Journal of Constructional Steel Research 144 (2018) 65–80

Contents lists available at ScienceDirect

Journal of Constructional Steel Research

Experiments on the global buckling and collapse of built-up cold-formed

steel columns
David C. Fratamico a,⁎, Shahabeddin Torabian a,b, Xi Zhao c, Kim J.R. Rasmussen d, Benjamin W. Schafer a
a
Department of Civil Engineering, Johns Hopkins University, Baltimore, MD, USA
b
School of Civil Engineering, University of Tehran, Tehran, Iran
c
Department of Mathematics and Computer Science, West Virginia State University, USA
d
School of Civil Engineering, University of Sydney, Sydney, Australia

a r t i c l e i n f o a b s t r a c t

Article history: This paper reports on experiments addressing the buckling and collapse behavior of back-to-back lipped channel
Received 5 August 2017 built-up cold-formed steel (CFS) columns assembled using 16 different CFS lipped channel sizes. The lipped
Received in revised form 10 January 2018 channel sections are connected at the web using a pair of self-drilling screw fasteners at a specified spacing
Accepted 11 January 2018
along the column length of 1.83 m (6 ft). These experiments aim to quantify the effect of two web fastener
Available online xxxx
layouts on composite action for each section size, study member end fixity, observe buckling and collapse
Keywords:
behavior, and provide benchmarks for design that includes specific considerations for thin-walled member
Column buckling. A total of 32 monotonic, displacement-controlled, concentric compression tests are completed with
Built-up up to 17 position transducers monitoring displacements at key locations. All tests are conducted with the
Cold-formed steel built-up member seated in CFS tracks, as would be found in practice. Local–global interaction is shown to be a
Buckling prevalent failure mode, and the stud-to-track end condition is determined to be semi-rigid, but generally closer
Collapse to a fixed condition. End rigidities are estimated using a Southwell approach. Rational design approaches
DSM extending the application of the Direct Strength Method (DSM) and employing current state-of-the-art
numerical modeling techniques are proposed and validated with test data. In addition, the development of
definitive design recommendations that help reduce the complexity of fastener designs and incorporates the
DSM framework when predicting built-up member strength is underway.
© 2018 Elsevier Ltd. All rights reserved.

1. Introduction Current design of built-up CFS columns is highly simplified. The

maximum fastener spacing for the inter-connection between the webs
1.1. Background and current design is determined by ensuring that flexural buckling of the individual
studs between fasteners will not occur prior to flexural buckling of the
Cold-formed steel (CFS) framing provides a structural system that built-up section in both the Australia/New Zealand 4600 Standard [1]
consists of repetitively placed columns (studs) or beams (joists) to and in North America with AISI S100 [2]. In addition, AISI S100
frame out the building. When higher local rigidity, or axial or flexural Section I1.2 requires the calculation of axial capacity using a modified
capacity is required, such as in shear wall chords or around openings slenderness ratio approach, which was adopted from AISC 360 [3]. The
in headers or jambs, multiple members are connected together principle is that the individual studs in a built-up section cannot be
to form a built-up CFS member. The back-to-back lipped channel, assumed to be fully connected such that shear forces are continuously
or “I” section is the most traditional built-up CFS section as it is easy to transferred from one stud to another. The modified slenderness,
assemble and produces a doubly symmetric cross section. Individual which is a function of the slenderness of the fully built-up section and
CFS studs are placed back-to-back and fastened together using the fastener spacing, assumes a loss of total shear rigidity at the fas-
self-drilling screws, welds, or bolts. Depending on the detail composite teners, but considers only minor-axis flexural buckling in the estimation
action can be developed via these inter-connections, giving axial or of elastic buckling stress, which is calculated using the Euler buckling
bending capacities for the built-up member that are ideally greater formula. This method increases the slenderness ratio of the built-up sec-
than the sum of the capacities of the individual sections, when global tion to reduce its flexural buckling capacity. However, the modified
buckling controls. slenderness approach cannot predict the effects of fastener layouts
on torsional, flexural-torsional, local, or distortional buckling modes.
⁎ Corresponding author. As is typical in design, it is assumed that the back-to-back fasteners
E-mail address: fratamico@jhu.edu (D.C. Fratamico). are always employed in pairs. In addition, AISI S100 also requires a

https://doi.org/10.1016/j.jcsr.2018.01.007
0143-974X/© 2018 Elsevier Ltd. All rights reserved.
66 D.C. Fratamico et al. / Journal of Constructional Steel Research 144 (2018) 65–80

prescriptive fastener grouping: an array of screw pairs that must be lon-

gitudinally spaced no more than 4 diameters apart and for a distance
equal to 1.5 times the maximum width of the built-up section along
the length, at either end of the column. This grouping is henceforth
denoted as an end fastener group (EFG), and their function is to reduce
the relative slip between two connected studs as they sustain flexural
buckling; this reduction of slip boosts composite action and hence, the
buckling load of a column. The impact of the EFG on the modified slen-
derness is not treated directly. For a complete column design, nominal
strengths for modes other than minor-axis flexure are required to be
determined using either the Effective Width Method (EWM) or the
Direct Strength Method (DSM) on either the individual sections
or the built-up cross section, although guidelines in the specifications
[1, 2, 4, 5] are not clear.
Recent research on built-up CFS columns has worked to address lim- Fig. 1. The back-to-back section with web screws shown (left) and placement of
itations in current design specifications [1, 2, 5]. Tests on back-to-back EFGs (right).
lipped channels showed AISI's modified slenderness approach is conser-
vative and the importance of end conditions were highlighted in Stone
and LaBoube [6]. More complex CFS built-up columns including combi-
nations of Zee, sigma, and track sections were tested, studied by numer- Fig. 2 shows the MTS rig setup used for all tests. Studs were
ical analyses, and a DSM design approach including bucking mode installed within tracks, with stud flanges screwed to track lips, and
interaction was offered in Georgieva et al. [7–10]. Back-to-back built- tracks were placed on fixed platens made of 12.7 mm (0.5 in.) thick
up CFS sigma sections were tested and simplified models developed low carbon steel. Studs were installed parallel to each other with a
for predicting the partially composite buckling response in Zhang and maximum relative inclination error of 0.05° from the horizontal.
Young [11]. Tests on back-to-back and toe-to-toe built-up sections Fig. 2 (section B-B′) shows the placement of the specimens in
compared to AISI [2] and Eurocode [5] were shown to agree best the rig. Tracks were positioned and held in place on the platens
when fixed ends were assumed in the comparisons [12]. Tests on using clamps prior to testing and in the absence of applied loads.
built-up CFS sections with intermediate stiffeners focusing on local Eccentricity and out-of-plumbness were recorded for each specimen,
and distortional buckling demonstrated that using only single section and maintained to be less than 0.64 mm (0.025 in.). A load cell installed
properties (hence, employing a non-composite assumption) were ade- at the top crosshead measured axial force and a built-in LVDT measured
quate for design [13]. Others have also explored modifications to DSM applied axial displacements. A LabVIEW program coordinated all data
for built-up sections [14], although the effect of end conditions and fas- acquisition and National Instruments hardware was used to establish
tener layouts were not explicitly addressed in the new design approach. a control loop.
Other testing has shown that if individual sections are longer relative to To track specimen deformations, 17 position transducers (PTs) were
others within the same built-up section, the longer studs can attract installed in the test setup. Lateral bi-planar displacements and rotation
load earlier and buckle before the other studs, reducing overall column of the cross section at mid-height were tracked throughout the
capacity [15]. EFG are intended to attenuate this effect, but the authors tests using 11 PTs at mid-height as shown in section A-A′ of Fig. 2.
do not comment on their effectiveness. To monitor stud engagement to the track during the tests, a PT
The experimental work and design method developments pre- was installed between the column web and track web at the top and
sented herein follow a previous phase of sheathed and bare built-up again at the bottom. In addition, 1 PT was installed on the top and
CFS column tests reported in Fratamico et al. [16]. The goal here is to ex- bottom tracks to record out-of-plane deformations of the webs due to
perimentally assess whether current design methods are adequate for local buckling or localized failures at the column ends. Lastly, for some
an all-steel design of non-perforated and un-braced built-up CFS col- specimens with wider flange widths of 41.3 mm (1.625 in.), a PT
umns buckling mostly in flexure. True end conditions are recreated in was placed near the top and bottom of the column to monitor the
the lab by installing stud ends in tracks, thereby also allowing for a relative slip of the channel webs seated in the track at the ends. The
quantitative assessment of column end rigidities which heavily influ- setup for these PTs is shown in Fig. 3. The PTs were firmly attached
ence global buckling capacities. These tests are part of an ongoing effort with magnets, but they did not disrupt cross section deformations
to understand and quantify flexural, torsional, and shear rigidities of during the tests.
built-up CFS sections. The focus in this paper is to demonstrate the
effects of column end rigidity, web fastener layout, and cross section 2.2. Specimen selection and assembly
size on the achievable composite action, interactive buckling modes,
and collapse behavior of back-to-back CFS columns. Test results A total of 16 lipped channel section types were selected based on
will also augment benchmarks for the assessment of appropriate the capacity to include a wide range of global slenderness and cross
design methods. section shapes commonly used for columns in CFS structures. These
section types are listed in the test matrix in the following section
2. Design of experiments and have a nominal steel thickness ranging from 0.84 mm (33 mil) to
1.72 mm (68 mil). Nominal web depths range from 63.5 mm (2.5 in.)
2.1. Test setup to 152 mm (6 in.). Two nominal flange widths are chosen: 34.9 mm
(1.375 in.) and 41.3 mm (1.625 in.), with lip lengths specified according
All 32 test specimens are composed of back-to-back lipped channel to flange widths. All columns were unsheathed and tested at a length of
sections (henceforth referred to as studs) with screw pairs connecting 1.83 m (6 ft). Although this is shorter than the typical 2.44 m (8 ft) stud
the channels through the webs as shown in Fig. 1. Monotonic, length in residential construction, the specimens are still long enough to
displacement-controlled, concentric compression loading was ap- mobilize the global buckling mode of failure.
plied to each column using a 445 kN (100 kip) MTS universal testing The lipped channel sections were fastened together and to their
rig which has a servo-controlled hydraulic actuator. The load rate did corresponding track sections with Simpson #10 self-drilling screws
not exceed 0.76 mm/min (0.03 in/min). for sections with plate thicknesses of 1.37 mm (54 mil) and 1.72 mm
 D.C. Fratamico et al. / Journal of Constructional Steel Research 144 (2018) 65–80 67

