Professional Documents
Culture Documents
An
Introduction to
Mathematical
Fire Modeling
Preface xi
lntroduction xv
References 25 1
Index 257
PREFACE
Marc L. Janssens
San Antonio, TX
September 1999
INTRODUCTION
This thought provided motivation for the procedure taken during the
development of the model developed herein.
Indeed, the initial approach followed during the development of the
model presented was the upgrading of an existing model. That model was
ASET. ASET is perhaps the most basic mathematical compartment fire
model available today that accounts for time-varying conditions. However,
the model developed here is not simply a modification of ASET, but a
completely new model. Initial motivation for the simplicity and structure
of the model was so that it could be used as a tool in the investigation and
reconstruction of fires. With this in mind the model developed herein was
titled FIRM, an acronym for Fire Investigation and Reconstruction Model.
As with ASET, the model is certainly not limited to those uses suggested
by its name. The ASET model did have an important role in the
development of FIRM. As such, Chapter 3 will address the ASET model
with primary emphasis on the development of the governing mathematical
equations.
The more complex models mentioned above also influenced the
development of FIRM. It is beyond the scope of this book to present these
models. However, the physical concepts that are at the basis of these
complex models will be reviewed in Chapter 2. This will provide the
reader with a good introduction to the discussion of the ASET and FIRM
models in subsequent chapters. The numerous references listed should be
consulted for a more complete discussion of the individual models and fire
modeling in general. A comprehensive list of fire models, with a concise
summary of and references to each model, can be found in the survey by
Friedman 151.
CHAPTER 1
Basic Compartment Fire Theory
INTRODUCTION
I.l
I l FUEL I
1 Conduction heat transfer through the fuel
2 Radiative heat losses f r o m the fuel surface
3 Convection a n d radiation from the flame
4 Flow of combustible fuel volatiles
5 Entrained a i r flow
6 Plume mass flow
enclosure. Just prior to and afker flashover, the thermal radiation fkom the
hot layer can also have a pronounced effect on the burning rate because the
increased heat flux incident on the fuel surface will increase the pyrolysis
rate and thus the burning and heat release rates. The concepts and
importance of flashover and the effects of hot-layer radiation to fuel
surfaces will be discussed in greater detail in Section 4.5.
Not all of the heat energy released by the fire remains within the
enclosure. At any given time, a portion of the energy released by the f r e
is radiated away from the combustion region and rising plume. A major
hction of this energy is received by the interior surfaces and is conducted
away fiom the interior of the enclosure through the bounding surfaces of
the enclosure 161.
The energy not radiated away fiom the combustion region and plume
is convected up through the plume into the forming hot layer. A fraction
of this energy is lost fiom the hot layer to the bounding surfaces through
convection and radiation. Once again, the energy imparted to the walls is
conducted througb the walls from the hot interior surface towards the cool
rear surface. The energy transferred to the bounding walls also results in
the elevation of the temperature of the walls. The energy remaining in the
hot layer is directly responsible for the increase in gas temperature of the
smoky hot layer. As will be shown, the energy and mass of the hot layer
are the unknowns that must be found in order to calculate the size and
temperature of the hot layer within the enclosure.
The other major source of energy loss from the hot layer is due to
hot-layer venting. Hot-layer venting also accounts for mass loss from the
hot layer. As a hot layer forms, it will descend. If a vent is present within
an enclosure, the hot layer may eventually drop below the top (or soffit) of
the vent. At this time, hot gas will flow out of the compartment due to the
pressure differential created by the temperature differences of the interior
and exterior (ambient) gases. Associated with this hot gas flow is the loss
of mass and energy that is important in the calculation of the hot-layer size
and temperature.
