Fire Safety Journal 36 (2001) 679691

Structural performance of rotationally restrained

steel columns in re
Faris Ali*, David OConnor
Fire Research Center, University of Ulster, Jordanstown, Co. Antrim BT37 0QB, UK
Received 24 August 2000; received in revised form 8 December 2000; accepted 28 March 2001

Abstract

The paper represents the outcomes of a parametric experimental investigation on the

performance of rotationally restrained steel columns in re. This experimental programme is a
part of major research project performed at the Fire Research Center, University of Ulster in
collaboration with the University of Sheeld. As a part of steel frames, half scale steel
columns were tested in re under two values of rotational restraint 0.18 and 0.93 and one value
of axial restraint of approximately 0.29. Each case of rotational restraint was tested under ve
loading levels 0, 0.2, 0.4, 0.6 and 0.8. The paper includes a comparison with the behavior of a
steel column previously tested in re under axial restraint only. The paper represents also a
method of estimating the eective length of xed end (partial xity) columns tested under re.
r 2001 Elsevier Science Ltd. All rights reserved.

Keywords: Steel columns; Steel frames; Fire; Rotational Restraint; Eective length

1. Introduction

When a re takes place in a building and remains contained in a section of the

building other unexposed parts remain relatively cool. This variance in thermal
deformations imposes restraint on the structures elements under re. In columns,
the imposed restraint can be axial and rotational. Axial restraint opposes thermal
axial expansion of the column while rotational restraint resists columns end
rotation. The imposed restraint can generate substantial unforeseen forces in the
columns during re adding another hazard that may cause uncalculated structural
failure and increased losses and casualties. Eect of axial restraint on steel columns
exposed to re was investigated over the last twenty years by many researchers

*Corresponding author. Tel.: +44-28-9036-8302; fax: +44-28-9036-6826.

E-mail address: f.ali@ulst.ac.uk (F. Ali).

0379-7112/01/$ - see front matter r 2001 Elsevier Science Ltd. All rights reserved.
PII: S 0 3 7 9 - 7 1 1 2 ( 0 1 ) 0 0 0 1 7 - 0
680 F. Ali, D. OConnor / Fire Safety Journal 36 (2001) 679691

[1,46,810] and others. These studies added a vital database to the literature both in
the theoretical and the experimental elds of research. However, in practice cases of
pure axial restraint can rarely be found. In most cases, in real structures rotational
restraint conjuncts axial restraint. This point raises concerns about the necessity of
studying the performance of steel columns subjected to both axial and rotational
restraint during re. Despite the fact that this issue has been studied theoretically by
many researchers including [3,4,7], experimental investigation has been poor.
Therefore a parametric investigation was conducted at the Fire Research Center,
University of Ulster, UK (a joint research project with the University of Sheeld,
UK) on the behavior of axially and rotationally restrained steel columns under high
temperatures focusing on the forces generated during re. Two parameters were
involved in the study: the degree of rotational restraint and the loading level imposed
on columns. Two extreme cases of rotational restraint were chosen, very high and
very low.
This paper presents a parametric experimental investigation of steel columns
restrained axially and rotationally during re. The paper discusses the main
outcomes of the research and demonstrates measurements of generated forces,
temperatures, lateral and axial displacements of the columns. It also includes a
comparison with the behavior of a steel column previously tested in re under axial
restraint only. The paper represents a method of estimating the eective length of
xed end columns (partial xity) tested under re.

2. Rotational restraint

The degree of rotational restraint r imposed on a column is dened as the ratio

of the rotational stiness of the surrounding (to the column) structure rs to the
summation of rotational stiness of the column rc and the structure rs :
rs
r : 1
rs rc
According to Eq. (1) the following boundary conditions can be applied:
when rs crc ; then rE1-the column end is fully xed
when rs E0; then r 0-the column is pin ended.
In the experimental investigation which has been carried out, two degrees of
rotational restraint were involved. Low value of restraint r1 0:186 and high
restraint r2 0:936 (see Table 1). The rotational stiness was worked out according
to BS 5950, Appendix (E) where a value of column rotational stiness=256 kN  mm
was calculated. The rst restraint value r1 0:186 was achieved by connecting two
steel plates 200  10 mm (cross section) and 750 mm length to the top and the bottom
of the column specimen (Fig. 1). The second degree of rotational restraint r2 0:936
was produced by connecting two heavier steel plates of 200  40 mm2 (cross section)
and the same length to the top and the bottom of the column. The two steel plates
were connected to steel blocks using light bolts to simulate simple supports. The
resulted steel frame is shown in Fig. 1.
 F. Ali, D. OConnor / Fire Safety Journal 36 (2001) 679691 681

