ACI STRUCTURAL JOURNAL TECHNICAL PAPER

Title No. 113-S40

Reliability of Temperature-Dependent Models for Analysis

of Reinforced Concrete Members Subjected to Fire
by Fady ElMohandes and Frank J. Vecchio
A computational modeling procedure has been developed for better concrete members. A high degree of sensitivity is illustrated
estimating the behavior of reinforced concrete members subjected through analyzing various reinforced concrete members
to fire. The highly nonlinear procedure of combined thermal and that were tested under fire in combined thermal and structural
structural analysis integrates the transient conductive, boundary- analyses using the models provided by three different codes
convective, and boundary-radiative heat transfer analysis with the
and comparing the analytical results to the experimental
structural analysis of reinforced concrete members. The effect of
results. The codes compared are the ASCE Manual of Practice,1
the models selected for the various temperature-dependent prop-
erties of concrete and steel reinforcement on the overall response the former version of the Eurocode (ENV 1992-1-2:19952), and
of reinforced concrete members with various loading and fire its current version (EN 1992-1-2:20043).
scenarios was studied. For concrete materials, the temperature- The combined thermal and structural analyses are carried
dependent properties investigated and included the density, thermal out using VecTor3,4-6 a finite element analysis computer
conductivity, specific heat capacity, thermal expansion strain, peak program that uses the advanced computational modeling
compressive stress and the corresponding strain, initial modulus procedure presented by ElMohandes.4 On the structural
of elasticity, tensile strength, and the shape of the stress-strain analysis level, the modeling procedure uses the Modified
curve. For steel reinforcing bars, they included the thermal expan- Compression Field Theory (MCFT),7 a well-established
sion strain, yield stress, ultimate stress, and Young’s modulus. In concrete model with corroborated performance and intrinsic
addition to the heat development phase, the cooling phase was
capabilities in modeling shear mechanisms in reinforced
also considered as a fire scenario, where the residual capacity of
concrete. Also, VecTor3 accounts for numerous advanced
concrete columns subjected to fire was investigated.
behavioral mechanisms experienced by reinforced concrete
Keywords: ASCE; cooling; Eurocode; fire; heat; structures; thermal. under various loading conditions. In addition, variations
of the thermal, physical, and mechanical temperature-
INTRODUCTION dependent properties experienced by concrete and steel rein-
Fire causes extensive financial losses due to property forcement when subjected to fire are taken into account.
damage. It also endangers the life and well-being of occu- The combined heat and structural analysis procedure is
pants. This has driven much interest toward the develop- time stepped, where the results are generated at predefined
ment of fire-resistant structures. However, with the low time steps. The procedure, illustrated in Fig. 1, involves two
probability of fire occurrence, prescriptive design concepts distinct steps. At first, the finite element model is subjected
provided by many codes and standards around the world can to a specific temperature-time curve along certain surfaces
lead to substantively uneconomic designs. Also, the neglect to simulate the exposure of the member to fire. The proce-
of the interaction between the different reinforced concrete dure follows a transient conductive, boundary-convective,
members in an assembly or an entire building subjected to and boundary-radiative heat transfer analysis that is solved
fire can lead to unsafe designs. Therefore, performance-based through an iterative finite element scheme on the model
design has increasingly become a major means of achieving level and an iterative finite difference analysis scheme for
safe, yet economic, designs for fire-resistant structures. This the time discretization. The procedure takes into account
is why advanced modeling of reinforced concrete structures the temperature-dependent properties of concrete, including
subjected to fire has been a main focus of researchers and density, thermal conductivity, and specific heat capacity.
structural engineers for many decades. The transient heat transfer analysis provides the tempera-
Advanced modeling of reinforced concrete structures tures at all the nodes comprising the finite element model,
subjected to fire has been treated by researchers in various which are translated to temperatures of the finite elements.
ways. The most efficient way is combined thermal and struc- These element temperatures are then used to determine
tural analysis, which is a time-stepping procedure where the thermal expansion strain and the mechanical properties of
transient heat transfer analysis is performed at certain time the concrete and steel reinforcing bars. The mechanical
intervals and the results are instantaneously used to perform temperature-dependent properties of concrete include the
the structural analysis. thermal expansion strain, peak compressive stress and the
Key elements in combined thermal and structural analysis corresponding strain, initial modulus of elasticity, tensile
are the specific models used to define the values of the ACI Structural Journal, V. 113, No. 3, May-June 2016.
thermal and mechanical temperature-dependent proper- MS No. S-2014-176.R1, doi: 10.14359/51688605, received June 27, 2015, and
reviewed under Institute publication policies. Copyright © 2016, American Concrete
ties of both concrete and steel reinforcing bars. This paper Institute. All rights reserved, including the making of copies unless permission is
presents a study of the effect of making different choices obtained from the copyright proprietors. Pertinent discussion including author’s
closure, if any, will be published ten months from this journal’s date if the discussion
among these models on the overall response of reinforced is received within four months of the paper’s print publication.