The inflow of relatively cool, ambient-temperature air, through the
lower portion of an opening (vent) in the enclosure boundaries, is partially
responsible for providing the owgen required for combustion within the
fire compartment. It is assumed that the ambient air supply is
uncontaminated, at a constant temperature, and that the supply available
is sufficient to provide the necessary volume needed throughout the course
of the fire. If insufficient amounts of air enter the compartment, the fire
4 BASIC COMPARTMENT FIRE THEORY
may become oxygen starved. This condition results in reduced heat release
rates and an increase in the products of incomplete combustion. Therefore,
it is quite important to consider the flow of air into an enclosure and the
oxygen needed for complete combustion. Figures 1-2(a)- (d) depict the
typical stages of a developing compartment fire with respect to
compartment venting. A more detailed explanation and discussion of the
various venting regimes is presented in Section 2.6.2.2. Oxygen starvation
is discussed in Section 4.2.
Zone models are the most common type of physical compartment fire
models in use today. Zone models predict fxe development within an
8 BASIC COMPARTMENT FIRETHEORY
source code is provided for every program, to allow the user to customize
the software to suit his specific needs. Stand-alone executables are also
provided for the programs described in this book.
More recent versions of BASIC, such as Visual Basic, would have
made it easier to include efficient and attractive user interfaces. To
illustrate this point, the programs are also provided in Visual Basic. This
version of the software is discussed in Appendix E.
Thus, the user can examine, customize, and use the models that are
developed in this book without any additional investment by using the
QBasic programs. Alternatively, he can purchase a Visual Basic compiler,
and use the Visual Basic code to develop a customized compartment fire
simulator with a professional user interface.
CHAPTER 2
Introduction to Mathematical
Compartment Fire Modeling
The rising plume gases collect below the ceiling and form a hot smoky
layer. This layer may continue to grow while the f ~ exists. e The space
between the floor and the hot layer consists of cool uncontaminated air.
Numerous pre-flashover full-scale room f i e tests have shown that the
interface between both layers is relatively sharp, while the composition and
temperature of the layers are reasonably uniform. Consistent with these
experimental observations, zone models are based on the assumption that
the room gas volume comprises two distinct and uniform layers or zones:
a lower layer of cold air and an upper layer of hot gases. The resulting
idealized geometry of the gas volume inside the compartment is shown in
Figure 2- 1.
Heat Release Rate of the Fire
Some zone fae models have the capability to predict heat release and
mass loss rate fi-om the fire as a function of calculated conditions within
the enclosure. However, this capability is available only for a few simple
fuel geometries. For example, FIRST has a growing fire subroutine that
predicts the size and heat release rate of a horizontal fuel slab such as a
mattress [l]. FIRST also includes a subroutine to model flammable liquid
pool fres.
In general, the heat release and mass loss history must be supplied by
the model user. It is a common misconception that zone models predict fire
growth, while, apart from the few exceptions cited above, they only
14 INTRODUCTION TO MATHEMATICAL COMPARTMENT FIRE MODELING
determine the effects of a fire that is specified by the user. There are
several methods and many data sources that can be used to estimate heat
release and mass loss rates from burning objects. This topic is addressed
in more detail in Section 4.4.
The radiation from the flame is transferred to the floor, wall, and
ceiling surfaces, and is partly absorbed by the upper hot layer gases. In
addition, the bounding surfaces of the enclosure exchange heat internally
by radiation, and with the gas layers by radiation and convection. At each
surface, there is a balance between the total radiative and convective heat
flux that is received, and the heat flux that is transferred by conduction into
the solid. It is clear, therefore, that it is not trivial to calculate the heat
losses or gains of the gas layers. An enclosure-wide heat balance is needed
that includes the three modes of heat transfer. An engineering approach
will be used in Section 2.6.3.1 to arrive at an approximate, but reasonably
accurate solution.
Zone models predict how the state of the upper and lower gas layers
change with time by solving the conservation equations of mass and energy
with the appropriate boundary conditions. To illustrate the technique of
zone fire modeling, a generic set of equations will be developed, and a
procedure for solving the equations will be presented below. However,
since there are eight properties (four for each layer), and only four
conservation equations, four additional equations are needed. These
additional equations are derived first.
V, + V'= V = WLH
where
The static pressure difference between floor and ceiling level is equal
to the hydrostatic pressure at the bottom of an air column of room height.
