Professional Documents
Culture Documents
www.elsevier.com/locate/engstruct
Received 8 March 2006; received in revised form 1 December 2006; accepted 4 December 2006
Available online 10 January 2007
Abstract
This paper outlines a research program on an experimental investigation to determine the failure time of unprotected steel columns subjected
to various axial restraint ratios. Axial restraints were applied to simulate the thermal restraint effects due to adjacent cooler parts of a steel
framed structure in fire. All columns had an effective length of 1.74 m, and were divided into 4 groups according to their minor-axis slenderness
ratios of 45, 55, 81 and 97, respectively. The columns were axially loaded and exposed to a monotonically increasing heating condition. Initial
imperfections such as column crookedness and load eccentricity were measured by a specially designed facility. The test results show that axial
restraints, as well as initial imperfections, significantly reduce the failure times of axially-loaded steel columns. By contrast, bearing friction
substantially retards column failure times. These effects are represented by flexural springs at both ends of a column specimen. Besides, a simple
but reliable Rankine approach is used to compute the column failure times. The Rankine predictions agree well with experimental results.
c 2006 Elsevier Ltd. All rights reserved.
Keywords: Axial restraint; Buckling; Failure time; Elevated temperature test; Steel columns
of 9.2 m×3 m×3 m (height), with a movable bulkhead that can The magnitudes of P0 are shown in Table 2. It is observed
be adjusted to accommodate columns up to 4.2 m long. Loading that P0 of RS55 4, RS81 2, RS97 1, RS97 2 and RS97 4 are
was applied horizontally. Knuckle bearings were used at both less than their peers in the same series. This is because (1)
ends of a specimen to provide pinned end conditions. Columns another pair of bearings, being ground and greased to minimise
were orientated such that they bent about the weak axis. the end friction effect, were used for these 5 specimens; and
(2) the imposed working load P0 took 50% of Pu20 predicted
2.2. Test specimens numerically which excluded end friction effects. However, it is
Four series of bare steel columns were tested to failure felt that exclusion of friction effects on the calculation of Pu20
in this study, namely, RS45, RS55, RS81 and RS97; the for these five columns somewhat underestimated actual load
numerals refer to their slenderness ratios. Four different hot- capacities. Through numerical study the authors were aware
rolled sections were selected, namely, UC152 × 152 × 37, that for the ground knuckle bearings, friction effects werr still
UB203 × 133 × 25, UB152 × 89 × 16 and UB127 × 76 × 13, significant albeit every effort has not been spared to reduce
corresponding to minor axis slenderness ratios λ of 45, 55, it. This is confirmed by the FE models incorporating friction
81 and 97, respectively. In each series, a column was tested effects, which agree better with test results compared with the
to failure at ambient temperature to obtain the column load same FE models without friction effects.
capacity Pu20 . For elevated temperature tests, these four series Although the physical dimension of columns is 1.5 m long,
of pinned–pinned steel columns were subjected to four working they have an effective length of 1.74 m when measured from
axial loads P0 , which were fixed at 50% of the respective Pu20 . the knuckle bearings at both ends. A summary of the end plate
1644 K.-H. Tan et al. / Engineering Structures 29 (2007) 1641–1652
Table 1
Test specimen data
RS55 1 3.19 203.3 135.4 7.7 5.41 0.8 1.0 0.9 205 357
RS55 2 3.43 203.1 135.5 7.6 5.41 −0.1 −0.3 −0.7 205 357
UB203 × 133 × 25/(55)
RS55 3 5.05 203.2 135.5 7.7 5.41 −0.7 −0.5 −1.1 205 357
RS55 4 3.67 204.4 133.7 7.46 5.40 0.3 0.0 −0.3 221 344
RS81 1 2.38 152.4 88.9 8.1 4.78 0.1 −0.4 −0.8 200 312
RS81 2 0.65 154.0 90.1 7.0 5.16 −0.3 0.2 0.2 215 295
UB152 × 89 × 16/(81)
RS81 3 1.99 154.7 90.5 7.9 4.54 −0.5 −0.6 −0.8 208 332
RS81 4 6.29 154.1 89.9 7.1 5.16 −0.9 0.2 0.1 215 305
RS97 1 4.08 127.9 76.2 7.09 4.54 −0.4 −0.6 −0.1 200 320
RS97 2 4.16 128.3 76.8 7.2 4.36 −0.3 0.0 0.0 200 320
UB127 × 76 × 13/(97)
RS97 3 1.54 127.8 76.0 7.10 4.54 −0.2 −0.2 −0.8 200 316
RS97 4 1.65 129.9 76.4 7.94 4.54 −0.2 −1.0 −0.8 200 316
Table 2
Column test results
Slender. ratio λ Column no. P0 (kN) kc (kN/mm) kl (kN/mm) βl test (min)
tcr RK (min)
tcr eTRK
Mean 0.152
eccentricities e, section dimensions (width b, height H , flange where kc = E 020 A/lc denotes the column axial stiffness at
thickness t f and web thickness tw ) and initial crookedness room temperature and kl is the axial restraint stiffness. In the
along the column length measured prior to testing (i 1 , i 2 and computation of kc , the parameters E 020 , A and lc denote elastic
i 3 denote imperfections at L/4, L/2 and 3L/4, respectively) modulus at room temperature, measured cross-sectional area
are shown in Table 1. In the nomenclature, the 2-digit number and column effective length, respectively.