Table 1
Parameters involved in the testing program of rotationally restrained columnsa

Degree of Applied rot. Column ref., loading levels and loads

rot. restraint r restraint rs
kN mm

Group 1 P3UB1 P3UB2 P3UB3 P3UB4 P3UB5

0.936 3746 0 0.2 0.4 0.6 0.8
(H.R.R) (0) (97 kN) (145 kN) (205 kN) (284 kN)

Group 2 P3UB6 P3UB7 P3UB8 P3UB9 P3UB10

0.186 58.5 0 0.2 0.4 0.6 0.8
(L.R.R) (0) (97 kN) (145 kN) (205 kN) (284 kN)
a
H.R.R.=High rotational restraint, L.R.R.=low rotational restraint; for r see Eq. (1).

3. The experimental programme

In the research programme 10 columns 127  76UB13 were tested. All the columns
were half scale with 1800 mm length. Table 1 shows the parameters involved in the
investigation. As mentioned before, two degrees of rotational restraint were tested.
Each degree was tested under ve loading levels from 0% to 80% of the estimated
compressive design load BS5950, Part 1. All the columns were subjected to an axial
restraint degree of approximately 0.29 (imposed axial stiness E57 kN/mm) [1,2].
The test programme was designed to be compatible with previously performed tests
on columns with axial restraint only. Details of these tests including the rig and
instrumentation can be found in references [1,2]. The rig used in the tests is the same
as that used previously in the programme of tests on the axially restrained columns
[1,2]. The rig was modied in order to impose rotational restraint in conjunction with
axial restraint. The furnace used in these tests is the same furnace that was used in
reference [1,2] with some alterations. The rig and the furnace layout are shown in
Fig. 2. The top and the bottom plates of the steel frame were located outside the
furnace (see Fig. 2) which allowed heating of the column only, leaving other parts of
the steel frame at room temperature. The axial restraint imposed on the column was
produced by the rigs own stiness (see Refs. [1,2]). The heated length of the column
was E1750 mm.
All the tests were performed following the same procedure to ensure repeatability.
The axial load was rst applied to the column by 5% load increments until the
desired level was reached and this load was kept constant during the re test. The
applied loads were measured using two load cells of capacity 200 kN each. The nuts
responsible for imposing axial restraint were then tightened to apply the axial
restraint (Fig. 2), [1,2]. Then the re was ignited until failure took place by losing
stability (column buckling). All the columns were heated in the same rate. Typical
timetemperature curves for the furnace (average of three points top, bottom and
middle) and the steel column are shown in Fig. 3. The curve representing the column
temperature is the average temperature of 20 thermocouples distributed on four
levels over the column length at height 300, 600, 1200 and 1500 mm. At each level
682 F. Ali, D. OConnor / Fire Safety Journal 36 (2001) 679691

Fig. 1. Steel frame involved in the tests.

ve thermocouples were used (see Fig. 2). Four of the thermocouples were attached
to the column anges and one to the center of the web as shown in Fig. 2. During the
re test the column expansion was monitored and recorded using LVDTs (Linear
Variable Displacement Transducer) at the top and the bottom of the column. The
lateral displacements were also recorded using LVDTs on both sides at the middle of
the steel column. During the tests, the axial forces generated in the columns (due to
the imposed restraint) were measured using two load cells (capacity 400 kN each). A
reasonably good temperature distribution along the column length was achieved.
The maximum dierence between the average temperature at the top and the bottom
of the column was between 10 and 15%. Curves representing the temperature prole
along one of the heated columns are shown in Fig. 4.

