You are on page 1of 47

This may be the author’s version of a work that was submitted/accepted

for publication in the following source:

Kesawan, Sivakumar, Mahendran, Mahen, & Sivaprakasam, Thananjayan


(2020)
Numerical study of the section moment capacity of complex-shaped alu-
minum mullions.
Journal of Architectural Engineering - ASCE, 26(4), Article number:
04020033.

This file was downloaded from: https://eprints.qut.edu.au/205258/

c ASCE

This work is covered by copyright. Unless the document is being made available under a
Creative Commons Licence, you must assume that re-use is limited to personal use and
that permission from the copyright owner must be obtained for all other uses. If the docu-
ment is available under a Creative Commons License (or other specified license) then refer
to the Licence for details of permitted re-use. It is a condition of access that users recog-
nise and abide by the legal requirements associated with these rights. If you believe that
this work infringes copyright please provide details by email to qut.copyright@qut.edu.au

License: Creative Commons: Attribution-Noncommercial 4.0

Notice: Please note that this document may not be the Version of Record
(i.e. published version) of the work. Author manuscript versions (as Sub-
mitted for peer review or as Accepted for publication after peer review) can
be identified by an absence of publisher branding and/or typeset appear-
ance. If there is any doubt, please refer to the published source.

https://doi.org/10.1061/(ASCE)AE.1943-5568.0000429
1 Numerical Study of the Section Moment Capacity of Complex-Shaped
2 Aluminium Mullions
3
4 Sivakumar Kesawan1, Mahen Mahendran2 and Thananjayan Sivaprakasam3
5
6 Queensland University of Technology (QUT), Brisbane, Australia
7
8 Abstract: Curtain wall systems made of aluminium frame and infilled with glass panes
9 are extensively used in present day high-rise buildings. The vertical members of the
10 aluminium frame, known as mullions, are complex-shaped and are primarily subjected to
11 bending action when the curtain wall panels are exposed to wind pressure or suction loading.
12 This paper investigates the section moment capacity of 6063-T6 aluminium alloy mullion
13 sections subject to wind actions using a numerical study. An experimental study conducted
14 to determine the section moment capacity of the mullions is summarized first in this paper.
15 The results from an experimental study on the mechanical properties of 6063-T6 aluminium
16 alloy mullions are then presented including a suitable stress-strain model. Finite element
17 models were developed to determine the section moment capacities of tested mullion
18 sections, and the details of the finite element modelling procedure and the results are
19 presented in this paper. The developed models included mullions used in the structural and
20 captive glazing systems subject to both wind pressure and suction loading, and were
21 validated using the available test results. Finite element analysis results were compared with
22 the predictions from three design methods, the limiting stress method, direct strength
23 method and total moment capacity approach, based on which the most suitable design
24 method is recommended.
25
26 Keywords: Mullions, Curtain walls, Aluminium, Bending, Section moment capacity,
27 Limiting stress method, Direct strength method, Total moment capacity approach
28
29
30
1
31 Research Fellow, Queensland University of Technology (QUT), Brisbane, Australia
2
32 Professor, Queensland University of Technology (QUT), Brisbane, Australia
3
33 PhD student, Queensland University of Technology (QUT), Brisbane, Australia
34
35 Corresponding author’s email address: m.mahendran@qut.edu.au

1
36 1. Introduction

37 Unitised curtain wall systems (Fig. 1) made of aluminium frame skeleton with glass panel
38 infills are commonly used in mid- to high-rise commercial buildings at present. Their facade
39 panels are factory assembled and then brought to site where they are interconnected to form
40 the building envelope. A good workmanship in a factory controlled environment, and
41 quicker construction time at site have contributed to their popularity (Russell, 2006, Allana
42 and Carter, 1999). To enable easier connectivity between the adjacent curtain wall panels,
43 the vertical members of the aluminium frame known as mullions are made of a pair of
44 extruded male and female mullions (Fig. 2). Façade engineers prefer not to screw fasten the
45 male and female mullions together in order to allow thermal and building movements, easily
46 remove the panel from the façade assembly, if needed, and to give aesthetically pleasing
47 appearance.
48
49 The male and female mullions have a complex cross-section geometry (Fig. 2). They are
50 made of such shapes in order to allow thermal movements between the adjacent façade
51 panels, include screw flutes and support for fixing the sunshade devices, and enable easier
52 connectivity between the two mullions. In facade systems, single or double glazing panels
53 are used, and mullions are made of different cross-sections to suit them (Fig. 3). Furthermore,
54 architects and energy consultants do not prefer aluminium protruding through the facade
55 panels due to aesthetic and energy efficiency concerns. In these instances structural glazing
56 system is used instead of the commonly used captive glazing system (Fig. 3). Extruadabilty
57 characteristics of aluminium facilitates the manufacturing of such vast array of complex
58 cross-sections. High strength to weight ratio, high durability, energy efficiency and
59 sustainability of aluminium have added value to its usage in the supporting frames of unitised
60 curtain walls. These aluminium mullions must transfer the wind loading acting on the glazed
61 panels to the main structural system while satisfying the ultimate and serviceability criteria.
62 The mullions do not carry any other axial load other than their self-weight, which is
63 negligible. During the wind actions, the aluminium mullions are subject to bending action,
64 and thus they must have adequate bending moment capacity.