ACI Structural Journal/May-June 2016 481

 MODELS COMPARISON SCHEME
Because the combined heat and structural analysis proce-
dure involves two distinct steps, a comparison of the different
models of temperature-dependent properties also needs to be
carried out in two steps. The first comparison involves the
physical and thermal temperature-dependent properties of
concrete that affect the transient heat transfer analysis. These
properties include the density, thermal conductivity, and
specific heat capacity of concrete. The second comparison
involves the mechanical temperature-dependent properties of
both concrete and steel reinforcing bars. For concrete, these
properties include the thermal expansion strain, peak compres-
sive stress and the corresponding strain, initial modulus of
elasticity, tensile strength, and the shape of the stress-strain
curve. For steel reinforcing bars, they include thermal expan-
sion strain, yield stress, ultimate stress, and Young’s modulus.
For a fair comparison of the models defining the mechanical
properties of concrete and steel, the same temperature distri-
bution needs be used for all the structural analyses in which
the different mechanical-properties models are used. Thus,
after completing the comparison of the models affecting the
transient heat transfer analysis, a decision is made regarding
which ones yield the most accurate results compared to the
experimental results. These models are then used for all the
structural analyses carried out afterward. A full list of the
models and assumptions of the properties of both concrete
and steel reinforcement is provided by ElMohandes.4
Three different loading cases and fire scenarios are inves-
tigated through three different experimental series conducted
Fig. 1—Flowchart of time-stepping procedure for combined by the National Research Council of Canada (NRC):
thermal and structural analysis. Case I: Reinforced concrete columns under sustained
strength, and the shape of the stress-strain curve. For steel loading during the event of fire, investigated through the
reinforcing bars, they include thermal expansion strain, experimental series of Columns 10 to 12.8
yield stress, ultimate stress, and Young’s modulus. Case II: Reinforced concrete columns under sustained
The second step in the procedure, repeated during each axial loading during the event of fire, taking into account the
iteration at each time step, is the structural analysis. The effect of the lateral expansion of slabs, investigated through
thermal expansion strain and the updated mechanical prop- the testing of Column 1582.9
erties of the finite elements are used in the structural Case III: Reinforced concrete columns under sustained
analysis, along with all the other external and internal axial loading during the event of fire and through the
loading conditions. The analysis is solved through an iter- cooling phase, investigated through the experimental series
ative finite element scheme. The time-stepping structural of Columns A and B.10
analysis procedure provides a full structural response of the
structure at predefined time intervals, including the failure EXPERIMENTAL SERIES
mode. More details are provided by ElMohandes.4 All three experimental series were conducted on geometri-
cally and structurally identical full-scale column specimens.
RESEARCH SIGNIFICANCE The columns had a 305 mm (12 in.) square cross section and
Due to the lack of standardized fire tests for concrete and a height of 3810 mm (12.5 ft). They were reinforced using
steel as construction materials, the models available in the four 25 mm (1 in.) diameter longitudinal steel bars with a
literature for defining their various temperature-dependent clear cover of 48 mm (1.9 in.), tied using 10 mm (0.4 in.)
properties when subjected to fire show evident scatter. diameter ties at a spacing of 305 mm (12 in.). Figure 2 shows
This paper uses a computational modeling procedure the reinforcement and cross section of a typical specimen.
recently developed by ElMohandes4 to show the effect of For the steel reinforcing bars, only one test was carried out
the variations among these models on the overall response for all the specimens, where the yield stress and the ultimate
of reinforced concrete members subjected to fire. Models strength of the longitudinal bars were reported as 444 and
presented by the ASCE Manual of Practice,1 the former 730 MPa (64.4 and 105.9 ksi), respectively, and the yield
version of the Eurocode (ENV 1992-1-2:19952), and its stress and the ultimate strength of the ties were reported as
current version (EN  1992-1-2:20043) are used to analyze 427 and 671 MPa (61.9 and 97.3 ksi), respectively.
reinforced concrete members tested under various fire The simulation of natural fire conditions in the three
scenarios and different loading conditions. experimental series was done in the Column Furnace

482 ACI Structural Journal/May-June 2016

 at midheight of the column at two diagonally opposite quad-
rants of the square cross section.