At ambient temperature this is approximately 12 Pa per meter of room
height, or 35 Pa for a typical room height of 3 m. Because atmospheric
pressure is close to 100,000 Pa, the pressure difference between the two
layers can be neglected and
where
where
This analysis assumes that the process is adiabatic, the specific heat is
constant, the gas behaves like an ideal gas, hydrostatic pressures are
negligible, and the heat addition rate Q is constant.
Zukoski presented an example using Equation (2.3) in his paper. The
same example is repeated here. Consider a small fire that steadily releases
100 kW in a room that has a total volume of 28.5 m3. In ten seconds
(t = 10), the pressure would rise approximately 0.07 bar (1 psi). This
pressure increase would be sufficient to cause the breaking of a window
that would effectively vent enough gas to prevent any W e r rise, thus
limiting the pressure rise to a negligible amount. Other building
construction materials would also fail at such pressure increases (see
Reference [15], page 3-327). Table 2-1 lists damages to structures exposed
to various overpressures.
A Generic Compartment Zone Fire Model 17
where
For practical purposes, moisture may be ignored and the lower layer air
considered dry.
Based on the discussion above, pressure at ceiling level may also be
considered equal to atmospheric pressure. The upper layer temperature
seldom exceeds 1500 K. Therefore, the upper layer also behaves as an ideal
gas so that
where
The main constituents of the upper layer are N,, 4,H20, CO2, and CO.
Under some conditions appreciable amounts of other species such as HC1,
A Generic Compartment Zone Fire Model 19
HCN, and unburnt hydrocarbons may also be present. The ideal gas
constant for the upper layer can be expressed as
all species
where
where
The terms on the right hand side of Equation (2.8) account for all flows
that add mass to or remove mass from the layer. They consist of the mass
loss rate of the fuel and rate of entrainment into the flame and plume (see
Figure 1-l), and mass flows through the vent (see Figures 1-2(a)- (d)).
The mass loss rate of the fuel, riz, is small compared to the other mass
flow terms and is often neglected. For the purpose of developing a generic
zone model, it is assumed that m, is specified. The remaining terms on the
right hand side of Equation (2.8) can be written as an algebraic function
of the primary variables. These functional relationships will be developed
fmt, before casting Equation (2.8) into mathematical form and applying
it to the two gas layers.
Air is entrained from the lower layer into the flre flame and plume over
a height between the fuel surface and the hot layer interface. This provides
a mechanism for transfer of mass from the lower layer to the upper layer.
Hence, the mass conservation equations for upper and lower layer include
an entrainment term that is of the same magnitude, but opposite in sign.
A Generic Compartment Zone Fire Model 21
where
For circular pool fires, D is the diameter of the fuel pan. For fres with a
rectangular or square surface, D is usually the equivalent diameter, i.e., the
diameter of a circle with the same area as the fuel surface.
A commonly used plume model of the form of Equation (2.9) is the
following correlation developed by Zukoski, Kubota, and Cetegen in 1980
12l]:
where
K = a constant (kglk~"~-rn~'~*s)
L, = radiative loss fraction
The fuel mass flow rate has usually been ignored in the derivation of
plume correlations. Therefore, Equation (2.9) can also be used to predict
the entrainment rate over a particular height above the fuel surface (see
Figure 2-2). Hence, the flow of lower layer air into the flame and plume
can be expressed as follows
22 INTRODUCTION TO MATHEMATICAL COMPARTMENT FIRE MODELING
where
Equation (2.11) takes the following form, based on the plume model of
Zukoski et al.
The fire is considered to be a point source of heat, i.e., the burning fuel
is considered to release its heat fiom a point and not fiom an area.
Variations of density in the flow are considered small when compared
to the ambient density.
Air entrainment into the plume is considered to be proportional to the
velocity of the plume at each location.