65 Since the male and female mullions are not screwed or clipped together, the bending moment
66 capacity of the mullion couple is taken as the sum of the individual bending moment
67 capacities of the male and female mullions. The cost of designing, manufacturing and
68 constructing the unitised curtain wall system is normally more than 15% of the total project

2
69 cost of high-rise buildings. Thus there is a need for accurate design methods to determine
70 the structural capacity of aluminium mullion sections subject to wind actions.
71
72 In the past, some research studies (Lai and Nethercot, 1992, Kim and Pekoz, 2003,2010,
73 Wang, 2006, Zhu and Young, 2009, Su et al, 2014,2016, Kesawan et al, 2018) were
74 conducted on the bending moment capacity of mullion sections. Except (Wang, 2006,
75 Kesawan et al, 2018) others studies were limited to the section moment capacity of
76 aluminium box and I-sections. Wang (2006) investigated the structural behaviour of E type
77 mullions, but the number of tests performed in his study is inadequate to make reliable
78 conclusions. Kesawan et al, (2018) conducted an experimental study to develop the section
79 moment capacity data of complex-shaped mullions. About 30 tests were performed where
80 the mullions were tested under both wind pressure (positive loading) and suction (negative
81 loading) (Fig. 4). This experimental study is summarised in this paper. It is also important
82 to know the accurate mechanical properties of aluminium mullion sections. For this purpose,
83 more than 30 tensile tests were performed on different 6063-T6 aluminium alloy sections,
84 and their results are presented in this paper. The measured stress-strain curves were
85 compared with the models given in Eurocode 9 Part 1.1 (ECS, 2007), and based on this
86 comparison a suitable stress-strain model is recommended in this paper.
87
88 Experimental study requires considerable time and resources, and developing a validated
89 finite element model is very useful to further investigate the section moment capacity of
90 mullions. In the past, Kim and Pekoz (2003,2010), Zhu and Young (2009) and Su et al.
91 (2014,2016) have developed finite element models to determine the section moment capacity
92 of aluminium sections. Their studies were limited to I-sections and box sections with and
93 without internal stiffeners. They validated their models using the failure loads and modes of
94 aluminium sections obtained from experimental studies. Wang (2006) performed numerical
95 investigations of E type mullion sections, but his study was limited to single mullion couple.
96 These demonstrate the need for the development of validated finite element models of
97 complex-shaped mullion sections. This paper addresses this shortcoming and presents the
98 developed finite element models capable of predicting the section moment capacity of
99 complex-shaped mullion sections used in unitised curtain wall systems. The developed
100 models were validated using the test results of 15 different mullions subject to negative and
101 positive wind loading. This paper presents the finite element modelling procedure,
102 mechanical properties used in the models, and finite element analysis (FEA) results. Two

3
103 types of model were developed. In the first one, back to back connected mullions were
104 modelled – full model, and in the latter only one mullion was modelled – idealised model.
105 The FEA results were also compared with the design rule predictions, where three different
106 design methods were considered, limiting stress method (LSM), total moment capacity
107 approach (TMCA) and direct strength method (DSM), and this paper recommends the most
108 suitable design method. A finite element analysis based parametric study was conducted to
109 investigate the section moment capacities of mullions subject to local buckling, but with
110 varying cross-sectional slenderness, and the FEA results are then compared with the DSM
111 design predictions.
112
113 2. Experimental Study

114 An experimental study was performed to investigate the section moment capacities of
115 complex shaped mullion sections. Fifteen mullions used in the structural and captive glazing
116 systems, and single and double glazing units were considered. These aluminium mullion
117 sections were made of 6063-T6 aluminium alloy, and supplied by G. James Glass and
118 Aluminium, Australia. The tensile tests were conducted to obtain the actual mechanical
119 properties of these mullions, and their details are presented in Section 3. In buildings, glazing
120 systems can be subject to either positive or negative wind loading. Due to the asymmetry of
121 the mullions, their section moment capacities are different under these two loading cases.
122 Therefore, the moment capacities of the mullions were determined under both positive and
123 negative load cases using a series of four-point bending tests. The schematic diagram of the
124 test set-up is shown in Fig. 5. The mid-span region between the loading points of the mullion
125 specimen was subject to uniform bending with no shear forces. This forced the failure to
126 occur in this region during testing.
127
128 Two male or female mullions were connected back to back with T-shaped stiffeners at the
129 loading and support locations to simulate symmetry about the line of loading as shown in
130 Figs. 4 and 6. If a single section was tested by applying the load on its flange or web, torsional
131 stresses would have been induced in the mullions due to the location of the load away from
132 the shear centre. The provision of T-shaped stiffener avoided the occurrence of web crippling
133 failure at the support and loading points. Top and bottom flanges of the back to back mullions
134 were connected by angle brackets as shown in Fig. 6, which along with the T-shaped
135 stiffeners prevented the lateral movement of the mullions. In real buildings, the lateral

4
136 movement of the mullion is prevented by the restraints provided by the glass panels. Special
137 clamping brackets were fabricated and used as stiffeners at locations where the M16 high
138 tensile bolts were used to connect the T-shaped stiffeners and the mullions (Fig. 6). This is
139 to avoid the occurrence of local bearing failure at those locations.
140
141 A 300 kN Instron testing machine was used to apply the load, where the load was transferred
142 to the mullions through the half rounds connected to the spreader beam, which was attached
143 to the top jaw of the Instron machine. The load was applied at a constant displacement rate
144 of 1 mm/min until the mullion specimen failed. The failure point was taken as when the
145 mullion was not able to carry any further load, and the applied load began to drop. The
146 section moment capacities obtained from the tests are summarized in Table 1.

147 3. Tensile Tests of 6063-T6 aluminium mullion sections

148 Tensile coupons were cut from a series of mullion sections given in Table 2. At least two
149 tensile coupon tests were performed for each mullion section. The test coupon dimensions
150 were based on AS/NZS 1391 (SA, 2007), and are shown in Fig. 7(a). The coatings of the
151 aluminium specimens were removed prior to the testing, and the actual dimensions were
152 measured. The tests were performed using a 50 kN Instron machine available at the
153 Queensland University of Technology, and the test set-up is shown in Fig. 7(b). A clip-on
154 extensometer was used to measure the strain. In total, 30 tests were performed with the load
155 application based on displacement control method at a rate of 1 mm/min (AS/NZS 1391 (SA,
156 2007)). Fig. 8 shows the measured stress versus strain curves of the 650-027 and 650-028
157 mullion specimens. The curves show a non-linear stress-strain behaviour with considerable
158 strain hardening beyond the yield point. The same type of curves was obtained in the other
159 tests as well. Table 2 presents the measured values of average 0.2% proof (yield) strength,
160 ultimate strength, ultimate strain, fracture strain, elastic modulus and 0.1% proof strength.
161 The yield and ultimate strength values are higher than the values specified in AS/NZS 1664
162 (SA, 1997) for design, which are 172 and 207 MPa, respectively. It is important to use the
163 correct mechanical properties and material models in the finite element models to obtain
164 accurate results. An investigation was conducted to determine the suitable model to represent
165 the strain-strain behaviour of the tested 6063-T6 aluminium alloy, and is presented next.

