therefore enforced at the middle of the member to ensure kinematic stability of the FE model.
Displacement-based loading is applied at the two ends simultaneously through a reference node located at the cen-
troid of the cross-section at each end; see [30] for further details.

5.5. Initial geometric imperfections

To seed a particular geometric imperfection, either a recorded imperfection Gr (x, y, z) or a geometric imperception
model Gm (x, y, z), into a computational model, the nodal coordinates need to be identified. For Gr (x, y, z), to match the
geometry of measured data with FE mesh, appropriate data processing techniques and geometric transformations–that
depend on the measurement platform; see [18, 19, 52, 54]–are required. For Gm (x, y, z), depending on the choice of
di (x, y, z) employed, di↵erent approaches need to be adopted. To seed a geometric imperfection shape di that mimics
anl
the analytical buckling modes i we use our in-house Python code to directly adjust the positions of the individual
nodes in FE models to comply with spatial variations associated with such modes. Buckling modes obtained from FE
FE
eigen analysis i on the other hand, are directly superimposed to the perfect model. For buckling modes obtained
FS
from FSM i we extract the distribution of imperfection across the cross-section from CUFSM. The longitudinal
distribution will then follow a sinusoidal function with a half-wavelength associated with the particular cross-sectional
mode shape to be seeded (note the global mode is seeded as is). Alternatively, we can perturb the perfect model
dec
with imperfections that comply with pure cross-sectional and global imperfections i obtained using cFSM. As
mentioned before the ability to categorize and reduce the complicated modal deformations into basis forms is crucial
in that, on one hand, it allows for direct application of buckling modes as geometric imperfection patterns and, on
the other hand, it makes it easier to connect the magnitude of geometric imperfection ↵i to measured data. This is of
particular importance, considering the fact that geometric imperfections are a result of fabrication and other processes
and as such any attempt to include them in the analysis process needs to be informed by data. We note, however, that
regardless of the approach adopted, the imperfect member is always assumed stress-free at the onset of analysis.

22
6. Comparative study

Despite many studies on the influence of geometric imperfections on load carrying capacity of CFS members, the
combinatorial role of di↵erent geometric imperfection shapes in Gm (x, y, z) for a member with an arbitrarily dominant
buckling mode has not been fully examined. In what follows (see Section 6.1) we undertake such an examination.
We then turn our attention back to the geometric imperfection modeling strategies discussed in this paper and provide
a brief comparative study where we demonstrate the noticeable variability in the load carrying capacity and post-
buckling behavior that could stem from adaptation of a multitude of geometric imperfection models available to the
analyst.

Figure 5: KDF curves and identified imperfection magnitudes corresponding to 80% of the ultimate load for the perfect model; Case 2 in Sec-
tion 2.1.5 with L = 668 mm

6.1. Coupled role of imperfection shape and combination coefficient

The coupled e↵ect of imperfection shape di (x, y, z) and combination coefficient ci on the load carrying capacity
is investigated by excluding the role of imperfection magnitude ↵i . A lipped channel cross-section with a distortional
dominant buckling mode (Case 2 in Section 2.1.5) with a length L = 668 mm is considered. The buckling load for the
member without geometric imperfection is predicted to be 164 KN in an ABAQUS simulation.
We first identify, imperfection magnitudes corresponding to 80% of load carrying capacity of the perfect model.
This is done through constructing the Knock Down Factor (KDF) curves for geometrically imperfect models each
involving only one of the traditional geometric imperfection shapes di : bow, camber, twist, local, and distortional;
see Figure 5. Even though each of these imperfection shapes can be considered with a positive or negative sign, here,
for all the imperfections except bow, both choices of sign result in identical load carrying capacities. Therefore, we
23
examine both signs for only bow imperfections and only one sign for others in constructing the KDF curves. The
insensitivity to sign for local and distortional imperfections is due to the existence of multiple half-waves where either
positive or negative sign always triggers the worse of the two. Note, however, for imperfections with odd number of
half-waves it can matter whether positive or negative sign is in the middle. The sign matters also for short members
with a length equal to a single half-wave. For the cross section considered here, the half-wavelengths are 84 mm and
334 mm for local and distortional imperfections, respectively, which both result in even number of half-waves along
the length. The insensitivity to sign for camber and twist imperfections is simply because of the symmetry.
The force-displacement curves obtained from nonlinear analysis of all combinations of these five imperfection
shapes di (x, y, z) and three combination coefficients ci 2 { 1, 0, 1} (a total of 35 = 243 combinations) are then
obtained. These are depicted in Figure 6 along with the histogram of the ultimate loads. From the inset in this figure,
it can be observed that the coupled e↵ect of di (x, y, z) and ci results in 31% variation (absolute values vary between
92 to 133 KN) in the load carrying capacity indicating the importance of the combination strategy. The ultimate
load for the case when all five imperfection shapes are included with positive (negative) sign is 102 KN (108 KN).
Also, the worst four imperfect models with combination coefficients (CT = 1, C B = 1, CC = 1, C D = 1, C L = 0),
(CT = 1, C B = 1, CC = 1, C D = 1, C L = 0), (CT = 1, C B = 1, CC = 1, C D = 1, C L = 0), and (CT = 1, C B =
1, CC = 1, C D = 1, C L = 1) result in minimum values in the range of 92 95 KN. The numerical values of the
load carrying capacity also indicate that geometric imperfection models affine to only the first buckling mode shape
may result in capacity estimations that are tangibly non-conservative.

