TECHNICAL NOTE

Temperature Analysis Using ABAQUS

by Jonas Palm
Palm is a research assistant at the Department of Building Technology at the
Swedish National Tesfing and Research Institute in Boräs, Sweden.

Abstract
This paper describes the use of the general-purpose finite-element code ABAQUS to tem-
perature analysis of concrete structures exposed to fire. Various structural members are stud-
ied, such as rectangular beams, slabs with circular enclosures, and steel-concrete composite
slabs. The thermal analysis allows us to establish the temperature distribution in the struc-
tural members throughout the fire history. Influences of phase changes and other nonlineari-
ties are taken into account. The results are compared with results from the computer codes
TASEF and SUPER-TASEE

Introduction
The development and acceptance of fire codes and regulations have traditionally
been based on fire tests. However, the development of computer software based on
either the finite-difference or the finite-element methods of numerical analysis have
shown that it is possible to simulate fire tests. ~
Fire tests are expensive to perform. Normally, the cost of a computer-based analy-
sis is considerably less. Some experimental investigations are necessary, however,
in order to study the overall structural response and to identify important material
properties.
The basis for evaluation has been obtained from standard fire tests. As Amdahl
and Hellan show, ~-this has been unsatisfactory in many cases for several reasons.
First, the temperature-time relationships used in standard fire tests are not repre-
sentative of the development of real fires. Second, the use of standard fire tests has
led to concepts of critical temperatures, such as 500°C for steel structures. These are
often conservative, especially for structures in which significant moment redistrib-
ution capability exists--that is, in which the forces in a failing member can be trans-
ferred to intact members in the adjacent structure. And finally, for practical reasons,
only single components or a few members can be tested simultaneously in a fur-
nace. This yields significant uncertainties with respect to modeling boundary con-
ditions. Using numerical analysis, it is possible to reduce the number of tests and to
close the gap between the previously mentioned results of single components and
the results for larger structural members or structures. 3~ Indeed, one objective of
numerical analysis is to calibrate material models, constitutive equations, and other
data, thereby equalizing calculated and experimental results.
292 Fire Technology Third Quarter 1994

According to Amdahl and Hellan, the fire engineering design comprises at least
three aspects: fire-exposure history, transient temperature states within the struc-
ture, and mechanical response of the structure. In the past, these topics have been
considered more or less individually. However, some computer codes can handle
coupled analysis, meaning that the transient temperature states within the structure
and the mechanical response of the structure are integrated and analyzed simulta-
neously. An example of such a code is the ABAQUS finite-element code?
Computer codes that can handle the transient temperature states within a structure
include TASEF ~ and SUPER-TASEF5 This paper deals only with this aspect. The
influences of various time-integration techniques and the modeling techniques of
phase changes are studied by comparing results from the previously mentioned
computer codes. Milke made similar comparisons. ~
The objectives of this paper are to study temperature distributions in concrete
structures exposed to fire. The thermal analysis is primarily performed using the
general-purpose finite-element code, ABAQUS. Examples are also presented.

General Analysis
In the analysis, the ABAQUS finite-element code ~ was used. The code is a general
pur-pose FE code for nonlinear analysis. The element library provides a complete
geometric modeling capability, and the material library contains several different
constitutive models.
The analysis may consist of several steps, each of which defines how the
response should be simulated. The code has a built-in, automatic, and adaptive
choice of time incrementation. The ABAQUS system also includes preprocessing
and postprocessing capabilities.
The ABAQUS code can handle coupled analysis, meaning that temperature and
displacements are integrated and analyzed simultaneously. Two computer codes
that can handle temperature analysis are TASEF and SUPER-TASER The latter
originates from the TOPAZ code developed by A. Shapiro and others at the
Lawrence Livermore Institute. The basic development that has been included in
SUPER-TASEF is the capacity to run a thermal-stress analysis of steel and rein-
forced-concrete beams.
This paper will only address transient temperature analysis. The influences of
various time-integration techniques and modeling techniques of phase changes are
studied by comparing results from the previously mentioned computer codes.
Milke made similar comparisons of TASEF to the code FIRES-T3. ~°Using
ABAQUS, it is possible to perform three-dimensional analysis; the other codes
studied are restricted to two-dimensional analysis.
The ABAQUS code uses implicit time-integration, while TASEF uses explicit
time-integration. Using the SUPER-TASEF code, both time-integration techniques
can be studied. The implicit method is unconditionally stable, while the explicit
method is a so-called conditionally stable method because the time-step is limited
TemperatureAnalysis UsingABAQUS 293