166

5
167 Since aluminium exhibits a non-linear stress-strain behaviour with considerable strain
168 hardening beyond the yield point, the most widely used isotropic material model (Eq. 1),
169 which is initially proposed by Ramberg and Osgood (1943), and then modified by Hill
170 (1944) would be the most suitable one. The strain hardening exponent (n) in the Ramberg-
171 Osgood equation (1943) has a significant influence on the characteristic behaviour of the
172 stress-strain curve. The evaluation of exponent (n) requires a reference stress (fx) and a
173 conventional limit of elasticity (fe) (Eq. 2). De Martino et al. (1990) conducted an
174 investigation on the stress-strain behaviour of four different aluminium alloys where they
175 considered six different cases to evaluate the strain hardening exponent (n) (De Martino, et
176 al, 1990).
𝜎 𝜎 𝑛
𝜀 = + 0.002 ( ) --------------(1)
𝐸 𝑓0.2
𝜀𝑜,𝑥 𝑓𝑥
𝑛 = log ( ) / log ( ) --------------(2)
𝜀𝑜,𝑒 𝑓𝑒

177 Based on the error percentages evaluated by comparing the experimental and the simulated
178 curves using the different cases, they suggested that the choice of conventional limit of
179 elasticity does not have a significant influence on the approximation of the curve (De
180 Martino, et al, 1990). Also, the second reference parameter (fx) can be chosen depending on
181 the criteria that needs to be considered, i.e., if the analysis involves small deformations such
182 as buckling, f0.1 can be chosen as the second reference stress, and if the analysis involves
183 large deformations, the use of fmax will give a good approximation. Eurocode 9 Part 1.1 (ECS,
184 2007) provides various analytical models, such as piecewise model which can be either bi-
185 linear or tri-linear, and continuous model for aluminium alloys. Gardner and Ashraf (2006)
186 investigated the aluminium hollow sections under compression and bending actions, and
187 found that the use of bi-linear material model for aluminium leads to inaccurate and
188 conservative predictions. Eurocode continuous model is the same as Ramberg and Osgood’s
189 (1943) model (Eq. 1), and is summarized next.
190
191 In the continuous model given in Eurocode 9 Part 1.1 (ECS, 2007), the determination of the
192 strain hardening exponent (n) is suggested in different ways depending on the scope of the
193 analysis to be performed, that is, whether the analysis involves elastic or plastic deformations.
194 The general form of equation for the determination of n is given by Eq. 2. Eq. 2 can be
195 expressed as Eq. 3 when 0.2% proof strength (f0.2) and the corresponding residual strain

6
196 (0.002) are taken as the conventional limit of elasticity (fe) and its corresponding residual
197 strain (εo,e), respectively.

ln(0.002/𝜀𝑜,𝑥 ) --------------(3)
𝑛=
ln(𝑓0.2 ⁄𝑓𝑥 )

198 When the analysis is about elastic deformations, Eq. 4 can be used to determine n by taking
199 f0.1 as the second reference stress (fx). When the analysis is about plastic deformations, Eq. 5
200 can be used to determine n by taking fmax as the second reference stress (fx).
ln(2)
𝑛= --------------(4)
ln(𝑓0.2 ⁄𝑓0.1 )
ln(0.002/𝜀𝑜,𝑚𝑎𝑥 )
𝑛= --------------(5)
ln(𝑓0.2 ⁄𝑓𝑚𝑎𝑥 )
201 In the absence of actual data, u can be determined using Eq. 6, which is a function of 0.2%
202 proof strength (f0.2), and Eq. 7 (Eurocode 9 Part 1.1 (ECS, 2007)).

𝑓0.2 (𝑁/𝑚𝑚2 ) ---------------(6)


𝜀𝑢 = 0.30 − 0.22 𝑓0.2 < 400𝑁/𝑚𝑚2
400

---------------(7)
𝜀𝑢 = 0.08 𝑓0.2 ≥ 400𝑁/𝑚𝑚2
203 The equation for the calculation of ultimate strain (max) corresponding to ultimate tensile
204 strength (fmax) proposed by Su et al. (2014) is given in Eq. 8 , where fmax/f0.2 >1.01.

𝜀𝑚𝑎𝑥 = 0.13(1 − 𝑓0.2 ⁄𝑓𝑚𝑎𝑥 ) + 0.06 ---------------(8)

205 Ultimate fracture strains obtained from tensile tests (εu,test) were compared with the values
206 (εu,pred) obtained using the predictive models given in Eurocode 9 Part 1.1 (ECS, 2007) (Eq.
207 6). It was found that the ultimate strain values predicted by Eurocode 9 Part 1.1 (ECS, 2007)
208 are considerably higher than the experimental results with significant scatter. Therefore, a
209 new predictive model to determine the ultimate fracture strain (εu) value was proposed, and
210 is given in Eq. 9. The coefficients of the proposed model were obtained using the tensile test
211 results by means of least-squares regression method. The proposed model gives more
212 accurate results than those found from Eq. 6 (Fig. 9) for 6063-T6 aluminium alloy. However,
213 the applicability of the proposed model for the ultimate fracture strain prediction of other
214 types of aluminium alloy should be verified in the future. Furthermore, the model (Eq. 8)
215 proposed by Su et al. (2014) to determine the strain corresponding to ultimate strength (εmax)

7
216 was compared with the values obtained from tests. It gave better predictions for the ultimate
217 strain values as shown in Fig. 10.