Figure 6: Force-displacement curves for 243 combinatorial geometric imperfection models; The inset represents histogram the ultimate loads; Case
2 from Section 2.1.5 with L = 668 mm

6.2. The impact of adopted geometric imperfection model

Figure 8 shows the force-displacement curves for 15 geometric imperfection models summarized in Table 4.
The imperfection shapes and magnitudes used to build these models along with the load carrying capacity values
are provided. The curves pertain to Case 2 from Section 2.1.5 with L = 668 mm and the imperfection shapes and
24
magnitudes are borrowed from the extensive review presented in Section 4. Both FE (ABAQUS) and cFSM (CUFSM)
are used to extract the eigenmodes–see the first 5 FE modes in Figure 7–and the analysis framework laid out in
Section 5 is used to obtain the load-displacement curves. The noticeable variability in the load carrying capacity
stemming from the adaptation of di↵erent geometric imperfection models (see the histogram in the inset) highlights
an urgent need for reliable probabilistic frameworks (see e.g. [76]). We will show in a companion paper (see [1]) how
such frameworks can pave the path towards data-informed validation of imperfection models used by researchers and
practitioners.

Figure 7: Buckling modes for Case 2 from Section 2.1.5 with L = 668 mm extracted using ABAQUS: (a) mode 1 (b) mode 2 (c) mode 3 (d) mode
4 (e) mode 5

Figure 8: Force-displacement curves for 15 di↵erent geometric imperfection models; Case 2 from Section 2.1.5 with L = 668 mm

25
 Table 4: Summary of load carrying capacity from collapse analysis with di↵erent imperfection
models; Case 2 from Section 2.1.5 with L = 668 mm

Models Shape Magnitude Load Carrying Capacity (KN)

FE
GI 1 1 0.333t 131.89
FE
GI 2 2 0.333t 132.66
FE
GI 3 3 0.333t 133.45
FE
GI 4 4 0.333t 139.49
FE
GI 5 5 0.333t 135.76
FE FE
GI 6 1 to 5 0.333t 126.53
FE
GI 7 1 ( y / cr ) 145.64
L
GI 8 d ( y/ cr ) 145.99
FE
GI 9 1 0.006w 132.25
FE
GI 10 1 t 113.03
L
GI 11 d 0.006w 133.65
D
GI 12 d t 121.64
L D B C T a
GI 13 d, d d, d, d
, 50%ile 131.82
L D
GI 14 d, d 50%ilea 130.60
B
GI 15 d L/960 158.59

a
50%ile exceedance values from [52]

7. Concluding remarks

This paper is the first in a two-part e↵ort that examines the role of geometric imperfection models in nonlinear
collapse analysis of Cold-Formed Steel (CFS) structural members. In this paper, a review of di↵erent geometric
imperfection models used in shell Finite Element (FE) analysis of CFS members was provided. The categorization of
elastic buckling mode shapes as the main source of inspiration for geometric imperfection shapes was systematically
presented. A unified expression for imperfection profile in its full three-dimensional spatially varying form was
provided. This expression can represent nearly all available imperfection models in the literature and is comprised
of three components: imperfection shape (frequently chosen to be in the form of eigenmode shapes), magnitude, and
combination coefficient. It was shown how higher eigenmode solutions separate significantly from the full stability
solution and the lowest decomposed (or pure) buckling modes can be used to efficiently capture the full solution; a
characteristic that makes them the most suitable choices for approximating the geometric imperfection shapes. The
need to tie a particular imperfection shape to imperfection magnitudes obtained from measurements further motivated
an in-depth study of categorized eigenmodes.
An in-house set of Python and MATLAB scripts coupled with ABAQUS and CUFSM software packages were
developed to perform a numerical study on the impact of choice of geometric imperfection model on nonlinear re-

