Fire Safety Journal 100 (2018) 67–75

Contents lists available at ScienceDirect

Fire Safety Journal

journal homepage: www.elsevier.com/locate/firesaf

Modeling creep of steel under transient temperature conditions of fire T

a,∗ a b
Elie G. Hantouche , Karim K. Al Khatib , Mohammed A. Morovat
a
Department of Civil and Environmental Engineering, American University of Beirut, Lebanon
b
Department of Civil, Architectural and Environmental Engineering, The University of Texas at Austin, United States of America

A R T I C LE I N FO A B S T R A C T

Keywords: This paper presents a methodology for explicit modeling of the time-dependent behavior or creep of structural
Thermal creep steel in the Abaqus finite element (FE) models under transient temperature conditions representative of building
Time-dependent fires. In this methodology, available creep models based on conventional creep tests under constant or steady-
Steady-state temperature state temperature conditions were used to predict the time-dependent or creep behavior of steel under variable
Transient-state temperature
or transient-state temperature conditions. Specifically, user-defined CREEP subroutines were created in Abaqus
Fire
Shear tab connection assemblies
to define time-dependent or viscoplastic deformation of structural steel under changing temperatures. To de-
Finite element monstrate the robustness of the proposed methodology, the time-dependent behavior of structural steel at ele-
vated temperatures was modeled under three distinct conditions. These were: steady-state temperature condi-
tions, stepwise steady-state temperature conditions, and transient-state temperature conditions. This paper
further presents and discusses the application of the methodology in studying the creep behavior of a shear tab
connection assembly under the transient temperature conditions of a fire. Through this application, the im-
portance of explicit consideration of creep in predicting the response of connection assemblies at different stages
in the evolution of a structural-fire is also emphasized.

1. Introduction and background
1.1. Characterization of thermal creep of structural steel
The stress-strain behavior of structural steel at elevated tempera-
tures is shown to be time-dependent for stresses and temperatures re-
presentative of building fires [1,2]. The time-dependent or creep be-
havior of steel is defined as the time-dependent inelastic strain of
structural steel resulting from the application of both stress and tem-
perature. Thermal creep of structural steel is commonly established by
conducting steady-state temperature creep material tests in tension. In
these tests, the phenomenon of creep is observed through the increase
in mechanical strain under the conditions of sustained mechanical
stresses and constant temperatures. The resulting inelastic strain-time
curves usually exhibit three distinct stages: the primary stage char-
acterized by decreasing creep strain rates, the secondary stage char-
acterized by constant or steady creep strain rate, and the tertiary stage
characterized by increasing creep strain rates [2]. Depending on stress
and temperature levels, the increasing creep strains in the tertiary stage
may result in fracture of the steel coupon specimen [2].
Using the data from creep tests, material creep models have been
developed to predict the time-dependent strain behavior of different
grades of structural steels for steady-state temperature conditions
[2–5]. For example, Harmathy [4,6] and Fields and Fields [5] devel-
oped creep models for ASTM A36 steel, Morovat [2] proposed a creep
model for ASTM A992 steel, and Wang et al. [3] suggested a creep
model for Q460 high-strength steel.
1.2. Creep behavior of steel members and assemblies in fire
The influence of thermal creep of structural steel on the fire re-
sponse of steel members and assemblies has been recognized in pre-
vious studies [2,7–11]. It was shown in these studies that the behavior
of structural steel elements can be highly time-dependent depending on
the load conditions, duration of exposure to fire, and temperature
magnitudes. For instance, in a series of creep buckling tests on ASTM
A992 steel columns, Morovat [2] showed how the buckling strength of
steel columns became time-dependent as a result of thermal creep of
steel. Similar time-dependent buckling phenomenon was observed by
Yang and Yu [8] during experiments on columns made of SN490FR fire-
resistant steel. In addition, better agreements between experimental
and computational predictions were observed when the thermal creep
was explicitly modeled in FE simulations [2,7–11]. As an example,
studies by Kodur and Dwaikat [9] showed that the explicit

∗
Corresponding author. Department of Civil and Environmental Engineering, American University of Beirut, P.O.Box 11-0236, Riad El-Solh/Beirut, 1107 2020,
Lebanon.
E-mail addresses: elie.hantouche@aub.edu.lb (E.G. Hantouche), kka05@mail.aub.edu (K.K. Al Khatib), morovatma@utexas.edu (M.A. Morovat).