NRC COLUMN 1582 TEST

This test, carried out by Mostafaei et al.,9 involved testing
a column specimen for fire resistance assessment under both
axial and lateral loads. This type of loading profile aims at
imitating the loading conditions of columns as parts of build-
ings, rather than individual members. A column in a building
that is exposed to fire would experience differential lateral
displacement at its ends due to the thermal expansion of the
slabs it connects to at its top and bottom, in turn, inducing
significant levels of lateral shear forces.
To estimate the lateral displacement acting on the column
specimen, the commercial thermal analysis program SAFIR14
was used to carry out a finite element analysis for a six-story
prototype building for which a compartment fire scenario
was assumed. The building had six 9 m (29.5 ft) spans in
one direction and four 5 m (16.4 ft) spans in the other. Each
of the six stories was 3.8 m (12.5 ft) high, resulting in a
total height of 22.9 m (75.0 ft) for the entire building. The
compartment selected for the fire scenario was on the first
floor in one of the middle spans of one of the shorter edges.
The compartment was exposed to the CAN/ULC S10111
standard temperature-time curve. Column 1582 was selected
for the experimental testing as a worst-case scenario, with
the maximum axial and lateral load combination.
Fig. 2—Reinforcement and cross section of typical NRC The concrete was mixed with calcareous aggregates and
column specimen. had a compressive strength of 55.0 MPa (8.0 ksi). The spec-
Facility of the National Research Council of Canada (NRC). imen had one group of thermocouples at midheight of the
The tests followed the CAN/ULC S10111 temperature-time column. On the day of testing, the relative humidity at the
model, which is similar to the ASTM E11912 model, for the center of Column 1582 was reported as 72.8%, which is
ascending temperature branch (fire development phase) and equivalent to a moisture content of 3.07% by weight. The
the ISO 83413 model for the descending temperature branch column was loaded to an axial load of 1590 kN (357.4 kip).
(fire decay phase). This amounted to 31.1% of the column capacity and was
applied from the bottom, prior to the start of fire. Rotation
NRC COLUMNS 10 TO 12 TESTS was restrained at both ends.
These three columns were constructed and tested by The setup of the test required that the column’s top and
Lie and Lin.8 On the day of testing, the relative humidity bottom edges be covered by insulation. Thus, only the middle
at the centre of Columns 10, 11, and 12, was reported as 3175 mm (10.42 ft) length of the 3810 mm (12.5  ft) long
75%, 75%, and 76%, respectively, which is equivalent to a column was subjected to fire. The mechanical and thermal
moisture content of 3.16%, 3.16%, and 3.20% by weight, loading setup of Column 1582 is shown in Fig. 3. To allow
respectively. The concrete was mixed with calcareous aggre- for a longer fire exposure duration, the lateral displacement
gates and the compressive strengths were 40.9, 36.9, and applied to the top of the column was approximated and
40.0 MPa (5.9, 5.4, and 5.8 ksi) for Columns 10, 11, and 12, capped at 50 mm (2 in.), following the profile shown in Fig. 4.
respectively. Column 10 was loaded to 800 kN (179.8 kip), At 120 minutes from the start of fire and lateral loading appli-
Column 11 to 1067 kN (239.9 kip), and Column 12 to cation, the maximum lateral displacement of 50 mm (2 in.)
1778  kN (399.7  kip), which amounted to 20.4%, 29.7%, was reached. The column was left exposed to fire under the
and 46.7% of their capacity, respectively, based on a mono- maximum lateral displacement until failure.
tonic loading analysis carried out using VecTor3. The test
involved loading the columns to the target axial load, then, NRC COLUMNS A AND B TESTS
1 hour later, subjecting them to the CAN/ULC S10111 stan- This experimental series, carried out by Lie et al.,10 was
dard temperature-time curve until failure. aimed at assessing the residual strength of reinforced concrete
To monitor the temperature of concrete through the depth columns after exposure to fire. The objective for such an assess-
of the columns, four groups of thermocouples were installed ment is the determination of the feasibility of repair of fire-dam-
at three levels through the height. A group of thermocou- aged structures. Columns A and B were loaded, then subjected
ples was installed at one-fourth of the height of the columns to fire, following the CAN/ULC S10111 standard tempera-
from the top and another at the same distance from the ture-time curve, in the testing furnace for 1 and 2 hours, respec-
bottom. Two other groups of thermocouples were installed tively. They were both allowed to cool naturally (in air at room

ACI Structural Journal/May-June 2016 483

 Fig. 4—SAFIR-estimated and experimental lateral loading
profile for NRC Column 1582.
generated using the models of the respective type of aggregate.
The models provided by the current version of the Eurocode
(EN 1992-1-2:20043), however, do not distinguish between
concrete mixed with different types of aggregate; hence, the
experiment results of all the experimental series are compared
to the analytical results generated using these models.
Fig. 3—Mechanical and thermal loading setup of NRC Another major difference between the models is that the
Column 1582. former version of the Eurocode (ENV 1992-1-2:19952)
and the current version (EN 1992-1-2:20043) recognize the
temperature) for approximately 24 hours until they reached effect of the moisture content of concrete on its specific heat
room temperature and then were axially loaded until failure. capacity, while the ASCE Manual of Practice1 does not. A
The concrete was mixed with siliceous aggregates. On moisture content of 4% by weight is chosen for concrete
the day of testing, the compressive strength was 38.9 MPa in all the analyses. This moisture content is a reasonable
(5.6 ksi) for Column A and 41.8 MPa (6.1 ksi) for Column B. estimate for concrete at a relatively young age and closely
The columns had one group of thermocouples at midheight. matches the values measured on the days of testing.
The relative humidity at the center of Columns A and B was In all the analyses, for concrete, the convective heat transfer
reported as 87% and 82%, respectively, which is equiva- coefficient and the emissivity (for the radiative heat transfer
lent to a moisture content of 3.66% and 3.45% by weight, coefficient) were chosen as 25 W/m2°C and 0.7, respectively,
respectively. Both columns were loaded axially 1 hour prior according to the recommendations of the current version of
to the fire test. Column A was loaded to 992 kN (223 kip) the Eurocode (EN 1992-1-2:20043). The initial density of
and Column B to 1022 kN (229.8 kip), which amounted to concrete was taken as 2400 kg/m3 (150 lb/ ft3).
22.0% and 21.4% of their axial capacity, respectively.
HEAT TRANSFER IN CALCAREOUS-
TRANSIENT HEAT TRANSFER ANALYSIS AGGREGATE SPECIMENS
Only one-fourth of the cross section of the column speci- In the experimental series of Columns 10 to 128 and the
mens was analyzed because the section is symmetric in two experimental series of Column 1582,9 calcareous aggre-
directions. The model employed only one element, 25 mm gates were used in the concrete mixture. The results of these
(1 in.) thick, in the longitudinal direction because the heat two experimental series were used to compare the different
transfer analysis is transversely bidirectional through the models defining the thermal properties of concrete mixed
depth of the column. The finite element mesh was chosen with calcareous aggregates at elevated temperatures.
such that there was a node at the location of every thermo- Figures 5 and 6 show the results of the analysis at depths of
couple. The distances of the nodes from the surface of the 25.4 and 152.5 mm (1 and 6 in.), respectively. For Columns
model were 4.4, 6.4, 9.1, 12.8, 17.8, 25.4, 26.7, 38.1, 44.5, 10 to 12, the results shown represent the average measure-
63.5, 71.3, 101.6, 108, and 152.5 mm (0.175, 0.25, 0.36, 0.5, ments of the thermocouples at the same location for the four
0.7, 1, 1.05, 1.5, 1.75, 2.5, 2.8, 4, 4.25, and 6 in.). The element groups installed. More results are presented by ElMohandes.4
discretization distances were identical in both directions. The plots show that the experimental results of
The ASCE Manual of Practice1 and the former version Columns 10, 11, and 12 are relatively similar, whereas those
of the Eurocode (ENV 1992-1-2:19952) provide different of Column 1582 are different. This can be attributed to the
models for the thermal properties of concrete mixed with fact that Columns 10, 11, and 12 were cast and tested as a
calcareous aggregates and concrete mixed with siliceous part of one test series; hence, the same materials and same
aggregates. Therefore, the experiment results of the different test technique were used, while Column 1582 was a part of
experimental series are compared to the analytical results a different series that was cast and tested many years later.