The profiles of the vertical velocity and buoyancy force in horizontal
sections are similar at all heights.
where
AZ, = distance between the fuel surface and the virtual origin (m)
where
C = orifice coefficient
PO&) = static pressure outside the compartment at height Z (Pa)
P,(Z) = static pressure inside the compartment at height Z (Pa)
p,@ = gas density in the vent at height Z (kg/m3)
Subsequently, the neutral plane rises above the sill and ambient air
starts to flow into the compartment at the bottom of the vent. The neutral
plane height continues to increase, and quickly exceeds the height of the
interface between the gas layers (see Figure 2-4c). The resulting flow
conditions prevail for almost the entire pre-flashover fire period, since the
duration of the preceding piston flow regime is typically less than one
minute.
where
It is assumed that the composition of the upper layer is close enough to that
of dry air, so that Equation (2.15) is applicable to both layers.
2, _c 2,and Zir 2, Typical pressure profiles for this case are shown in
Figures 2-5(a) (T, = TJ and 2-5(b) (T, > TJ. Pressure difference as a
-
function of height is given by
T, = Ta:A P (2) P
X
, 2 )- Pin(2) = constant and 2,- -m (2.16a)
28 INTRODUCTIONTO MATHEMATICALCOMPARTMENT FIRE MODELING
or
where
FIGURE 2-5. Piston flow pressure profiles (2, I Z,, and Zi 2 2,)
A Generic Compartment Zone Fire Model 29
and
ni,=m,=O (2.17b)
where
, 2'' Z, 4.
2, S-2and Typical pressure profies for this case are shown
in Figures 2-6(a) (T, = Td and 2-6(b) (T, TJ. The pressure difference
below the interface (Z i ZJ can be calculated fiom Equation (2.16a)
(T,= Th or Equation (2.16b) (T, > T,). Between the interface and the soffit
of the vent, the pressure difference is given by
FIGURE 2-6. Piston flow pressure profiles (2, S Z, and Z,, < Z < ZJ
30 INTRODUCTION TO MATHEMATICAL COMPARTMENT FIRE MODELING
z,
~ , = W v l p , ~ ( Z ) d Z ~ 2 6 . 6 W v ~ C ~ ~ d Z
(2.19a)
Zb Zb
and
The mass flow of ambient air into the compartment is still zero (ma = 0).
Zb c 2, S Ziand Zb < Zi< 2, Typical pressure profiles for this case are
shown in Figures 2-7(a) (T, = TA and 2-7(b) (T, TJ. The pressure
difference equations are the same as for the previous case, i.e., Equation
(2.16a) (T,= TJ or Equation (2.16b) (T,> TJ below the interface (Z S 23,
and Equation (2.18) above the interface. Note that the constant in Equation
(2.16a) in this case is equal to zero, and that 2, is equal to Zi. Equation
(2.19b) is still applicable for the upper layer flow, m,. However, ambient
air flows into the compartment below the neutral plane, and lower layer
gases leave the compartment between Z, and 3. The equations for ni, and
m, are given below. Equation (2.19b) is still valid for calculating the upper
layer flow, ni,.
and
A Generic Compartment Zone Fire Model 31
FIGURE 2-7. Pressures during transition to main flow regime (2,< 2, < Zi)
Zi< 2, < Z, and 2,< Zi< 2, Typical pressure profdes are shown in
Figures 2-8(a) (T, = T h and 2-8(b) (T, > Td. The pressure difference is
given by
and
2,
1
f i a = W,, pJv(Z)ldZ=26.6Wy C
zb 7 \iar::z)..
zb
32 INTRODUCTlON TO MATHEMATICAL COMPARTMENT FIRE MODELING
FIGURE 2-8. Pressure profiles during main flow regime (Zi IZ, ZJ
and
Furthermore, to relate density to temperature for the lower and upper layer
gases, the equation of state for dry air, Equation (2.15) is used.
Consequently, the mass of the lower layer can be expressed as
A Generic Compartment Zone Fire Model
It is shown below that the temperature derivative on the right hand side of
Equation (2.25) is a function of the primary variables. Hence, since m,, ni,
(see Section 2.6.2.2), and m, (see Section 2.6.2.1) are also functions of the
primary variables, the lower layer mass conservation equation is of the
following form
where
1 The moisture content of the incoming and lower layer air is neglected.
(Typically, the moisture content of ambient air is of the order of 1% by
mass.) With this assumption, the air consists of oxygen (23.2% by
mass) and a balance of inert gases, primarily nitrogen.