by a stability restriction. Despite the need of a small time-step, the method can be
computationally efficient because no matrices of the complete finite-element
assemblage need be calculated. Using the implicit method, the time-step can be
longer than that of the stability restriction previously mentioned. However, equi-
librium iterations are normally performed to ensure a realistic solution.
In ABAQUS and SUPER-TASEF the latent-heat option can be used to simulate
the phase change of water in concrete. (The phase change causes an energy con-
sumption that delays the temperature increase in the structure.) By this means, the
free water in concrete is mainly considered, here 5%. The behavior is modeled by
adding the amount of the heat of the evaporation of the water to the specific heat-
temperature relationship at a certain temperature interval, here 100 to 115°C. On
the other hand, TASEF models the phase change in the specific volumetric
enthalpy-temperature relationship. This is accomplished by a sudden increase in
specific volumetric enthalpy in the pre,dously mentioned temperature interval.
In ABAQUS, it is possible to write auser subroutine to provide internal heat gen-
eration and possibly state variable update. This routine can handle phase changes
occurring during the solution and the influences of stare variables, such as the frac-
tion transformed, which themselves evolve with the solution. Thus, moisture diffu-
sion can be handled.

Thermal Analysis
Various concrete structural members have been studied, such as rectangular beams
(Structure A), slabs with circular enclosures (Structure B), and steel-concrete com-
posite slabs (Structure C). No slab is attached to the beam in A, which means that
only one-fourth of it has to be modeled (see Figure 1).
The structures analyzed were modeled by four-node, linear-interpolation, heat-
transfer elements (DC2D4). This concerns both the concrete and the steel. Heat
transfer to the exposed assembly is characterized in terms of boundary conditions.
Three types of boundary conditions are permitted within ABAQUS, specified by
total heat flux and specified surface temperature. Furthermore, different ways of
describing convection and radiation can be studied. The heat transfer at a boundary
is defined, using the CONVECTION and RADIATION options, by the time-depen-
dent fire temperature surrounding the structure.
Moreover, the heat transfer can be modeled by interface elements using the GAP
CONVECTION and GAP RADIATION options. The time-dependent fire temper-
ature in the gap of the interface element defines the heat transfer. (Auser subrou-
tine can be written that altows the gap characteristics to be a function of state vari-
ables and time.) The interface element DINTER2 was used. The time history of the
fire temperature is given in tabular form, and a comparatively large number of
points should be defined for nonlinear problems. Using another boundary condi-
tion, however, it is possible to define time functions as a nonuniform film coeffi-
cient and associated sink temperatures. The user can define a rate of change for the
294 Fire Technology Third Quarter 1994

Analyzed quadrant

A ) H e a t transfer in a concrete b e a m

IS0834 Standard fire

B) H e a t transfer in a concrete
slab w i t h a circular enclosure

IS0834 Standard fite

Concrete 5% water
i . . . .

C) H e a t transfer in
a steel-concrete
composite slab

Steel profile

ISO834 Standard fire

Figure 1. Structures analyzed, and nodes used for

temperature-time plots.
 Temperature Analysis Using A B A QUS 295

TABLE 1
Material Properties
i
Concrete Steel
L
Temperan4re Specific Heat Conductivi O, ~ Temperature Specific' Heat i Conductivity :
.......... I . . . . . . . . . . .

oC J/kgOC J/ms°C °C J/kg°C i J/ms°C

0 1210 1.94 0 469 i 60.0
! 100 1210 1.37 200 525 5t.8
!! tt5 1000 1.28 400 611 I 43,5
!I 200 1000 1.14 500 665 i~
l
39.4
600 1000 0.94 600 727 35.3
I000 1000 0.66 700 796 31.1
1200 I000 0.52 800 873 27.0
1500 1000 0.3I 2000 873 27.0

film coefficient with respect to surface temperatures. (Defining this value improves
the rate of convergence during the nonlinear solution, especially when the film
coefficient is a strong function of surface temperature.) In this study, both convec-
tion and radiation were taken into account.
Resultant emissivities of 0.8 were used for both the exposed and unexposed
faces. A convection factor and a convection power of 1.0 W/m~°C and 1.33 were
used for the exposed face, and 2.2 W/m~°C and 1.25 were used for the unexposed
face. The structures were exposed to the standard ISO 834 fire.
Several references to the temperature-dependence of the material properties for
concrete and steel can be found in the literature surveys. H.,_,:~3The properties used
here were obtained from the RILEM handbooks. ~- '~ Details of the material prop-
erties are given in Table 1. The density is assumed to be constant at 2,300 kg/m ~
for concrete, and the moistur- dependence is included in the "Specific Heat" and
"Latent Heat" options. For steel, the density is assumed to be 7,800 kg/m 3 and
constant.