26
sponse and load carrying capacity. The study revealed the choice of combination coefficients–that is how di↵erent
geometric imperfection patterns are combined to form the initial imperfection profile–has a tangible impact on the
ultimate strength. Imperfections affine to only a particular buckling mode shape were shown to have the potential
to lead to non-conservative estimates of load carrying capacity. The observed variability in load carrying capacity
from the adaptation of various geometric imperfection models was significant, warranting the need for more rigorous
analysis frameworks, based (for example) on probabilistic metrics, for data-driven validation of such models. Such
frameworks which are the subject of the companion paper (see [1]) are expected to be valuable tools in the path to
fully enable analysis-based design procedures for CFS members.

Acknowledgments

This work was supported by grants CMMI-1351742, CMMI-1235238 and CMMI-1235196 from National Science
Foundation. The authors gratefully acknowledge this support.

References

[1] S. Farzanian, A. Louhghalam, B. W. Schafer, M. Tootkaboni, Geometric imperfection models for CFS structural members, Part II: Validation
through data-driven stochastic modeling, under review.
[2] B. H. Smith, A. Chatterjee, S. R. Arwade, C. D. Moen, B. W. Schafer, System reliability benefits of repetitive framing in cold-formed steel
floor systems, Journal of Structural Engineering 144 (6) (2018) 04018061.
[3] A. R. Prabowo, S. J. Baek, H. J. Cho, J. H. Byeon, D. M. Bae, J. M. Sohn, The e↵ectiveness of thin-walled hull structures against collision
impact, Latin American Journal of Solids and Structures 14 (7) (2017) 1345–1360.
[4] S. Farzanian, M. Tootkaboni, Collapse analysis of thin cylindrical shell structures: Extending single perturbation load approach via surrogate
modeling, in: Engineering Mechanics Institute Conference, San Diego, CA, 2017.
[5] E. Azzuni, S. Guzey, Comparison of the shell design methods for cylindrical liquid storage tanks, Engineering Structures 101 (2015) 621–630.
[6] A. Chatterjee, A. Algara, F. Castano, C. Moen, Thin-walled, cold-formed, steel-welded tube design in a long span dome, in: Structures
Congress 2014, 2014, pp. 2699–2706.
[7] T. A. Schaedler, A. J. Jacobsen, A. Torrents, A. E. Sorensen, J. Lian, J. R. Greer, L. Valdevit, W. B. Carter, Ultralight metallic microlattices,
Science 334 (6058) (2011) 962–965.
[8] L. Salari-Sharif, S. Godfrey, M. Tootkaboni, L. Valdevit, The e↵ect of manufacturing defects on compressive strength of ultralight hollow
microlattices: A data-driven study, Additive Manufacturing 19 (2018) 51–61.
[9] E. T. Filipov, T. Tachi, G. H. Paulino, Origami tubes assembled into sti↵, yet reconfigurable structures and metamaterials, Proceedings of the
National Academy of Sciences 112 (40) (2015) 12321–12326.
[10] P. M. Reis, F. Lpez Jimnez, J. Marthelot, Transforming architectures inspired by origami, Proceedings of the National Academy of Sciences
112 (40) (2015) 12234–12235.
[11] T. von Karman, The strength of thin plates in compression, Trans. ASME 54 (1932) 53–57.
[12] G. Winter, Lateral bracing of columns and beams, Journal of the Structural Division 84 (2) (1958) 1–22.
[13] Steel Framing Industry Association.
URL http://cfsteel.org
[14] H. Foroughi, C. Moen, A. Myers, M. Tootkaboni, L. Vieira, B. Schafer, Analysis and design of thin metallic shell structural members-current
practice and future research needs, in: Proceedings of the Annual Stability Conference Structural Stability Research Council, Toronto, CA,
2014.