Section A-A’

A A’

B B’ Section B-B’

Fig. 2. MTS testing rig with specimen installed (left), position transducer arrangement at column mid-height (Section A-A′), and specimen positioning on the platens (Section B-B′).

Fig. 3. PT setup to measure shear slip in an undeformed (left) and a deformed state (right).
68 D.C. Fratamico et al. / Journal of Constructional Steel Research 144 (2018) 65–80

Fig. 4. Cross section linear dimensions (left), and angles and radii (right).

(68 mil). For sections with 0.84 mm (33 mil) and 1.09 mm (43 mil) of the 3D reconstructed “true” geometry from the point cloud.
plate thickness, smaller sized Posi-Grip #8 self-drilling screws Fig. 4 provides the cross-section dimension nomenclature and
were used. Tables 1 and 2 contain summary dimension data per stud type. More
detailed data is available in Fratamico [19].
2.3. Laser scanning for quantification of geometric imperfections
2.4. Coupon testing for material characterization
Measurements of specimen cross section dimensions were com-
pleted using a novel laser scanning method [17]. Full-field 3D geometric A series of 42 coupon tests were completed to quantify basic
information was obtained as a point cloud of stitched longitudinal scan material properties of each channel and track section used. Coupon
readings from multiple scan angles [18]. The average plate thickness for preparation and testing were completed in accordance with ASTM
each specimen was measured by hand using a calibrated micrometer A370-12a [20]. All coupons were longitudinally cut from the webs and
and incorporated into the reconstruction of the stud geometry from flanges and milled to shape with a computer numerical control (CNC)
point clouds. Typical specimens consist of a point cloud of over milling machine. The protective zinc coating was removed from the
100,000 points defining the member. Fig. 5 provides an example steel using a 1 M HCl solution at each end of the coupons (outside of
their gauge length) to measure coated and uncoated thicknesses.
Summary material results are provided in Table 3. The yield stress was
determined using the 0.2% offset method as indicated in Table 3 and
illustrated in Fig. 6. Yield stresses Fy for the sections with nominal
0.84 mm (33 mil) and 1.09 mm (43 mil) thicknesses, manufactured
with a specified yield stress of 228 MPa (33 ksi), were determined to
be 320 MPa (46.4 ksi) and 333 MPa (48.3 ksi), respectively. For the
sections with nominal thicknesses of 1.37 mm (54 mil) and 1.73 mm
(68 mil), Fy was determined to be 395 MPa (57.3 ksi) and 353 MPa
(51.1 ksi), respectively. These sections were manufactured with a
specified yield stress of 345 MPa (50 ksi). Young's modulus was not
extracted from the data and is assumed to be 203 GPa (29,500 ksi) as
prescribed in AISI S100 [2].

3. Experimental study

3.1. Test matrix

The test series is intended to explore the effect of changing cross sec-
tional shape, plate thicknesses, and web fastener layout on composite
action, observed buckling modes, and strength. No sheathing, bracing,
or stud perforations are included in these test specimens. Two fastener
Fig. 5. Full-field 3D reconstruction of true geometry, showing color contours as spatial layouts are used, as shown in Fig. 7: the first case is a full, AISI-based
deviations from the nominal shape of a 362S137-68 built-up stud. layout with EFGs superimposed on an even spacing of L/4 as calculated
 D.C. Fratamico et al. / Journal of Constructional Steel Research 144 (2018) 65–80 69

Table 1
Summary cross section dimensions for studs with web depths less than 152 mm (6 in.)

Section Left stud: Hleft B1 B2 D1 D2 R1a R1b R2a R2b θ1a θ1b θ2a θ2b

Right stud: Hright B4 B3 D4 D3 R4a R4b R3a R3b θ4a θ4b θ3a θ3b

(mm) (mm) (mm) (mm) (mm) (mm) (mm) (mm) (mm) (deg.) (deg.) (deg.) (deg.)

250S137-33 Nominal 63.50 34.93 34.93 9.53 9.53 1.94 1.94 1.94 1.94 90.00 90.00 90.00 90.00
Mean (left) 65.87 32.12 32.10 12.88 12.26 4.49 2.83 4.23 3.04 83.46 89.12 83.38 88.96
Mean (right) 65.82 31.08 32.82 11.66 12.40 4.49 2.78 4.78 3.53 76.60 89.39 87.92 89.48
250S137-43 Nominal 63.50 34.93 34.93 9.53 9.53 1.81 1.81 1.81 1.81 90.00 90.00 90.00 90.00
Mean (left) 66.24 33.12 31.78 13.34 11.96 4.36 3.13 4.43 2.68 88.37 90.96 80.36 89.58
Mean (right) 66.17 31.71 32.67 11.50 12.61 4.62 3.16 5.11 3.68 79.09 89.58 85.59 90.63
250S137-54 Nominal 63.50 34.93 34.93 9.53 9.53 2.16 2.16 2.16 2.16 90.00 90.00 90.00 90.00
Mean (left) 63.91 33.27 34.21 9.78 9.98 4.59 4.13 4.52 4.16 79.79 90.06 82.75 89.17
Mean (right) 63.99 33.76 33.32 9.41 9.23 5.02 4.44 5.01 3.97 81.27 88.90 78.16 90.96
250S137-68 Nominal 63.50 34.93 34.93 9.53 9.53 2.72 2.72 2.72 2.72 90.00 90.00 90.00 90.00
Mean (left) 62.79 32.96 31.19 10.04 9.94 4.59 4.44 4.66 3.98 85.04 88.33 81.44 89.25
Mean (right) 62.73 31.28 32.51 9.75 9.49 4.93 4.73 4.42 4.43 79.88 89.16 81.13 87.77
362S137-33 Nominal 92.08 34.93 34.93 9.53 9.53 1.94 1.94 1.94 1.94 90.00 90.00 90.00 90.00
Mean (left) 93.12 33.63 31.97 12.14 11.61 4.62 2.91 4.61 2.75 77.08 89.68 74.22 89.17
Mean (right) 93.18 31.96 33.38 11.34 11.95 4.36 2.79 4.71 3.03 75.05 89.43 76.21 89.56
362S137-43 Nominal 92.08 34.93 34.93 9.53 9.53 1.81 1.81 1.81 1.81 90.00 90.00 90.00 90.00
Mean (left) 93.34 33.78 32.21 13.34 11.55 4.80 3.56 4.40 3.08 83.46 90.07 77.72 89.10
Mean (right) 93.38 32.22 33.39 11.81 12.77 4.74 3.04 5.01 3.88 77.20 89.12 81.30 90.36
362S137-54 Nominal 92.08 34.93 34.93 9.53 9.53 2.16 2.16 2.16 2.16 90.00 90.00 90.00 90.00
Mean (left) 92.43 34.06 33.04 9.53 9.39 4.31 3.84 4.38 3.94 80.54 88.82 78.65 88.77
Mean (right) 92.49 33.01 33.82 8.98 8.64 4.59 3.68 4.48 4.21 76.65 89.01 78.31 88.80
362S137-68 Nominal 92.08 34.93 34.93 9.53 9.53 2.72 2.72 2.72 2.72 90.00 90.00 90.00 90.00
Mean (left) 92.27 32.26 31.33 9.29 8.74 4.75 4.15 5.46 4.23 82.39 88.00 69.59 88.98
Mean (right) 92.27 31.42 32.29 8.87 8.27 4.97 4.26 4.76 4.41 78.36 89.14 79.84 88.61

Table 2
Summary cross section dimensions for studs with web depths of 152 mm (6 in.)