218 𝜀𝑢 = 0.052 + 0.11(𝑓0.2 ⁄400) ----------------(9)


219
220 The tested mullions are used in the aluminium frame of unitised curtain wall systems. As
221 discussed, these mullions are primarily subject to bending actions. Therefore, their flexural
222 capacity is important for design. They can fail by either section yielding with or without
223 local and distortional buckling, or flexural torsional buckling. Therefore, both the elastic and
224 plastic regions of the stress-strain curve are important. At first, Eq. 4, which was given to
225 determine the value of n for the analysis involving elastic deformation was evaluated, where
226 the predicted models using this n value, were compared with the test results. It is evident
227 from Fig. 11 that these predictive models did not agree with the test results, especially in the
228 strain hardening region. Thereafter Eq. 5, which was given to determine the n value for the
229 analysis involving plastic deformations, was considered. For all the mullion sections, the
230 predicted stress-strain models using this n value matched well with the experimental results
231 in both elastic and plastic regions. Fig. 12 compares the stress-strain curves of 650-027 and
232 650-028 sections. Therefore, Ramberg and Osgood’s (1943) model with the n value
233 calculated based on Eq. 5 was used in the finite element models developed in this study.
234

235 4. Development of finite element models to predict the section moment capacity of
236 mullion sections

237 This section presents the finite element models developed to predict the section moment
238 capacities of tested mullion sections (refer Section 2). ABAQUS/CAE Version 6.14-2 was
239 used in the modelling. The tested mullions have a complex geometry, and are made of thin-
240 walled elements with their thickness varying across the cross-section. Shell elements are the
241 most suitable elements to model thin-walled members. Therefore, the mullion sections were
242 idealized into different elements of various thicknesses (Fig. 13). The section properties of
243 the idealized sections were compared with those of the original sections, and the deviations
244 between them were small. The cross-sections of the original and idealized 650-027 mullion
245 sections are shown in Fig. 13. More details can be found in (Kesawan and Mahendran, 2018).
246 These idealized models were also used in the finite strip analysis using CUFSM to determine
247 the critical buckling mode and the corresponding buckling load of mullions subject to a
248 uniform bending moment (Kesawan and Mahendran, 2018). The measured mechanical

8
249 properties and the thickness of the flat elements, along with the nominal corner dimensions
250 provided by the manufacturers were used in the modelling.
251
252 Due to the symmetry of the test set-up, a half model was developed (Figs. 5 and 14). Two
253 different types of half model were developed. In the first one, Model Type A, the mullion
254 pair was modelled (Mullions 1 and 2 in Fig. 4) along with the T-shaped stiffener, while in
255 the latter model only one of the mullion sections (Mullion 1 or 2 in Fig. 4) was modelled.
256 Fig. 15(a) shows the Model Type A. Rigid R3D4 elements were used for the stiffener while
257 S4R shell elements were used for the aluminium mullion section. S4R is a four node
258 quadrilateral general purpose conventional shell element with reduced integration to avoid
259 shear and membrane locking. It accounts for finite member strains and arbitrary large
260 rotations. The mesh size used was 4 mm x 4 mm. The same element types were used by
261 others (Kesawan and Mahendran, 2015, Anapayan et al, 2011, Siahaan et al, 2016) in the
262 finite element models developed to predict the section moment capacity of thin-walled cold-
263 formed steel beams. Previously shell element S4R had also been used by others (Kim and
264 Pekoz, 2003,2010, Wang, 2006, Zhu and Young, 2009, Su et al, 2014,2016) to model the
265 bending behaviour of aluminium members. The mullion models were partitioned and the
266 rigid plates were connected to them by using tie constraints. A reference node is defined at
267 the centroid of the rigid plates. In the reference point of the rigid plate at the support, the
268 following boundary conditions were applied – the displacements along X- and Y-axes, and
269 the rotation about the longitudinal axis (Z-axis) were restrained. On the other end of the
270 mullion symmetric boundary conditions were defined at the element edges, where the
271 rotation about X- and Y- axes and the displacement along Z-axis were restrained. These
272 boundary conditions are shown in Fig. 15(a). The lateral restraints were simulated at the
273 locations where the angles were connected, by restraining the X-axis displacement of the
274 corresponding nodes. In actual buildings, the displacement along the X-axis is restrained by
275 the glass panels. The load was applied at the reference node of the loading plate.
276 Furthermore, at the loading point, twisting about the Z-axis and displacement along the X-
277 axis were restrained.
278
279 It is not necessary to model the back to back mullion sections as in Model Type A. The
280 purpose of connecting the mullions back to back is to avoid any eccentric loading since
281 torsional stresses can develop in the mullions if the load acts away from the shear centre.
282 This condition can be simulated by modelling one mullion with the load applied at its shear

9
283 centre. This simplified approach was found to be adequate to simulate the section moment
284 capacity tests of cold-formed lipped and hollow flange sections in the past (Anapayan et al,
285 2011, Siahaan et al, 2016, Pham et al, 2014, Pham and Hancock, 2010). In Model Type B
286 (Fig. 15(b)), this method was used where only one mullion was modelled and the load was
287 applied at its shear centre. A 10 mm thick rigid plate was attached to the mullion to represent
288 the T-shaped stiffeners at the loading and support locations. R3D4 elements were used for
289 the rigid plates, while S4R shell elements were used to model the mullions (Fig. 15(b)). The
290 mesh size used was 4 mm x 4 mm. The corresponding reference node of these rigid plates
291 was defined at the shear centre of the mullions. At the reference point of the support plate,
292 the displacements along X- and Y- axes were restrained while the rotation about the
293 longitudinal axis (Z-axis) was restrained. On the other end of the mullion, the rotation about
294 the X- and Y- axes and the displacement along Z-axis were restrained at the element edges.
295 Lateral restraints provided by the angles to the mullions were simulated by restraining the
296 displacement along the X-axis. At the reference point of the loading plate, displacement
297 along X-axis and rotation about Z-axis were restrained (Fig. 15(b)).
298
299 The measured mechanical properties given in Section 3 were used in the modelling.
300 ABAQUS requires a stress-strain relationship of material in terms of true stress and
301 logarithmic plastic strain. Since the used models give the engineering stress-strain curves,