2 Air is entrained below the interface into the flame and plume at a rate
that exceeds the rate needed for complete combustion of the fuel
volatiles. This condition is generally met in the early stages of a fie.
A Generic Compartment Zone Fire Model 35
It may not be valid if the distance between the fuel surface and the
interface is too small, i.e., because the fuel is located far above floor
level, or because the layer interface has descended too close to the fuel
surface. Underventilated fires that occur in these cases are discussed
in more detail in Section 4.2.
3 The flame can be modeled as a Simple Chemical& Reacting System
(SCRS), in which 1 kg fuel reacts with s kg dry air to form 1 + S kg
products of combustion (see Section B.4.1 S). The air to fuel ratio, S,
is usually larger than the stoichiometric ratio, S,,,,. If the composition
of the fuel volatiles vary during the course of the fire (as they do for
realistic fuels), s may not be constant.
If more lower layer air is entrained into the flame and plume than needed
for stoichiometric combustion, only a fi-action (S,,, h ) is used. The
remaining fraction of the entrained air (m, - S, liz,) is theoretically not
needed for combustion, and is referred to as excess air. The excess air is
not affected by the combustion reactions, and acts as a diluent of the
products of combustion. The stoichiometric air to fuel ratio, S,,,,, is
usually not known. However, the stoichiometric mass flow of air can be
determined fi-om the heat release rate on the basis of the oxygen
consumption principle. In 1917, Thomton showed that for a large number
of organic liquids and gases, a nearly constant amount of heat is released
by complete combustion per mass unit of oxygen consumed 1251. Huggett
found this to be also true for organic solids [26], and determined an
average value for this constant of Ah,,Jr,,, = 13,100 kT/kg of 0 , where
Ah,,, is the net heat of combustion and rdCh is the stoichiometric oxygen
to fuel ratio. This generic value may be used for most practical
applications, and is accurate, with very few exceptions, to within &5%.
Consequently, the stoichiometric air flow rate can be estimated fi-omthe
heat release rate via
sstozch
. mf =
where
36 INTRODUCTION TO MATHEMATICAL COMPARTMENT FIRE MODELING
where
Subtracting Y, times the overall upper layer mass balance Equation (2.27),
after rearrangmg, leads to
A Generic Compartment Zone Fire Model 37
The remaining two primary variables, T, antiT,, are obtainec fiom the
solution of the lower and upper layer energy conservation equations.
Conservation of energy of a layer can be expressed in general terms as
= { enthalpyfrow in }
- { enthalpyflow out }
+ { heat transferred to the layer )
- { work done by the layer ) (2.34)
38 INTRODUCTION TO MATHEMATICAL COMPARTMENT FIRE MODELING
where
The net radiative heat flux entering surface j is equal to the heat flux
removed fiom surface j by means other than radiation to keep it at
temperature I;. The set of radiation transfer equations is obtained by
integrating similar equations for monochromatic radiation derived by
Siege1and Howell [27] over all wavelengths, assuming that the gas and all
surfaces are grey, and taking flame radiation into account. With grey
surfaces and a grey gas, zjkand E,, may also be written as
where
Equation (2.36) is also valid for radiation from the flame. Because the
calculation of the mean beam length in Equation (2.36) is rather complex,
40 INTRODUCTION TO MATHEMATICAL COMPARTMENT FIRE MODELING
where
where
Since the upper layer has the shape of a parallelepiped, V, and A, are given
by
and
A Generic Compartment Zone Fire Model 41
where
where
Both C and n depend on the orientation of the plate, the flow regime
(laminar or turbulent) and whether the plate is cooled or heated by the
fluid. Appropriate values can be found in a textbook on heat transfer (e.g.,
Reference [29]). The Nusselt, Grashof, and Prandtl numbers in Equation
(2.41) are calculated using properties of air evaluated at the film
temperature, i.e., the average of fluid and surface temperature. The length
42 INTRODUCTION TO MATHEMATICAL COMPARTMENT FIRE MODELING
scale, I,, is equal to (LW)'" for horizontal surfaces, and to the height for
vertical surfaces.