Section Left stud: Hleft B1 B2 D1 D2 R1a R1b R2a R2b θ1a θ1b θ2a θ2b

Right stud: Hright B4 B3 D4 D3 R4a R4b R3a R3b θ4a θ4b θ3a θ3b

(mm) (mm) (mm) (mm) (mm) (mm) (mm) (mm) (mm) (deg.) (deg.) (deg.) (deg.)

600S137-33 Nominal 152.4 34.93 34.93 9.53 9.53 1.94 1.94 1.94 1.94 90.00 90.00 90.00 90.00
Mean (left) 153.3 34.30 32.40 11.38 10.20 4.61 2.96 4.35 2.86 79.52 87.92 77.38 89.20
Mean (right) 153.5 32.48 34.31 10.69 10.99 4.18 2.88 4.67 3.36 76.60 89.39 79.37 87.55
600S137-43 Nominal 152.4 34.93 34.93 9.53 9.53 1.81 1.81 1.81 1.81 90.00 90.00 90.00 90.00
Mean (left) 153.2 34.73 32.92 12.23 11.17 4.65 3.70 4.50 3.14 83.41 89.09 78.37 88.41
Mean (right) 153.2 33.09 34.71 11.06 12.16 4.66 3.05 4.77 3.76 78.26 88.67 82.65 89.03
600S137-54 Nominal 152.4 34.93 34.93 9.53 9.53 2.16 2.16 2.16 2.16 90.00 90.00 90.00 90.00
Mean (left) 152.2 33.86 32.44 10.30 9.46 4.28 3.84 4.45 4.30 81.49 87.92 77.22 88.21
Mean (right) 152.2 33.73 32.63 10.23 9.72 4.98 3.91 4.61 3.51 82.89 87.91 77.83 88.51
600S137-68 Nominal 152.4 34.93 34.93 9.53 9.53 2.72 2.72 2.72 2.72 90.00 90.00 90.00 90.00
Mean (left) 151.2 30.68 31.97 10.18 10.85 4.90 4.00 4.61 4.07 80.37 88.56 85.43 87.24
Mean (right) 151.2 31.74 30.89 10.66 9.70 4.72 4.33 5.22 4.41 83.89 87.28 79.71 88.76
600S162-33 Nominal 152.4 41.28 41.28 12.70 12.70 1.94 1.94 1.94 1.94 90.00 90.00 90.00 90.00
Mean (left) 152.6 40.45 40.23 11.04 11.64 3.59 4.43 3.53 3.93 83.89 89.49 84.03 89.05
Mean (right) 152.8 40.01 40.33 12.08 10.15 4.14 3.89 3.84 4.35 82.67 88.91 82.96 89.30
600S162-43 Nominal 152.4 41.28 41.28 12.70 12.70 1.81 1.81 1.81 1.81 90.00 90.00 90.00 90.00
Mean (left) 152.6 41.04 41.39 10.93 12.01 3.54 5.05 3.64 4.37 84.87 89.97 84.55 90.48
Mean (right) 152.7 41.12 40.93 11.33 11.05 3.90 5.10 4.11 4.15 83.18 90.46 83.10 90.19
600S162-54 Nominal 152.4 41.28 41.28 12.70 12.70 2.16 2.16 2.16 2.16 90.00 90.00 90.00 90.00
Mean (left) 153.1 41.30 41.45 11.91 11.62 3.66 4.98 3.99 4.89 86.32 90.26 86.52 89.70
Mean (right) 153.0 41.27 41.41 11.63 10.47 3.90 4.55 4.03 4.66 86.30 89.68 85.08 89.96
600S162-68 Nominal 152.4 41.28 41.28 12.70 12.70 2.72 2.72 2.72 2.72 90.00 90.00 90.00 90.00
Mean (left) 153.5 40.66 38.99 10.68 12.62 4.09 5.17 4.38 5.20 85.26 93.54 81.28 89.05
Mean (right) 153.8 39.14 41.05 11.97 9.77 4.54 6.22 4.58 6.09 80.85 89.07 84.41 93.39

Table 3
Summary results from coupon tests, grouped by plate thickness.

Nom. thickness (mm) [mil] Base thickness t (mm) [in.] Yield stressa Fy (MPa) [ksi] Tensile strength Fu (MPa) [ksi] Strain at ultimate εu (mm/mm)

0.84 [33] Mean 0.88 [0.0345] 320 [46.4] 383 [55.6] 0.215
CoV (%) 1.86 4.45 3.67 9.14
1.09 [43] Mean 1.12 [0.0441] 333 [48.3] 403 [58.5] 0.208
CoV (%) 0.92 2.69 7.68 11.7
1.37 [54] Mean 1.40 0.0553] 395 [57.3] 482 [69.9] 0.173
CoV (%) 2.39 5.01 4.85 6.16
1.73 [68] Mean 1.83 [0.0719] 353 [51.1] 474 [68.7] 0.157
CoV (%) 1.17 8.75 6.72 7.03
a
The 0.2% offset method is used.
70 D.C. Fratamico et al. / Journal of Constructional Steel Research 144 (2018) 65–80

700 100 Table 4

Test matrix.
90
600
Specimen Section EFG Specimen Section EFG
80 ID installed? ID installed?
500
70 A1-33 250S137-33 Yes C1-33 600S137-33 Yes
A1-43 250S137-43 Yes C1-43 600S137-43 Yes
Stress (MPa)

Stress (ksi)
400 60
A1-54 250S137-54 Yes C1-54 600S137-54 Yes
50 A1-68 250S137-68 Yes C1-68 600S137-68 Yes
300 A2-33 250S137-33 C2-33 600S137-33
40 A2-43 250S137-43 C2-43 600S137-43
A2-54 250S137-54 C2-54 600S137-54
200 30 A2-68 250S137-68 C2-68 600S137-68
B1-33 362S137-33 Yes D1-33 600S162-33 Yes
20
100 Engineering B1-43 362S137-43 Yes D1-43 600S162-43 Yes
True 10 B1-54 362S137-54 Yes D1-54 600S162-54 Yes
0.2% Offset
B1-68 362S137-68 Yes D1-68 600S162-68 Yes
0 0 B2-33 362S137-33 D2-33 600S162-33
0 0.05 0.1 0.15 0.2 0.25
B2-43 362S137-43 D2-43 600S162-43
Strain (mm/mm)
B2-54 362S137-54 D2-54 600S162-54
B2-68 362S137-68 D2-68 600S162-68
Fig. 6. Example of stress vs. strain relationship from a test on a coupon cut from a 600S137-
54 stud; only engineering stress and strain are included in the results table.

using AISI S100 [2], and the second case is with the even fastener
spacing, but without the EFGs. Fastener spacing is determined per AISI specimens. EFG lengths were calculated as 1.5 times the maximum
S100-16 sections I1.2 and J4.2 to be L/4, or 45.7 cm (18 in.) for all width of the built-up section: 105 mm (4.13 in.), 138 mm (5.44 in.),
and 229 mm (9.00 in.) for the specimens with 63.5 mm (2.5 in.),
92.1 mm (3.625 in.), and 152 mm (6 in.) web depths, respectively.
The 16 section types (4 cross sections and 4 thicknesses) were selected
based on the capacity of the testing rig and with the intent of including a
wide range of studs commonly used in CFS framing. Table 4 provides
the test matrix.