302 these curves (  eng and  eng ) were converted to true stress-logarithmic plastic strain (  true

and  true ) curves using Eqs. 10 and 11 for input to ABAQUS.


pl
303

 true =  eng (1 +  eng ) ---------------(10)

 true
 true
pl
= ln(1 +  eng ) − ---------------(11)
E
304 Aluminium metric standard (2013) specifies the dimensional tolerance for the aluminium
305 extrusion sections. The local geometric imperfection tolerance for the flat plate elements of
306 open extruded mullion section is given as 0.004b (where b is the width). However, if the
307 plate width is less than 25 mm, the allowable tolerance should be taken as 0.1 mm. The
308 geometric imperfection values used in the models were based on these specifications. A
309 bifurcation buckling analysis was performed first in ABAQUS, and the imperfection was
310 applied in the shape of the critical buckling mode (Schafer and Pekoz, 1998). Nonlinear
311 analyses were then conducted with and without the imperfections. Residual stresses are

10
312 negligible in the extruded aluminium sections, and thus are not considered in the finite
313 element models (Mazzolani, 1995). The same assumption was used in the numerical studies
314 reported in (Kim and Pekoz, 2003,2010, Wang, 2006, Zhu and Young, 2009, Su et al,
315 2014,2016).
316
317 The mullion sections subject to bending actions involve buckling and post-buckling
318 behavioural characteristics. Modified Riks method is widely used to analyze problems
319 showing unstable response, e.g buckling deformations. Riks method uses true static
320 equilibrium equation (Eq. 12) to derive the solutions, where Newton’s method is the base.
321 Quasi-static analysis is also sometimes used when numerical convergence issues are
322 encountered with Riks method, and thus it was also investigated in one case for
323 comparison. Quasi-static analysis refers to solving the static problems using dynamic
324 equilibrium equation given in Eq. 13, by considering the structural deformations as very
325 small dynamic events, where the inertia and damping forces can be neglected. Natario et al.
326 (2014) and Janarthanan (2017) successfully used the quasi-static analysis approach to predict
327 the web crippling capacity of cold-formed steel beams. However, when using the quasi-static
328 analysis technique the kinetic energy to internal energy ratio of the model should be less than
329 5% to ensure that the inertial effects are minimal.
𝐹 = 𝐾𝑈 ---------------(12)

𝐹 = 𝑀𝑈̈ + 𝐶𝑈̇ + 𝐾𝑈 ---------------(13)

330 The load versus vertical displacment curves of 650-027 mullion subject to positive loading
331 obtained using Riks method of analysis, and quasi-static analysis using the explicit and
332 implicit integration schemes are presented in Fig. 16(a). They agree well with each other.
333 The moment and the kinetic to internal enery ratio versus time curves from the quasi-static
334 implicit and explicit analyses are given in Figs. 16(b) and (c), respectively, which show that
335 the kinetic energy to internal energy ratio is less than 1% until failure. The failure moments
336 from Riks, and quasi-static implicit and explicit anlyses are 6.105, 6.084 and 6.067 kNm,
337 respectively. This demonstates that the quasi-static analysis method can also be employed to
338 predict the section moment capacity of complex shaped aluminium mullions. In this study,
339 the modified Riks method was used in all other analyses.
340
341 Two different mesh patterns were used to define the geometry of the cross-section as shown
342 in Fig. 17 – Mesh Types 1 and 2. Mesh Type 2 is more refined at the edges and stiffener

11
343 locations. Analyses were performed for 650-027 sections subject to negative and positive
344 loadings. The results presented in Fig. 18 show that they are about the same, where the
345 difference is only 0.6 and 0.1% for the negative and positive loadings, respectively. Since
346 the differences are small, Mesh Type 1 was used in all other analyses.

347
348 As discussed earlier, two different types of half model, Model Type A and Model Type B
349 were developed. Fig. 18 compares the FEA predictions of 650-027 section subject to
350 negative and positive loadings based on Model Type A and Model Type B. The ultimate
351 moment capacities obtained for the negative loading are 4.741 and 4.743 kNm for Model
352 Types A and B, respectively. They are 6.105 and 6.101 kNm, respectively, for the positive
353 loading case. Also, the moment versus vertical displacement curves of Model Types A and
354 B agree well with each other. Figs. 19(b) and (c) present the failure modes of 650-027
355 mullion section obtained based on Model Types A and B for the negative loading case. The
356 failure modes from the two models are the same, where the yielding of flange elements is
357 observed. These failure modes are very similar to that observed in the test as shown in Fig.
358 19(a). Fig. 20 presents the von Mises stress diagram at failure for Model Types A and B,
359 which are also very similar. These findings and good agreements demonstrate that both
360 Model Types A and B can be used to predict the section moment capacity of mullions.
361 Model Type B takes less time to develop, and requires less memory and time for the analysis.
362 Therefore, Model Type B was used to simulate the section moment capacity tests of all other
363 mullion sections.
364
365 The finite element analysis results of all the tested mullions subjected to positive and
366 negative loadings are summarized in Table 1. These results agree reasonably well with the
367 experimental results. As seen in Table 1, the effect of geometric imperfection on the section
368 moment capacity is about 1% on average. The ratios of the section moment capacities from
369 tests and FEA are also summarized in Table 1. For predictions using the model with and
370 without geometric imperfection, the mean ratios are 1.06 and 1.05, respectively. The models
371 without geometric imperfection gave a slightly better agreement with the test results. During
372 the manufacturing of mullion sections, the sections are extruded first and are then pulled
373 apart to straighten them. This could have eliminated any geometric imperfections in the
374 mullions.
375

