Professional Documents
Culture Documents
1 s2.0 S0141029623012737 Main
1 s2.0 S0141029623012737 Main
Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct
A R T I C L E I N F O A B S T R A C T
Keywords: The model fire test of reinforced concrete components typically requires a furnace temperature-rising curve that
Fire testing is significantly higher than that of the prototype, according to traditional similarity theory. However, this
Model testing requirement can sometimes render the model fire test infeasible due to surpassing the heating capacity of the
Reinforced concrete columns
furnace. To address this challenge, the present study conducts experiments and finite element analysis (FEA) on
Similarity
Fire resistance
geometrically similar, reinforced, square concrete columns of various sizes exposed to fire from two or three
sides, using the ISO 834 standard fire curve for all columns. The fire resistance results from experiments and FEA
are utilized to derive similarity relationships for the time scale of the model to prototype columns. The similarity
in the thermal–mechanical response of the model and prototype columns under the proposed time scale is
evaluated and validated using experimental data from this study and previous research. The findings indicate
that good similarity between the mechanical response and fire resistance of the model and prototype columns can
be attained as long as the model and prototype columns exhibit similar average sectional temperatures and
similar temperatures near the rebar. By normalizing the time scale using s1.19 for 2-sided heating and s1.29 for 3-
sided heating, where s represents the geometric scale factor of model to prototype, the data from the model
column under standard fire effectively predicts the axial deformation development and fire resistance of the
prototype column. Although there is a certain deviation in the temperature near the fire exposure surface and
concrete center of the model and prototype columns under the proposed time scale factor, this discrepancy does
not significantly affect the similarity of their mechanical response.
* Corresponding author.
E-mail address: bowu@scut.edu.cn (B. Wu).
https://doi.org/10.1016/j.engstruct.2023.116858
Received 30 March 2023; Received in revised form 28 August 2023; Accepted 3 September 2023
Available online 9 September 2023
0141-0296/© 2023 Elsevier Ltd. All rights reserved.
H. Zhang et al. Engineering Structures 295 (2023) 116858
Ng reported [3] the results of fire tests on two 1:2.23 and 1:3 scale Adjusting the temperature-rise curve for the model fire test can
models of reinforced concrete (RC) columns, and compared the model provide a good thermal similarity between the model and prototype RC
column fire responses under time-scaled fire curve with that of full-scale members, but increase the difficulty of fire tests, even sometimes
columns under ASTM standard fire curve. Due to the fact that the times requiring a heating rate that exceeds the capacity of the furnace. If
of the fire curves for the models were scaled by s2, a heating rate far similarity between the fire resistance of prototype and model can be
higher than that of the prototype is required for the model. For example, maintained using the same temperature-rise curve (Fig. 1 (b)), model
the furnace temperature reached by the prototype in 60 min has to be fire testing is more practical, although in this case the temperatures of
reached by the 1:2 scale model in 15 min, as shown in Fig. 1 (a). The the prototype and the model may be not completely similar. Reddy
intense fire for the 1:3-scale model column exceeded the furnace’s ca simulated numerically geometrically-similar square columns exposed to
pacity, resulting in some discrepancies between the experimental a standard fire on all four sides [9], and endorsed a relative fire duration
furnace temperature curve with the theoretical curve, and thus the fire to relative size relationship of s1.46 for square columns, based on the
endurance of the prototype predicted from the 1:3-scale model test data criteria that of the temperatures at every geometrically equivalent layer
using that time scaling (s2) had to be corrected. Consequently, it can be being the same or as close as possible, and the average temperatures in
concluded that Ng’s methodology is not applicable in cases where there the cross-sections with different dimensions being the same. However,
is a substantial disparity in scale between the model and the prototype. Reddy’s concept was not experimentally validated. If this approximate
O’Connor also derived the s2 scaling between prototype and model (average) thermal similarity criteria can be applied to model fire test,
through dimensionless analysis [4]. However, he reported that apart the difficulty of model fire tests will be greatly reduced, by adopting the
from scaling the time axis using s2, the temperatures of the fire curves for same fire curve for both full- and reduced-scale RC members.