https://doi.org/10.1016/j.firesaf.2018.07.006
Received 8 February 2018; Received in revised form 11 May 2018; Accepted 9 July 2018
Available online 17 July 2018
0379-7112/ © 2018 Elsevier Ltd. All rights reserved.
E.G. Hantouche et al.
consideration of thermal creep using ANSYS creep models resulted in
more accurate predictions of restraint beams response under fire, and
studies were also done to investigate the creep effect under different
fire scenarios. Torić et al. [1] showed an example of unrestrained steel
beam subjected to fire temperatures up to (600 °C) for about 110 min.
The results showed that including creep effect, using implicit creep
model, improve the predictions of the beam response compared to ex-
perimental results. The National Institute of Standard and Technology
(NIST) in its study of WTC building collapse in September 11, 2001
recognized the importance of thermal creep of steel [12,13]. Therefore,
creep was considered in all simulations of the WTC building collapse to
model both columns and floor trusses. Further, El Ghor and Hantouche
[14], and Morovat et al. [11] carried out FE simulations to investigate
the effect of creep on isolated flush endplate connections at elevated
temperatures, and proposed a methodology to quantify the creep be-
havior of these connections in the form of isochronous curves.
Recent studies [9,15] showed that explicit consideration of creep is
more appropriate and result in more accurate representation of the real
case. Kodur et al. [16] made a comparison between different models
available in literature including Eurocode 3 [17] temperature-stress-
strain relationships. The results show that when effect of creep is ex-
plicitly considered, the predicted deflections compared well with the
experimental results. Also, when using the Eurocode temperature-
stress-strain relationships, which takes into account the effect of creep
implicitly, the analysis showed a conservative predictions of beam de-
flections when compared with the experimental results.
As indicated in previous studies, implementing creep material
models in computational tools allows the explicit evaluation of the
time- and temperature-dependent response of steel members and as-
semblies at elevated temperatures. This quantification of structural
behavior in terms of both temperature and time is of utmost importance
in developing performance-based design frameworks for the fire safety
of steel structures. However, since the time-dependent behavior of steel
is obtained under constant temperature in a creep test, arguments exist
against the applicability of creep test results for structural-fire appli-
cations where temperatures are changing throughout the course of a
building fire. Therefore, the primary goal of this paper is to show how
material creep models developed using creep tests can be utilized in
analysis of steel structures exposed to the transient-state temperature
conditions of a fire.
1.3. Contribution of this study
This paper describes a methodology to consider the creep effects on
the behavior of steel structures exposed to changing temperatures of
fire. To consider the thermal creep of structural steel under transient-
state temperature conditions, user-defined subroutines were developed
and implemented in computational analyses using the general-purpose
FE software Abaqus. The constitutive model proposed by Fields and
Fields [5] for the creep of ASTM A36 steel under steady-state tem-
perature conditions were chosen in the development of the metho-
dology. To show the capability of the proposed methodology in pre-
dicting time-dependent behavior of steel structures under varied
temperatures, three distinctive temperature environments were con-
sidered: steady-state, stepwise steady-state, and transient-state tem-
perature environments. The application of the developed methodology
in considering thermal creep of steel under the transient-state tem-
perature conditions of fire is further shown through the study of creep
response of a steel shear tab connection assembly. Finally, the im-
portance of explicit consideration of creep in predicting the response of
connection assemblies at different stages in the evolution of a struc-
tural-fire is further indicated and discussed.
2. Development of the methodology
This section describes the proposed methodology for explicit
68
Fire Safety Journal 100 (2018) 67–75
modeling of the thermal creep of steel under transient-state tempera-
ture environment of structural fires. Specifically, a procedure is devel-
oped and implemented as user-defined subroutines in Abaqus to modify
and extend a steady-state temperature creep model for structural-fire
applications.
2.1. Formulation of thermal creep of steel
In order to develop a procedure to explicitly account for the creep of
structural steel under changing temperature environment of a building
fire, a computational material creep model was formulated. The con-
stitutive material law in the form of power-law creep by Fields and
Fields [5] was utilized in the formulation of the computational creep
model and its implementation as a user-defined subroutine CREEP in
Abaqus. The creep material model by Fields and Fields [5] was ori-
ginally developed for the ASTM A36 steel in the temperature range of
350 °C–600 °C, and for creep strains up to 6%. To develop the proce-
dure, it was further assumed that the total strain under constant tem-
perature could be divided into three independent components:
εT = εe + εp + εc (1)
In Eq. (1) εT is the total strain, εe is the time-independent elastic
strain, εp is time-independent plastic strain, and εc is the time-dependent
plastic or creep strain. Therefore, the two components εe and εp are
temperature- and stress-dependent, whereas the creep strain compo-
nent is time-, temperature-, and stress-dependent.
The empirical equation proposed by Fields and Fields [5] in the
form of power-law creep (Norton-Bailey power-law creep) is shown in
Eq. (2).
εc = at bσ c (2)
In Eq. (2), t is time, σ is stress, and coefficients a , b , and c are
temperature-dependent material constants. Formulas for the calculation
of these material constants are presented in Eqs. (3)–(5).
b (T ) = b0 + b1 T (3)
where b0 = −1.1 and b1 = 0.0035
c (T ) = c0 + c1 T (4)
where c0 = 2.1 and c1 = 0.0064
a (T ) = (0.145c )10−(a0 + a1 T ) (5)
where for T < 500°C , a0 = 8.1 and a1 = 0.00573, and for T > 500°C ,
a0 = 15.25 and a1 = −0.00851.
In Eqs. (3)–(5), T is in °C, t is in minute, σ is in MPa, and εc is in mm/
mm (unitless).
To formulate the computational material creep model and calculate
creep strains, an equation for the creep strain rate is necessary. Creep
strain rate can be represented in two different formats: strain- and time-
hardening formats. In the strain hardening formulation, the creep strain
rate is defined as a function of the creep strain (as shown in Eq. (6)),
whereas in the time hardening formulation, the creep strain rate is
defined as a function of time (as shown in Eq. (7)). Fig. 1 further depicts
these two representations of the creep strain rate. It has been shown in
previous studies [2,6] and confirmed in this study that the strain
hardening formulation yields better and more accurate results when
computing the time-dependent strains for variable stress history.
Therefore, in this study, the Fields and Fields [5] equation is described
in a strain-hardening formulation.
1 b−1 c
ε˙ c = F (εc ; σ ; T ) = a b bεcb σb
(6)
ε˙ c = F (t; σ ; T ) = abt b − 1σ c (7)
In Eqs. (6) and (7), ε̇c is in per minute. Eq. (6) can be further sim-
plified, through the change of the constants (a , b , and c ), as shown in
Eq. (8):
E.G. Hantouche et al. Fire Safety Journal 100 (2018) 67–75

