FDS2FTMI User’s Guide

An automated code to one-way coupling

between FDS and FEM using FTMI

Julio Cesar Silva

October 21, 2017

FDS Version 6.6.0
Revision: FDS6.5.3-3014-g53fffa5ea
Contents

1 Introduction 1

2 Getting Started 3
2.1 Adiabatic Surface Temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2 Capturing the Mesh Information from FEM . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.3 Executing the code . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

3 Verification Cases 9
3.1 Simple panel exposed to radiation and convection (simple_panel_hot) . . . . . . . . . . . 9
3.2 H profile column exposed to a localizes fire (h_profile) . . . . . . . . . . . . . . . . . . . . 10
3.3 Summary of Verification Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

Bibliography 13

i
ii
Chapter 1

Introduction

This document explains the developed strategy to evaluate the behavior of structures under fire conditions by
a coupled analysis including Fire Dynamics Simulator (FDS[1]) and Finite Element Method (FEM) codes.
Common practice is to extract the results from a fire simulation, and apply those as boundary conditions
at the exposed surfaces for the thermomechanical analysis. This kind of approach is commonly referred
as one-way coupling. In a two-way coupling strategy, the thermomechanical results, e.g., displacements,
collapses, etc. are transposed back to the fire simulation. The two-way approach can lead to a more complex
simulation, increasing the amount of data to be transferred between the models. Its advantages are related
to cases where displacements can change the fire source or the fluid flow pattern, creating a different fire
scenario. Using one-way coupling it is possible to establish each model separately, which means that the
same fire simulation results can be used on slightly different thermomechanical models, contributing to
development of optimized structures.
The Fire-ThermoMechanical Interface [2] is a one-way coupling procedure that uses the Adiabatic Sur-
face Temperature (TAST ) concept [3, 4] to establish an interface between FDS and FEM codes. The procedure
recreates the net heat flux evaluated in FDS, in the FEM environment by transferring TAST and the heat trans-
fer coefficient (h) from FDS results and adopting the wall temperature (Tw ) provided by the FEM code at
each time step (h can also be adopted as a constant). This heat flux can be calculated by FEM over the
elements faces and applied into each node of the exposed surface. FDS2FTMI is a code created to perform
these tasks in an automated way, allowing the simulation of the behavior of global structures, discretized
with shell and/or solid elements, exposed to fire conditions. FTMI can be used with any FEM code, at this
time, FDS2FTMI is implemented between FDS and ANSYS or between FDS and LS-DYNA 1 .
The code FDS2FTMI is included in the FDS-SVM repository [5] under Utilities/Structural_
Interaction/fds2ftmi. A makefile is provided to help in the compilation process with different oper-
ational systems and compilers. A shell and a bat scripts (scripts/build_fds2ftmi.sh, scripts/
build_fds2ftmi.bat) are also provided to build this manual, run the Verification cases and check the
results 2 .
Some validation work done with this methodology is provided at Zhang et. al [6].

1 Certain commercial entities, equipment, or materials may be identified in this document in order to describe an experimental

procedure or concept adequately. Such identification is not intended to imply recommendation or endorsement by the author or the
National Institute of Standards and Technology, nor is it intended to imply that the entities, materials, or equipment are necessarily
the best available for the purpose.
2 Two environment variables are needed to run these scripts: FDS_GITROOT, which is the path for fds-smv git folder; and

ANSYS, which is the path to the ANSYS executable (including the executable).

1
2
Chapter 2

Getting Started

Heat can be transferred from flames and hot gases to structures surfaces by radiation and convection. The
total heat flux (q̇00tot ) is defined by the sum of these two parcels:

q̇00tot = q̇00r + q̇00c = ε einc − σ Tw4 + h(Tg − Tw )

(2.1)

where q̇00r is the radiative heat flux, q̇00c is the convective heat flux, ε is the emissivity, einc is the radiative ther-
mal energy incident on the exposed surfaces, σ is the Stefan-Boltzmann constant, Tw is the wall temperature
and Tg is the gas temperature.
Advanced fire simulation models provide results that characterize the three dimensional evolution of
the fire, the radiative thermal energy incident on the exposed surfaces and the gas temperatures. However,
the models are not capable of accurately characterizing the temperature distribution on solids, and thus the
surface temperature. Consequently, the total heat flux, as presented on Eq. 2.1, cannot be precisely calculated
at the end of the fire simulation. Therefore, an additional approach is necessary to cross this barrier.