Results
Results are shown in Figures 2 to 7. The influence of phase changes on the tem-
perature history is evident at about 100 to 115°C and can be seen already at the sur-
face temperatures. The influence of conductivity becomes more significant with the
depth of the structures. Due to symmetry, generally only a half of the structure
needs to be analyzed (one-fourth for the beam). The contours are shown for
ABAQUS and SUPER-TASEF temperatures. In the latter case, the RRVA option is
used for Structures B and C, meaning that the plots are reflected around the verti-
cal axis. For TASEE however, only the maximum temperatures are shown, exem-
plified on the rectangular beam.
296 Fite Technology Third Quarter 1994

ABAQUS TASEF SUPER-TASEF

contour levels

TEMP
VALUE

17 9.45E+02 a = 0.102E+04
18 1.00E+03 b = 0.105E+04
19 1.06E+03 c = 0.107E+04
20 1.12E+03 d = 0.110E+04
21 1.18E+03 e = 0.113E+04

Figure 2 . Temperature distributions of Structure A, time_-4.0

hours. No contour plots for TASEF, only maximum tempera-
tures. Temperatures in °C.
Temperature Analysis UsingABA QUS 297

ABAQUS SUPER-TASEF

TEMP
VALUE contour levels

1 5.00E-05
2 5,00E+01 a = 0.500E+02
3 1,00E+02 b = 0.145E+03
4 1.50E+02
5 2.00E+02 c = 0.240E+03
6 2.50E+02
7 3.00E+02 d = 0,335E+03
8 3,50E+02 e = 0.430E+03
9 4.00E+02
10 4.50E+02 f = 0.525E+03
I1 5.00E+02
12 5.50E+02 g = 0.620E+03
13 6.00E+02 h = 0,715E+03
14 6.50E+02
15 7,00E+02 i = 0.810E+03
16 7.50E+02
17 8.00E+02 J = 0.905E+03
18 8.50E+02 k = 0.100E+04
19 9.00E+02
20 9.50E+02
21 1.00E+03

Figure 3 . Temperature distribution of Structure B, t i m e - - 2 . 0

hours. No results are shown for TASEF. Note contour plots indi-
cating adiabatic boundary at the enclosure using ABAQUS.
Temperabures in °C.
298 Fire Technology Third Quarter 1994

ABAQUS SUPER-TASEF

a = 0.500E+02
b = 0.100E+03
TEMP c = 0.150E+03
VALUE d = 0.200E+03
e = 0.250E+03
1 1.00E-04
f = 0.300E+03
2 1.00E+02
g = 0.350E+03
3 2,00E+02
h = 0.400E+03
4 3.00E+02
5 4.00E+02 i = 0.450E+03
6 5.00E+02 j = 0.500E+03
7 6.00E+02 k = 0.550E+03
8 7.00E+02 1= 0.600E+03
9 8.00E+02 m = 0.650E+03
l0 9.00E+02 n = 0,700E+03
I1 1.00E+03
o = 0.750E+03
p = 0.800E+03
q = 0.850E+03
r -- 0 . 9 0 0 E + 0 3

Figure 4. Temperature distributions of Structure C, t i m e - - l . 0

hour. No results are shown for TASEF. Note contour plots for
temperatures of the interface elements using ABAQUS.
Temperatures in °C°
TemperatureAnalysis UsingABAQUS 299

Only results from ABAQUS using the GAP CONVECTION and GAP RADIA-
TION options are shown, particularly in Figure 4, where the temperature contour
of the interface elements have not been suppressed. Normally, the temperatures
using the CONVECTION and RADIATION options were about 5 to 10% lower. In
some calculations, the symmetry at the adiabatic boundary falls using ABAQUS.
However, the behavior shows a strong mesh dependence and declines with mesh-
refinement and the number of contour lines specified. Similar results have been
found using computer codes that are not studied here. ''»
In the enclosure of Structure B, all three models of heat transfer--conduction,
convection, and radiation--affect the results. Using ABAQUS, the enclosure was
assumed to be a perfect insulator (view factors are lacking in this computer code).
However, some comparisons can be made to the other codes, since two cases have
been studied for these--those with heat transfer, and those in which which heat

I I I r I J i I 1

I. 10E+03

1.00E+03

9,00E+02

8.00E+02

b
7.00E+02

E 6.00E+02

E r~
'C,
5.00E+02
'õ

4.00E+02 g

3.00E+02
,I
/.; . . . .
ABAQUS
SUPER-TASEF
2.00E+02 ~ - ~ TASEF
-d

1.00E+02

-I l I [ [ r I I I I [

time

Figure 5. Temperature-time histories of Structure A.

