Professional Documents
Culture Documents
Thin-Walled Structures: Ashkan Shahbazian, Yong Chang Wang
Thin-Walled Structures: Ashkan Shahbazian, Yong Chang Wang
Thin-Walled Structures: Ashkan Shahbazian, Yong Chang Wang
Thin-Walled Structures
journal homepage: www.elsevier.com/locate/tws
art ic l e i nf o a b s t r a c t
Article history: This paper investigates the applicability of a simple fire resistance design method for axially loaded thin-
Received 11 October 2013 walled steel studs in wall panel assemblies when exposed to parametric fires from one side. The simple
Received in revised form method includes calculations of cross-section temperatures and ultimate load carrying capacities at
2 December 2013
elevated temperatures. The simplified calculation method for heat transfer in the cross-section is based
Accepted 2 December 2013
on dividing the cross-section into a number of segments. The thermal properties of these layers are based
on weighted averages of the thermal properties of the components contained within. The structural
Keywords: capacity calculation method is based on the Direct Strength Method. Results from the design method are
Thin-walled steel studs compared with the results from Finite Element simulations for heat transfer and structural analysis (236
Wall panel
models). The calculation results are in good agreement with the simulation results and the proposed
Parametric fire
method may be used in performance-based fire engineering design of such construction.
Design method
Fire resistance & 2013 Elsevier Ltd. All rights reserved.
Performance-based fire engineering
0263-8231/$ - see front matter & 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.tws.2013.12.001
68 A. Shahbazian, Y.C. Wang / Thin-Walled Structures 77 (2014) 67–76
Fast fire
For heating phase : θg ¼ θa þ 445 log 10 ð8t þ 1Þ; t max ¼ 20 min
ð1aÞ
Fig. 2. Wall panel exposed to fire from one side. For cooling phase : θg ¼ θmax 625ðt t max Þ ð1bÞ
approach which is based on simple finite difference solution of one Medium fire
dimensional heat transfer in the panel thickness direction and uses For heating phase : θg ¼ θa þ 345 log 10 ð8t þ 1Þ; t max ¼ 40 min
the weighted average of thermal resistances in the panel width
ð2aÞ
direction. This temperature calculation method has been demon-
strated to be accurate for application under different fire conditions For cooling phase : θg ¼ θmax 250ð3 t max Þ ðt t max Þ ð2bÞ
and this method will be summarised in Section 5.1.
Combining the proposed temperature calculation method and
Slow fire
structural load carrying capacity method allows the fire resistance
to be easily evaluated in design. The applicability of the combined For heating phase : θg ¼ θa þ 245 log 10 ð8t þ 1Þ; t max ¼ 60 min
thermal–structural performance method has been assessed under ð3aÞ
the standard fire condition (heating stage) and the results are very
accurate [11]. This paper investigates applicability of this method For cooling phase : θg ¼ θmax 250ð3 t max Þ ðt t max Þ ð3bÞ
under parametric fire conditions which are now becoming more
widely used in practical performance-based fire engineering
design. The effects of cooling will also be included. In the above equations: θg is the fire temperature; θa is the
ambient or room temperature; θmax is fire temperature at the
maximum heating time and tmax is the maximum heating time.
2. Assumed parametric fire temperature–time curves Fig. 3 shows the three representative fire temperature–time curves
assumed in this study. These three fire curves produce a range of
EN 1991-1-2 [3] gives an equation for evaluating the parametric desired different temperature distributions in the cross-section.
fire curves which depend on the thermal properties of the fire
enclosure, the opening factors (related to the area of openings 3. Methodology
relative to and total area of the enclosure) and the fire load. For the
pffiffiffiffiffiffiffiffiffiffi standard fire
so-called enclosure with a thermal property value of Fig. 4 shows the procedure used in validating the proposed
ðρcλÞ ¼ 1160 J=m2 s1=2 K and an opening factor of 0.04 m1/2, the method for calculating the fire resistance of thin-walled steel studs
A. Shahbazian, Y.C. Wang / Thin-Walled Structures 77 (2014) 67–76 69
Comparison of results
(Section 6)
coefficient on the fire exposed side ¼25 W/m2 K and on the in this way. To do so, the structure temperatures at any time were
unexposed side ¼10 W/m2 K. The element type is DC2D4 (4-node applied and then the mechanical load was applied until the steel
linear heat transfer quadrilateral) [11]. The temperature distribu- stud reaching its ultimate load carrying capacity. The thermal
tion in the steel studs from the heat transfer FE models were used expansion coefficient for steel was assumed to be 0.000014/1C. The
in the structural analysis model using Abaqus. mechanical properties at elevated temperature were based on EN
1993-1-2 [13] with a steel yield stress and elastic modulus 350 and
205,000 N/mm2 respectively at ambient temperature.