3.2. Test results

All columns were expected to buckle in global, minor-axis flexure

potentially with local buckling interaction. From general test observa-
tions, this hypothesis holds. However, local–global interactive buckling
was prevalent, particularly in the trials with deeper webs and thinner
steel. Although the effect of buckling interactions on column strength
is not the main scope of this work, the effect of the interactions on
composite action is briefly discussed below.
Load vs. axial displacement data is provided in Fig. 8. For thinner
specimens, namely those with 0.84 mm (33 mil) thick plates, localized
failures near the flanges and lips governed post-peak deformations. In
particular, for smaller web depths of 63.5 mm (2.5 in.) and 92.1 mm
(3.625 in.), a modest increase in capacity is observed when EFGs are
added to the columns. For example, a 20% boost in capacity is observed
when comparing strengths of specimens A1-54 and A2-54, as shown in
Table 5. In thinner columns, local–global interaction is more prevalent,
as shown in Fig. 9. In addition, these thinner sections were also observed
to have a capacity and deformation mode which was more sensitive to
localized geometric imperfections in the flanges and lips; examples of
this are shown in Figs. 9 and 10. In some cases, such as trials B1-33
(with EFG) and B2-33 (no EFG), no increase in strength was observed
with the addition of the EFGs, and a stronger presence of localized
failure in lips and flanges (usually prior to flexural buckling) was instead
noted during the tests. EFGs are only effective against global buckling,
not local buckling.
As the thickness increases, less interaction of local buckling
with global (flexural) buckling is observed, and the effect of the
EFG on increasing capacity (in fact, up to 42% increase for B1-68
and B2-68) and stiffness is also greater when single-mode global
(flexural) deformations prevail. This behavior is shown in Table 5
and both plots in Fig. 8 for 1.73 mm (68 mil) thick specimens.
Table 6 provides results for all sections with web depths of 152 mm
(6 in.). In general, stiffness and capacity increase with the addition of
Fig. 7. Web fastener layouts with and without EFG as used for each section type, showing EFGs in the thicker sections, as shown in the plots in Fig. 11. For
examples for columns with web depths of 92.1 mm (3.625 in.) and 152 mm (6 in.) example, a 33.3% increase in capacity is observed in trial C1-68 when
 D.C. Fratamico et al. / Journal of Constructional Steel Research 144 (2018) 65–80 71

Fig. 8. Load vs. axial displacement curves for A (left) and B (right) series specimens, with solid and dashed lines representing specimens with and without EFG, respectively.

compared with C2-68 which did not have EFG installed. This boost the track once again. This is due to the rotation of the stud ends rel-
in capacity, as with the sections with shorter web depths, is not as ative to the track web as the column undergoes flexural deforma-
prevalent in thinner sections, which are dominated by local–global tions and increasing lateral displacements.
buckling. When comparing peak loads for the C and D series trials
(with differing flange widths), the ratios of capacity normalized to 3.3. Assessment of column end rigidity and shear slip
squash load remain around 0.4, and less than 0.55 for stockier sections.
For these 16 trials, local–global interaction was observed, and in some End conditions were closely monitored with cameras and position
cases, a local-distortional-global interaction occurred and was notice- transducers throughout the entire loading phase. After stud engagement
able at peak load, as shown in Fig. 12. to track, an additional source of end condition nonlinearity was observed
Figs. 8 and 11 show an approximately bi-linear stiffness before in most trials. Occurring just after buckling when flexural deformations
peak. This is attributed to the early stages of loading in which the col- are amplified by P-δ effects, lips and a portion of flanges on the concave
umn end conditions change as the ends of the studs bear down onto side of the column lift off from the tracks, changing not only the stress
the webs of the track (i.e. as the stud is seated in the track). This type distribution on the cross section, but also reducing the rigidity of the
of end condition nonlinearity was most prevalent in the B, C, and D end condition and column strength. Fig. 14 shows an example of the
series trials with cross-sections with web depths of 3.625 in. (92.1 semi-rigid end condition and the change in cross section bearing on
mm) and 6 in. (152 mm). The effect is intentional in all specimens the track web. This effect is not present in many other built-up CFS
to ensure a more realistic stud loading condition. Position transduc- column tests that employ welded plates at the stud ends.
ers measured the relative displacement of the tracks at the top and In Fig. 15, plots of load vs. the lateral deflection at mid-height are
bottom of the studs to monitor this stud seating in the track (PT dis- shown, and plots with curves for all section types are made for each
placement ceases to change when studs fully bear down on the track plate thickness. Lateral deflection was measured using the average of
webs), as shown in Fig. 13. At P/Py = 0.35, the stud begins to lift off PTs 1 and 3, whose positions are shown in Fig. 2. In many trials, the

Table 5
Test results for sections with web depths 63.5 mm (2.5 in.) and 92.1 mm (3.625 in.)

Trial ID Section EFG installed Stiffness, k kN/mm [kip/in]a Observed buckling modeb Tested strength, Pu kN [kips] Pu/Py Pu,EFG/Pu,noEFG

A1-33 250S137-33 Yes 20.44 [116.7] L-G 71.56 [10.38] 0.57 1.11
A2-33 250S137-33 – 21.94 [125.3] L-G 64.66 [9.378] 0.51
A1-43 250S1374 Yes 26.24 [149.9] G 103.5 [15.01] 0.61 1.06
A2-43 250S137-43 – 25.11 [143.4] G 97.61 [14.16] 0.57
A1-54 250S134 Yes 34.42 [196.5] G 175.9 [25.51] 0.71 1.20
A2-54 250S137-54 – 29.23 [166.9] G 146.2 [21.20] 0.59
A1-68 250S137-68 Yes 42.44 [242.3] G 184.6 [26.77] 0.67 1.14
A2-68 250S137-68 – 38.74 [221.2] G 162.3 [23.54] 0.59
B1-33 362S137-33 Yes 20.90 [119.4] L-G 68.70 [9.964] 0.45 0.99
B2-33 362S137-33 – 15.41 [88.01] L-G 69.12 [10.03] 0.46
B1-43 362S137-43 Yes 30.30 [173.0] L-G 109.6 [15.89] 0.54 1.16
B2-43 362S137-43 – 21.68 [123.8] L-G 94.56 [13.71] 0.46
B1-54 362S137-54 Yes 37.55 [214.4] L-G 159.6 [23.15] 0.53 1.15
B2-54 362S137-54 – 28.96 [165.4] L-G 139.1 [20.17] 0.46
B1-68 362S137-68 Yes 41.66 [237.9] G 179.7 [26.06] 0.54 1.42c
B2-68 362S137-68 – 22.51 [128.6] G 126.6 [18.37] 0.38
a
The initial elastic stiffness, interpolated from raw data at the point after the studs engage with the tracks.
b
L = web local buckling and G = minor-axis flexural buckling.
c
B2-68 had a distinct initial imperfection as shown in Fig. 5.
72 D.C. Fratamico et al. / Journal of Constructional Steel Research 144 (2018) 65–80

Fig. 9. Local–global (flexural) buckling deformation in A1-33 (left) and single-mode minor
axis flexural buckling deformation in A2-68 (right) just after peak load.

Fig. 10. Localized failures in B2-43 (left) and single-mode minor axis flexural buckling in
lateral deflection changed sign through the initial loading phase due to B2-68 (right) after peak load
the interaction of local and global buckling waves at the location of the
PT, before measuring the progression of the flexural buckling deflection.
In Fig. 15, all flexural buckling-based lateral deflections are shown to be
positive. Since flexural and local–global (flexural) buckling interaction Pcr is the approximate elastic minor-axis flexural buckling load
controlled in the whole test series, the built-up cross sections with the (since local, distortional, and torsional deformations can alter Pcr) and
deepest webs and widest flanges (the D series specimens), which had δo is the initial mid-height imperfection in the column. The estimated
the largest minor-axis moment of inertia, reached the highest capaci- critical load is the slope of a plot of δ/P vs. δ. The slope is recorded, and
ties. The plots also indicate that the varying column stiffnesses and ini- Pcr estimated only when the stud is fully seated, i.e. in the pre-peak
tial flexural imperfections, coupled with the nonlinear end conditions, range when the stud-to-track movement is fixed as shown in Fig. 13.
are present across all specimens and must be considered when estimat- Once the critical load is estimated, an effective length factor K can be
ing column strength. Nevertheless, these tests provide excellent bench- back-calculated from the Euler buckling formula, where the Euler
marks with realistic end conditions for developing future analytical and critical load is set equal to Pcr calculated using the Southwell approach.
design methods. Fig. 16 provides back-calculated K values for all trials, assuming non-
As described for an earlier phase of column tests [16], the composite or fully composite action when calculating the moment of
Southwell method [21] was employed to back-calculate the effective inertia to use in the Euler buckling formula; Eq. (2) is for the non-
length factor K from the Euler equation for flexural buckling using composite moment of inertia INC and Eq. (3) is for fully composite IFC
the test data. This method was selected as a simple approach to esti- and employs the parallel axis theorem. A is the single channel section
mate the elastic global buckling load for a column buckling in almost area, Io is the single channel section weak axis moment of inertia,
pure minor-axis flexure, provided reliable data of load and mid- and x is the distance from the outside of the web to the centroid
height lateral displacement can be extracted with minimal noise along the strong axis.
from a column test before the peak load has been reached. The lateral
displacement is measured as the average of PTs 1 and 3 at mid-height
(i.e. the data shown in Fig. 15) and is used as δ in Eq. (1) along with INC ¼ 2Io ð2Þ
the applied load P.
IFC ¼ 2I o þ 2Ax2 ð3Þ
 
δ Although the level of composite action is challenging to estimate ac-
δ ¼ P cr − δo ð1Þ
P curately, a partially-composite condition is achieved through the web
 D.C. Fratamico et al. / Journal of Constructional Steel Research 144 (2018) 65–80 73

Table 6
Test results for sections with web depths of 152 mm (6 in.)