the model fire tests must be increased due to the contributions from The present investigation involved conducting fire tests and nu
thermal convection and radiation to the heat transfer. He proposed a merical simulations on columns of full and reduced scales, which were
purely convective model, a purely radiative model and a combined exposed to non-uniform heating (3- and 2-sided fire exposure) and
model to enhance the furnace temperature of the model fire tests [5]. subjected to ISO 834 standard fire. The study aimed to establish a sim
Among the three models, the purely radiative model was strongly rec ilarity relationship between the fire resistance of different scale col
ommended (Fig. 1 (a)) for the case of using small furnace, in which the umns, and subsequently evaluate the similarity of the thermal and
furnace wall was so close to the specimen that the burner system pro mechanical responses during the fire. The study’s results offer a practical
vided little forced convection in the furnace environment [4–6]. model fire testing method, which can accurately predict the fire resis
O’Connor’s tests on small-scale brick walls, reinforced concrete slabs tance of RC columns at full scale and mitigate the need for costly large-
and steel columns [5,6] verified his theories. scale fire testing.
Lv applied O’Connor’s purely radiative model in his fire tests on
reduced-scaled model RC beams [7]. To reduce the heating rate 2. Experimental program
requirement in model fire tests, a designed fire curve lower than ISO 834
standard fire curve (Fig. 1 (a)) was used for the prototype beam. The test 2.1. Test specimens
results showed that good similarity between the temperatures at each
normalized position in the model and prototype beams could be ach The present study’s fire tests focused on six square RC columns,
ieved with s2 time scaling, but the similarity relationship between fire constructed at scales of 0.5, 0.7 and full scale. They were exposed to fire
resistance of model and prototype beams was better modified to s1.7 or on two or three sides. The specimens were denoted as 0.5-S2, 0.5-S3,
s1.8, due to size effect and nonlinear mechanical behavior which devel 0.7-S2, 0.7-S3, 1.0-S2 and 1.0-S3, where the number before the hyphen
oped in the beams. Zhuang conducted similar fire tests and performed is the geometric scale and number after the hyphen refers to the number
numerical simulation of full-scale and model-scale RC columns exposed of sides exposed to fire. Prototype columns (1.0-S2, 1.0-S3) measured
to fire on all four sides [8]. His data supported an s1.8 relationship for 350 mm × 350 mm in cross-section and were 3810 mm in height, with
time scale of fire resistance as long as no significant concrete spalling 2600 mm exposed to the fire. The longitudinal reinforcement consisted
occurred during the fire exposure. However, in Lv’s and Zhuang’s fire of four ϕ20 deformed rebars with a yield strength of 437 MPa. The
tests, the fire resistance of the prototype members predicted from the stirrups were ϕ10 plain rebar spaced 200 mm in the middle 2400 mm
models is under a reduced design fire curve, rather than a standard and 100 mm apart at each end. The concrete cover over the stirrups was
temperature-rise curve. 30 mm thick. The geometric dimensions of the model columns are 0.7
2
H. Zhang et al. Engineering Structures 295 (2023) 116858
and 0.5 times of that of the prototype columns. The longitudinal rein 2.3. Test procedure
forcement ratio in the model-scale columns was almost same (0.54–0.56
%), and the volumetric stirrup ratios were also almost equal (1.08–1.12 All of the columns were subjected to an axial load equal to 35 % of
% in the middle). The thickness of the concrete cover was scaled their bearing capacity at ambient temperature. The bearing capacities
accordingly. The columns’ detailed geometry is summarized in Table 1 were calculated using the actual material strength tested on the day of
while Table 2 outlines the reinforcement details. Fig. 2 provides a the fire test. Table 1 listed the values of the bearing capacities and
clearer illustration of the sizes and reinforcement. applied loads of each column. The load was applied to the column at
All of the columns were fabricated from the same batch of com least 20 min prior to starting the heating and maintained until there was
mercial siliceous aggregate concrete. To minimize spalling during the no further increase in axial deformation. The load was then kept con
fire exposure, normal strength concrete with a designed strength grade stant throughout the test.
of C30 was used. The columns were cured indoors and at ambient All of the columns were exposed to the ISO 834 standard fire. Fig. 5
temperature for 210 days before the testing. The average compressive compares the measured temperature curves with the standard fire curve.
strength of six 150 mm concrete cubes was 36.1 MPa at the time of the The real fire curves of all of the columns followed the standard curve
fire tests. closely.