Column RS45 4 was not tested because it was not feasible
following RS (for restraint steel) denotes the respective column
to achieve the required restraint level for the stocky column
slenderness ratio about the minor axis. The 1-digit number based on the existing test rig. Material tests were conducted on
following the underscore refers to columns subjected to an axial coupons at ambient temperature and measured values of elastic
restraint ratio βl which is defined in Eq. (1). With the number modulus and yield strength are shown in Table 1.
increasing from 1 to 4, βl increases from zero to around 0.16
(see Table 2). 2.3. Initial crookedness and load eccentricity measurement
Initial out-of-plane deflections of specimens were measured
βl = kl /kc (1) by the scanning system as shown in Fig. 2(a), which consisted
K.-H. Tan et al. / Engineering Structures 29 (2007) 1641–1652 1645
Fig. 2. (a) Out-of-plane crookedness measurement set-up. (b) Initial crookedness of RS97 1.
the significant P–δ effect on the columns before buckling. for a higher axial restraint, the build-up of induced column
However, there are a few unexpected test results. internal force P becomes faster. This in turn leads to a lower
The behaviour of Column RS81 2 is a departure from the buckling time and lower associated thermal expansion.
norm. From Fig. 11(c), RS81 2 with βl = 0.074 survived As discussed in Section 4.2, the test result of Column
longer than RS81 1 without restraint. Here the column failure RS81 2 is outside expectation. At the end of heating, the
time tcr is defined as the moment when the column internal displacement at the column end was retarded due to the
axial force P decreases to its beginning value P0 . This is presence of end friction and lower force N exerted on the
because compared to the other columns in the RS81 series, column. In contrast, other columns in the same series contacted
the actual external force N applied on RS81 2 was lower. abruptly (Fig. 12(c)).
Moreover, the end friction has greater effect on this column by A general trend between the axial restraint ratio and column
delaying the time to collapse. Fig. 12(c) shows that beyond 57 failure time tcr can be observed from the curves plotted in
min of heating, the rapidly growing u was retarded at the end of Figs. 11 and 12 and Table 2 (detailed values of tcr ). That is,
heating on RS81 2. This can explain why in Fig. 13(c) there is for columns of the same slenderness ratio λ, an increase in
a noticeable plateau for the column internal actual force P on axial restraint ratio reduces their tcr . For RS97 series columns,
RS81 2. It should be noted that the difference between P and N Table 2 shows that their failure time tcr reduces from 73.2 to
(external loading) was contributed by the transverse steel beam 38.7 min when βl increases from zero to 0.155.
and was proportional to the column end movement u. That is,
4.4. Effect of axial restraint
P − N = kl (u − u slk ) (2)
where kl is the transverse beam stiffness, u is the longitudinal The development of column internal axial force P will be
expansion of the heated column (Fig. 16), and u slk the examined in this section to study the axial restraint effect.
mechanical slack (constant). Clearly, Eq. (2) illustrates that if u Fig. 13(a) to (d) show the relationships between P/P0 and time
does not change, the beam reaction P − N will be kept constant. for all columns, where P0 denotes the measured value of P at
the beginning of heating (the subscript ‘0’ denotes t = 0).
4.3. Axial deformation The axial compressive force from the transverse restraint
beam built up in the columns when they are heated up. At the
Fig. 12(a) to (d) show the column axial displacement-versus- beginning of every elevated temperature test, P/P0 started from
time relationships for all specimens. Generally, with an increase unity. As temperature increased, the restraint force developed
in axial restraint, the maximum axial displacement attained and P/P0 increased in value. After some time, P/P0 reached
decreases. This indicates an inverse relationship between a maximum as the restraint force approached a peak value.
maximum axial deformation and axial restraint. This is because This was followed by an abrupt drop of P/P0 , implying
1650 K.-H. Tan et al. / Engineering Structures 29 (2007) 1641–1652
spring stiffness kl and mechanical slack are considered in the contrast, for more stocky columns, i.e., Series RS45 and RS55,
analyses. Nevertheless, the friction effect is not accounted for Fig. 17 shows that tcrRK is greater than t test . This can be due to the
cr
and the temperature distribution along and across a column fact that EC3 model significantly over-predicts the steel yield
section is assumed to be uniform. The failure temperatures are f yT at the range 400–500 ◦ C (see Fig. 14 in [12]), where Series
1652 K.-H. Tan et al. / Engineering Structures 29 (2007) 1641–1652