Trial ID Section EFG installed Stiffness, k kN/mm [kip/in]a Observed buckling modeb Tested strength, Pu kN [kips] Pu/Py Pu,EFG/Pu,noEFG

C1-33 600S137-33 Yes 22.88 [130.7] L-G 77.87 [11.29] 0.38 0.96
C2-33 600S137-33 – 17.42 [99.46] L-G 80.82 [11.72] 0.40
C1-43 600S137-43 Yes 31.87 [182.0] L-G 122.0 [17.69] 0.44 1.11
C2-43 600S137-43 – 35.07 [200.3] L-G 109.8 [15.92] 0.40
C1-54 600S137-54 Yes 48.52 [277.0] L-G 159.8 [23.18] 0.39 1.14
C2-54 600S137-54 – 45.36 [259.0] L-G 140.6 [20.39] 0.35
C1-68 600S137-68 Yes 58.69 [335.1] L-G 234.2 [33.97] 0.52 1.33
C2-68 600S137-68 – 56.59 [323.1] L-G 175.7 [25.49] 0.39
D1-33 600S162-33 Yes 21.50 [122.8] L-G 76.25 [11.06] 0.35 1.01
D2-33 600S162-33 – 20.69 [118.2] L-G 75.20 [10.91] 0.34
D1-43 600S162-43 Yes 30.77 [175.7] L-G 111.5 [16.17] 0.37 0.95
D2-43 600S162-43 – 32.61 [186.2] L-G 117.2 [17.00] 0.39
D1-54 600S162-54 Yes 48.26 [275.6] L-G 213.5 [30.97] 0.49 1.22
D2-54 600S162-54 – 38.06 [217.3] L-G 174.7 [25.34] 0.40
D1-68 600S162-68 Yes 65.43 [373.6] L-G 270.5 [39.24] 0.55 1.06
D2-68 600S162-68 – 52.83 [301.7] L-G 254.9 [36.97] 0.52
a
The initial elastic stiffness, interpolated from raw data at the point after the studs engage with the tracks.
b
L = web local buckling and G = minor-axis flexural buckling.

fasteners in all trials. When comparing tests with and without EFG, the built-up members. DSM assumes that if the elastic buckling modes
effect of EFG on end rigidity is small. Nevertheless, using both upper and for local, distortional, and global (including flexure, torsion, and
lower bounds of composite action, and noting the effect of cross-section combinations of both) are determined and appropriately combined
type, approximate end rigidities are calculated as follows: assuming with the squash load that these parameters can be used to empirically
non-composite action, the average K factor is 0.72 and assuming fully- determine member strength. Thus, the design problem in DSM consists
composite action, the average K factor is 0.89, as illustrated in Fig. 16. of determining elastic buckling loads, followed by application of the
Although load-axial displacement data shows the effect of compos- existing strength formulae and assessment of their adequacy. In the sec-
ite action on stiffness and strength, direct measurements of shear tion that follows, multiple rational approaches to determine the elastic
slip between the screw-connected webs were also made using a buckling load of a built-up member with discrete fastening, varying end
special PT setup on all D series specimens. As an example, results for boundary conditions, and multiple stability modes are presented. This is
the 1.09 mm and 1.37 mm (43 mil and 54 mil, respectively) thick spec- followed by an evaluation of strength, which depends in part on the
imens using PT data are shown in Fig. 17. When fastener groups are accuracy of the elastic buckling predictions.
installed at the points of maximum shear slip in members in flexure,
the slip between the webs is greatly reduced, as shown by the reduced 4.1. Elastic buckling
travel in the blue curves in Fig. 17.
Numerical modeling was completed to determine the elastic buck-
4. Comparison with predictions for nominal strength ling loads of the tested specimens. Two approaches were employed:
a finite strip method (FSM) approach using CUFSM and a more involved
The Direct Strength Method (DSM) of AISI S100 [2] provides a basic shell finite element method (FEM) approach employing Abaqus. For the
framework for determining the strength of CFS members, including first approach, finite strip models in the semi-analytical elastic buckling

Fig. 11. Load vs. axial displacement curves for C (left) and D (right) series specimens, with solid and dashed lines representing specimens with and without EFG, respectively.
74 D.C. Fratamico et al. / Journal of Constructional Steel Research 144 (2018) 65–80

(denoted here as FSM) was employed for clamped end models. The
cross section “mesh” consisted of 4 strips per corner, lip, and flange,
and 8 strips per web. Web fasteners were modeled with master–slave
nodal constraints at the location on the webs where screws were
installed in the tests, as shown in Fig. 18.
A second set of models were created in Abaqus [24] with pinned and
fixed end boundary conditions. These models employ discrete spring
models for the web screws. Using an empirical backbone proposed in
Moen et al. [23], shear spring stiffnesses for the steel-steel screws are
calculated and are provided in Table 7. A typical model is shown in
Fig. 19 in which a web screw layout with EFG was used. For pinned
end cases, the load was applied as concentrated forces on each node
of each cross section at either end of the column, and lateral translation
of those nodes was restrained. Axial translation was restrained only
at mid-length, enabling a globally pinned, warping free end condition
which allowed for a more appropriate comparison with CUFSM
pin-ended models. To model the fixed ends, one end of the column
was fully clamped and warping fixed while the load was applied at
the other end. At the loaded end, all cross-section nodes were tied via
multi-point constraints to a reference node, which was restrained in
all degrees of freedom except for axial translation. A unit axial load
was applied to that reference node. The S9R5 element (a nine-noded,
quadratic, isoparametric shell finite element) was employed in Abaqus.
The same cross-sectional mesh density from the FSM models was
also maintained, and a total of 288 elements along the column length
was selected to achieve an element aspect ratio less than 2:1. Poisson's
ratio is 0.3 and Young's modulus was specified as 203.4 GPa (29,500 ksi)
for CFS.
Several rational approaches were developed based on current
state-of-the-art modeling tools and tested observations of composite
action and column end rigidity. For determining local and distortional
buckling capacities, as in Fratamico et al. [16], end boundary conditions
and level of composite action (with or without modeling fasteners as
smeared constraints in CUFSM or discrete springs in Abaqus) are not
significant. Therefore, local and distortional buckling capacities are de-
termined using a pin-ended non-composite (NC) approach in CUFSM
(for a lower bound) and following test observations, fixed ended, par-
Fig. 12. Local–global (flexural) buckling in C1-54 (left), local-distortional-flexural
tially composite (PC) models in CUFSM and Abaqus are also employed
interaction in C2-54 (right) after peak load.
as shown in Table 8. For global buckling, combinations of the following
have been considered: end conditions as pinned, semi-rigid (assuming
software CUFSM [22] were created. Studs were modeled with nominal K = 0.7, with guidance from the fully-composite (FC) average calcu-
dimensions and measured thicknesses, and both pinned and clamped lated in the Southwell approximations), and fixed; levels of composite
end analyses were employed per [22]. A semi-analytical finite strip action as analytical non-composite and fully composite, and partially
analysis (SAFSM), or specifically a signature curve analysis, was used composite (CUFSM and Abaqus); and a final case using the modified
for pin-ended models, and a general boundary condition approach slenderness approach with K = 1.0 as required in current AISI S100
[2] design.
Although summary values are shown here, elastic buckling values for
each assumption for each specimen are provided in Fratamico [19].
For local and distortional buckling, the end boundary conditions
(fixed/pinned) are not influential as the mean Pcrl,2/Pcrl,1 and Pcrd,2/Pcrd,1
are 1.00 and 1.06, respectively. Further, for local and distortional
buckling, modeling discrete springs (as in the shell FE model) or
smeared constraints (as in the FSM model) is not influential as the
mean Pcrl,3/Pcrl,2 and Pcrd,3/Pcrd,2 are 1.00 and 1.02, respectively. Global
buckling, however, has a much greater sensitivity to the parameters
considered in this work. Theoretically, for global (flexural) buckling,
the fixed end condition is 4 times that of the pinned end condition.
However, considering partially composite action and potentially elastic
modal interactions, the shell FE fixed/pinned comparison has a mean
Pcre,3c/Pcre,3a ratio of 2.62, and for FSM fixed/pinned, Pcre,2c/Pcre,2a is
3.26. A comparison can also be made in the level of composite action de-
veloped in FSM and FEM solutions based on the type of fastener models
used. For discrete fastener models in shell FE, the PC/NC ratio with
pinned ends (Pcre,3a/Pcre,1a) is 1.24 and for fixed ends (Pcre,3c/Pcre,1c) is
0.81, which is less than 1 due to the presence of local and distortional
Fig. 13. Example of stud-to-track engagement in specimen C1-54 using PT data. buckling interactions in FE analysis and a general difficulty in choosing
 D.C. Fratamico et al. / Journal of Constructional Steel Research 144 (2018) 65–80 75