To determine the concrete moisture content, four 150 mm concrete During the tests, the applied load, upper end displacement and
cubes were dried in an oven set at 105 ◦ C for several days until their temperature were recorded in real-time. According to Chinese national
weight no longer decreased. The reduction in weight was taken to code GB/T 9978.1-2008 [10], when a column’s axial deformation rea
represent the water loss, thereby allowing the average moisture content ches h/100 (where h is the heated length) or the deformation rate rea
to be calculated. Based on the measurement, it was found that the ches 3 h/1000 mm/min, the specimen should be considered to have
average moisture content of the concrete was 2.8 % (66.7 kg/m3). failed. At that point the fire test was terminated and specimen was
During the test, the model-scale column was mounted in the furnace allowed to cool naturally to room temperature.
on a supporting column as shown in (Fig. 3). To ensure that their stiff
ness was significantly greater than that of the specimens, the supporting 3. Test results
columns had a larger cross-section (see Fig. 2 (b)(c)), and were
completely covered with three layers of heat-insulating cloth. This 3.1. Failure mode
approach ensured that any deformation in the supporting columns could
be disregarded when measuring the displacements of the model Of the six columns tested, five, all except 1.0-S2, failed in shear-
columns. compression mode (see Fig. 6 and Fig. 7). Column 1.0-S2 did not fail
During the 2-sided heating tests, adjacent sides of the columns were after 5 h of exposure. The failure of 0.5-S3 occurred near the lower end,
exposed to fire, while the unheated sides were covered with three layers but for the other columns it was near the upper end. Inclined major
of heat insulating cloth that was secured with molybdenum wire to cracks and falling-off of concrete can be observed where the failure
provide effective heat insulation. To isolate the length to be exposed to occurred, exposing the rebar to the fire and resulting in its local buck
the fire, both ends of the columns were also covered with heat insulating ling. Longitudinal bars on all sides bent out at the failure position,
cloth. It should be noted that the ratios of the heated height to the total indicating a small eccentric compression failure. The eccentricity-
height were the same for all of the columns. induced tensile cracks on the unheated sides reported by Jin [11]
were not observed, perhaps because the load ratio was higher in the
current tests. The compressive stress exceeded the tensile stress pro
2.2. Test apparatus duced by eccentricity due to uneven heating.
No major inclined crack was observed in column 1.0-S2 which did
The fire tests were conducted using the vertical component furnace not fail, but some concrete falling-off was observed along the longitu
at South China University of Technology. The furnace can heat and dinal bars on the heated sides. Such falling-off was also observed from
apply compressive loading at the same time, as shown in Fig. 3. The columns 0.7-S3 and 1.0-S3. Because there was no sudden temperature
internal dimensions of the furnace measure 2.5 m × 2.5 m × 3.0 m in increase, the falling-off probably occurred in the final stage of fire
height, and it has a maximum load capacity of 5000 kN. The columns’ exposure as temperature-induced deterioration of the concrete failed to
upper ends were pinned with a loading platen, and the lower ends were resist the expansion of the rebar at high temperatures.
pinned with the ground using steel plates and screws. Fig. 6 and Fig. 7 show that no significant concrete explosive spalling
The columns were instrumented with thermocouples and displace occurred on any of the columns, which the temperature data confirm,
ment transducers. Twenty-seven thermocouples designated as T1 to T27 since there was no sudden increase in temperatures. According to
were installed at 1/4, 1/2 and 3/4 of each column’s height. Thermo Eurocode 2, explosive spalling is unlikely to occur when the moisture
couple T28 (and T29 for 2-sided heating columns) evaluated the insu content of concrete is less than 3 % by weight [12]. The content here was
lating effect of the cloth insulation on the unheated sides. Fig. 4 shows 2.8 %.
the locations of the thermocouples.
Axial deformation was measured at the top of the columns using four 3.2. Thermal response
symmetrically distributed linear variable differential transducers
(LVDTs). Fig. 8 (a) and (b) plot the temperature evolution in the concrete and
Table 1
Summary of the specimen parameters and test results.