calculated and introduced in the subroutine when developing the FE

model. To account for the expected change in the element stresses due
to different reasons (e.g. stress relaxation, geometrical nonlinearity …)
during the FE analysis, the strain hardening formulation is used in the
subroutine.

2.3. Stepwise steady-state temperature environment

As a first step towards studying the behavior of the system under

transient-state heating, a consecutive steady-state creep steps are con-
ducted while including the creep effect. This method is done by as-
signing a heating step, where temperature is increased by constant in-
crements without the creep effect, followed by a steady-state creep step
for a specific duration. This procedure is repeated, in a step-by-step
manner, until reaching the desired temperature. The smaller the size of
steps is, the more accurate the results will be. Since only one user-de-
fined subroutine can be assigned to a single simulation (job), the sub-
routine created for this stage includes conditional statements that
control which coefficients should be used dependent on the tempera-
ture reached in each step.

2.4. Transient-state temperature environment

In real fire scenarios, a transient-state conditions are more likely to

occur than steady-state conditions. Therefore, it is essential to carry out
numerical studies on steel structures subjected to temperature change
while including transient creep strain especially when reaching high
Fig. 1. Creep response when changing stress and/or temperature using (a) time
temperatures (above 400 °C). Transient-state creep analysis is defined
hardening formulation; (b) strain hardening formulation.
when both temperatures and stresses are considered as variables. As
mentioned before, the change in stress is taken into consideration by
1
ε˙ c = [Aσ n (εc (m + 1))m] m + 1 (8) using the strain hardening formulation as a criterion. However, the
change in temperature implies that the three temperature-dependent
where A = ab , m = b − 1, and. n = c coefficients ( A , m , and n ) also vary with time. Therefore, the sub-
The creep strain difference per iteration can be written as follows: routine is modified so that it can perform transient-state analysis within
1 1+m a single step. The equations of temperature-dependent coefficients are
1
⎡ Aσ n ⎞ 1 + m ⎤ included in the subroutine. This modification enables automated com-
Δεc = ⎢ ⎛ × Δt + εc1 + m ⎥ − εc
⎝ 1 + m ⎠ (9) putation of these coefficients at the beginning of each incremental step
⎣ ⎦
based on the new temperature. To inhibit the formation of any creep
In case of implicit creep integration, the derivative of the strain strains below 350 °C, A is set to zero for all temperatures below this
difference with respect to stress needs to be included in the subroutine temperature. A summary of the subroutine calculation process is illu-
as per Abaqus documentation [18] and can be written as follows: strated as a flowchart in Fig. 2. In numerical solutions, such as finite
1 1 m element (FE) analysis, the time is discretized into increments where
1
∂Δεc Δtn ⎛ Aσ n ⎞ m + 1 ⎡ ⎛ Aσ n ⎞ m + 1 ⎤ each increment is treated under steady-state conditions (temperature,
= ⎢Δt + εc1 + m ⎥
∂σ σ ⎝m + 1⎠ ⎝ m + 1⎠ (10) stress …). That is, in numerical solutions, the transient-state analysis is
⎣ ⎦
basically a chain of steady-state increments linked together. The same
Note that it is more convenient to use implicit integration scheme applies on how the subroutine works [18], in incremental steps, at each
since it is unconditionally stable. However, a tolerance is assigned to iteration the temperature and stress are assumed constant during the
reach accurate results by controlling the difference in the change of increment. At the start of each increment, and that is what the flow-
creep strain between the iterative steps ( Δεi + 1 − Δεi < tolerance ). If the chart is representing, the subroutine is called to estimate the change in
difference is not within the predefined tolerance, the iteration is re- creep strain (Eq. (9)) based on the instant stress, temperature, and creep
peated with smaller Δt until the tolerance is satisfied. Based on the strain history of each mesh element in the FE model and the conditions
presented equations, user-defined subroutines are developed to con- are assumed constant throughout the increment.
sider different temperature environments. It is noted that implicit and explicit integration schemes follow the
backward and forward Euler integration methods, respectively. Thus,
2.2. Steady-state temperature environment the temperature used in calculating the temperature-dependent coeffi-
cients is the temperature at the end of each increment for implicit in-
In steady-state creep analysis, both the temperature and the load tegration scheme and vice versa. However, this difference in the out-
applied are held constant. Such conditions are mainly provided in come can be hardly noticed if the increment is very small, where the
material creep tests to directly measure the creep strains with respect to difference between the beginning and the end temperature is almost
time. To better understand the time-dependent behavior of steel negligible. Note that only the implicit integration scheme is used in this
structures under elevated temperature without the interference of any study for all the cases hereafter.
other parameter, the steady-state creep analysis is chosen to be the first
stage of this study. The coefficients ( A , m , and n ) associated with the 2.5. Application of the methodology
subroutine are solely dependent on the instant temperature of steel,
which is maintained constant. Therefore, these coefficients will remain In this section, the application of the proposed methodology is il-
constant throughout the simulation, and thus can be manually lustrated through the analysis of creep behavior of a shear tab