If the momentum of the fire is sufficiently high when the plume hits
the ceiling, a forced flow is generated, commonly referred to as the ceiling
jet. Cooper developed an approach to account for the forced convection
heat transfer fiom the ceiling jet to the ceiling and upper wall sections.
Algorithms implementing this approach for a two-zone room fire
environment are described in Reference [30]. Beller included a ceilingjet
algorithm into the WPI fire code [3 l], which is an extended and modified
version of FIRST, developed at Worcester Polytechnic Institute [32]. For
the purpose of the generic model developed in this chapter, ceilingjet heat
transfer is not considered. Additional discussion of the subject can be
found in Section 4.3.
Heat Conduction.The net heat flux to a surfacej, q;,,, is the sum of the
net radiative and convective heat fluxes, which are calculated as described
above.
where
The heat loss from the unexposed side, qy,,ms,may be a function of the
unexposed surface temperature, or it may be a constant (zero for a well-
insulated segment, approximating adiabatic boundary conditions).
Equations (2.43a) to (2.43d) are too complex for an analytic solution.
Most often, finte difference techniques are used to solve the equations
numerically [33]. Additional complexities, such as multilayered walls or
temperature-dependent thermal properties, can also be addressed without
too much difficulty if a finite-difference technique is used.
where
where
where
Because internal energy of the lower layer and enthalpy of the lower Iayer
are related via m,h, = mlul + P,&, and because P, is assumed to be
constant (see Equation 2.2), Equation (2.46) can be rewritten as
where
Because ambient air and lower layer gases have identical composition,
their reference enthalpies are identical. The energy balance of the lower
layer, therefore, can be rewritten as
where
The first law of thermodynamics applied to the upper layer yields the
following equation, similar to Equation (2.46) for the lower layer:
or
where
The reference enthalpies of fbel volatiles, entrained air, and plume gases
at the layer interface are related by the following equation
where
A Generic Compartment Zone Fire Model 47
The upper layer mass balance Equation (2.27) can be used to eliminate the
time derivative of m, on the left hand side of Equation (2.59). Therefore,
if he1 mass flow is neglected, Equation (2.59) can be rewritten as
or
where
Equations (2.26) for mass conservation of the lower layer, (2.33) for
upper layer species conservation, (2.53) for lower layer and (2.62) for
upper layer energy conservation form a set of ordinary differential
equations (ODEs). Equation (2.29) for mass conservation of the upper
layer, and auxiliary Equations (2.35), (2.40), (2.42), (2.43) (in a finite
difference form), (2.44), and (2.45) form a set of, primarily non-linear,
algebraic equations. The system of equations is very complex, so that an
analytic solution cannot be obtained. Numerical techniques have to be used
to estimate the values of the primary and auxiliary variables at discrete
times At, 2At, etc.
One approach to solve this complex combined system of ODEs and
non-hear algebraic equations is by using a canned solver that can handle
this type of problems. An example of such a solver is described in
Reference [34].
An alternative approach consists of an iterative sequence by which
groups of equations are solved in a predetermined order. If values are
available for the primary and auxiliary variables at time t, the following
steps are taken to obtain values at t + At:
If the time step At is small (e.g., of the order of 1 second) and the user-
specified tolerances are reasonable (e.g., of the order of 0.1 K for
temperatures), the solution usually converges rapidly and steps 6- 10 have
to be repeated not more than two or three times.
CHAPTER 3
ASET-QB: A Simple
Room Fire Model
3.1 INTRODUCTION
where t,, is the time into the simulation at which the user-defined
hazardous conditions are reached, and t, is the time at which the user-
defined conditions for fire detection are reached (i.e., the time that egress
is initiated). The concept of estimating the available safe egress time
according to Equation (3. l), which is at the basis of the ASET model, is
discussed in detail in Reference [36].
The original ASET program by Cooper and Stroup was written in
FORTRAN and contained over 1500 program lines. When it was released,
52 ASET-QB: A SIMPLE ROOM FIRE MODEL
Due to assumptions 3 and 10, there are only two primary variables, Z and
T,. Furthermore, assumption 9 greatly simplifies the heat transfer
calculations, eliminating the need for any auxiliary variables.