27
[15] H. Amouzegar, B. Amirzadeh, X. Zhao, B. Schafer, M. Tootkaboni, Statistical analysis of the impact of imperfection modes on collapse
behavior of cold-formed steel members, in: Proceedings of the Annual Stability Conference Structural Stability Research Council, Nashville,
TN, 2015.
[16] M. Tootkaboni, L. Graham-Brady, B. Schafer, Geometrically non-linear behavior of structural systems with random material property: An
asymptotic spectral stochastic approach, Computer Methods in Applied Mechanics and Engineering 198 (37-40) (2009) 3173–3185.
[17] H. Amouzegar, B. Schafer, M. Tootkaboni, An incremental numerical method for calculation of residual stresses and strains in cold-formed
steel members, Thin-Walled Structures 106 (2016) 61–74.
[18] X. Zhao, M. Tootkaboni, B. W. Schafer, Development of a laser-based geometric imperfection measurement platform with application to
cold-formed steel construction, Experimental Mechanics 55 (9) (2015) 1779–1790.
[19] X. Zhao, M. Tootkaboni, B. W. Schafer, Laser-based cross-section measurement of cold-formed steel members: Model reconstruction and
application, Thin-Walled Structures 120 (2017) 70 – 80.
[20] B. W. Schafer, Developments in research and assessment of steel structures: Highlights from the perspective of an American researcher,
ce/papers 1 (2-3) (2017) 95–114.
[21] S. P. Timoshenko, J. M. Gere, Theory of elastic stability. 1961, McGrawHill-Kogakusha Ltd, Tokyo 109.
[22] V. Vlasov, Thin-Walled Elastic Beams (English Translation), National Science Foundation, Oldbourne Press Washington, DC, London, 1961.
[23] A. Chajes, G. Winter, Torsional-flexural buckling of thin-walled members, Journal of the Structural Division 91 (4) (1965) 103–124.
[24] Y. K. Cheung, The finite strip method in the analysis of elastic plates with two opposite simply supported ends, in: Proc. Am. Soc. Civ. Eng.,
Vol. 94, 1968, pp. 1365–1378.
[25] G. J. Hancock, Local, distortional, and lateral buckling of I-beams, Journal of the Structural Division 104 (11) (1978) 1787–1798.
[26] CUFSM: elastic buckling analysis of thin-walled members by finite strip analysis. CUFSM v4.05; 2012,
url=http://www.ce.jhu.edu/bschafer/cufsm,.
[27] B. W. Schafer, S. Ádány, Understanding and classifying local, distortional and global buckling in open thin-walled members, in: Tech. Session
and Mtg., Structural Stability Research Council, Citeseer, 2005.
[28] S. Ajeesh, S. A. Jayachandran, A constrained spline finite strip method for the mode decomposition of cold-formed steel sections using GBT
principles, Thin-Walled Structures 113 (2017) 83–93.
[29] M. Khezri, M. Gharib, M. Bradford, Z. Vrcelj, Analysis of thick and orthotropic rectangular laminated composite plates using a state-space-
based generalised RKP-FSM, Composite Structures 133 (2015) 691–706.
[30] B. Ahmadi, S. Farzanian, A. Asadpoure, B. W. Schafer, M. Tootkaboni, A Chebyshev-Ritz finite strip method for buckling analysis of
cold-formed steel members with general boundary conditions, in submission.
[31] S. Ádány, B. W. Schafer, A full modal decomposition of thin-walled, single-branched open cross-section members via the constrained finite
strip method, Journal of Constructional Steel Research 64 (1) (2008) 12–29.
[32] N. Silvestre, D. Camotim, First-order generalised beam theory for arbitrary orthotropic materials, Thin-Walled Structures 40 (9) (2002)
755–789.
[33] N. Silvestre, D. Camotim, Second-order generalised beam theory for arbitrary orthotropic materials, Thin-Walled Structures 40 (9) (2002)
791–820.
[34] S. Ádány, N. Silvestre, B. W. Schafer, D. Camotim, GBT and cFSM: two modal approaches to the buckling analysis of unbranched thin-walled
members, Advanced Steel Construction 5 (2) (2009) 95–223.
[35] S. Ádány, Shell element for constrained finite element analysis of thin-walled structural members, Thin-Walled Structures 105 (2016) 135–
146.
[36] S. Ádány, Constrained shell Finite Element Method for thin-walled members, Part 1: constraints for a single band of finite elements, Thin-
Walled Structures 128 (2018) 43–55.
[37] S. Ádány, D. Visy, R. Nagy, Constrained shell Finite Element Method, Part 2: application to linear buckling analysis of thin-walled members,
Thin-Walled Structures 128 (2018) 56–70.