69
E.G. Hantouche et al.
Fig. 2. Flowchart representing the subroutine incremental solution.
connection assembly under the transient temperature conditions of a
fire. FE models of a full-scale shear tab connection assembly was de-
veloped in Abaqus to study its creep behavior in fire. This shear tab
connection assembly was used in the experimental work performed by
Wang et al. [19], and was selected to validate computational results
against the experimental ones. Details of the FE model of the shear tab
connection assembly are described below. Results of the simulations in
predicting the time-dependent beam axial forces, and mid-span de-
flections are further presented and discussed.
3. Modeling the shear tab connection assembly
3.1. Geometry of the shear tab connection assembly
The shear tab connection assembly used in the analysis consists of
one shear tab plate (150 × 130 × 10 mm), which connects a UC
254 × 254 × 73 column to a UB 178 × 102 × 19 beam. The shear tab
plate is welded to the column and bolted to the beam. Four shear bolts
of diameter 20 mm (M20) are used with bolt holes of diameter 22 mm
(standard size holes). Details of the shear tab connection assembly are
shown in Fig. 3.
70
Fire Safety Journal 100 (2018) 67–75
Fig. 3. The shear tab connection detail [19].
3.2. Material properties
An idealized bilinear model is used for the steel material. The am-
bient material properties used for the structural parts in the FE model
are the same as the material properties reported in the experiment [19].
For the estimation of mechanical properties at elevated temperatures,
retention factors proposed by Lee et al. [20] are used for the structural
steel material, whereas the retention factors proposed by Hu et al. [21]
are used for the structural bolts. The fire test conducted by Wang et al.
[19] follows the standard temperature-time curve which is known for
its fast heating regime. Further, some parts (e.g. insulated beam top
flange) of the heated frame did not reach high temperatures, thus not
allowing for creep strains to develop. Therefore, creep is not included in
the validation simulation.
3.3. Model discretization
All the components of the connection model in Abaqus were dis-
cretized using eight-node brick elements with reduced integration
(C3D8-R). To account for the high stresses present in the connection
region, where the failure was likely to happen, a finer mesh was used
around the connection region. Moreover, to avoid stress concentration
around the bolt holes, a mapped meshing technique was used to dis-
cretize bolts and their surrounding areas. Surface-to-surface contact
with a finite sliding coefficient was used to reproduce contact surfaces.
A friction coefficient of 0.25 was utilized to model friction between the
contact surfaces, while allowing separation, sliding, and rotation of the
contact surfaces.
3.4. Applied loads and boundary conditions
Two loading steps were defined in the simulations; one for the pre-
tensioning of the shear bolts and one for applying two equal con-
centrated loads of 40 kN. Due to the symmetry of the connection as-
sembly, only half of the frame was modeled in Abaqus and symmetry
was applied at the beam mid span section. In addition, column ends
were horizontally restrained yet free to expand, due to heating, in the
vertical direction. Tie constraint was assigned between the column and
the shear tab plate interface in order to model the weld used in the
experiment. Horizontal restraint was also assigned to the beam top
flange to prevent lateral torsional buckling of the beam. Details of the
FE model are shown in Fig. 4.
E.G. Hantouche et al.
Fig. 4. Detail of the full-scale steel frame FE model assembly.
3.5. Temperature profile
In the experiment, the beam top flange was wrapped with isolating
material to include the heat-sink effect of the concrete slab. The furnace
size, as specified, was 3000 × 1600 × 900 mm, hence part of the
column was present inside the furnace and exposed to fire tempera-
tures. Moreover, some difficulties were encountered throughout the test
in maintaining uniform distribution of temperature in the furnace and
among the structural components of the frame. Therefore, different
temperature profiles were considered for different structural compo-
nents to closely model temperature histories measured during the ex-
periments. Specifically, the FE model was divided into six regions with
uniform temperature distributions.
4. Analysis schemes and results
4.1. Time-independent simulations: transient-state temperature environment
The results obtained from the FE model are first validated against
the experimental data and shown in Fig. 5. The beam experiences in-
duced thermal axial compressive forces due to the restrained expansion.
At the beginning of the heating phase, linear increase in the internal
beam axial forces and slight increase in the mid-span deflection oc-
curred. This is due to the insignificant degradation in the material