Specimen Scale Number of heated Side length Total height Heated length Bearing capacity Applied load Fire resistance
No. sides (mm) (mm) (mm) (kN) (kN) (min)
3
H. Zhang et al. Engineering Structures 295 (2023) 116858
Table 2
Reinforcement in the prototype and model columns.
Specimen Cover thickness Longitudinal rebar Yield strength of the Stirrups Longitudinal reinforcement/stirrups Yield strength of the
scale (mm) (mm) longitudinal rebar (MPa) volumetric ratio stirrups (MPa)
Fig. 2. Geometric sizes and reinforcement of the prototype and model columns.
the longitudinal rebar in the prototype and model columns subjected to columns with different scales shows that the heating mode has a larger
2-sided and 3-sided fire. Thermocouples T1, T10 and T12 were at the influence on small-scale columns than on the prototype columns. For
concrete surface, 1/4 depth away from the surface and the concrete’s example, at the 100th minute of fire exposure, the maximum tempera
center, respectively; and thermocouple T3 was on the longitudinal ture difference at the rebar (T3) between columns heated on three and
rebar, at the mid-height section. two sides is 233 ◦ C for the 0.5-scale column, 198 ◦ C at 0.7-scale and
It shows that under the same fire curves, the surface temperatures 143 ◦ C at full scale.
(T1) of columns with different scales are close, but the inner tempera At 1/4-depth (T10) and at the center (T12) there is a plateau in the
tures of the smaller-scale specimens significantly exceeded those of temperature–time curve at about 100 ◦ C. The plateau at the center of the
larger-scale specimens, consistent with the principle of heat flow input prototype column lasted longer than those of the smaller-scale columns.
and temperature increase. Although the small-scale columns received a That was due to the larger amount of moisture involved and longer
lower total heat flow input when subjected to identical heat flux in migration path to the surface for the water vapor in the prototype
tensities, the reduced mass resulted in more rapid temperature increases column.
within the concrete columns.
Clearly, heating three sides of a column generates higher tempera 3.3. Structural response and fire resistance
ture at a given location than heating only two sides. However, the
temperature difference between 2- and 3-sided heating columns at the Fig. 9 presents the mean axial deformation development of the six
concrete’s center is smaller. Comparing the temperature curves of columns in the fire measured by the four LVDTs on the top of the
4
H. Zhang et al. Engineering Structures 295 (2023) 116858
(c) Insulation of column side (d) Pinned support between model and
supporting column through screws
Fig. 3. The test setup.
columns. As the figure shows, the axial deformation first increased in the Heating more faces gives greater heat input, causing larger thermal
positive (tensile) direction due to thermal expansion and then gradually expansion and faster temperature development, resulting in faster
decreased to negative (compressive deformation) as the high tempera deterioration and shorter fire resistance.
ture degraded the strength of the concrete and steel. When a column But the effects differ at different scales. The fire resistance of the 0.5-
could no longer bear the load, the axial deformation increased rapidly scale column exposed to fire on three sides (107 min) was 36 % less than
and the column reached its fire resistance. The fire resistances of all of that of a similar column with 2-sided heating (166 min). At 0.7-scale,
the columns are listed in Table 1. however, the difference was 33 %. And at full scale it was 28 %. The
Comparing the deformation development of columns with the same heating mode is less important for the fire resistance of larger columns,
scale but different heating modes in Fig. 9 shows that columns heated similar as the effect on the thermal response.
from three sides exhibited more significant expansion, briefer expansion
and shorter fire resistance than the columns heated from two sides only.
5
H. Zhang et al. Engineering Structures 295 (2023) 116858
for modeling the concrete. The bond between the steel bars and the
concrete was defined as an “embedded region”, thus the slip between
steel bars and concrete was neglected. The boundary conditions for both
column ends were se as pinned connections, allowing axial translation
and rotation at the upper end, but only rotation at the lower end. Lie’s
published data [13,14] were used for the specific heat and conductivity
of the concrete and the steel. The thermal expansion was that of Euro
code 2 [12]. A concrete damage plasticity model [15] was used for the
concrete and an elastic-plasticity model was adopted for the steel bars.