484 ACI Structural Journal/May-June 2016

Fig. 5—Temperature change at depth of 25.4 mm (1 in.) for Fig. 6—Temperature change at depth of 152.5 mm (6 in.) for
NRC Columns 10, 11, 12, and 1582. NRC Columns 10, 11, 12, and 1582.
Also, Column 1582 was subjected to lateral loading, which
caused significant spalling of the concrete cover over the
entire face that was subjected to tension. This would effec-
tively cause the higher temperatures observed in the plots.
Averaging the experimental results of the four different
specimens, the mean analytical-to-experimental ratio for the
temperatures at a depth of 25.4 mm (1 in.) generated by the
models provided by the ASCE Manual of Practice,1 the former
version of the Eurocode (ENV 1992-1-2:19952), and its current
version (EN 1992-1-2:20043) are 1.04, 1.17, and 1.12, respec-
tively, with coefficients of variation of 6.9%, 11.5%, and 11.8%,
respectively. For the temperatures at the centers of the columns,
the mean analytical to experimental ratio is 0.92, 1.02, and 0.94,
for the three models in the same order, with coefficients of vari-
ation of 23.3%, 18.6%, and 18.1%, respectively.
It can be observed that the models provided by the ASCE
Fig. 7—Temperature change at depth of 25.4 mm (1 in.) for
Manual of Practice1 for concrete mixed with calcareous
NRC Column A.
aggregates manage to accurately estimate the temperature of
concrete at a depth of 25.4 mm (1 in.), but not at the centers of fully restored to their pre-fire values when the concrete
the columns where the dispersion is also significantly high. returns to room temperature.
The models provided by the latest version of the Eurocode Figures 7, 8, 9, and 10 also show the analytical estimates
(EN 1992-1-2:20043) are capable of estimating the exper- presented by Lie et al.,10 denoted ‘NRC’. These estimates
imental temperatures to a lesser extent, but consistently at were calculated using the models presented by the ASCE
both depths. However, they manage to capture the kink in Manual of Practice1 in a finite difference method, based on
the plot at 100°C (212°F) caused by the evaporation of the the procedure presented by Lie and Allen15 and Lie et al.16
evaporable water inside the concrete, which is neglected by The experimental results show that the models presented
the models provided by the ASCE Manual of Practice.1 by the current version of the Eurocode (EN 1992-1-2:20043)
are the ones most capable of estimating the experimental
HEAT TRANSFER IN SILICEOUS- results. The average analytical-to-experimental ratios at the
AGGREGATE SPECIMENS two depths presented were 1.10, 1.04, and 1.02 for the three
In the experimental series of Columns A and B,10 siliceous aforementioned standards, respectively. This is based on
aggregates were used in the concrete mixture. The results of the comparison of the maximum temperatures measured at
this experimental series were used to compare the different the various depths through the columns sections with their
models defining the thermal properties of concrete mixed respective values estimated by VecTor3.
with siliceous aggregates at elevated temperatures. However, all the models presented tend to significantly
Figures 7 and 8 show the results of the analysis of Column underestimate the post-peak temperatures, while the NRC
A at depths of 25.4 and 152.5 mm (1 and 6 in.), respectively. computed values overestimate them but to a lesser extent.
Figures 9 and 10 show the results of Column B. More results Because the NRC estimates were based on analyses using
are presented by ElMohandes.4 Because the standards under the models presented by the ASCE Manual of Practice,1 yet
comparison do not present models for the post-fire thermal these estimates differ from the analytical results calculated
properties of concrete, it was assumed that the properties by VecTor3 using the same models, one may assume that the
follow the same models in the cooling phase until they are models are not responsible for this difference. Lie et al.10