28
[38] T. B. Pekoz, G. Winter, Torsional-flexural buckling of thin-walled sections under eccentric load, Journal of the Structural Division 95 (5)
(1969) 941–963.
[39] W.-W. Yu, R. A. LaBoube, Cold-formed steel design, John Wiley & Sons, 2010.
[40] R. D. Ziemian, Guide to stability design criteria for metal structures, John Wiley & Sons, 2010.
[41] S. Ádány, B. W. Schafer, Generalized constrained finite strip method for thin-walled members with arbitrary cross-section: Primary modes,
Thin-Walled Structures 84 (2014) 150 – 169.
[42] B. W. Schafer, Distoritonal Buckling of Cold-formed Steel Columns, Reseacrh Report sponsored by AISI (2000).
[43] S. C. Lau, G. J. Hancock, Distortional buckling formulas for channel columns, Journal of Structural Engineering 113 (5) (1987) 1063–1078.
[44] B. Schafer, T. Peköz, Laterally braced cold-formed steel flexural members with edge sti↵ened flanges, Journal of Structural Engineering
125 (2) (1999) 118–127.
[45] Z. Li, M. T. Hanna, S. Adany, B. W. Schafer, Impact of basis, orthogonalization, and normalization on the constrained finite strip method for
stability solutions of open thin-walled members, Thin-Walled Structures 49 (9) (2011) 1108 – 1122.
[46] AISI S200, North American Standard for Cold-Formed Steel Framing - General Provisions, American Iron and Steel Institute, Washington,
D.C. (2012).
[47] B. Schafer, Local, distortional, and Euler buckling of thin-walled columns, Journal of structural engineering 128 (3) (2002) 289–299.
[48] S. Ádány, B. W. Schafer, Generalized constrained finite strip method for thin-walled members with arbitrary cross-section: secondary modes,
orthogonality, examples, Thin-Walled Structures 84 (2014) 123–133.
[49] Z. Li, J. C. B. Abreu, J. Leng, S. Ádány, B. W. Schafer, Review: Constrained finite strip method developments and applications in cold-formed
steel design, Thin-Walled Structures 81 (2014) 2 – 18.
[50] Z. Li, Finite strip modeling of thin-walled members, Ph.D. thesis (2011).
[51] B. W. Schafer, T. Peköz, Computational modeling of cold-formed steel: characterizing geometric imperfections and residual stresses, Journal
of constructional steel research 47 (3) (1998) 193–210.
[52] V. Zeinoddini-Meimand, Geometric imperfections in cold-formed steel members, Ph.D. thesis (2011).
[53] B. Young, K. J. R. Rasmussen, Measurement techniques in the testing of thin-walled structural members, Experimental Mechanics 43 (1)
(2003) 32–38.
[54] K. D. Peterman, B. W. Shafer, Experiments on the stability of sheathed cold-formed steel studs under axial load and bending, Research Report
(2012).
[55] S. Farzanian, A. Louhghalam, B. Schafer, M. Tootkaboni, Geometric Imperfection Models in Shell FE Analysis of CFS Members: Validation
through Data-Driven Stochastic Models, in: Engineering Mechanics Institute Conference, Cambridge, MA, 2018.
[56] A. Lama Salomon, D. Fratamico, B. W. Schafer, C. Moen, Full field cold-formed steel column buckling measurements with high resolution
image-based reconstruction, in: Proc. of the Annual Stability Conf., Structural Stability Res. Co., Orlando, FL, 2016.
[57] L. McAnallen, D. Padilla-Llano, X. Zhao, C. Moen, B. W. Schafer, M. Eatherton, Initial geometric imperfection measurement and charac-
terization of cold-formed steel c-section structural members with 3D non-contact measurement techniques, in: Proceedings of the Structural
Stability Research Council, Toronto, Canada, 2014.
[58] X. Zhao, M. P. Tootkaboni, B. W. Schafer, High fidelity imperfection measurements and characterization for cold-formed steel members, in:
Proceeding of the 7th International Conference on Coupled Instabilities in Metal Structures, 2016.
[59] S. Farzanian, A. Louhghalam, B. W. Schafer, M. Tootkaboni, Geometric imperfections in shell finite element models of CFS members - A
review of current state of practice , in: Proceedings of the Annual Stability Conference Structural Stability Research Council / NASCC: The
Steel Conference, Baltimore, MD, 2018.
[60] R. G. Dawson, A. C. Walker, Post-buckling of geometrically imperfect plates, Journal of the Structural Division 98 (1) (1972) 75–94.
[61] G. Winter, Strength of thin steel compression flanges, Trans. ASCE 112 (1947) 527–554.
[62] J. Dutheil, The theory of instability through disturbance of equilibrium, in: Proceedings of the 4th Congress of IABSE, Association Interna-
tionale des Pont et Charpentes, 1952, pp. 275–295.