properties at low temperatures. A sudden slip occurs, at about 200 °C, in
the shear plane (between the shear tab plate and the beam web) which
causes loss in the thermal energy, thus releasing some beam axial
stresses. The slip is initiated when the beam axial force and gravity
loads applied exceed the static frictional bond between the plate and
the beam web imposed by the clipping force of bolts pre-tensioning.
The slip is then arrested by bearing between bolt shanks and the edges
of the bolt holes at around 270 °C. Then, as temperature increases, a
nonlinear increase in axial force and deflection is observed due to
material strength and stiffness degradation. Also, it can be seen that the
beam end rotation increases due to the loss in the beam flexural stiff-
ness. As shown in Fig. 3, the gap between the beam and the column is
10 mm. And due to excessive rotation, contact occurred between the
beam bottom flange and the column flange at a temperature of 770 °C.
At 780 °C, the beam loses its flexural stiffness due to the formation of
plastic hinge at the mid-span and experiences rapid increase in the mid-
span deflection. This rapid loss in beam stiffness causes the FE model to
go through unstable zone. The post-ultimate response of the structure
cannot be reached in this simulation, since the implicit (ABAQUS/
Standard) solver is used to run the simulation without the interference
of any stabilizing options or the inclusion of material fracture proper-
ties. High stresses are observed at the shear tab plate near the vicinity of
the weld which indicate a potential weld fracture.
71
Fire Safety Journal 100 (2018) 67–75
Fig. 5. Transient-state heating of full-scale shear tab connection: (a) beam axial
force; (b) midspan deflection.
4.2. Time-dependent simulations: steady-state temperature environment
After obtaining the validated scenario, a steady-state creep study
was conducted with some modifications. Unlike the validated condi-
tions, and for simplicity, a uniform distribution of temperature was
assigned to all the heated structural components. In addition, due to the
limitations imposed through the application of the creep model by
Fields and Fields [5], the maximum temperature of 600 °C was con-
sidered.
Based on the equations provided by Fields and Fields [5], a user-
defined subroutine was developed and assigned to both the beam and
the shear tab. Creep effect was not considered in modeling bolts since
they did not contribute much to the overall ductility of the connection,
and thereby did not have major impacts on the time-dependent re-
sponse of the connection. To focus on the beam behavior and to insure a
constant end-restrain stiffness for all cases, creep was also not in-
corporated in the steel column material.
The model was first heated up to the desired temperature (400 °C,
450 °C, 500 °C, 550 °C, and 600 °C) with load applied and without in-
cluding creep effect (very fast heating rate). The temperature was as-
sumed to increase linearly with time and uniformly distributed in the
heated parts of the structure. After the heating step was completed, the
final temperature reached was maintained constant and a steady-state
creep step (VISCO) for 120 min was then started with creep effect in-
cluded in the beam and shear tab plate as mentioned before. The con-
stants ( A , m , and n ), available in Eqs. (9) and (10), were calculated
based on the final temperature reached in the heating step and were
E.G. Hantouche et al.
Fig. 6. Beam axial force of steady-state creep at different temperatures versus
(a) temperature; (b) time.
included in the subroutine. To control the accuracy of the results, a
small tolerance of 10−6 was assigned.
The variation in the beam axial force and the mid-span deflection
during the steady-state temperature simulations are shown in Figs. 6
and 7, respectively. As can be seen, after the steady-state creep step is
initiated, the beam axial force starts to decrease gradually. During the
first few minutes, the primary stage takes place by rapid decrease in the
axial force which is then followed by secondary stage associated with
slow and steady decrease in the axial force. This decrease in the axial
force is due to the stress relaxation caused by the restrained thermal
expansion. Similarly, as shown in Fig. 8, the mid-span deflection in-
creases with time during the steady-state temperature step.
For high temperatures (greater than 500 °C), where the creep strain
rate is much higher, a severe impact of creep on the behavior of the
connection assembly starts to take place. The loss in axial force is ob-
served in a shorter duration. The mid-span deflection also experiences a
rapid increase, especially during the first few minutes, caused by the
generated moment of the applied load and P-Δ effect. The degradation
in the material properties in terms of both strength and stiffness along
with the development of significant time-dependent plastic strains re-
sult in permanent deformations.