2.1 Adiabatic Surface Temperature

In order to establish an accurate interface between the fire simulation and the thermomechanical analysis, the
Adiabatic Surface Temperature concept is used [3]. A real surface can be replaced by a perfectly insulated
surface exposed to the same heating conditions; the total heat flux to this ideal surface is by definition zero :
4


ε einc − σ TAST + h(Tg − TAST ) = 0 (2.2)

and the temperature of this idealized surface can be obtained as an output from the fire simulation.
Since the real and the hypothetical surfaces are exposed to the same heating conditions, the total heat
flux that will be applied to the thermomechanical model can be obtained by the solution of the system formed
by Eqs. (2.1) and (2.2):
q̇00tot = εσ TAST
4
− Tw4 + h(TAST − Tw )


(2.3)
This single variable (TAST ) is considered capable of reducing the complexity of the fire simulation into
one simple scalar [7, 8, 9]. To achieve a correct definition of the total heat flux, accounting correctly both
the convective and radiative heat fluxes, it is also necessary to add the convective heat transfer coefficient
(h).
During the FDS simulation, TAST and h can be recorded via DEVC or BNDF outputs. To use DEVCs, the
points (or nodes) where the information will be collected need to be determined in advance, before the fire
simulation. On other hand, BNDF generates outputs at all exposed surfaces (boundary cells marking the
interface between gas and solid) into the domain. FDS2FTMI extracts information stored in BNDF files, so

3
the coupling procedure can be achieved for different discretization levels (at FEM). Small modifications or
dimensioning in FEM does not imply restarting the fire simulation.

2.2 Capturing the Mesh Information from FEM

In FDS, the domain is divided into orthogonal cells in a rectilinear grid and geometries are approximated by
blocks (OBSTs). For FEM thermomechanical analysis, the structure is usually discretized with solid (tetrahe-
drons or hexahedrons) or shell (3D plane) elements. Also, the element size needed at the thermomechanical
model can lead to an unfeasible fire simulation. Even for modest structures, with simple geometries, the
models generated for each side of this interface may not match perfectly. Therefore, the communication
between the two models is based on the spatial position of the nodes on the exposed surfaces (FEM) and
the cell faces (FDS). The geometries created for both models need to share a common Cartesian coordinate
system, i.e., they need to have the same origin and they need to be consistent. FDS2FTMI main function is
to read the BNDF files and extract TAST and h at the exposed surfaces. This code also generates the files that
need to be read by the FEM code to apply correct boundary conditions. The source code and a makefile can
be found at Utilities/Structural_Interaction/fds2ftmi/scripts.
The input parameters for FDS2FTMI are the spatial position of the exposed surfaces at FEM model.
These parameters are divided in 2 files, “nodes.dat” and “elements.dat”, and include the node positions
(Cartesian coordinates) and the node numbers at the exposed faces ordered using right-hand rule with normal
pointing outwards.

ANSYS
The interface with ANSYS is developed using surface effect elements, so the information presented in the
files “nodes.dat” and “elements.dat” are related to those surface effect elements. To help in the generation
of the surface effect elements and these two input files, some ANSYS macros were created. It is possible
to export the mesh information using the macros “ansys2ftmi.geom” or “ansys2ftmi_shell.geom”. The
first can be used for solid elements and the second one is devoted to shell elements. The shell element
file creates interfaces on both sides (top and bottom layers) of the shell elements 1 . These macros were
designed to create interface elements attached to the boundary element faces between solid and gas. To
select just the exposed surfaces related to a specific scenario, those surfaces can be selected in ANSYS before
running the macros (Select Entities > Areas). The macros are available at Utilities/Structural_
Interaction/fds2ftmi/scrips.

LS-DYNA
To create this interface with LS-DYNA, only the information of the actual mesh needs to be presented in the
files “nodes.dat” and “elements.dat”.

Examples of these two files are given below:

nodes.dat
The first line of this file has the number of nodes and the highest node number in the FEM model. All other
lines have just the node number and the Cartesian coordinates (x, y and z).

100 2400
1 1.000 0.000 0.000
1 Tohelp convert an ANSYS model geometry into OBST, an ANSYS macro (“ansys2fds_areas_obst.geom”) that convert areas
into OBST is available at scripts.