Fig. 14. Side view of a semi-rigid condition at peak load (left) and lip/flange lift-off for C2-54 (right).

a suitable global buckling eigenmode. Contrasting with smeared for global buckling, E3 for local–global interactive buckling, and E4 for
constraints employed in FSM, the PC/NC ratio for pinned ends distortional buckling are employed. The elastic buckling values for
(Pcre,2a/Pcre,1a) is 1.49 and for fixed ends (Pcre,2c/Pcre,1c) is 1.23. The cor- local, Pcrl, distortional, Pcrd, and global, Pcre, buckling follow Table 9
responding PC/NC ratio for the modified slenderness approach which and Py is assumed to be AgFy where Ag is the nominal steel cross-
employs pinned ends (Pcre,mod/Pcre,1a) is 1.49. Although almost equiva- sectional area typically used in design, and Fy the mean yield stress,
lent to the PC/NC ratios obtained from the numerical solutions, the provided in Table 3 for all thicknesses.
FSM and FEM solutions may be regarded as more adapt mechanically In Tables 9a and 9b, all results for nominal strength are normal-
for the estimation of buckling capacity. When comparing analytical ized by the tested capacity (Pu,test ) for each stud. At the lower
bounds of composite action, the mean ratio of FC/NC for pinned and bound, if one assumes non-composite and pinned end conditions
fixed ends (Pcre,4a/Pcre,1a and Pcre,4c/Pcre,1c, respectively) is 1.64. (Pn,1a ), the mean tested strength is 2.22 times this prediction.
Composite action exists, and the pinned end condition (commonly
4.2. Strength predictions assumed in practice) is inadequate for estimating strength. On the
other hand, if the built-up section is assumed fully composite and
Seven rational design approaches for built-up CFS column design with fixed ends, the mean ratio of Pu,test/Pn,4 is 0.92, (and individual
were considered as summarized in Tables 9a and 9b. The DSM predic- test-to-predicted ratios are as low as 0.79) which is modestly non-
tion equations for axial strength provided in AISI S100 in Section E2 conservative for design. When EFG are used, the ratio is 0.97;

Fig. 15. Plots of load vs. mid-height lateral deflection for all specimens: plot (a) for 0.84 mm (33 mil), plot (b) for 1.09 mm (43 mil), plot (c) for 1.37 mm (54 mil), and plot (d) for 1.73 mm
(68 mil) thick specimens.
76 D.C. Fratamico et al. / Journal of Constructional Steel Research 144 (2018) 65–80

Fig. 16. Estimations of column end rigidity for all test trials.

without EFG, the method under-predicts capacity to a greater extent with an analytical global buckling capacity with K = 0.7 is also too
with a ratio of 0.87. Implementing the current modified slenderness conservative.
ratio (with K = 1.0) still underestimates the capacity, with a mean The numerical methods (FSM and FEM) tend to deliver more
ratio Pu,test/Pn,mod of 1.61 (the ratio with EFG is 1.70 and without accurate predictions of strength under the assumption of partially
EFG is 1.52). composite action and fixed end conditions. In the “PC Fixed FSM”
In the “PC K = 0.7” approach, a partially composite condition is approach, the mean test-to-predicted ratio Pu,test/Pn,2c is 1.00 (the
achieved in CUFSM and a nearly-fixed end condition of K = 0.7 is ratio with EFG is 1.05 and without EFG is 0.94), though the scatter
employed analytically for global buckling predictions; while challenging is fairly high. In the parallel approach employing shell FEM, the
to combine a partially composite condition with a semi-rigid end ratio Pu,test/Pn,3 is 1.11 (the ratio with EFG is 1.16 and without EFG
condition, both are considered separately and thus, the maximum of is 1.05). A numerical approach employing an efficient model for fas-
the global buckling capacities: Pcre,2a and Pcre,1b, is used. Although this teners and employing fixed end conditions can yield more accurate
method is based on test observations, it is conservative with a mean strength predictions, and using the current release of CUFSM and
test-to-predicted ratio of 1.32 (the ratio with EFG is 1.39 and without its modeling capabilities prove useful in implementing a design
EFG is 1.25). The approach using a non-composite assumption, paired method for built-up CFS columns.

Fig. 17. Load vs. shear slip curves showing the amount of slippage between webs from initial loading to collapse (blue and red lines represent specimens with and without EFG, respectively
for both specimen sizes).
 D.C. Fratamico et al. / Journal of Constructional Steel Research 144 (2018) 65–80 77

5. Discussion

5.1. Tested behavior

Local–global (flexural) buckling controlled in most cases, and

single-mode global buckling was observed only in thicker studs,
i.e. thicknesses greater than 1.1 mm (43 mil). In these thicker sections,
a greater level of composite action can be achieved with the installation
node-to-node of an EFG, as flexural deformations were more pronounced and less
local interaction was observed before peak loads were reached. In thin-
constraints ner sections, local buckling reduced the global capacity and limited the
effectiveness of EFGs to boost flexural rigidity. Boundary condition non-
linearity was present before studs were fully seated with the tracks
during the loading process. A change in stud axial stiffness is seen in
all tests, and error in stud end cut quality can alter response. This behav-
ior is consistent with installed behavior, yet once dead load is present
and the stud is fully seated in the track, it may not be as relevant to
ultimate strength.
Column end rigidity and composite action are shown to compete to
boost column capacity. This coupled effect is observed in the Southwell
Fig. 18. Example of a CUFSM model showing cross section mesh and constraints as estimations, where lateral column displacements are used to back-
web fasteners.
calculate K factors which inherently contain the contribution of the
composite action via the web screws which affect the global (flexural)
buckling deformations. A rational design approach considers not only
Table 7 partially composite action, but also an end condition which is observed
Shear spring stiffnesses used to model the web screws, per steel plate thickness using data
locally at the column ends. In future work, a test setup which can
from Moen et al. [23].
properly measure end rigidities local to column ends (and thus,
Plate thickness (mm) [mil] Shear stiffness (kN/mm) Shear stiffness (kip/in.) a function mostly of the stud ends and stud-to-track seating) should
0.84 mm (33 mil) 2.96 16.91 be considered. However, for the work presented here, a fixed end
1.09 mm (43 mil) 5.98 34.13 condition assumption appears reasonably adequate.
1.37 mm (54 mil) 7.68 43.88
1.73 mm (68 mil) 12.7 72.29
5.2. Design approaches

The approaches described in Section 4.2 of this paper were selected

as rational methods to design built-up CFS columns using tested
behavior and means by which current CFS building designers have ac-
Table 9b also reports resistance factors ϕ and reliability indices β cess, namely numerical analysis programs. The most accurate method
which were calculated following guidelines in AISI S100 (2016) was shown to employ fixed-end conditions (K = 0.5) and partially
Section K2.2.1. Resistance factors were calculated assuming a composite sections using CUFSM, with a mean test-to-predicted ratio
target reliability index of 2.5, and reliability indices were calculated (Pu,test/Pn,2c) of 1.00 (but with a fairly high variance). Modeling fasteners
assuming ϕ = 0.85. Although the mean test-to-predicted ratio discretely in the shell FE approach, also employing a partially composite
in the “PC Fixed FSM” approach is 1, the spread of data is rather condition and using fixed ends, gave an average Pu,test/Pn,3 of 1.11.
large. With a large coefficient of variation: ϕ = 0.77 and β = 2.19 Design approaches using pinned ends were shown to be too conserva-
for all specimens, ϕ = 0.85 and β = 2.49 for EFG specimens, tive, and the upper bound approach using fully-composite with fixed
and ϕ = 0.71 and β = 1.93 for non-EFG specimens. Specimen B2-68 is ends can overestimate capacity.
an outliner with a rather low test-to-predicted ratio of 0.56 when Fig. 20 shows column curves comparing test strengths using
this “PC Fixed FSM” method is used. If results for B2-68 are not the elastic buckling solutions determined for the “PC Fixed FSM”
considered in the reliability analysis, the resistance factor increases and “PC Fixed FEM” design approaches. Slenderness values are thus
to ϕ = 0.76 for non-EFG specimens. In LRFD design, ϕ = 0.85 in calculated using the squash load Py, the buckling values from FSM
AISI S100 (2016). Clearly, the presence of the EFG not only increases (Pcrl,2 and Pcre,2c), and the buckling values from FEM (Pcrl,3, and Pcre,3c)
strength modestly, but increases reliability: resistance factors calcu- when appropriate. Both local–global and isolated global (flexural) col-
lated for EFG specimens are on average 16% higher and reliability in- umn curves are plotted. Clearly, the local–global strength predictions
dices are on average 20% higher across all seven design approaches using either numerical approach to determine elastic buckling loads
when compared with corresponding reliability parameters for the are adequate, although scatter is present. The pure global (flexural)
non-EFG specimens. buckling predictions tend to be non-conservative, as slight local buck-
ling interaction with the global (flexural) buckling in the tests may
have reduced capacity slightly. Also, comparing results for specimens
with EFG (blue) and without EFG (red), local–global capacities tend
to be greater when the EFG are installed.