Based on Eq. (4), c is always greater than unity, meaning that the
distribution of inelastic stresses will change with time. The nonlinear
relation between stress and creep strain rate (Eq. (6)) imposes a re-
distribution of the stresses over the cross section of the beam. The FE
results (creep starts at 500 °C) of the axial stresses distribution over the
beam depth (at the mid-span) versus time are shown in Fig. 8(a). At the
72
Fire Safety Journal 100 (2018) 67–75
Fig. 7. Beam mid-span deflection of steady-state creep at different temperatures
versus (a) temperature; (b) time.
end of the loading step, the distribution of axial stresses over the beam
depth can be seen in Fig. 8(b). After the heating phase takes place, the
neutral axis shifts towards the compression zone (downward) as can be
seen in Fig. 8(c). In the steady-state creep step, the beam is subjected to
constant bending moment and additional compression stresses due to
restraint thermal expansion. These additional stresses expedite the de-
velopment of creep strains in the compression zone of the beam section.
In response, the beam neutral axis shifts towards the tension zone
(downward) to maintain equilibrium. Moreover, at the end of the creep
step, the redistribution of axial stresses can be seen (Fig. 8(d)) as dis-
cussed previously.
Following the mid-span deflection, the connection exhibits ex-
cessive rotation because of the loss in the flexural stiffness of the beam
through time. Also note that despite the constant temperatures, the
stresses are constantly changing due to relaxation. Therefore, the de-
crease in slope of axial relaxation is caused not only by the transition
from primary to secondary stage, but also by the drop in the axial force,
which in return results in a decrease in the creep strain rate.
Furthermore, at a temperature of 600 °C, the creep effect did not only
reduce the compressive stresses, but did also generate the tensile
stresses caused by the beam catenary action. The loss of the beam
flexural stiffness leads to the resistance of the applied load through
alternative equilibrium configuration, namely the catenary action. The
presence of beam end axial restraints enables the development of the
catenary action to resist the vertical load. Thus, the beam survives for
longer durations even after the development of plastic hinges. At a
certain point, the tensile forces reach a maximum value, upon which
E.G. Hantouche et al.
Fig. 8. Axial stress distribution over the cross-section of the beam at the mid-span (creep starts at 500 °C): (a) at different depths versus time; (b) at the end of loading;
(c) at the end of heating; (d) at the end of creep.
most of the applied load is resisted through the beam axial stiffness. On
the other hand, mid-span deflection continues to increase with a
smaller rate.
4.3. Time-dependent simulations: stepwise steady-state temperature
environment
Two creep steps durations, both with temperature increase rate of
50 °C/step, are chosen: 5 min/step and 10 min/step which are equiva-
lent to heating rate of 10 °C/min and 5 °C/min, respectively. The re-
sulted beam axial load and the beam mid-span deflection versus tem-
perature are plotted in Fig. 9(a) and (b), respectively. After the load is
applied, the temperature is increased from 20 °C to 400 °C, and then
consecutive steady-state creep steps are applied as explained previously.
It can be seen that the effect of creep is greater for creep duration of
10 min/step, and the higher the temperature is the more reduction in
compressive stresses is observed. The same can also be applied for the
mid-span deflection, which shows a similar behavior.
4.4. Time-dependent simulations: transient-state temperature environment
Transient-state conditions are applied in the model. The tempera-
ture is increased from 20 °C to 600 °C while considering material re-
tention factors as function of temperature. Four cases of heating rate are
conducted: very fast heating rate (without creep), 5, 10, and 20 °C/min.
Fig. 10 shows the beam axial force and mid-span deflection for different
heating rates. Creep effect is considered to be negligible at temperatures
below 350 °C, so the heating rate will not take effect below this tem-
perature. After the specimen temperature reaches 350 °C, the time-de-
pendent strains start to take place, and stress relaxation is observed
throughout the heating phase. Moreover, stress relaxation increases
with lower heating rate, the exponential decrease in the compressive
stresses is noticed since the creep strain rate increases as the tempera-
ture becomes higher. The gradual development of catenary action im-
plicitly contributes to the reduction of the axial stresses. The catenary
action is caused by the rapid increase in the mid-span deflection due to
creep, as seen in Fig. 10(b).
73
Fire Safety Journal 100 (2018) 67–75
For the transient-temperature environment, the response of the