4
 2 1.000 0.100 0.000
3 1.000 0.200 0.000
4 1.000 0.300 0.000
.
.
.
100 1.000 1.000 1.000
In this example, the exposed surface has 100 nodes and the highest number at the FEM model is 2400. So,
when the code generates nodes for interface elements it starts in 2401. All additional lines have the nodes
coordinates.

elements.dat
The first line of this file has the number of interface elements that compose the exposed surface, and 3
additional flags. First flag says the type of element used in FEM model: 0 means solid elements and 1
means shell elements. If the thermomechanical model uses shell elements, the interface is defined in both
top and bottom layers. The second flag defines in which layer the heat flux is prescribed: 1 means top layer
and 2 means bottom layer. The last flag defines the number of the interface element type in ANSYS.

200 0 0 2
801 120 2 148 117 121 149 139 118
802 126 128 130 151 127 144 125 150
803 118 120 117 135 119 118 157 136
804 128 130 128 126 129 145 127 144
.
.
.
1000 178 163 41 69 179 164 10 80
END

In this example, there are 200 interface elements in this model, they are attached to solid elements (so
the first and second flag are 0) and they correspond to the element type 2 in the FEM model. The code
FDS2FTMI can work with bilinear or linear elements, which have a different number of nodes. In the case
of using linear elements, the non-used positions are filled with zeros.
In the next example there are 72 interface elements attached to top layer of shell elements (from 73 to
144) and another 72 applying the heat flux to bottom layer of the shell elements (145-216). The input file
structure remains the same, but it is split into two parts, each one with its own flags.

72 1 1 2
73 1 8 10 2 0 0 0 0
74 2 10 9 3 0 0 0 0
75 4 11 12 6 0 0 0 0
76 2 10 11 4 0 0 0 0
.
.
.
144 82 91 90 84 0 0 0 0
72 1 2 3
145 8 1 2 10 0 0 0 0

5
 146 10 2 3 9 0 0 0 0
147 11 4 6 12 0 0 0 0
148 10 2 4 11 0 0 0 0
.
.
.
216 91 82 84 90 0 0 0 0
END

2.3 Executing the code

After the definition of the files nodes.dat and elements.dat, the code FDS2FTMI is executed to extract
information from FDS boundary files, this code requires some information about the FDS model:

CHID CHID drives the naming convention of FDS output files. It is needed to give the path to the code
get information from .smv and .bf files.

Cell size This is used to create a virtual cell over the node position to search for results in BNDF files. It is
generally adopted as the same value as the cell size in FDS simulation.

Number of variables The number of variables (1 or 2) is used to determine if the interface will get h from
FDS simulation or use a constant value. If the number of variables is defines as 2, the code will extract both
TAST and h from FDS simulation. If the number of variables is defined as 1, the code will get just the results
for TAST and a constant value of h is prescribed in ANSYS. So, just in this case, the next value on this same
line is h.

Time To address the time boundary, the code needs the starting, ending time and the time interval. The
time interval that the results are written into BNDF files can be controlled by DT_BNDF at DUMP line, the
default is given by 2∆t/NFRAMES.

Number of average points In some scenarios, a constant exposure condition can be assumed. To calculate
a constant steady-state exposure condition from a short time history, a number of average points should be
provided. If NO average is wanted, 0 (zero) should be prescribed.

Time average The time step at the FEM model can be bigger than the time interval when FDS write
results. In this case, an average needs to be performed to calculate the correct thermal exposure of the
surfaces. If NO average is wanted, 0 (zero) should be prescribed.

FEM Code This is used to tell FDS2FTMI which FEM code will be used in the thermomechanical anal-
ysis. At this moment, ANSYS and LS-DYNA is supported, and this variable should be equal to 1 for ANSYS
and 2 for LS-DYNA.

Output file The code will generate an output file containing all the commands that need to be read by
ANSYS (File > Read input from). This ouput file will create all elements, parameters and loads necessary to
input the heat flux from the flames and hot gases into the nodes of the exposed surfaces.