The stress–strain relationship for the concrete was that of Guo [16] and
for the steel it was that of Eurocode 2 [12]. The mesh sizes were pro
portional to the columns’ dimensions to keep the mesh quantities of all
of the columns identical. Sequential coupling was adopted for the
thermal–mechanical analysis, with the mechanical analysis based on the
results of the thermal analysis. Mesh and boundary conditions are shown
Fig. 5. Measured and ISO 834 standard fire curves.
in Fig. 10.
Six FEA models were built representing the 6 columns tested. Mesh
4. Finite element modeling and parametric study sizes for 1.0-scale, 0.7-scale, 0.5-scale columns were 43.75 mm, 30.625
mm and 21.875 mm, respectively. Fig. 11 shows the temperature-
4.1. Building and validation of FEA model exposure time curves for thermocouples T1, T3, T10 and T12 on col
umns 0.5-S2, 0.5-S3, 1.0-S2 and 1.0-S3 predicted from the FEA and the
Finite element analysis (FEA) was used to extend these observations experimental observations. Apart from the measure point T1 at the fire
to other scales. It used ABAQUS commercial software. Truss elements exposure surface, the calculated curves agree well with the observations.
were adopted to model the steel bars and C3D8R elements were chosen Fig. 12 compares the predicted and observed axial deformations.
6
H. Zhang et al. Engineering Structures 295 (2023) 116858
Based on the fire resistance test results and the FEA, a similarity
relationship quantifying the relative fire resistance of prototype and
model columns can be obtained by regression analysis. This similarity
relationship, as the time scale, should also apply to the development of
the temperature distribution and axial deformation.
4.2. Parametric study where tm and tp are the fire resistances in model and full scale; lm and
lp are their characteristic sizes and x is called the time-similarity index.
The FEA models were used in parametric studies varying the geo Reddy proposed that x is 1.46 with square columns and 4-sided heating
metric scale. 0.6-, 0.8- and 0.9-scale columns, were modeled. Their di [9]. Using the form of Eq. (1) and the data in Table 3, the time-similarity
mensions were proportional to those of the prototype columns, while the indexes for 2-sided and 3-sided heating can be estimated by regressing
7
H. Zhang et al. Engineering Structures 295 (2023) 116858
Fig. 11. Comparison of the temperatures predicted by the FEA and the experimental observations.
Fig. 12. Comparison of the axial deformations predicted by the FEA and the experimental observations.
the fire resistances against the scale. Fig. 13 presents the results for both This implies that the size-induced difference in fire resistance of square
2-sided and 3-sided heating. The regressions suggest indexes for 2- and RC columns decreases with the number of sides heated.
3-sided heating of 1.19 and 1.29. The R2 values are 0.990 and 0.987,
respectively, so the regressed curves fit the data points well, as shown in
Fig. 13. A published curve and data points for 4-sided heating [9] are 5.2. Similarity of thermal response
also plotted. It shows that as the number of heated sides decreases, the
value of the time-similarity index decreases, and the relationship be The similarity relationship of fire resistance is commonly regarded as
tween the relative fire resistance and the relative size tends to be linear. the time scale [2]. Using this similarity relationship to scale the time
axis, the thermal and mechanical response of the model columns can be
8
H. Zhang et al. Engineering Structures 295 (2023) 116858
9
H. Zhang et al. Engineering Structures 295 (2023) 116858
Fig. 14. Normalized temperature –time curves of columns with 3-sided heating.
Fig. 15. Sectional average temperature evolution with 2- and 3-sided heating.
Fig. 18 shows the calculated sectional temperature distribution from deviation are illustrated in Fig. 18 (d) and (h) for 3- and 2-sided heating
FEA on normalized 150th minute for the six tested columns, where the columns, respectively, in which deeper color indicates larger deviation
sectional sizes of columns are also normalized. For each normalized and white color indicates small deviation. From Fig. 18 (d) and (h),
position, the standard deviation in temperatures of three columns with deviations in descending order are deviations at surface (less than
same heating modes can be calculated. The contour plots of the standard 100 ◦ C), near the concrete center (about 50 ◦ C) and at the rebar (less
10
H. Zhang et al. Engineering Structures 295 (2023) 116858
Fig. 17. Axial deformation evolution in Wang’s [17] experiment and predicted using proposed method.