4.2. Structural analysis FE model For the FE structural analysis, the non-uniform structural
temperatures were taken from the output of the FE heat transfer
The structural analysis is based on the validated Abaqus model analysis (Section 4.1). Fig. 7 shows the mechanical boundary and
[5–7] by the authors. The objective of this study is to compare the loading conditions: the column was simply supported (preventing
FE results with calculation results using the proposed method. lateral displacement, allowing longitudinal displacement at each
Since the proposed method gives the load carrying capacity of the node except for the centre of one end) at the top and bottom of the
structure at elevated temperatures, the FE analysis was conducted column (Fig. 7b); the axial loads were applied on each node at the
two ends of the column (Fig. 7b); it was restrained at 300 mm
Fire temperature around the surface
Convection and Radiation spacing in the lateral (flange) direction to simulate the restraining
2
h = 25 W/m .K ε = 0.3 effect of the gypsum boards (Fig. 7c). It should be noted that in
reality, the gypsum boards may lose their restraining effects on the
steel studs in fire. However, since the objective of this study is to
assess the proposed simple method, this effect was not considered
and the assessment was considered valid as long as the same
restraining conditions were applied in both the finite element
model and the proposed method. The length of the column is
3000 mm. Abaqus shell element type S4 was used and the global
mesh size was 2.5 mm. The FE structural analysis model incorpo-
rated initial imperfection. The initial imperfection mode corre-
sponded to the expected final failure mode of the column. For
distortional buckling, the maximum initial imperfection was equal
to half of the section thickness; for pure local and pure global
buckling, the maximum initial imperfection was equal to the
thickness; for interactive local/global buckling, the maximum
initial imperfection for the local buckling mode was equal to the
Convection and Radiation
20 °C thickness and the maximum imperfection for the global buckling
h = 10 W/m2.K ε = 0.8
mode was equal to H/500 (where H is the column height)
Fig. 6. Thermal boundary conditions. respectively.
Length = 3000 mm
Fig. 7. Mechanical boundary conditions. (a) column boundary, (b) loading condition and (c) restraints.
A. Shahbazian, Y.C. Wang / Thin-Walled Structures 77 (2014) 67–76 71
5. Proposed design method times component thermal conductivity) was used. The total thermal
capacitance (density times specific heat) of the layer was assumed to
5.1. Thermal analysis be the sum of all the components based on their proportion to the
total mass. Based on these assumptions, a 1D finite difference
In the research study by Feng et al. [10], it was found that for program has been written to calculate the temperatures of all the
steel studs in wall panels exposed to fire attack from one side, the layers. This method is much quicker than using 2D finite element
temperature distribution in the steel stud may be assumed to be heat transfer analysis, but more importantly, it can be easily adopted
linear between the hot and cold sides, with the flanges having in design.