300 Fire Technology Third Quarter 1994

transfer was not considered, as the enclosure was assumed to be a perfect insulator.
In cases in which heat transfer is not considered, close agreement is noted,'but this
cannot be expected generally. Some differences can be found between the TASEF
results and the SUPER-TASEF results, as shown in Figure 6. This may be due to
different ways of modeling the heat transfer in the enclosure. The latter code takes
radiation into account, while convection is included in the former, as well. '7 The
analyses of the steel-concrete composite slabs are mainly of theoretical interest. In
reality, an air-gap caused by the steel profile deformations has a great influence on
the heat transfer from profile to concrete. '~

Discussion a n d Conclusions
The results of the ABAQUS runs were encouraging in s0me cases. Reasonable
agreement with results of the reference codes TASEF and SUPER-TASEF were
found for two of the concrete structures studied, namely Structures A and C, which

L . I O E + 0 3 ~ I.I0E+03
1.00E+03B . ;~~ ~ . ~ b ~ b'~" 1.00E+03 B.@~,./1~b/
9.00E+02I ,~°\/ / 9.00E+02
#,4@G b ~
~.ooE+o2[- ~")~ / ~ ' ~.ooE+o2

| 7.00E+02f~/ ~? ~ 7.00E+02 ~_.._~~~~~«ASEF

b .-7

4.00E+02 ~ 4.00E+02
3'00E+02V]/ ~ ABAQUS
~-~b]a - - - - SUPER-TASEF
2,00E+02]Il ~ - ~ TASEF
3.00E+02
2.00E+02
g~
,Y ./."
! ..~2;."t , c " "
1,00E+02[/1 1.00E+02
a
g g g g g g g g

a) time b) time

Figure 6. Temperature-time histories of Structure B. a)

Enclosure modeled as perfect insulator for ABACIUS and the
TASEF codes, b) Enclosure radiation for SUPER.TASEF, and radi-
ation and convection for TASEF.
Temperature Analysis UsingABAQUS 301

were a concrete beam and a steel-concrete composite slab, respectively. However,

all codes have their respective advantages and disadvantages. In studying Structure
B, a concrete slab with a circular enclosure, it became evident that ABAQUS
lacked view factors. Hence, the temperatures due to the heat transfer in the enclo-
sure could not be realistically modeled. Moreover, agreement with the reference
codes was primarily found using interface elements for ABAQUS. Thus, the time-
dependent fire temperature in the gap of the interface element defined the heat
transfer. In the initial parts of the time-histories, the temperatures inside the struc-
tures normally showed slower increases for ABAQUS than for the reference codes.
The interface elements complicated the input handling, making ABAQUS com-
paratively less user-friendly. On the other hand, the powerful element-generation
technique and hode numbering flexibility--they need not increase uniformly--and
the possibility of conducting 3-D analysis were nice features.
In addition to the reference codes, ABAQUS can use solid elements, such as reg-
ular four-sided quadrilateral elements, as weil as triangles. In ABAQUS, however,
8-node elements (20 nodes in the 3-D case), as weil as beam and shell elements
using quadratic interpolation, can be studie& One can quite easily use a mesh or

9.00E+02 i t i i n a

8.00E+02 r ~'c°'+
C,¢:"~e
a b
7.00E+02

6.00g+02

0.
5.00E+02
E
4.00E+02 / - - ABAQUS
/
/ .... SUPER-TASEF
t ---- TASEF
3.00E+02

2.00E+02 / .'~-'5
/ ~~3/"
ù~~;-
I.OOE+02 / _ ..~s ~ ~
..~":~~."t,. . . . _--~-_"---- ' ~

time

Figure 7. Temperature-time histories of Structure A.