4. Test results

All the tests showed that imposing axial and rotational restraint during the re
causes additional forces to be generated in the column. Fig. 5 shows curves of
 F. Ali, D. OConnor / Fire Safety Journal 36 (2001) 679691 683

Fig. 2. Rig, furnace and instruments used in the tests.

applied and generated forces, lateral and axial displacement of column reference
P3UB4 tested under high rotational restraint (r 0:936) and loading level=0.6 of
the design load (see Table 1). From Fig. 5 it can be seen that after smooth and low
rate of generation, the axial force reached a maximum value of 143 kN and then
started to decrease gradually as the mechanical properties of the column deteriorates
with time and rise of temperature. At the failure stage, the lateral displacement of the
column begins to increase slowly as the column starts losing its stability for a period
of time and then a sudden column buckling takes place as shown in Fig. 5. This can
684 F. Ali, D. OConnor / Fire Safety Journal 36 (2001) 679691

Fig. 3. Typical re curves adopted in the tests.

Fig. 4. Temperature prole in the four levels along the column.

be attributed to the fact that the rig does not allow for load redistribution beyond the
moment where the restraint force comes to zero. At this stage the buckling process
induces end moments in the steel plates at both ends of the column. These moments
are resisted by the rotational restraint imposed by the connected steel plates. In cases
where thin steel plates were used in the test (small rotational restraint) the end
moments were high enough to bend the thin plates in the pattern shown in Figs. 1
 F. Ali, D. OConnor / Fire Safety Journal 36 (2001) 679691 685

Fig. 5. Curves of applied and generated forces, lateral and axial displacements of column P3UB4 tested
under rotational restraint.

and 6. The plates did not suer any bending when heavy steel plates were used. The
curves shown in Fig. 5 are typical of the curves for all the columns tested. The force
generation and displacement curves had nearly the same trend in all the tests but
they diered in the values and their rates according to the test parameters. Curves
representing the evolution of the generated axial forces for the ten tests are shown in
Figs. 7 and 8 respectively. For comparison, the results of a column tested under axial
restraint only [2] are shown in Fig. 9. When the column behavior shown in this gure
is examined it can be seen that restraint forces generated in a column which is only
restrained axially have a trend similar to that in the case of rotationally restrained
column (Figs. 5, 7 and 8). The main dierences are in the values of these forces and
686 F. Ali, D. OConnor / Fire Safety Journal 36 (2001) 679691

Fig. 6. Deformation in the lower part of the steel frame after the re test (when thin plates were used).

Fig. 7. Axial forces generated in columns tested under high rotational restraint.

in the failure process. A column with only axial restraint had a sudden failure when
the lateral displacement increased suddenly and swiftly due to buckling (Fig. 9).
However, when rotational restraint is present the force generated starts to decrease
(after the peak point) at a slow rate as shown in Fig. 5. Then a second stage of
 F. Ali, D. OConnor / Fire Safety Journal 36 (2001) 679691 687

Fig. 8. Axial forces generated in columns tested under low rotational restraint.

sudden failure takes place when the lateral displacement suddenly increases and
column buckling occurs. However, a sudden failure pattern can still be seen in
rotationally restrained columns (Fig. 8) in cases where the applied loads are
signicantly higher than the generated forces (columns P3UB8, P3UB9 and
P3UB10). Table 2 shows failure temperatures and the maximum generated forces
recorded during the tests. By examining Table 2 it can be concluded that increasing
the load level decreases the maximum value of the generated forces. The other
important conclusion is that changing the degree of rotational restraint between two
extreme values did not signicantly aect the values of the generated forces.
Moreover, Table 2 shows that the generated forces could increase the total imposed
load to dangerous levels, which may exceed the columns design load.

5. Columns eective length

There is no methodology to determine the eective length of partially xed end

columns during the tests, but the shape of the cooled column (after test) gave the
possibility of calculating the columns eective length. By using the geometrical data
of the deformed column, the positions of contra-exure points were found. As shown
in Fig. 10 the distance between these two points represents the eective length of the
column. The eective lengths of the tested columns obtained in this way are shown in
Table 3. It is clear from Table 3 that columns with higher rotational restraint value
had shorter eective lengths.
688 F. Ali, D. OConnor / Fire Safety Journal 36 (2001) 679691

Fig. 9. Curves of applied and generated force, lateral and axial displacement of column 15UB13 tested
under axial restraint only [1].