uniform temperature distributions. This simplifies the tempera- This procedure has been implemented for panels exposed to the
ture field in the panel structure considerably and lends it to the standard fire from one side and comparison against the results of an
development of a simple method of heat transfer analysis. extensive set of finite element heat transfer simulations comprehen-
Recently the authors [11] have developed such a simple approach. sively validated this procedure [11]. This paper (Section 6) provides
In this approach, the wall panel is represented by a segment of further validation of this method for panels exposed to parametric
width We (Fig. 8), outside of which the panel temperature fires with cooling stage. The width of the segment (Fig. 8) for
distribution is not affected by the steel stud. This segment is then calculating the weighted average of thermal properties is
divided into 13 layers through the panel thickness, as shown in W e ¼ t i þ bf þ bi ¼ 45 þ0:85 bf ðin mm; t i ¼ bi Þ ð4Þ
Fig. 9. Heat transfer is assumed to be one dimensional in the
thickness direction of the panel, with each layer having its own
thermal conductivity and thermal capacitance (density specific 5.2. Extended direct strength method for fire applications
heat). The thermal properties of each layer are obtained from the
weighted averages of the all materials contained in the layer Direct strength method [8] is a recently adopted method [9] for
within the segment. For calculating the contributions of different calculating the load carrying capacity of thin-walled columns. It is
components within each layer to the overall thermal conductivity, based on regression equations using the elastic critical load and
this research applied the additive rule of thermal resistance the plastic resistance of the column. The elastic buckling load for
(thickness/thermal conductivity) in the direction of heat conduction local, distortional and global buckling can be obtained by using a
(conductors in series). In the direction perpendicular to heat numerical procedure such as CUFSM which is a Finite Strip
conduction (parallel conductors), the additive rule of thermal con- Method program [16]. The plastic resistance of the cross-section
ductivity (ratio of component thickness to overall layer thickness (squash load) is simply the gross sectional area multiplied by the
yield stress.
Recently the authors [5–7] have applied this method to fire
conditions, for both uniform and non-uniform temperature dis-
tribution in the cross-section.
For calculating the elastic buckling loads using CUFSM, the
different strips of the cross-section are assigned different Young's
modulus values according to their temperatures. With non-
uniform temperature distribution in the cross-section, the effects
Fig. 8. Proposed geometry for simplified 1D heat transfer calculation [11]. of thermal bowing and shift of the centre of resistance generate
bending moments in the cross-section. Therefore, the original
column under pure axial load becomes a beam-column with
combined axial load and bending moment. The squash load of
the column used in DSM is the axial load value that causes the
cross-section with non-uniform temperature to reach the full
plastic limit of the cross-section under combined axial load and
bending. This is termed “effective” squash load in this study and is
obtained using the following procedure [5]:
2
for λc r1:5; P ne ¼ ð0:495 λc ÞP y ð5bÞ
!
0:462
for λc 41:5; P ne ¼ Py ð5cÞ
λ2c
above); P cre is the critical elastic global buckling load; P crd is the
critical elastic distortional buckling load; P crl is the critical elastic
local buckling load; λc , λd and λl are the slenderness for global,
distortional and local buckling modes respectively; P ne , P nl and
P nd are the column axial strength for global, distortional and local
buckling modes respectively.
Fig. 12. Temperature distributions for LC100 56 15 2 at: (a) 30 min and Fig. 14. Position of centre of resistance for lipped channel 100 56 15 2 (web
(b) 60 min. dimension ¼ 100 mm, at ambient temperature the centre of the cross-section/
resistance is equal to 50 mm).
Fig. 15. Results of using different schemes to calculate the squash load for LC100
Fig. 13. Lateral deflection caused by thermal bowing for LC100 56 15 2. 56 15 2.
74 A. Shahbazian, Y.C. Wang / Thin-Walled Structures 77 (2014) 67–76
slightly conservative results before the temperature reaches their the same as the centre of the cross-section because the mechanical
maximum. Fig. 12 further shows that the agreement for tempera- properties are distributed uniformly in the cross-section. However,
ture distributions in the cross-section at different times (30 and when the mechanical properties are non-uniform in the cross-
60 min) is very good for all three fire curves. section due to non-uniform high temperatures, the centre of
resistance is different from the geometrical centre of the cross-
6.2. Comparison of column strength section. With slower heating, the temperature gradient in the
cross-section is smaller, hence the shift of the centre of resistance
Section 5.2 has provided a description of how the effects of is smaller. However, in all cases, the maximum shift of the centre
thermal bowing and shift of centre of resistance can be included in of resistance is quite significant. This shift gradually returns to the
the calculations for the elastic buckling loads and the “effective” minimum when the cross-section cools down and the temperature
squash load. In order to do so, it is necessary to calculate the in the cross-section return to ambient temperature. To demon-
thermal bowing and the shift of centre of resistance. For the strate the importance of including the effects of the thermal
thermal bowing deflection, Fig. 13 shows a comparison between bowing and shift of the centre of resistance, Fig. 15 compares the
Abaqus FEA results and the analytical equation ðαL2 ΔT=8dÞ [1]. squash loads with and without considering these effects.