ACI Structural Journal/May-June 2016 485

 Fig. 10—Temperature change at depth of 152.5 mm (6 in.)
Fig. 8—Temperature change at depth of 152.5 mm (6 in.) for
for NRC Column B.
NRC Column A.
properties of concrete and steel reinforcing bars. These models
will be used in the structural analysis stage that is undertaken
at the end of the heat transfer analysis stage at each time step.
Because the mechanical properties of the steel reinforcing
bars are very sensitive to temperature, different finite element
discretizations were used. They involved more elements in
the concrete cover area to ensure an accurate estimation of
the temperatures of the bars.
The deformation of the specimens is governed by a balance
between the thermal expansion of steel and concrete, and
the deterioration of their stiffnesses. In general, regardless
of the level of axial loading the specimens are subjected to,
the specimens expand during the first part of the test where
the behavior is mainly dominated by the thermal expansion
strains while the reduction of the stiffnesses of concrete and
Fig. 9—Temperature change at depth of 25.4 mm (1 in.) for steel is not substantial at lower temperatures. In this part,
NRC Column B. one can notice a kink in the time-displacement plot, which is
caused by the yielding of steel reinforcing bars, as the yield
stated that measurements were made of the furnace tempera- stress declines with the increase in temperature.
tures during the fire exposure and the cooling periods until the The expansion reaches a peak after a certain period of
average furnace temperature reached near-ambient tempera- time, and then decreases as time passes. After reaching the
tures, yet they failed to report these temperatures. Therefore, peak expansion, the stiffnesses of concrete and steel are
one may surmise that the reason for the difference between compromised to the extent that the contraction displace-
the analytical and experimental results can be attributed to a ment resulting from the axial loading exceeds the expansion
possible difference between the actual furnace temperatures displacement resulting from the increase in temperature. As
and the ISO 83413 model that was used in the analysis for the time passes and the temperatures of the concrete and steel
descending temperature branch (fire decay phase). increase, their stiffnesses keep declining until a certain point
where their strengths are not sufficient to withstand the axial
COMBINED THERMAL AND loads, and the specimen fails.
STRUCTURAL ANALYSIS
The models provided by the current version of the Euro- NRC COLUMNS 10, 11, AND 12 ANALYSIS
code (EN 1992-1-2:20043) proved to produce the most For the finite element discretization chosen for the anal-
accurate estimation of the temperatures through the depth yses of these columns, only one-fourth of the column section
of concrete, compared to their counterparts in the other two was modeled, taking advantage of the double symmetry of
standards. This applies to concrete mixed with both calcar- the geometrical, structural, and loading setup. The mesh used
eous and siliceous aggregates; hence, it applies to the three was symmetric, with twelve 5.04 mm (0.2 in.) thick elements,
experimental series chosen for comparison. through the depth of the concrete cover to the longitudinal
Therefore, the models provided by the Eurocode (EN 1992- reinforcing bars, and five 18.4 mm (0.72 in.) thick elements
1-2:20043) were used for the heat transfer stage of the for the core region. Figure 11 shows the mesh through a
combined thermal and structural analyses for all the experi- section between the ties, together with the location of the steel
mental series. This aims at isolating the effect of the different reinforcing bar, which is shown as a black circle (not to scale).
models provided by the three standards for the mechanical Longitudinally, 150 elements with a length of 25.4 mm (1 in.)

486 ACI Structural Journal/May-June 2016

 Fig. 12—Vertical displacement of NRC Column 10 from
start of fire to failure.

Fig. 11—Finite element discretization for structural analysis

for cross section of NRC Columns 10, 11, and 12 at section
occurring between ties.
each, were used. The ties were spaced at 305 mm (12 in.); that
is, a tie at every 12 layers of elements along the height. The
axial load was applied at the top of the column at the center of
the cross section. The models were restrained in all directions
at their bases. The axial load was kept constant until failure,
and a time step of 60 seconds was used for the analysis.
Figures 12, 13, and 14 show the experimental results
of the vertical displacement at the top of the columns for
Columns 10, 11, and 12, respectively. The figures also show
the analytical results estimated by VecTor3 using the models
provided by the ASCE Manual of Practice,1 the former Fig. 13—Vertical displacement of NRC Column 11 from
version of the Eurocode (ENV 1992-1-2:19952), and its start of fire to failure.
current version (EN 1992-1-2:20043). A positive displace-
ment means expansion along the height of the column, and a
negative displacement means contraction. For the analytical
results, the displacements shown only represent the change
in height from the start of the fire, neglecting the initial
displacements resulting from the axial load that was applied
an hour prior to the fire exposure.
Table 1 shows the maximum expansion displacement
reached, d, and the time from the start of fire to failure, t,
for Columns 10, 11, and 12. For the maximum expansion
displacement, all the models manage to estimate its value
with reasonable accuracy. A general trend of underestimation
is evident in the case of the Eurocode (ENV 1992‑1‑2:19952),
with a mean analytical-to-experimental ratio of 0.85 and a
coefficient of variation of 20%. With the ASCE Manual of
Practice,1 there is a general trend of overestimation, with
a mean analytical-to-experimental ratio of 1.26 and a coef- Fig. 14—Vertical displacement of NRC Column 12 from
ficient of variation of 14%. The models provided by the start of fire to failure.
Eurocode (EN 1992-1-2:20043), on the other hand, managed
to capture the maximum expansion displacement more start of fire to failure within a reasonable range of accu-
precisely, with a mean analytical-to-experimental ratio of racy, giving an unsafe overestimation with a mean analyt-
0.97 and a coefficient of variation of 2%. ical-to-experimental ratio of 1.25 and a coefficient of vari-
However, the models provided by the Eurocode ation of 8%. The models provided by the ASCE Manual of
(EN  1992‑1-2:20043) do not estimate the time from the Practice1 also overestimate the time to failure with a mean