6. Conclusions

1. Adding rotational restraint had a relatively minor eect on the value of generated
restraint forces but failure temperatures were greatly increased under the same
load.
2. Changing the value of the applied rotational restraint has an insignicant eect on
the values of generated restraint forces.
 F. Ali, D. OConnor / Fire Safety Journal 36 (2001) 679691 689

Table 2
Summary of the results of re tests

Column Loading Failure Max. force Total load Applied

ref. level temp. 1C generated in
columns (kN) Column strength Generated

Group one, High P3UB1 0 648 260 1.06 0

rotational restraint
P3UB2 0.2 589 220 1.10 0.44
P3UB3 0.4 525 179 1.13 0.81
P3UB4 0.6 421 142 1.18 1.44
P3UB5 0.8 277 69 1.08 4.11

Group two, low P3UB6 0 652 290 1.19 0

rotational restraint
P3UB7 0.2 509 256 1.25 0.37
P3UB8 0.4 379 192 1.19 0.75
P3UB9 0.6 271 143 1.18 1.43
P3UB10 0.8 200 97 1.19 2.92

Fig. 10. Measured deformed shape of column P3UB8.

3. The generated forces could increase the total imposed load to dangerous levels
which may exceed the columns design load.
4. Increasing the load level from 0.2 to 0.8 caused a signicant drop in the generated
restraint forces by up to 45%.
690 F. Ali, D. OConnor / Fire Safety Journal 36 (2001) 679691

Table 3
Eective lengths of the tested columnsa

Column ref. Eective length

P3UB1 N/A
Group one (high rotational restraint) P3UB2 0.56L
P3UB3 0.57L
P3UB4 0.58L
P3UB5 0.52L

Group two (low rotational restraint) P3UB6 N/A

P3UB7 0.61L
P3UB8 0.60L
P3UB9 0.65L
P3UB10 0.61L
a
(L=Column length=1800 mm).

5. The failure stage in rotationally restrained columns has no sudden drop in the
generated restraint force as in columns restrained axially only (excluding the cases
mentioned in 6 below).
6. A sudden failure pattern was noticed in few cases where the applied load is
signicantly higher than the generated forces.
7. Increasing the loading level also caused a signicant drop in the failure
temperature.
8. Calculating the eective length of steel columns (using the geometrical data of the
cool columns) gave an average value of 0.56L for highly restrained columns and
0.61 for low rotational restraint value.

References

[1] Ali FA, Shepherd P, Randall M, Simms IW, OConnor DJ, Burgess I. The eect of axial restraint on
the re resistance of steel columns. J Constructional Steel Res 1998;98:177.
[2] Ali FA, Simms I, OConnor D. Eect of axial restraint on steel columns behaviour during re.
Proceedings of the Fifth International Fire Safety Conference, Melbourne, Australia, 1997.
[3] Bailey CG. Simulation of the structural behavior of steel framed buildings in re. Ph.D Thesis.
University of Sheeld, 1995.
[4] Bailey CG, Wadee M, Baltzer K, Newman GM. The behavior of steel columns in re. Report RT524,
Version 1, Submitted by The Steel Construction Institute to the Department of Environment, UK,
March 1996.
[5] Cabrita-Neves I. The critical temperature of steel columns with restrained thermal elongation. Fire
Safety J 1995;24:21127.
[6] Franssen J-M. Axially restrained columns tests database. Report of The University of Liege, Belgium,
1994.
[7] Franssen J-M, Dotreppe JC. Fire resistance of columns in steel frames. Fire Safety J 1992;19
(23):15975.
 F. Ali, D. OConnor / Fire Safety Journal 36 (2001) 679691 691

[8] Lennon T. Local column heating. Proceedings of the First Cardington Conference Building Research
Establishment, 1994.
[9] Pettersson O, Witteveen J. On the re resistance of structural steel elements derived from standard re
tests or by calculation. Fire Safety J 1979;2:7387.
[10] Valente JC, Carbita N. Fire resistance of steel columns with elastically restrained axial elongation and
bending. J Constructional Steel Res 1999;52:31931.