The agreement is very good. Appendix demonstrates how the DSM method is applied, at
Fig. 14 shows the shift of the centre of resistance of the cross- given flange temperatures, through an example. Fig. 16 compares the
section in time. At ambient temperature, the centre of resistance is column strengths at different fire exposure times. The agreement is
Fig. 16. Comparing capacity at different times. Fig. 17. Comparing the results for global buckling mode.
Table 2
Minimum capacity comparison.
Abaqus DSM Temperature calculation Capcaity calculation Slow fire Medium fire Fast fire
(PFE) (Pu) method method (Pumin) (Pumin) (Pumin)
75 50 15 2.5 89,840.20 87,711.50 Abaqus FEA (Section 4.1) Abaqus FEA (Section 4.2) 39,658.10 31,195.70 30,425.20
Abaqus FEA (Section 4.1) Extended DSM (Section 5.2) 40,990.94 35,359.20 34,956.23
Thermal analysis (Section 5.1) Extended DSM (Section 5.2) 41,319.86 35,275.27 35,267.43
100 50 5 1.8 71,864.20 80,925.73 Abaqus FEA (Section 4.1) Abaqus FEA (Section 4.2) 34,846.40 27,291.60 26,637.90
Abaqus FEA (Section 4.1) Extended DSM (Section 5.2) 37,219.72 29,701.74 28,625.41
Thermal analysis (Section 5.1) Extended DSM (Section 5.2) 36,804.42 30,292.68 29,035.79
100 50 15 0.8 27,210.40 23,400.84 Abaqus FEA (Section 4.1) Abaqus FEA (Section 4.2) 10,433.40 72,64.61 6300.39
Abaqus FEA (Section 4.1) Extended DSM (Section 5.2) 9482.36 6966.33 6629.64
Thermal analysis (Section 5.1) Extended DSM (Section 5.2) 9397.19 7250.10 6785.57
100 50 15 2.5 144,323.00 130,114.40 Abaqus FEA (Section 4.1) Abaqus FEA (Section 4.2) 65,687.00 53,376.30 52,612.50
Abaqus FEA (Section 4.1) Extended DSM (Section 5.2) 68,381.32 58,476.38 58,203.11
Thermal analysis (Section 5.1) Extended DSM (Section 5.2) 68,791.99 59,599.22 58,473.07
100 56 15 2 115,000.00 112,873.66 Abaqus FEA (Section 4.1) Abaqus FEA (Section 4.2) 51,514.10 40,804.90 39,584.30
Abaqus FEA (Section 4.1) Extended DSM (Section 5.2) 55,332.65 46,746.98 45,376.24
Thermal analysis (Section 5.1) Extended DSM (Section 5.2) 55,959.89 46,897.32 45,306.95
100 75 15 0.8 24,827.70 27,458.33 Abaqus FEA (Section 4.1) Abaqus FEA (section 4.2) 8991.67 6067.26 5191.60
Abaqus FEA (Section 4.1) Extended DSM (Section 5.2) 9962.35 5537.07 4947.17
Thermal analysis (Section 5.1) Extended DSM (Section 5.2) 10,075.37 6090.63 4964.85
7. Conclusion
Appendix A. A worked example as illustration of the Fig. B2. Determination of the effecitve squash load .
calculation procedure
Second step: Determining the load-carrying capacity of the steel
Requirement: Calculate the load carrying capacity of a panel at channel stud by using the extended DSM.
50 min of the medium parametric fire exposure time. Using method described briefly in Section 5.2 and in detail
Basic input data: Steel section dimensions—lipped channel 75 in [5], the centre of plastic resistance is 32.93 mm from the lower
50 15 2.5; Interior insulation—ISOWOOL; Gypsum—One temperature flange side (37.5 mm at ambient temperature). There-
layer of 12.5 mm thick Fireline British gypsum on each side; fore, the shift of centre of resistance is 37.5–32.93¼ 4.57 mm.