ACI Structural Journal/May-June 2016 487

Table 1—Experimental and analytical results of
Columns 10, 11, and 12
Experimental ASCE Manual ENV 1992- EN 1992-1-
results of Practice1 1-2:19952 2:20043
d, d, d, d,
mm t, mm t, mm t, mm t,
Specimen (in.) min (in.) min (in.) min (in.) min
11.10 12.15 9.21 11.06
Col. 10 510 517 455 598
(0.44) (0.48) (0.36) (0.04)
7.90 8.94 6.32 7.71
Col. 11 365 411 359 443
(0.31) (0.35) (0.25) (0.3)
2.50 3.65 1.77 2.37
Col. 12 215 295 228 294
(0.1) (0.14) (0.07) (0.09)

analytical-to-experimental ratio of 1.17 and a coefficient of

variation of 16%. The models that manage to capture the Fig. 15—Vertical displacement of NRC Column 1582 from
time of failure more precisely are the ones provided by the start of fire to failure.
Eurocode (ENV 1992-1-2:19952), estimating it with a mean
analytical-to-experimental ratio of 0.98 and a coefficient of The models provided by the former version of the Euro-
variation of 9%. code (ENV 1992-1-2:19952) and the ones provided by its
current version (EN 1992-1-2:20043) tend to estimate the
NRC COLUMN 1582 ANALYSIS time of failure from the start of fire on the conservative side,
The quarter cross section finite element model created for with an analytical-to-experimental ratio of 0.76 and 0.82,
the analysis of Columns 10, 11, and 12 could not be used for respectively. On the other hand, the models provided by the
the analysis of Column 1582. The bending moment ensuing ASCE Manual of Practice1 tend to estimate a fire resistance
from the lateral loading required the entire depth of the that is longer in time than the experimental fire resistance
column to be modeled. Therefore, half of the cross section with an analytical-to-experimental ratio of 1.13.
was modeled, using the same finite element discretization The models provided by the former version of the Eurocode
described previously for Columns 10, 11, and 12. Longitu- (ENV 1992-1-2:19952) seem to generate the most accurate
dinally, the column was discretized into 75 elements with a deformations, with an analytical-to-experimental ratio of 0.96
height of 50.13 mm (1.97 in.) each, to achieve a total height for the maximum expansion displacement, followed by those
of 3810 mm (150 ft) for the model. The 13 ties were modeled provided by the current version (EN 1992-1-2:20043) with an
at 300 mm (11.8 in.), instead of the actual 305 mm (12 in.), analytical-to-experimental ratio of 0.95 for the same displace-
to fit in the selected finite element discretization. This means ment. The models provided by the ASCE Manual of Practice1
that a tie was located at every sixth element along the height. tend to estimate significantly higher expansion displacements,
A time step of 60 seconds was used for the analysis. The with an analytical-to-experimental ratio of 1.68.
axial load of 1590 kN (357.4 kip) was applied prior to the
start of the fire at the top of the column. The exact lateral NRC COLUMNS A AND B ANALYSIS
loading profile applied during the test was applied at the The finite element model created for the analyses of
top of the column in a displacement control loading scheme Columns 10, 11, and 12 was used in the analysis of Columns
until failure. The thermal loading of the fire and the lateral A and B. A time step of 60 seconds was used. The models of
loading were started at the same time, resembling the exper- Columns A and B were loaded to their respective axial load,
imental conditions. then subjected to the CAN/ULC S10111 standard tempera-
Figure 15 shows the experimental results for the vertical ture-time curve for 1 and 2 hours, respectively. Then, the
displacement at the top of Column 1582 compared to the fire was allowed to decay following the ISO 83413 standard
analytical results estimated using the models presented by temperature-time curve for the descending temperature
the ASCE Manual of Practice,1 the former version of the branch (fire decay phase). For the post-fire residual mechan-
Eurocode (ENV 1992-1-2:19952), and its current version ical properties of concrete, the models presented by Chang
(EN 1992-1-2:20043), respectively. et al.17 were used. For steel reinforcing bars, it was assumed
It should be noted that the test results showed an initial that the mechanical properties were fully recovered to their
lateral load of approximately 14 kN (7.2 kip) after applying initial pre-fire values when they cooled to room temperature.
the axial load but before the start of the fire or the appli- It should be noted that the literature presents models for
cation of the lateral displacement. This initial lateral load the mechanical properties of concrete and steel at elevated
might be a result of a slightly eccentric axial loading or some temperatures during the event of fire and other models for
imperfections in the specimen construction or in the loading the residual mechanical properties after the event of fire. Yet,
mechanism. To be able to compare the experimental results there is a void in the literature for the cooling period and the
to the analytical results calculated by VecTor3, this initial path that the mechanical properties follow, going from their
lateral load has been deducted from all the lateral loads values at the maximum reached temperature to their values
reported throughout the test. when they cool to room temperature (residual properties).