11
H. Zhang et al. Engineering Structures 295 (2023) 116858
Data availability
prototype and model columns was evaluated. Based on the results from This research was supported by National Natural Science Foundation
experiments and analysis, the following conclusions can be drawn: of China (grant No. 51978286) and the Natural Science Foundation of
Guangdong Province, China (grant No. 2020A1515010728). That sup
1. Achieving a completely similar sectional temperature distribution to port is gratefully acknowledged. Any opinions, findings, conclusions or
that of the prototype is not a prerequisite for conducting model fire recommendations expressed in this paper are those of the authors and do
tests. As long as the model and prototype columns have similar not necessarily reflect the views of the sponsors.
average sectional temperatures and similar temperatures near the
rebar, good similarity between the mechanical response of the model References
and prototype can be achieved. Therefore, the same fire curve can be
[1] McGuire JH. The scaling of dimensions in heat conduction problems. Fire Res
used for the model and prototype fire tests. Notes 1954;94.
2. Under the ISO 834 standard fire, the ratio of the fire resistance of a [2] McGuire JH, Stanzak WW, Law M. The scaling of fire resistance problems. Fire
model square RC column to that of a similar full-scale column varies Technol 1975;11(3):191–205.
[3] Ng AB, Mirza MS, Lie TT. Response of direct models of reinforced concrete columns
as about s1.19 for 2-sided heating and s1.29 for 3-sided fire exposure, subjected to fire. ACI Struct J 1990;87.
where s is geometric scale factor. [4] O’Connor DJ, Silcock GWH. A strategy for the fire testing of reduced scale
3. Under the proposed time scale factor, it has been observed that there structural models. Fire Technol 1992;28(1):48–69.
[5] O’Connor DJ, Silcock GWH, Morris B. Furnace heat transfer processes applied to a
is a certain deviation in the temperature near the fire exposure sur strategy for the fire testing of reduced scale structural models. Fire Saf J 1996;27
face and concrete center of the model and prototype columns. (1):1–22.
However, this deviation does not have a significant impact on the [6] O’connor DJ, Moris B, Silcock GWH. A model fire test for parametric testing of half
scale structural components. Fire Saf Sci 1997;5:997–1008.
similarity of their mechanical response. This is because that the outer [7] Lv HR. Similarity in thermal-structural response of reinforced concrete beams with
layer of concrete contributes little to the bearing capacity of the different scales in fire. Master: South China University of Technology; 2019 [in
columns in fire and no significant strength degradation occurs near Chinese].
the concrete center.
12
H. Zhang et al. Engineering Structures 295 (2023) 116858
[8] Zhuang HJ. Similarity of thermal-structural response of reinforced concrete [13] Lie T.T. , Lin T.D. Fire performance of reinforced concrete columns. In: Fire safety:
columns with different scales and influence of concrete spalling under fire. Master: science and engineering 1985.
South China University of Technology; 2020 [in Chinese]. [14] Lie TT. Fire resistance of circular steel columns filled with bar-reinforced concrete.
[9] Reddy DV, Sobhan K, Liu L, Young JD. Size effect on fire resistance of structural J Struct Eng 1994;120(5):1489–509.
concrete. Eng Struct 2015;99:468–78. [15] Lubliner J, Oliver J, Oller S, Oñate E. A plastic-damage model for concrete. Int J
[10] GB/T9978.1—2008. Fire-resistance tests-elements of building construction Part1: Solids Struct 1989;25(3):299–326.
general requirements. Beijing: Standardization Administration of China; 2008. [16] Guo Z, Shi X. Experiment and calculation of reinforced concrete at elevated
[11] Jin J. Fire performance of axially restrained high strength concrete (HSC) columns temperatures. Butterworth-Heinemann; 2011.
in uneven fire boundary. Master: Suzhou University of Science and Technology; [17] Chao W. Study on fire resistance of RC columns with different faces exposed to fire.
2014 [in Chinese]. Master: South China University of Technology; 2006 [in Chinese].
[12] EN1992-1-2. Eurocode2, Design of concrete structures - Part 1-2: general rules -
structural fire design. Brussels: European Committee for Standardization; 2004.
13