5.3. Future work

An attempt in this work to extract information on end rigidity was

carried out using the Southwell method, and it provides useful informa-
Fig. 19. Example of elastic buckling in an Abaqus model of column C1-54, as used in tion of fixity estimates for each type of column tested: single sections
Fratamico et al. [16]. and built-up sections with varying fastener levels. However, the
78 D.C. Fratamico et al. / Journal of Constructional Steel Research 144 (2018) 65–80

Fig. 20. DSM column design curves comparing tested strengths for columns which failed in single-mode global (flexural) buckling (left) and local–global (flexural) interactive
buckling (right).

end rigidities in the screw-connected sections are coupled with the currently used in finite element analyses that consider true geometry
composite action developed through the web screws. Using both and a material model (based on coupon tests) to create models with fi-
a non-composite and fully composite assumption for the built-up delity that validate well with test results. The motivation is to create and
sections when back-calculating the effective length factors, K varies by validate a finite element modeling protocol to continue to analyze more
approximately 0.2 for all section types, as shown in Fig. 16. Further complex built-up CFS section types without requiring more tests, study
experimental tests can be performed in which myriad cross sections the effect of increasing yield stress on collapse modes [25], and study
can be tested in two different setups which allow for isolated studies the behavior of perforated built-up sections in compression or bending.
of end conditions and composite action.
Fastener layouts, particularly the addition of EFGs to built-up CFS 6. Conclusions
columns, should be addressed directly in future work. At this point,
a numerical approach using shell FE models can accommodate the In this work, bare, unbraced, and unperforated back-to-back lipped
addition of EFG for elastic buckling and collapse modeling studies channel sections have been tested using realistic end conditions by
to investigate optimal EFG lengths for different section types. Clearly, installing each stud in track to study the buckling behavior and strength
the EFG have an effect of boosting unbraced column capacity by up to of multiple cross section sizes and varying web fastener layouts. Based
33% as reported in Table 6 and their presence increases member on the test results, insight is given on the new application of DSM for-
reliability as shown in Table 9b when compared to results for non- mulas for built-up CFS column design. Comparisons of tested capacities
EFG specimens. However, they are expensive details that required are made with DSM strength predictions using existing strength equa-
up to 13 rows of screw pairs on either end of the column in some tions provided in AISI S100-16, and the following general conclusions
test cases. A rational, engineering approach to the design of EFG are made.
lengths should be developed such that their effectiveness is balanced Composite action is developed through the web screws when more
with their ease of installation. isolated global (flexural) buckling occurs. Installing EFGs can boost
Geometric imperfections are not considered in the treatment of test capacity by up to 33% but can also increase member reliability indices
data, but laser scans of all tested sections were completed and are when compared to columns without EFG installed. However, when

Table 8
Definition of methods to determine elastic buckling capacities.

Local buckling Distortional buckling

End conditiona Pin Fix Fix Pin Fix Fix

Composite actionb NC PC PC NC PC PC
Methodc SAFSM FSM FEM SAFSM FSM FEM
Buckling ID Pcrl,1 Pcrl,2 Pcrl,3 Pcrd,1 Pcrd,2 Pcrd,3

Global buckling

End conditiona Pin Pin Pin Pin Pin Semi Semi Semi Fix Fix Fix Fix
Composite actionb NC PC PC PC FC NC NC FC NC PC PC FC
Methodc,d Ana (KL/r)m FSM FEM Ana Ana Test Ana Ana FSM FEM Ana
Buckling ID Pcre,1a Pcre,mod Pcre,2a Pcre,3a Pcre,4a Pcre,1b Pcre,test Pcre,4b Pcre,1c Pcre,2c Pcre,3c Pcre,4c
a
End conditions: pin = global & local plate pinned and warping free, semi = semi-rigid effective length factor K = 0.7, fix = globally & locally fixed and no warping.
b
Composite action levels: NC = non-composite, PC = partially composite, FC = fully composite, with sheathing added as applicable.
c
Method: SAFSM = semi-analytical finite strip (signature curve) analysis, FSM = general end boundary condition finite strip analysis, FEM = shell finite element analysis.
d
Method: Ana = analytical treatment of weak axis moment of inertia for non-composite (INC = 2I) and fully composite action (using parallel axis theorem), (KL/r)m = AISI modified
slenderness ratio approach [2].
 D.C. Fratamico et al. / Journal of Constructional Steel Research 144 (2018) 65–80 79

Table 9
Summary of test-to-predicted strength ratios for each test specimen using seven DSM-based design approaches.

Specimen Section Approach Identifiers Bare NC Bare NC K = 0.7 AISI Now (KL/r)m PC K = 0.7 PC Fixed FSM PC Fixed FEM Ideal Built-Up

End Condition Pin Semi-Rigid Pin Pin Fix Fix Fix

Composite Action NC NC PC PC PC PC FC

Local Buckling Pcrl,1 Pcrl,1 Pcrl,2 Pcrl,2 Pcrl,2 Pcrl,3 Pcrl,2

Distortional Pcrd,1 Pcrd,1 Pcrd,2 Pcrd,2 Pcrd,2 Pcrd,3 Pcrd,2

Buckling

Global Buckling Pcre,1a Pcre,1b Pcre,mod max(Pcre,2a, Pcre,1b) Pcre,2c Pcre,3c Pcre,4c

EFG Py (kN) [kips] Pu,test/Pn,1a Pu,test/Pn,1b Pu,test/Pn,mod Pu,test/Pn,2a Pu,test/Pn,2c Pu,test/Pn,3 Pu,test/Pn,4

(a)
A1-33 250S137-33 Yes 81.4 [18.3] 2.01 1.24 1.36 1.24 0.91 1.05 0.91
A2-33 250S137-33 81.4 [18.3] 1.82 1.12 1.23 1.12 0.82 0.95 0.82
A1-43 250S137-43 Yes 110 [24.6] 2.29 1.20 1.38 1.20 0.85 0.95 0.83
A2-43 250S137-43 110 [24.6] 2.16 1.13 1.30 1.13 0.80 0.90 0.79
A1-54 250S137-54 Yes 161 [36.2] 3.23 1.61 1.89 1.61 0.92 1.07 0.90
A2-54 250S137-54 161 [36.2] 2.68 1.34 1.57 1.34 0.76 0.89 0.75
A1-68 250S137-68 Yes 177 [39.9] 2.84 1.44 1.65 1.44 0.84 0.98 0.82
A2-68 250S137-68 177 [39.9] 2.50 1.26 1.45 1.26 0.74 0.86 0.72
B1-33 362S137-33 Yes 97.5 [21.9] 1.99 1.29 1.51 1.28 0.95 1.13 0.95
B2-33 362S137-33 97.5 [21.9] 2.00 1.30 1.52 1.29 0.96 1.14 0.95
B1-43 362S137-43 Yes 131 [29.5] 2.14 1.33 1.57 1.33 1.02 1.12 0.96
B2-43 362S137-43 131 [29.5] 1.84 1.15 1.36 1.15 0.88 0.97 0.83
B1-54 362S137-54 Yes 193 [43.4] 2.58 1.33 1.69 1.32 0.91 1.04 0.88
B2-54 362S137-54 193 [43.4] 2.25 1.16 1.48 1.15 0.80 0.91 0.77
B1-68 362S137-68 Yes 214 [48.1] 2.43 1.22 1.58 1.22 0.80 0.85 0.77
B2-68 362S137-68 214 [48.1] 1.72 0.86 1.11 0.86 0.56 0.60 0.54