488 ACI Structural Journal/May-June 2016

Fig. 16—Vertical displacement of NRC Column A from start Fig. 17—Vertical displacement of NRC Column B from start
of fire to failure. of fire to failure.
Therefore, in this study, for the lack of a better experimen-
Table 2—Experimental and analytical results of
tally proven method, it has been assumed that the mechan-
Columns A and B
ical properties of concrete and steel change linearly moving
between the values mentioned previously. Experimental ASCE Manual ENV 1992-1- EN 1992-1-
Spec- results, of Practice1, kN 2:19952, kN 2:20043, kN
Figures 16 and 17 show the vertical displacements of
imen kN (kip) (kip) (kip) (kip)
Columns A and B from the start of the fire, respectively. The
analytical results estimated using the three sets of models to be Col. A 1987 (446.7) 2684 (603.4) 2686 (603.8) 2330 (523.8)
compared are presented, together with the experimental results Col. B 2671 (600.5) 2700 (607) 2830 (636.2) 2908 (653.7)
and the analytical estimate provided by Lie et al.,10 denoted
‘NRC’. This analytical estimate was derived by using the Table 2 shows the experimentally determined residual
experimentally measured temperatures of concrete at different capacity of the columns and the analytically estimated
depths from the surface to divide the section into zones based values, based on the different models. The analyses that
on the maximum temperature reached. Then, the mechanical were carried out using the models presented by the ASCE
properties of these zones were determined using the available Manual of Practice1 produced a mean analytical-to-
models for residual properties. Finally, a finite element analysis experimental value of 1.18. With the models presented by
was undertaken for a model constructed with these residual the former version of the Eurocode (ENV 1992-1-2:19952),
mechanical properties and the failure load was determined. this value was 1.21. Finally, with the models presented by
While examining Fig. 16 and 17, one should recall that the current version of the Eurocode (EN 1992-1-2:20043),
the temperatures of concrete through the section that were this value came to 1.13. Hence, one can conclude that the
analytically calculated were consistently below the experi- models presented by the current version of the Eurocode
mental ones. This might explain why the analytically esti- (EN 1992-1-2:20043) provide the most accurate results for
mated displacements of Columns A and B decreased at a concrete mixed using siliceous aggregates.
steeper rate than the experimental displacements. Other
reasons may be the limited data used to develop the models SUMMARY AND CONCLUSIONS
describing the post-fire properties and the approximate The finite element structural analysis computer program
procedure adopted to fill the cooling phase properties void VecTor34-6 was used to study the consequences of making
that was explained previously. different choices among the models available for the various
Figure 16 shows that, for Column A, the ASCE Manual temperature-dependent properties of concrete and steel rein-
of Practice1 and the former version of the Eurocode (ENV forcement subjected to fire. Three sets of models from three
1992-1-2:19952) accurately estimate the residual displace- different codes defining these properties were compared,
ment after cooling, while the current version of the Euro- namely, the ASCE Manual of Practice,1 the former version
code (EN 1992-1-2:20043) overestimates it. However, for of the Eurocode (ENV 1992-1-2:19952), and its current
Column B, all models seem to significantly underestimate version (EN 1992-1-2:20043).
the residual displacement, as can be seen in Fig. 17. For both Thermal temperature-dependent properties of concrete
columns, the maximum expansion displacements are over- including density, thermal conductivity, and specific heat
estimated, with the Eurocode (ENV 1992-1-2:19952) esti- capacity, were studied through transient heat transfer analyses.
mating significantly higher displacements. Mechanical temperature-dependent properties of concrete and
The NRC estimate seems to capture the peak expansion steel reinforcing bars were studied through combined thermal
displacements accurately, yet the residual displacements and structural analyses. They included thermal expansion
estimates are highly inaccurate. It is unclear why the NRC strain, peak compressive stress and the corresponding strain,
estimate for Column B estimated that the specimen expanded initial modulus of elasticity, tensile strength, and the shape
again while it was cooling. of the stress-strain curve for concrete; and thermal expansion