302 Fire Technology Third Quarter 1994

the 8-node problem to analyze a 4-node problem and to study the influence of var-
ious interpolations. The importance of beam and shell elements is normally
revealed in a temperature-stress analysis. These elements usually show superior
computational efficiency compared to solid elements. The use of internal nodes
gives ABAQUS its node-numbering flexibility, and automatic bandwidth opti-
mization similar to SUPER-TASEF is provided.
One nice feature of SUPER-TASEF is its ability to plot results during a compu-
ration, and hence to stop a bad run. Using ABAQUS, one can get similar informa-
tion by reading the statistical file during a run. Both codes have a built-in automat-
ic and adaptive choice-of-time incrementation. However, at least twice as many
time-steps were usually needed for a specific, implicit time-integration problem
using SUPER-TASEF than were needed when using ABAQUS (default parameters
of the time-incremenation algorithms were used).
Only limited conclusions could be drawn about which code was the fastest. The
TASEF and SUPER-TASEF computations were performed on an IBM PC PS/2
Model L40 SX using a 387 processor, while a DEC Workstation 3100 was used for
the ABAQUS and SUPER-TASEE These factors influenced the computational
time. However, no significant differences in CPU time could be found when com-
paring ABAQUS and SUPER-TASER For the Structures A, B, and C, the times
were 3 to 4 minutes, 15 to 18 minutes, and 5 to 6 minutes, respectively. The TASEF
runs were normally at least 10 times faster than the problems studied.
No significant differences in temperature due to various time-integration tech-
niques or different definitions of the phase changes could be found. However, the
influences of phase changes on nodes inside the structures were more pronounced
when using SUPER-TASEF than they were when either ABAQUS or TASEF were
used. The smaller time-steps of SUPER-TASEF might contribute to this.
Finally, the numerical results obtained indicate that the present FE models can
satisfactorily analyze the temperature distributions of Structures A and C. With
respect to temperature analysis exclusively, no particular advantages were found
for using ABAQUS instead of the reference codes.

References
1. Mackay, D. C. and Thompson, D. B., "Simulation of Fire Tests by Finite
Element Analysis," StructuraI Design for Hazardous Loads, Institution of
Structural Engineers, E & FN Spon, London, 1992, pp. 415-424.
2. Amdahl, J. and Hellan, Ö., "Progressive Collapse Analysis of Steel and
Aluminum Structures Subjected to Fire Loads," Structural Design f(»" Hazardous
Loads, Institution of Structural Engineers, E& FN Spon, London, pp. 393-402.
3. Judge, R. C. B., "Modeling the Behaviour of Reinforced Concrete Structural
Members Subjected to Fire Loading," ABAQUS Users' Conference Proceedings,
Temperature Analysis Using ABAQUS 303

Stresa, pp. 147-t60.

4. Judge, R. C. B., "Analysis of Steel Structural Members Subjected to Fire
Loading," AEA Technology, Report RTS-M-081, Risley, 1989.
5. HKS-Hibbitt, Kartsson & Sorensen Inc., ABAQUS 5.2, User's Manual,
Providence, 1992.
6. Sterner, E. and Wickstr6m, U., TASEF Temperature Analysis of Structures
Exposed to Fire, Swedish National Testing and Research Institute, Bor~s, 1990.
7. Gustafsson, H., Carlsson, R and WickstrOm, U., SUPER-TASEF--
Temperature Analysis of Structures Exposed to Fire, Swedish National Testing and
Research Institute, SP Report, Bor~s, 1993.
8. Milke, J.A., "Temperature Analysis of Structures Exposed to Fire," Fire
Technology, National Fire Protection Association, Quincy, May, 1992, pp. 184-189.
9. HKS--ttibbitt, Karlsson & Sorensen, Inc., ABAQUS 4.9 User's Manual,
Providence, 1990.
10. Iding, R.H., Nizamuddin, Z. and Bresler, B., Computer Progl~m/[?)r the Fire
Response of Structures--Thermal Three-Dimensional Version, University of
California, Berkeley, I977.
1 l. IRB-Verlag, Brandverhalten yon Metallen, lnformationszentrum Raum und
Bau, Literaturauslese Nr 1996, Stuttgart, 1988, p. 109.
12. IRB-Vertag, Brandverhalten von Bemn, Informationszentrum Raum und Bau,
Literaturauslese Nr 1999, Stuttgart, 1988, p. 78.
13. IRB-Verlag, Brandverhalten yon Stahl und Spannbeton, Infotmatiotlszentrum
Raum und Bau, Literaturaustese Nr 2003, Stuttgart, 1988, p. 114.
14. Anderberg, Y., ed., RILEM, Properties of Materials at High Temperatures--
Steel, Lund Institute of Technology, 1983.
15. Schneider, U., ed., RILEM, Properties of Materials at High Temperatures--
Concrete, Kassel Universitfit, 1985.
16. Anderberg, 5q and Pettersson, O., Brandteknisk Dimensionering av
Betongkonstruktioner, Del 1, (in Swedish), T13:1992, Byggforskningsr~det
Stockholm, 1992, p t19.
17. WickstrOm, U., A Numerical Procedure for Calculating TempeJvture in
Hollow Structures Exposed to Fire, Lund Institute of Technology, Department of
Structural Mechanics, Report No. 79-3, Lund, Sweden, 1979.
18. Saito, H. et aI., Study on Behavior of Concrete with Steel Liner under High-
Temperature Condition, SMiRT 11 ~IYansactions, Vol. H, Tokyo, 1991, pp. 85-90.