29
 [63] A. Chilver, The stability and strength of thin-walled steel struts, The engineer 196 (5089) (1953) 180–183.
[64] A. Walker, The post-buckling behaviour of simply-supported square plates, The Aeronautical Quarterly 20 (3) (1969) 203–222.
[65] K. Sivakumaran, N. Abdel-Rahman, A finite element analysis model for the behaviour of cold-formed steel members, Thin-walled structures
31 (4) (1998) 305–324.
[66] G. J. Hancock, Non linear analysis of thin sections in compression, Journal of Structural Division 25 (1) (1981) 455–471.
[67] J. Sun, J. W. Butterworth, Behaviour of steel single angle compression members axially loaded through one leg, in: Proc. Australasian Struct.
Engrg. Conference, Auckland, 1998, pp. 859–866.
[68] S. Chou, G. Chai, L. Ling, Finite element technique for design of stub columns, Thin-Walled Structures 37 (2) (2000) 97–112.
[69] J. Zaras, J. Rhodes, Carefully controlled compression tests on thin-walled cold-formed sections, Applied solid mechanics 2 (1987) 519–551.
[70] S. Chou, G. Chai, Ultimate design load of thin–walled stub columns, International Journal of Computer Applications in Technology 10 (1-2)
(1997) 27–33.
[71] L. Gardner, A new approach to structural stainless steel design, Ph.D. thesis, Imperial College London (University of London) (2002).
[72] P. Dinis, D. Camotim, Post-buckling behavior of cold-formed steel lipped channel columns a↵ected by distortional/global mode interaction,
in: Proceedings of SSRC Annual Stability Conference, 2008, pp. 405–431.
[73] P. Dinis, D. Camotim, Local/distortional/global buckling mode interaction in cold-formed steel lipped channel columns, in: Proceedings of
SSRC Annual Stability Conference, 2009, pp. 295–323.
[74] P. B. Dinis, D. Camotim, Local/distortional mode interaction in cold-formed steel lipped channel beams, Thin-Walled Structures 48 (10-11)
(2010) 771–785.
[75] B. W. Schafer, S. Ádány, Buckling analysis of cold-formed steel members using CUFSM: conventional and constrained finite strip methods,
in: Eighteenth international specialty conference on cold-formed steel structures, Orlando, FL, US, October, 2006, pp. 39–54.
[76] V. Zeinoddini, B. W. Schafer, Simulation of geometric imperfections in cold-formed steel members using spectral representation approach,
Thin-Walled Structures 60 (2012) 105 – 117.
[77] M. Tootkaboni, S. Farzanian, B. W. Schafer, Geometric imperfection model validation through data-driven stochastic modeling for reliable
analysis of thin-walled structures, in: Engineering Mechanics Institute Conference, San Diego, CA, 2017.
[78] S. Torabian, B. Zheng, S. Yared, B. W. Schafer, Direct Strength Prediction of Cold-Formed Steel Beam-Columns, Reseacrh Report RP16-3
sponsored by AISI (2016).
[79] S. Torabian, H. Amouzegar, M. Tootkaboni, B. W. Schafer, Finite element modeling protocols and parametric analyses for short cold-formed
steel zee-section beam-columns, in: Proceedings of the Annual Stability Conference Structural Stability Research Council, Orlando, FL,
2016.
[80] Australian Standard (AS 4084): Steel storage rackling (2012).
[81] Eurocode 3: Design of steel structures, part 1–6: strength and stability of shell structures (2007), European Committee for Standardization
(CEN), Brussels, Belgium.
[82] M. Farzanian, F. Arbabi, R. Pak, PML solution of longitudinal wave propagation in heterogeneous media, Earthquake Engineering and
Engineering Vibration 15 (2) (2016) 357–368.
[83] S. Farzanian, R. Shahsavari, Mapping the coupled role of structure and materials in mechanics of platelet-matrix composites, Journal of the
Mechanics and Physics of Solids 112 (2018) 169–186.
[84] B. W. Schafer, Z. Li, C. Moen, Computational modeling of cold-formed steel, Thin-Walled Structures 48 (10) (2010) 752 – 762.
[85] M. Crisfield, A fast incremental/iterative solution procedure that handles ”snap-through”, Computers & Structures 13 (1-3) (1981) 55–62.

30

View publication stats