6
 These information can be passed to the code via arguments (command line), input file or just typing as
the code steps through execution. As arguments or input file, the parameters need to be given at the same
order they are presented above. For example:

fds2ftmi_linux_64 simple_panel_hot 0.05 2 0 600 2 0 0 1 simple_panel

or

fds2ftmi_linux_64 simple_panel_hot.fds2ftmi

with simple_fire_hot.fds2ftmi:

simple_panel_hot
0.05
2
0 600 2
0 0
1
simple_panel

start the code with simple_panel_hot as CHID, a cell size equal to 0.05, 2 variables, starting at 0s to 600s
with 2s as time interval, with no average points, no time average, exporting the results in ANSYS format and
writing the results at “simple_panel.dat”.
The code will search into the BNDF files for the results. Searching over every surface in the FDS simu-
lation can slow down this process. To reduce the size of the BNDF files, add BNDF_DEFAULT=.FALSE. in
MISC line and the add BNDF_OBST=.TRUE. at the exposed surfaces OBST lines.

7
8
Chapter 3

Verification Cases

This chapter contains the Verification cases that test the FDS2FTMI code. The SVN number on the figures
is related to the FDS repository [5] subversion control number used to compile the codes and run the cases.

3.1 Simple panel exposed to radiation and convection (simple_panel_hot)

This example verifies the FDS2FTMI ability to translate the heat transfer between solid and gas phase at fire
simulation into boundary conditions for a FEM model (ANSYS). It also demonstrates that the code can extract
correctly the results from BNDF files. This model is composed by 2 panels, one has a constant temperature
of Ts = 1000 °C and the other one has an initial temperature of Ts = 20 °C. The temperature evolution at the
later panel is the target of this exercise (the input files are provided in examples/simple_panel_hot).
The case is set to be 2D, with only one cell at y axis. These panels are 1 m height (z axis) and they are 1 m
from each other. Three DEVCs were setup to compare the results, they are placed at 0.775 m (1), 0.525 m
(2) and 0.275 m (3) in z axis. After the fire simulation, the boundary conditions are translated to ANSYS
by FTMI. A comparison between the Adiabatic Surface Temperature TAST and the surface temperature Ts
obtained by FTMI and FDS is provided in Fig. 3.1.

)'6JIIIDHD )'6JIIIDHD
 6XUIDFH7HPSHUDWXUHVLPSOHBSDQHOBKRW
 $GLDEDWLF6XUIDFH7HPSHUDWXUHVLPSOHBSDQHOBKRW 
7HPSHUDWXUH &

7HPSHUDWXUH &

 )'67


)'67
 )'67
 $16<67
)70,7  $16<67
)'67 $16<67
 
           
7LPHPLQ 7LPHPLQ

Figure 3.1: Comparison of the results obtained by FTMI (with ANSYS) and FDS.

9
3.2 H profile column exposed to a localizes fire (h_profile)
In this case a simple supported H-profile column is exposed to an adjacent hot panel (Ts = 1000 °C). The
steel column is 3 m tall (z axis). The cross section is a 0.3 m (flange, parallel to x) x 0.4 m (web, parallel
to y) with a 12.5 mm thick web and 16 mm thick flanges. The hot panel is parallel to the web and localized
30 cm from the web in x direction (or 15 cm from the end of the flange). The column is discretized with shell
elements (FEM) to verify the use of the code to model shell structures. Each side of these plane elements will
have a different thermal exposure, incident radiation and gas temperature around the surface. FDS2FTMI is
designed to prescribe the correspondent heat flux at the shell elements top and bottom layers. Points A, B
and C are located at the column middle span. A and C are located at the end of the flange (closest points to
hot panel), B is located at the middle of the web.
The target of this qualitative verification exercise is to check the thermal gradient due to heat fluxes
applied in top and bottom layers, the spatial temperature distribution and the thermomechanical behavior
due to heating. The evolution of the surface temperature is presented in Fig. 3.2. As the hot panel is parallel
to the web, the heat flux incident to its surface will be higher than the heat flux incident to the flanges, so the
surface temperature at A is higher than at B. Even with the high conductivity of steel is possible to observe
a little gradient between the exposed and back surfaces at A and B. The spatial distribution also presented in
Fig. 3.2 shows the expected profile. This profile is not exactly symmetric through column mid-span because
of the energy reflected by the ground and the variation on the gas velocities around the column (which
changes the heat transfer coefficient - h), so the temperature is lower at the top of the column.

)'6JIIIDHD
OCT 21 2017 OCT 21 2017
11:30:44 11:30:44
h_profile_M_ansys h_profile_M_ansys
PLOT NO. 1 PLOT NO. 2

 425 425

7HPSHUDWXUH &

485 485

 545 545

7VB$
605 605
 7VB$BEDFN
7VB% 665 665
7VB%BEDFN
6XUIDFH7HPSHUDWXUHKBSURILOH
 725 725
   
7LPHPLQ

Figure 3.2: Evolution of surface temperature and spatial distribution at 60 min.