connection assembly in the cooling (decay) phase of fire was also in-
vestigated through applying different cooling durations. Cooling dura-
tion is the time needed to cool the heated connection back to the am-
bient temperature (20 °C). Four cooling durations are chosen: very short
(without creep), 5, 10, and 20 min cooling durations. Fig. 11 shows the
beam axial force and mid-span deflection for different cooling durations
after two heating cases. It can be seen that a slower heating rate pro-
duces larger axial tensile forces at the end of the cooling phase. This can
be attributed to the additional loss of restraint beam axial force, during
the heating phase, due to the development of creep strains at slower
heating rates. For faster heating rates, the difference between the re-
sults of the four cooling durations can be clearly seen since the com-
pression axial force at the end of the heating phase is larger, causing
faster creep rate. When the tension force reaches almost the same
magnitude (about 60 kN) at which the first slip occurred (during the
heating phase), a reverse slip is observed during the cooling phase.
5. Conclusions
The purpose of this study is to develop a methodology to explicitly
include the creep effect in FE models developed in Abaqus. The steps of
including the time-dependent strains were explained in this study from
which a subroutine was developed and incorporated in the FE simula-
tions. An example of a full-scale shear tab available in literature was
provided to show an application of the developed subroutine in the FE
models, and to illustrate the time-dependent behavior of steel at fire
temperatures. To this aim, FE model was first developed and validated
against the experimental data available in the literature. FE analyses
were performed under both steady-state and transient-state tempera-
ture conditions including the creep effect. In the case of transient-state
creep analysis, different heating rates and cooling durations were
conducted to show the influence of time on the structural behavior
when providing similar conditions. This study should be extended to
develop more simplified and practical models for fire structural design
purposes [14]. More reliable creep models should be developed to in-
clude the creep effect in other structural components (columns, bolts,
E.G. Hantouche et al.
Fig. 9. Stepwise steady-state temperature with and without including creep
effect: (a) beam axial force; (b) mid-span deflection.
welds …) as to study the global behavior of the system, and the creep
effect should be studied at temperatures higher than 600 °C. It is worth
mentioning that most of the fire tests available in literature were con-
ducted under fast heating regimes (i.e. using standardized temperature-
time curves). That is, exposing the tested specimens to short time
period, during which creep strains can hardly develop [22]. This im-
plies that more experimental data and more detailed computational
studies are still needed to better understand the extent of which creep
plays a role in predicting steel behavior in fire. The following points are
concluded from this study:
• The developed user-defined subroutine is able to model transient
creep strains in Abaqus FE models. These strains are introduced as
explicit plastic time-dependent strains that can impact the total re-
•
sponse of the FE models.
Based on the presented cases, it can be concluded that the sub-
routine is able to account for the three factors affecting the creep
rate (temperature, stress, and time) when incorporated in the FE
models.
• Using a strain hardening formulation with implicit (backward Euler)
integration method insures unconditionally stable results. However,
small size increments are needed to get accurate results.
• Creep has a significant influence on the structure behavior at high
temperatures. Under transient-state heating, it causes axial stress
relaxation and increases the mid-span deflection of the beam.
Ignoring the time-dependent strains at high temperatures can lead
to under prediction of the beam deflection and over prediction of the
74
Fire Safety Journal 100 (2018) 67–75
Fig. 10. Transient-state analysis with and without including creep effect: (a)
beam axial force; (b) mid-span deflection.
axial thermal induced forces in restrained beams. This will affect the
overall behavior of the structure in performance-based structure-fire
analysis.
• The results show that as the heating rate decreases, the influence of
creep effect increases, making it crucial to conduct time-dependent
analyses when dealing with steel structures subjected to fire for long
durations.
• Slower heating rate and longer cooling duration produce higher
beam tensile force and larger mid-span deflection at the end of the
cooling phase.
• In severe conditions, creep can alter the stability configuration of
the structure. The load resisting mechanism can change from flex-
ural to catenary action at high temperatures and high stresses for