ACI Structural Journal/May-June 2016 489

strain, yield stress, ultimate stress, and Young’s modulus for aggregates; NRC: analytical results estimated by the National Research
Council of Canada (NRC).
steel reinforcing bars. Different loading cases and fire scenarios
were analyzed and compared to experimental results.
The results showed that, as a general rule for transient heat AUTHOR BIOS
ACI member Fady ElMohandes is a Structural Engineer at AECOM,
transfer analysis, the models provided by the current version Markham, ON, Canada. He received his PhD from the University of
of the Eurocode (EN 1992-1-2:20043) yielded the most Toronto, Toronto, ON, Canada, in 2013. His research interests include
computational modeling and performance assessment and analysis of rein-
accurate results compared to the experimental results. This forced concrete structures and their behavior under fire and extreme loads.
is a general recommendation for concrete members regard-
less of the type of aggregate used in the concrete mixture. Frank J. Vecchio, FACI, is a Professor of civil engineering at the Univer-
sity of Toronto. He is a member of Joint ACI-ASCE Committees 441, Rein-
However, the models provided by the ASCE Manual of forced Concrete Columns, and 447, Finite Element Analysis of Reinforced
Practice1 managed to estimate the results more accurately Concrete Structures. He received the 1998 ACI Structural Research Award,
than those provided by the current version of the Eurocode the 1999 ACI Structural Engineer Award, and the 2011 ACI Wason Medal
for Most Meritorious Paper. His research interests include advanced consti-
(EN 1992-1-2:20043) for concrete mixed with calcareous tutive modeling and analysis of reinforced concrete, assessment and reha-
aggregates. However, the models it provided for concrete bilitation of structures, and response under extreme loads.
mixed with siliceous aggregates were not as successful.
For the models defining the mechanical temperature-de- REFERENCES
pendent properties of concrete and steel reinforcing bars, 1. Lie, T. T., “Structural Fire Protection,” ASCE Manuals and Reports on
Engineering Practice No. 78, American Society of Civil Engineers, Reston,
the former version of the Eurocode (ENV 1992-1-2:19952) VA, 1992, 241 pp.
yielded the most accurate estimates for the time of failure 2. CEN, “ENV 1992-1-2:1995—Eurocode 2: Design of Concrete
from the start of fire for concrete mixed with calcar- Structures—Part 1-2: General Rules. Structural Fire Design,” European
Committee for Standardization (CEN), Brussels, Belgium, 1996, 63 pp.
eous aggregates. However, for the same type of concrete, 3. CEN, “EN 1992-1-2:2004—Eurocode 2: Design of Concrete Struc-
the models provided by the current version of the Euro- tures—Part 1-2: General Rules. Structural Fire Design,” European
code (EN  1992-1-2:20043) yielded better estimates for the Committee for Standardization (CEN), Brussels, Belgium, 2005, 97 pp.
4. ElMohandes, F., “Advanced Three-Dimensional Nonlinear Analysis
maximum deformation reached throughout the fire. The of Reinforced Concrete Structures Subjected to Fire and Extreme Loads,”
models provided by the ASCE Manual of Practice1 yielded PhD thesis, University of Toronto, Toronto, ON, Canada, 2013, 407 pp.
the least accurate estimates in terms of both the time of 5. ElMohandes, F., and Vecchio, F. J., “Vector3: A. User’s Manual,” Depart-
ment of Civil Engineering, University of Toronto, Toronto, ON, Canada, 2013,
failure and the maximum deformation reached. 81 pp.
For concrete mixed with siliceous aggregates, the models 6. ElMohandes, F., and Vecchio, F. J., “Vector3: B. Sample Coupled
provided by both the ASCE Manual of Practice1 and the Thermal and Structural Analysis,” Department of Civil Engineering,
University of Toronto, Toronto, ON, Canada, 2013, 68 pp.
current version of the Eurocode (EN 1992-1-2:20043) 7. Vecchio, F. J., and Collins, M. P., “The Modified Compression Field
provided reasonable estimates of the maximum deformation Theory for Reinforced Concrete Elements Subjected to Shear,” ACI Journal
reached as opposed to those provided by the former version Proceedings, V. 83, No. 2, Mar.-Apr. 1986, pp. 219-231.
8. Lie, T. T., and Lin, T. D., “Fire Tests on Reinforced Concrete Columns,
of the Eurocode (ENV 1992-1-2:19952). Also, using the Specimens No. 1-12,” 12 NRC Publications—DBR Internal Reports
same set of models for residual properties of concrete and No. 478-489, Division of Building Research, National Research Council of
steel reinforcement bars, the models provided by the current Canada (NRC), Ottawa, ON, Canada, 1983.
9. Mostafaei, H.; Leroux, P.; and Lafrance, P.-S., “Fire Endurance of a
version of the Eurocode (EN 1992-1-2:20043) managed to Reinforced Concrete Column under Both Axial and Lateral Loads,” NRC
give the best estimates of the residual post-fire strength of Report IRC-RR-327, National Research Council Canada (NRC), Ottawa,
the reinforced concrete columns tested. ON, Canada, 2012, 31 pp.
10. Lie, T. T.; Rowe, T. J.; and Lin, T. D., “Residual Strength of Fire-
As a general conclusion, although the models provided Exposed Reinforced Concrete Columns,” Evaluation and Repair of Fire
by the current version of the Eurocode (EN 1992-1-2:20043) Damage to Concrete, SP-92, American Concrete Institute, Farmington
have some deficiencies, they are the closest to being capable Hills, MI, 1986, pp. 153-174.
11. CAN/ULC S101-07, “Standard Methods of Fire Endurance Tests
of generating accurate estimates of the response of reinforced of Building Construction and Materials,” Underwriters Laboratories of
concrete structure subjected to fire. However, the models still Canada, Toronto, ON, Canada, 2007, 85 pp.
do not provide the level of confidence required for a reliable 12. ASTM E119-12a, “Standard Test Methods for Fire Tests of Building
Construction and Materials,” ASTM International, West Conshohocken,
performance-based design. Additional experimental and PA, 2012, 34 pp.
analytical research is required to develop more reliable models 13. ISO, “ISO 834-1:1999: Fire-Resistance Tests—Elements of Building
for both the thermal and the mechanical temperature-depen- Construction—Part 1: General Requirements,” ISO, Geneva, Switzerland,
1999, 25 pp.
dent properties of concrete and steel reinforcing bars. 14. Franssen, J.-M., “SAFIR: A Thermal/Structural Program Modelling
Structures under Fire,” North American Steel Construction Conference
NOTATION FOR FIGURES (NASCC), American Institute of Steel Construction, Chicago, IL, 2003.
EXP - 10: experimental results of NRC Column 10; EXP  -  11: experi- 15. Lie, T. T., and Allen, D. E., “Calculations of the Fire Resistance
mental results of NRC Column 11; EXP - 12: experimental results of NRC of Reinforced Concrete Columns,” NRCC 12797, Division of Building
Column 12; EXP - 1582: experimental results of NRC Column 1582; EXP - Research, National Research Council of Canada, Ottawa, ON, Canada,
A: experimental results of NRC Column A; EXP - B: experimental results of 1972.
NRC Column B; ASCE: analytical results estimated by VecTor3 using the 16. Lie, T. T.; Lin, T. D.; Allen, D. E.; and Abrams, M. S., “Fire Resis-
models provided by the ASCE Manual of Practice1; ENV: analytical results tance of Reinforced Concrete Columns,” DBR Paper No. 1167, Division of
estimated by VecTor3 using the models provided by the former version of Building Research, National Research Council of Canada (NRC), Ottawa,
the Eurocode (ENV 1992-1-2:19952); EN: analytical results estimated by ON, Canada, 1984, 20 pp.
VecTor3 using the models provided by the current version of the Eurocode 17. Chang, Y. F.; Chen, Y. H.; Sheu, M. S.; and Yao, G. C., “Residual
(EN 1992-1-2:20043). C: analytical results estimated by VecTor3 using the Stress-Strain Relationship for Concrete after Exposure to High Tempera-
models of concrete mixed with calcareous aggregates; S: analytical results tures,” Cement and Concrete Research, V. 36, No. 10, 2006, pp. 1999-2005.
estimated by VecTor3 using the models of concrete mixed with siliceous doi: 10.1016/j.cemconres.2006.05.029

490 ACI Structural Journal/May-June 2016

Copyright of ACI Structural Journal is the property of American Concrete Institute and its
content may not be copied or emailed to multiple sites or posted to a listserv without the
copyright holder's express written permission. However, users may print, download, or email
articles for individual use.