This simple supported column is subjected to a vertical load of 325 kN, which correspond to 1/50 of the
Euler’s buckling critical load. Under these loading conditions, the column will be close to a uniform stress
state and remain straight at the beginning of the simulation, then the horizontal displacements generated will
be due to the temperature distribution. The evolution of horizontal displacements is presented in Fig. 3.3. As
expected, the horizontal displacements in x axis are equal at A and C and represents the maximum achieved
at the column, as can be observed at the lateral view. In y direction the value of the displacement at these
two points are the same, but as column is expanding and they are moving away from each other the signal
of these displacements are opposite.

10
 )'6JIIIDHD
 
)'6JIIIDHD
'LVSODFHPHQWLQ[D[LVKBSURILOH OCT 21 2017


11:30:44
h_profile_M_ansys

'LVSODFHPHQWFP
'LVSODFHPHQWFP
PLOT NO. 3

 -.005

 -.100E-02

  'LVSODFHPHQWLQ\D[LVKBSURILOHG\B$

G\B& .003

G[B$
 .007

G[B%  .011
G[B&

         .015

7LPHPLQ 7LPHPLQ

Figure 3.3: Evolution of horizontal displacements in x and y axis and a lateral view of the plane xz (displace-
ments amplified by x15).

3.3 Summary of Verification Results

This section summarizes the results from Verification cases with the objective of evaluate the output pre-
dicted by FDS2FTMI code. The presented results can help identify errors in the source code.

Case Name Expected Predicted Type of Error Error Error Within

Metric Metric Tolerance Tolerance
simple_panel_hot 4.865e+02 4.842e+02 Relative 4.651e-03 0.01 Yes
simple_panel_hot 4.128e+05 4.131e+05 Relative -6.241e-04 0.01 Yes
h_profile > 7.213e+02 7.221e+02 Logic - - Yes
h_profile 1.277e-02 1.169e-02 Absolute 1.078e-03 0.01 Yes
h_profile 2.235e-03 2.238e-03 Absolute 4.473e-03 0.01 Yes

11
12
Bibliography

[1] K. McGrattan, S. Hostikka, R. McDermott, J. Floyd, C. Weinschenk, and K. Overholt. Fire Dynamics
Simulator, User’s Guide. National Institute of Standards and Technology, Gaithersburg, Maryland,
USA, and VTT Technical Research Centre of Finland, Espoo, Finland, sixth edition, September 2013.
1

[2] J.C. Silva, A. Landesmann, and F.L. Ribeiro. Fire-Thermomechanical Interface Model to Performance-
based Analysis of Structures Exposed to Fire. Submitted to Fire Safety Journal, -(-):–, -. 1

[3] U. Wickström, D. Duthinh, and K.B. McGrattan. Adiabatic Surface Temperature for Calculating Heat
Transfer to Fire Exposed Structures. In Proceedings of the Eleventh International Interflam Conference.
Interscience Communications, London, 2007. 1, 3

[4] U. Wickström, R. Jansson, and H. Tuovinen. Verification fire tests on using the adiabatic surface temper-
ature for predicting heat transfer. Technical Report 2009:19, SP Technical Research Institute of Sweden,
Bo̊ras, Sweden, 2009. 1

[5] https://github.com/firemodels/fds. 1, 9

[6] C. Zhang, J.C. Silva, C. Weinschenk, D. Kamikawa, and Y. Hasemi. Simulation Methodology for
Coupled Fire-Structure Analysis: Modeling localized fire tests on a steel column. Fire Technology,
52(1):239–262, 2016. 1

[7] D. Duthinh, K. McGrattan, and A. Khaskiac. Recent advances in fire structure analysis. Fire Safety
Journal, 43:161–167, 2008. 3

[8] J. Sandström, U. Wickström, and M. Velkovic. Adiabatic Surface Temperature: A sufficient Input Data
for a Thermal Model. In Proceedings of Application of Structural Fire Engineering, 2009. 3

[9] U. Wickström, A. Robbins, and G. Baker. The Use of Adiabatic Surface Temperature to Design Struc-
tures for Fire. In Proceedings of Structures in Fire (SIF), 2010. 3

13