12
376 The moment versus vertical displacement (tension flange) plots of a few mullions obtained
377 from tests and FEA are compared in Figs. 21(a) to (f). They agree reasonably well with each
378 other. Further, von Mises stresses and failure modes of the 450-044 mullions subjected to
379 negative and positive wind loadings are shown in Figs. 22(a) and (b), respectively, and at
380 the point of failure they exhibit local and distortional buckling. Similar buckling modes were
381 also observed in the failed mullion specimens as shown in Figs. 23(a) and (b). All of these
382 comparisons demonstrate the accuracy of the developed finite element models in predicting
383 the section moment capacities of asymmetric complex-shaped aluminium mullions
384 considered in this study.
385
386 Kesawan et al. (2018) considered three different design approaches to predict the section
387 moment capacity of aluminium mullions. They are, total moment capacity approach
388 (TMCA) (Kim and Pekoz, 2003,2010), limiting stress method (LSM) (ADM, 2010,2015)
389 and direct strength method (DSM) (ADM, 2015). More details of these methods including
390 their limitations are given in Kesawan et al. (2018). In this paper the mullion capacities
391 predicted by these three design methods are compared with the FEA results in Table 3. FEA
392 results of the mullion models without geometric imperfection were considered as they were
393 more closer to the test results. The ratios between the moment capacities obtained from FEA
394 and design methods are also summarised in Table 3. The mean ratios for the TMCA, LSM
395 and DSM are 1.15, 1.25 and 1.01, respectively. This demonstrates that the TMCA and LSM
396 methods can be used to predict the section moment capacities of mullions, but their
397 predictions will be conservative. DSM (ADM, 2015) predictions are considerably better than
398 the other two methods considered.
399
400 5. Parametric study
401
402 A finite element analysis based parametric study was performed to investigate the section
403 moment capacity of an open mullion (Fig. 24) with varying cross-sectional slenderness
404 subject to negative wind loading. The section was considered to be of uniform thickness,
405 where a thickness range of 1.0 to 4.8 mm was considered. This resulted in different cross-
406 sectional slendernesses, in the range of 23 to 113 (Table 4), thus compact sections, and
407 sections subject to elastic and inelastic local buckling were considered.
408
409 Finite element Model Type B discussed in Section 4 was used. The method used to develop
410 and analyze the finite element models was the same as described in Section 4. The yield

13
411 strength and elastic modulus of the section was taken as 172 MPa and 70, 000 MPa,
412 respectively (AS/NZS 1664 (SA, 1997)). The finite element analysis results are given in
413 Table 4, and are compared with the predictions of DSM given in ADM (2015). The ratios
414 between the section moment capacities obtained from FEA and DSM design equations are
415 also given in Table 4. The predictions using the DSM agree well with the FEA results of
416 compact mullion sections, and those subject to inelastic buckling, as reflected by the mean
417 FEA to design rule prediction ratio of 1.03. However, the predictions were conservative for
418 slender mullion sections with a mean FEA to design rule prediction ratio of 1.59. The mean
419 and COV for the FEA to design rule prediction ratios considering all the mullion sections
420 are 1.17 and 0.22, respectively. This shows that the DSM in ADM (2015) can be
421 conservatively used to predict the section moment capacity of aluminium mullion sections
422 subject to local buckling. However, the DSM equations to determine the section moment
423 capacity of slender sections should be further improved. Further extensive FEA based
424 parametric study should be performed to develop a comprehensive database of the section
425 moment capacity of aluminium mullion sections with different shapes and sizes and made
426 of different aluminium alloys. This can then be used to develop more accurate design
427 equations to predict the section moment capacity of aluminium mullions.
428
429 6. Conclusion

430 This paper has presented a detailed finite element modelling based investigation on the
431 section moment capacity of complex-shaped aluminium mullion sections used in high-rise
432 buildings. It also includes a summary of the section moment capacity tests of aluminium
433 mullion sections, and an experimental study conducted to determine the accurate mechanical
434 properties of 6063-T6 aluminium alloy mullion sections. The latter experimental study
435 showed that the yield and ultimate strength values are well above those specified in AS/NZS
436 1664 (SA, 1997). It was found that Ramberg and Osgood’s (1943) stress-strain model given
437 in Eurocode 9 Part 1.1 (ECS, 2007) with the strain hardening exponent (n) calculated based
438 on Eq. 5 of this paper, accurately predicted the measured stress-strain curves, which were
439 included in the finite element models used in this research.
440
441 In the finite element studies, two different finite element models based on back to back
442 mullions and single mullion with the load acting on its shear centre were used initially, but
443 since both models gave similar results, the simpler single mullion based finite element model

14
444 was recommended and used in all other analyses. Comparison of the section moment
445 capacities, moment versus deflection curves and failure modes showed that the developed
446 finite element models were capable of predicting the section moment capacities of complex-
447 shaped aluminium sections. Finite element analysis results were also compared with the
448 limiting stress method, total moment capacity approach and DSM predictions. The
449 agreement between the DSM predictions and the analysis results was superior to the other
450 design rule predictions, and thus DSM is recommended.
451
452 This paper has also presented a numerical parametric study performed to investigate the
453 section moment capacity of mullions with significant variations in cross-sectional
454 slenderness (23 to 113). The results were compared with the DSM design rule predictions.
455 It was found that the DSM method given in ADM (2015) predicted the capacity of compact
456 sections and sections subject to inelastic local buckling accurately, but its predictions were
457 over-conservative for slender sections. Therefore, the DSM given in ADM (2015) can be
458 successfully used to predict the section moment capacity of mullion sections subject to local
459 buckling, but the equations for slender mullion sections can be further improved.
460
461
462 Notations