long period of time.
• When subjected to bending moment, creep strains cause axial stress
redistribution over the beam cross section, relieving high axial
stresses at the top and bottom fibers and redistributing them to the
inner fibers to achieve equilibrium.
Consistent with the previous studies [2,7–11,15,16], explicit con-
sideration of creep is more appropriate and more representative of the
real case than implicit consideration. Additionally, studying the effect
of different heating rates and cooling durations is not possible if creep is
not introduced explicitly. Ignoring the creep effect in structural ana-
lyses may lead to unsafe predictions of structural response when sub-
jected to fire. Therefore, many efforts have being made to explicitly
include the creep in the performance based structure-fire analysis to
E.G. Hantouche et al.
Fig. 11. Beam axial force and mid-span deflection response for different cooling durations after (a) fast heating; (b) heating for 20 °C/min.
improve the representation of the true behavior of steel structures in
fire. In this regard, future work will be carried out to study the effect of
key geometric and loading parameters on the time-dependent response
of connection assemblies.
Declarations of interest
None.
Acknowledgments
The authors gratefully acknowledge the financial support provided
by the American University of Beirut Research Board under grant No.
23310-103371.
References
[1] N. Torić, A. Harapin, I. Boko, Modelling of steel creep at high temperatures using an
implicit creep model, Key Eng. Mater. 553 (2013) 13–22.
[2] M.A. Morovat, Creep Buckling of Steel Columns Subjected to Fire, Ph.D.
Dissertation (2014) Department of Civil, Architectural and Environmental
Engineering, The University of Texas at Austin.
[3] W. Wang, S. Yan, J. Liu, Studies on temperature induced creep in high strength
Q460 steel, Mater. Struct. 50 (1) (2017) 68.
[4] T.Z. Harmathy, A comprehensive creep model, J. Basic Eng 89 (3) (1967) 496–502.
[5] B.A. Fields, R.J. Fields, Elevated Temperature Deformation of Structural Steel,
National Institute of Standards and Technology, NIST, 1989.
[6] T.Z. Harmathy, Creep deflection of metal beams in transient heating processes with
particular reference to fire, Can. J. Civ. Eng. 3 (2) (1976) 219–228.
[7] G.Q. Li, C. Zhang, Creep effect on buckling of axially restrained steel columns in real
fires, J. Constr. Steel Res. 71 (2012) 182–188.
75
Fire Safety Journal 100 (2018) 67–75
[8] K.C. Yang, Z.H. Yu, Experimental research on the creep buckling of fire-resistant
steel columns at elevated temperature, Steel Compos. Struct. 15 (2) (2013)
163–173.
[9] V. Kodur, M. Dwaikat, Effect of high temperature creep on the fire response of
restrained steel beams, Mater. Struct. 43 (10) (2010) 1327–1341.
[10] M.A. Morovat, M.D. Engelhardt, E. Taleff, T. Helwig, Importance of time-dependent
material behavior in predicting strength of steel columns exposed to fire, Appl.
Mech. Mater. 82 (2011) 350–355.
[11] M.A. Morovat, A.H. El Ghor, E.G. Hantouche, Time-dependent response of flush
endplate connections to fire temperatures, J. Struct. Eng. 144 (4) (2018) 04018023.
[12] NIST, Final Report on the Collapse of the World Trade Center Towers, Report NIST
NCSTAR 1, National Institute of Standards and Technology, Gaithersburg, MD,
2005.
[13] W.E. Luecke, et al., Federal Building and Fire Safety Investigation of the World
Trade Center—Mechanical Properties of Structural Steels, Report NIST NCSTAR
1–3D, National Institute of Standards and Technology, Gaithersburg, MD, 2005.
[14] A.H. El Ghor, E.G. Hantouche, Thermal creep mechanical-based modeling for flush
endplate connections in fire, J. Constr. Steel Res. 136 (2017) 11–23.
[15] Z.-F. Huang, K.-H. Tan, S.-K. Ting, Heating rate and boundary restraint effects on
fire resistance of steel columns with creep, Eng. Struct. 28 (6) (2006) 805–817.
[16] V. Kodur, M. Dwaikat, R. Fike, High-temperature properties of steel for fire re-
sistance modeling of structures, J. Mater. Civ. Eng. 22 (5) (2010) 423–434.
[17] Eurocode 3, Eurocode 3: Design of Steel Structures - Part 1-2: General Rules -
Structural Fire Design, (2005).
[18] ABAQUS Ver 6.14 Documentation, Dassault Systemes Simulia Corporation, (2014).
[19] Y.C. Wang, X.H. Dai, C.G. Bailey, An experimental study of relative structural fire
behaviour and robustness of different types of steel joint in restrained steel frames,
J. Constr. Steel Res. 67 (7) (2011) 1149–1163.
[20] J. Lee, M.A. Morovat, G. Hu, M.D. Engelhardt, E. Taleff, Experimental Investigation
of Mechanical Properties of ASTM A992 Steel at Elevated Temperatures 50 (2013),
pp. 249–272 4.
[21] Y. Hu, J.B. Davison, I.W. Burgess, R.J. Plank, Comparative Study of the behaviour of
BS 4190 and BS EN ISO 4014 bolts in fire, Proc., ICSCS, 2007, pp. 587–592.
[22] N. Torić, A. Harapin, I. Boko, Modelling of the influence of creep strains on the fire
response of stationary heated steel members, J. Struct. Fire Eng 6 (3) (2013)
155–176.