(b)
C1-33 600S137-33 Yes 131 [29.5] 2.58 1.66 2.16 1.66 1.30 1.41 1.20
C2-33 600S137-33 131 [29.5] 2.68 1.72 2.24 1.72 1.35 1.46 1.25
C1-43 600S137-43 Yes 177 [39.9] 2.64 1.67 2.19 1.67 1.22 1.48 1.18
C2-43 600S137-43 177 [39.9] 2.38 1.50 1.97 1.50 1.10 1.33 1.06
C1-54 600S137-54 Yes 262 [58.8] 2.45 1.52 2.01 1.52 1.32 1.39 0.99
C2-54 600S137-54 262 [58.8] 2.15 1.34 1.77 1.34 1.16 1.22 0.87
C1-68 600S137-68 Yes 291 [65.4] 2.75 1.59 2.09 1.58 1.20 1.32 1.05
C2-68 600S137-68 291 [65.4] 2.06 1.19 1.57 1.19 0.90 0.99 0.79
D1-33 600S162-33 Yes 142 [32.0] 1.76 1.26 1.46 1.26 1.16 1.23 1.02
D2-33 600S162-33 142 [32.0] 1.73 1.24 1.44 1.24 1.15 1.21 1.01
D1-43 600S162-43 Yes 192 [43.2] 1.66 1.16 1.36 1.16 0.98 1.07 0.93
D2-43 600S162-43 192 [43.2] 1.74 1.22 1.43 1.22 1.03 1.13 0.97
D1-54 600S162-54 Yes 283 [63.7] 2.21 1.46 1.77 1.46 1.15 1.28 1.11
D2-54 600S162-54 283 [63.7] 1.81 1.20 1.45 1.19 0.94 1.05 0.91
D1-68 600S162-68 Yes 315 [70.9] 1.97 1.32 1.58 1.32 1.22 1.23 1.02
D2-68 600S162-68 315 [70.9] 1.86 1.24 1.48 1.24 1.15 1.16 0.96

Specimens with EFG Mean 2.35 1.39 1.70 1.39 1.05 1.16 0.97

Resistance factor ϕ (β = 2.5) 1.86 1.19 1.38 1.19 0.85 0.96 0.83
Reliability index β (ϕ = 0.85) 5.13 3.76 4.16 3.75 2.49 2.92 2.39

Specimens without EFG Mean 2.09 1.25 1.52 1.25 0.94 1.05 0.87
Resistance factor ϕ (β = 2.5) 1.71 1.04 1.21 1.03 0.71 0.81 0.69
Reliability index β (ϕ = 0.85) 4.94 3.20 3.67 3.20 1.93 2.33 1.79

All specimens Mean 2.22 1.32 1.61 1.32 1.00 1.11 0.92
CoV (%) 17.84 14.42 18.03 14.44 19.40 18.11 16.30
Resistance factor ϕ (β = 2.5) 1.76 1.10 1.28 1.10 0.77 0.88 0.75
Reliability index β (ϕ = 0.85) 4.95 3.43 3.87 3.42 2.19 2.60 2.06

local buckling interacts with global buckling, buckling capacity and accurate strength predictions, yielding an average test-to-predicted
flexural deformations are slightly reduced, and the efficacy of the ratio of 1.0, but with a high variance worthy of further study.
EFG is also diminished. Recorded PT data is provided to support
this claim. Acknowledgements
Column end rigidities compete with the composite action developed
through the web screws to boost column capacity when flexural buck- Special thanks to Johns Hopkins University, Department of Civil
ling controls. Current design methods do not consider this interaction, Engineering laboratory instrument designer Nick Logvinovsky, master's
and following the recommendations herein propose a suitable approach students Isaiah Sampson and Xiaomeng Li, and undergraduate students
to design which should consider a fixed end boundary condition. Matthew Brandes and Avi Gordon for their time and assistance in pre-
Further, modeling fasteners using a smeared constraint approach in paring the test setup. Research for this paper was conducted with partial
CUFSM to consider a partially composite condition is shown to deliver U.S. Government support under FA9550-11-C-0028 and awarded by the
80 D.C. Fratamico et al. / Journal of Constructional Steel Research 144 (2018) 65–80

Department of Defense, Air Force Office of Scientific Research, National [11] J. Zhang, B. Young, Compression tests of cold-formed steel I-shaped open sections
with edge and web stiffeners, Thin-Walled Struct. 52 (2012) 1–11.
Defense Science and Engineering Graduate (NDSEG) Fellowship, 32 CFR [12] H.D. Craveiro, J. Paulo, C. Rodrigues, L. Laím, Buckling resistance of axially loaded
168a. Lastly, thanks to ClarkDietrich Building Systems and Simpson cold-formed steel columns, Thin-Walled Struct. 106 (2016) 358–375.
Strong-Tie for graciously providing all CFS sections and screw fasteners, [13] B. Young, J. Chen, Design of cold-formed steel built-up closed sections with interme-
diate stiffeners, J. Struct. Eng. 134 (2008) 727–737.
respectively. Any opinions, findings, and conclusions or recommenda- [14] Y. Lu, T. Zhou, W. Li, H. Wu, Experimental investigation and a novel direct strength
tions expressed in this material are those of the author(s) and do not method for cold-formed built-up I-section columns, Thin-Walled Struct. 112 (2017)
necessarily reflect the views of the sponsors or other participants. 125–139.
[15] Y. Li, Y. Li, S. Wang, Z. Shen, Ultimate load-carrying capacity of cold-formed
thin-walled columns with built-up box and I section under axial compression,
References Thin-Walled Struct. 79 (2014) 202–217.
[16] Fratamico, D.C., Torabian, S., Rasmussen, K.J.R., Schafer, B.W. “Experimental study on
[1] AS/NZS 4600, Cold-formed Steel Structures, Australian/New Zealand Standard, the composite action in sheathed and bare built-up cold-formed steel columns,”
Sydney, Australia, 2016. Thin-Walled Struct. (under review)
[2] AISI S100, North American Specification for the Design of Cold-formed [17] X. Zhao, M. Tootkaboni, B.W. Schafer, Laser-based cross-section measurement of
Steel Structural Members, American Iron and Steel Institute, Washington, cold-formed steel members: model reconstruction and application, Thin-Walled
D.C., 2016 Struct. 120 (2017) 70–80.
[3] AISC 360, Specification for Structural Steel Buildings, American Institute of Steel [18] X. Zhao, M. Tootkaboni, B.W. Schafer, High fidelity imperfection measurements and
Construction, Chicago, IL, 2010. characterization for cold-formed steel members, 7th International Conference on
[4] AISI S240, North American Standard for Cold-Formed Steel Structural Framing, Coupled Instabilities in Metal Structures, Baltimore, MD, 2016.
American Iron and Steel Institute, Washington, D.C., 2015 [19] D.C. Fratamico, Experiments, Analysis, and Design of Built-Up Cold-Formed Steel
[5] EN 1993-1-3, Eurocode 3 - Design of Steel Structures - Part 1–3: General Ruls - Columns(Ph.D. Thesis) The Johns Hopkins University, 2017.
Supplementary Rules for Cold-formed Members and Sheeting, European [20] ASTM A370-15, Standard Test Methods and Definitions for Mechanical Testing of
Committee for Standarization, Brussels, Belgium, 2006. Steel Products, ASTM International, West Conshohocken, PA, 2015.
[6] T. Stone, R. LaBoube, Behavior of cold-formed steel built-up I-sections, Thin-Walled [21] R.V. Southwell, On the analysis of experimental observations in problems of elastic
Struct. 43 (12) (2005) 1805–1817. stability, R. Soc. London 135 (828) (1932) 601–616.
[7] I. Georgieva, L. Schueremans, L. Pyl, Composed columns from cold-formed steel [22] Z. Li, B.W. Schafer, Buckling analysis of cold-formed steel members with general bound-
Z-profiles: experiments and code-based predictions of the overall compression ary conditions using CUFSM: Conventional and constrained finite strip methods, 20th
capacity, Eng. Struct. 37 (2012) 125–134. International Specialty Conference on Cold-formed Steel Structures, St. Louis, MO, 2010.
[8] I. Georgieva, L. Schueremans, G. De Roeck, L. Pyl, Experimental investigation of the [23] C.D. Moen, F. Tao, R. Cole, Monotonic and cyclic backbone response of single
deformation of built-up members of cold-formed steel profiles, Appl. Mech. Mater. shear cold-formed steel screw-fastened connections, International Colloquium on
70 (2011) 416–421. Stability and Ductility of Steel Structures, Timisoara, Romania, 2016.
[9] I. Georgieva, L. Schueremans, L. Pyl, L. Vandewalle, Numerical study of [24] Abaqus 6.14-5, Simulia, Dassault Systèmes, Providence, RI, 2014.
built-up double-Z members in bending and compression, Thin-Walled Struct. [25] H. Foroughi, B.W. Schafer, Simulation of conventional cold-formed steel sections
60 (2012) 85–97. formed from advanced high strength steel (AHSS), Annual Stability Conference,
[10] I. Georgieva, L. Schueremans, L. Vandewalle, L. Pyl, Design of built-up cold-formed Structural Stability Research Council, San Antonio, TX, 2017.
steel columns according to the direct strength method, Procedia Eng. 40 (2012)
119–124.