463 The following symbols are used in this paper:


464 σ = stress
465 ε = strain corresponding to the stress σ
466 fe = conventional limit of elasticity
467 εo,e = residual strain corresponding to the stress fe
468 fx = second reference stress
469 εo,x = residual strain corresponding to the stress fx
470 f0.2 = proof strength by means of 0.2% offset method
471 n = strain hardening exponent
472 E = modulus of elasticity
473 f0.1 = proof strength by means of 0.1% offset method
474 fmax = characteristic value of ultimate tensile strength
475 εo,max = residual strain corresponding to the stress fmax
476 εmax = strain corresponding to the stress fmax

15
477 εu = ultimate strain at fracture
478 U = nodal displacement
479 𝑈̇ = nodal velocity
480 Ü = nodal acceleration
481 K = global stiffness matrix
482 M = mass matrix
483 C = damping matrix
484 F = applied external force matrix
485

16
486 Acknowledgement
487 The authors would like to thank QUT for providing all the necessary support with testing
488 and computing facilities, G. James Glass and Aluminium for providing the specimens and
489 their engineers for the technical support, and Australian Research Council for providing the
490 financial support to conduct this research project.
491

492 Data Availability Statement


493 Some or all data, models, or code that support the findings of this study are available from
494 the corresponding author upon reasonable request.
495
496 References

497 1. ADM (2010). Aluminium Design Manual, Specification for Aluminium Structures. The
498 Aluminium Association. Washington, D.C., USA.
499 2. ADM (2015), Aluminium Design Manual, Specification for Aluminium Structures, The
500 Aluminium Association. Washington, D.C., USA.
501 3. Anapayan, T, Mahendran M, Mahaarachchi, D. (2011) Section moment capacity tests
502 of LiteSteel Beams. Thin-Walled Structures, 49(4), pp. 502-512.
503 4. Allana, P. K. and Carter, D. (2012). Curtain walls issues, problems and solutions. Proc.
504 of the Building Envelope Technology Symposium, RCI Inc., pp. 97-111.
505 5. Aluminium Association (2013). Aluminium standards and data Metric SI.
506 6. De Martino, A., Landolfo, R. and Mazzolani, F. M. (1990). The use of the Ramberg-
507 Osgood law for materials of round-house type. Materials and Structures, 23, pp. 59-67.
508 7. Gardner, L. and Ashraf, M. (2006). Structural design for non-linear metallic materials.
509 Engineering Structures, 28, pp. 926-934.
510 8. European Committee for Standardization (ECS) (2007). Eurocode 9: Design of
511 aluminium structures - Part 1–1: General structural rules. BS EN 1999-1-1:2007,
512 Brussels, Belgium.
513 9. Hill, H. N. (1944). Determination of stress-strain relations from "offset" yield strength
514 values. Washington, D.C., National Advisory Committee for Aeronautics, USA.
515 10. Janarthanan, B. (2017). Structural Behaviour and Design of Cold-Formed Steel Floor
516 Systems. PhD Thesis, Queensland University of Technology, Brisbane, Australia.
517 11. Kesawan, S. and Mahendran, M. (2015). Predicting the performance of LSF walls made
518 of hollow flange sections in fire. Thin-Walled Structures, Vol. 98(A), pp. 111-26.

17
519 12. Kesawan, S., Mahendran, M., Baleshan, B. (2018). Section moment capacity tests of
520 complex-shaped aluminium mullions. Thin-walled structures (submitted).
521 13. Kim, Y. and Pekoz, T. (2003). Behavior and Design of Aluminium Members in Bending.
522 Research report. Cornell University. NY, USA.
523 14. Kim, Y. and Pekoz, T. (2010). Ultimate flexural strength of aluminium sections. Thin-
524 walled structures, Vol. 47(10-11), pp. 857-65.
525 15. Lai, Y. F. W. and Nethercot, D. A. (1992). Strength of aluminium members containing
526 local transverse welds. Engineering Structures, Vol. 14(4), pp. 241-254.
527 16. Mazzolani, F. M. (1995). Aluminium alloy structures. 2nd ed. London, UK: E & FN
528 Spon.
529 17. Natario, P., Silvestre, N. and Camotim, D. (2014). Web crippling failure using quasi-
530 static FE models. Thin-walled Structures, Vol. 84, pp. 34-49.
531 18. Pham, C. H. and Hancock, G. J. (2010). Experimental investigation of high strength
532 cold-formed CC -sections in combined bending and shear. Journal of structural
533 engineering, ASCE, Vol. 136(7), pp. 866-878.
534 19. Pham, C. H., Davis, A. F., Bonney, R. and Emmet, R. (2014). Numerical investigation
535 of cold-formed lapped Z purlins under combined bending and shear. Journal of
536 construction steel research, Vol. 95, pp. 116-125.
537 20. Ramberg, W. and Osgood, W. (1943). Description of stress–strain curves by three
538 parameters. Technical note No. 902. National Advisory Committee for Aeronautics.
539 Washington, D.C., USA.
540 21. Russell, M. S. (2006). Curtain walls: not just another pretty façade. Journal of
541 Architectural technology, Vol. 23(1), pp. 1-18.
542 22. Siahaan, R., Keerthan, P. and Mahendran, M. (2016). Finite element modeling of rivet
543 fastened rectangular hollow flange channel beams subject to local buckling. Engineering
544 Structures, 126, pp. 311-327.
545 23. Standards Australia (SA) (1997). Aluminium Structures, Part 2: Allowable stress design.
546 AS/NZS 1664.2, Sydney, Australia.
547 24. Standards Australia (SA) (2007). Metallic materials-tensile testing at ambient
548 temperature. AS 1391, Sydney, Australia. Reissued incorporating Amendment No. 1
549 (August 2012).
550 25. Schafer, B. and Pekoz, T. (1998). Computational modeling of cold-formed steel:
551 Characterizing geometric imperfections and residual stresses. Journal of Construction
552 Steel Research, Vol. 47(3), pp. 193-210.