Panel height—3 m. Using equation in Fig. 12 gives a thermal bowing [5] value of
39.88 mm. Therefore, the eccentricities at the top and bottom and
First step: Calculate the steel temperatures in the flanges on the at the centre of the column are 4.57 mm and |4.57–39.88| ¼
fire exposed and unexposed sides. The equivalent panel width for 35.31 mm respectively. The column axial load–bending moment
calculating the average steel temperature is interaction curve is obtained based on Section 5.2 (http://j.mp/
We ¼45þ 0.85(50) ¼87.5 mm CALESL), as shown in Fig. A2. From the two eccentricities above,
Using the proposed method which should be implemented and the effective squash loads are 113.14 kN (top and bottom) and
solved numerically by computer program (http://j.mp/TTRCW1), 67.81 kN (middle) respectively. Therefore, the effective squash
the temperature–time results in Fig. A1 can be obtained. From this load is min(113.14, 67.81) ¼ 67.81 kN.
figure, the steel flange temperatures at 50 min on the exposed and Using the CUFSM program (http://bit.ly/CUFSM), with input
unexposed sides are 454.84 1C and 264.91 1C respectively. of the elevated temperature Young's modulus values of the steel
76 A. Shahbazian, Y.C. Wang / Thin-Walled Structures 77 (2014) 67–76
section, the elastic critical loads for global, distortional and local [5] Shahbazian A, Wang YC. Calculating the global buckling resistance of thin-
buckling are: walled steel members with uniform and non-uniform elevated temperatures
under axial compression. Thin-Walled Struct 2011;49:1415–28.
[6] Shahbazian A, Wang YC. Application of the Direct Strength Method to local
Pcre ¼ 72.96 kN buckling resistance of thin-walled steel members with non-uniform elevated
Pcrl ¼410.22 kN temperatures under axial compression. Thin-Walled Struct 2011;49:1573–83.
Pcrd ¼ 345.81 kN [7] Shahbazian A, Wang YC. Direct Strength Method for calculating distortional
buckling capacity of cold-formed thin-walled steel columns with uniform and
Using Eqs. (5)–(7): non-uniform elevated temperatures. Thin-Walled Struct 2012;53:188–99.
λc ¼ 0.9640; Pne ¼35.27 kN [8] Schafer BW. Local, distortional, and Euler buckling of thin-walled columns.
λl ¼0.2932; Pnl ¼ 35.27 kN J Struct Eng 2002;128:289–99.
λd ¼0.4428; Pnd ¼67.81 kN [9] AISI. North American specification for the design of cold-formed steel
structural members. Washington, DC: American Iron and Steel Institute; 2007.
Pn ¼ min(35.27, 35.27, 67.81) ¼35.27 kN [10] Feng M, Wang YC, Davies JM. Axial strength of cold-formed thin-walled steel
For comparison, the finite element solution is 31.19 kN. channels under non-uniform temperatures in fire. Fire Saf J 2003;38:679–707.
[11] Shahbazian A, Wang YC. A simplified approach for calculating temperatures in
axially loaded cold-formed thin-walled steel studs in wall panel assemblies
exposed to fire from one side. Thin-Walled Struct 2013;64:60–72.
References [12] A.U. FEA. Abaqus Unified FEA.
[13] CEN. Eurocode 3: design of steel structures – Part 1-2: general rules –
structural fire design. EN1993-1-2:2005. Brussels: European Standard; 2005.
[1] Wang YC. Steel and composite structures: behaviour and design for fire safety.
[14] Rahmanian I, Wang YC. A combined experimental and numerical method for
London: Spon; 2002.
[2] Wang YC, Burgess I, Wald F, Gillie M. Performance-based fire engineering of extracting temperature-dependent thermal conductivity of gypsum boards.
structures. London: CRC Press; 2012. Constr Build Mater 2012;26:707–22.
[3] CEN, Eurocode 1: actions on structures – Part 1-2: general actions – actions on [15] Salhab, B, Wang, YC. A study of the thermal performance of cold-formed thin-
structures exposed to fire. EN1991-1-2:2002. Brussels: European Standard; walled perforated steel studs (thermal studs) in fire, In: Proceedings of the
2002. Earth & Science 2004 conference, Houston, USA: 7-10 March 2004.
[4] Zhang C, Li GQ, Wang YC. Sensitivity study on using different formulae for [16] Schafer B.W., Adany S., Buckling analysis of cold-formed steel members using
calculating the temperature of insulated steel members in natural fires. Fire CUFSM: conventional and constrained finite strip methods. In: 18th interna-
Technol 2012;48:343–66. tional specialty conference on cold-formed steel structures; 2006.