18
553 26. Su, M. N., Young, B. and Gardner, L. (2014). Deformation-based design of aluminium
554 alloy beams. Engineering Structures, Vol. 80, pp. 339-49.
555 27. Su, M. N., Young, B. and Gardner, L. (2016). Flexural response of aluminium alloy
556 SHS and RHS with internal stiffeners. Engineering Structures, Vol. 121, pp. 170-180
557 28. Wang, Y. (2006). Structural behavior and design of two custom aluminum extruded
558 shapes in custom unitized curtain wall systems. MSc thesis. University of Cincinnati.
559 Ohio, USA.
560 29. Zhu, J. and Young, B. (2009). Design of aluminium flexural members using direct
561 strength method. Journal of Structural Engineering ASCE, Vol. 135(5), pp. 558-566.

19
1

Figure 1
1

Male Female Connected


mullion mullion together

Figure 2
1

M F M F M F M F

(a) (b) (c) (d)

Figure 3
1

Back to back
connected mullions

Mullion 1 Mullion 2 Mullion 1 Mullion 2

Load Load

(a) (b)
Figure 4
1

Applied load

Figure 5
1

Angle brackets

Web side plate


T-shaped
stiffeners M16 bolt

(a)

Loading arm and spreader beam

LVDT 1

Applied loads

LVDT 2

T-shaped stiffener

(b)
Figure 6
1

(a)

6063-T6 aluminium
tensile coupon

Upper and
lower grips

Clip-on extensometer

(b)
Figure 7
1

250

200

150
Stress (MPa)

100

50

0
0 0.02 0.04 0.06 0.08 0.1 0.12
Strain (%)
650-027-A 650-027-B
(a)

250

200

150
Stress (MPa)

100

50

0
0 0.02 0.04 0.06 0.08 0.1
Strain (%)
650-028-A
(b)

Figure 8
1

25

Eurocode 9
20 Part 1.1
Proposed
Proposed
eu,test (%)

15

10

0
0 5 10 15 20 25
eu,pred (%)

Figure 9
1

25

Su et al. (2014)

20

15
εmax,test (%)

10

0
0 5 10 15 20 25
εmax,pred (%)

Figure 10
1

250

Stress (MPa) 200

150

100

50

0
0 0.02 0.04 0.06 0.08 0.1 0.12
Strain (%)
Test results Ramberg-Osgood model

(a)

250

200
Stress (MPa)

150

100

50

0
0 0.02 0.04 0.06 0.08 0.1 0.12
Strain (%)
Test results Ramberg-Osgood model

(b)
Figure 11
1

250

200

150
Stress (MPa)

100

50

0
0 0.02 0.04 0.06 0.08 0.1 0.12
Strain (%)
Test results Ramberg-Osgood model
(a)

250

200

150
Stress (MPa)

100

50

0
0 0.02 0.04 0.06 0.08 0.1
Strain (%)
Test results Ramberg-Osgood model

(b)
Figure 12
1

(a) (b)
Figure 13
1

L1 = 400 mm, L2 = 300 mm (see Fig. 5)

Figure 14
1

Stiffeners between
two mullions-R3D4

Lateral
restraints

Boundary of symmetry
at mid-span
(U3=UR1=UR2=0)
Loading point
(U1=UR3=0)

Support reference point


Mullions-S4R
(U1=U2=UR3=0)
elements
(a)

Boundary of symmetry at mid-span


(U3=UR1=UR2=0)
Lateral restraints
(U1=0)

Loading
point
Stiffeners-
(U1=UR3=0)
R3D4 elements

Mullion-S4R Support
elements (U1=U2=UR3=0)

(b)
Figure 15
1

8
7
6
5
Moment (kNm)

4
3
2
1
0
0 5 10 15 20 25
Displacement (mm)
Riks method Quasi-static implicit Quasi-static explicit
(a)

8 50

7
40
6
30
Moment (kNm)

K.E*100/I.E
4 20

3
10
2
0
1

0 -10
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Time (s)
Moment K.E*100/I.E
(b)
2

8 50
7 45
40
6
35
5
Moment (kNm)

K.E*100/I.E
30
4 25
3 20
15
2
10
1 5
0 0
0 0.1 0.2 0.3 0.4 0.5 0.6
Time (s)
Moment K.E*100/I.E
(c)

Figure 16
1

Refined
Refined
stiffeners
corner areas

(a) (b)

Figure 17
1

Moment (kNm) 5

0
0 2 4 6 8 10 12 14 16 18 20
Displacement (mm)
Model Type B - Mesh Type 1 Model Type A - Mesh Type 1
Model Type A - Mesh Type 2

(a)

5
Moment (kNm)

0
0 1 2 3 4 5 6 7 8
Displacement (mm)
Model Type B - Mesh Type 1 Model Type A - Mesh Type 1
Model Type A - Mesh Type 2

(b)
Figure 18
1

(a)

Back to back mullion

(b)

Single mullion

(c)
Figure 19
1

(a)

(b)

Figure 20
1

5
Moment (kNm)
4

0
0 4 8 12 16 20
Displacement (mm)
FEA - no imp FEA - imp Test
(a)
8

5
Moment (kNm)

0
0 2 4 6 8 10 12 14 16
Displacement (mm)
FEA - no imp FEA - imp Test

(b)
2

12

10

8
Moment (kNm)

0
0 10 20 30 40 50
Displacement (mm)
FEA - no imp FEA - imp Test

(c)

4
Moment (kNm)

0
0 5 10 15 20 25 30
Displacement (mm)
FEA - no imp FEA - imp Test

(d)
3

5
Moment (kNm)

0
0 5 10 15 20 25 30 35 40 45
Displacement (mm)
FEA - no imp FEA - imp Test

(e)
7

5
Moment (kNm)

0
0 5 10 15 20 25 30
Displacement (mm)
FEA - no imp FEA - imp Test

(f)

Figure 21
1

Local buckling

Distortional
buckling

(a) (b)

Figure 22
1

Local buckling Distortional buckling

(a) (b)

Figure 23
1

Negative
loading

Figure 24: Mullion section considered in the parametric study

You might also like