Consequently, the system of ASET model equations only consists of two
conservation equations.
Formulation of the ASET Equations 55
where
where
This is the form that Cooper et al. used in the derivation of the ASET
model equations [6,371.
Equation (3.2) is valid as long as AZi > 0 (or Zi > Z,). When the
interface drops below the fuel surface, the same equation can still be used
with rir, equal to zero. However, when Zi S 2 , it is likely that the user-
specified heat release rate will be affected by the reduced oxygen that is
entrained into the flame and plume. Therefore, it is recommended to
terminate the calculations when the interface drops below the fuel surface.
*
Therefore, Equation (2.61) for conservation of energy of the upper layer,
with T,= T, = T, and constant specific heat, can be written as
Formulation of the ASET Equations 57
An expression for the outflow of lower layer gases can be derived from
the mass balance Equation (2.25), with T, = ;c:
To apply the equations shown above, they must be solved at each time
step throughout the simulation. However, the equations cannot be solved
explicitly, i.e., they are not simple algebraic equations. Therefore,
numerical methods for providing solutions to the equations for dZJdt, and
dTJdt, must be used. It was mentioned earlier that the original ASET
model incorporated a fourth-order Runge-Kutta solution to the differential
equations, while ASET-B (and subsequent versions of ASET) used the
improved Euler method.
The Euler method is one of the simplest stepwise methods for solving
differential equations [43]. It is commonly referred to as the tangent line
method. Given an equation, the slope of the tangent line at a given point
can be found. With this, a tangent line can be constructed and used to
estimate the solution to the equation at the next point. The improved Euler
method, as the name implies, is a modified version of the standard Euler
method. The modification, and added accuracy, lies in the method's use of
an averaging scheme between the points of interest. This approach was
used in the ASET-B programs to solve the differential equations at each
time step. The Runge-Kutta method was used in the original ASET
program. The fourth-order Runge-Kutta method is equivalent to a
five-term Taylor formula. This method is the most complex of the three
mentioned, but is also the most accurate. For a detailed discussion of the
numerical solution methods mentioned, the reader is directed towards
Section B.5.2, or any text on the subject (e.g., [43]).
With regard to the ASET-B model, the program was written to
consider the solution to the differential equations reached if the difference
between the predicted and corrected values for Z and T, is less than
0.001 m and 0.3 K respectively. Note that the equations considered by
Walton were the non-dimensional equations derived by Cooper, and that
the corresponding limit for the transformed interface height and upper
layer temperature is equal to 0.001. Also, to avoid possible i n f i t e loops
by the solution subroutine, the program allows 30 iterations for a solution.
If a solution is not reached within the limit, a warning is printed and the
program uses the last corrected value to proceed to the next time step. If
this occurs, the value used is an unconverged value and likely to introduce
an unknown error into the solutions. The user could change the program
source code to allow for greater solution tolerance and iteration limit
values.
The ASET-QB Computer Program 59
In the first edition of this book, Birk simplified the ASET-B model by
using the standard Euler method to solve the ODES. The corresponding
computer program, ASETB-S, gave nearly identical results for three
sample cases. On this basis, Birk recommended the simplified approach
and used it in F'IRM, an extended version of the ASET model that includes
algorithms for calculating the flow through a wall vent.
ASET models, including ASET-QB, require that the heat release rate
history of the fire be provided as input. The data is entered in two parts: the
heat release rate and the time at which it occurs. ASET-QB, like the other
ASET models, uses a linear interpolation routine to provide heat release
data at times other than those given as input. Since the routine is based on
the equation of a straight line, it is important for the user to provide enough
data sets to adequately describe the heat release history of the fire. An
example of this procedure can be found in Section 3.S.
The values that are perhaps the most difficult to determine are those of
the radiative and total heat loss fractions, i.e., L, and L, respectively. The
values for L, and L, vary throughout the course of a fire as the
characteristics of the radiating source (flame) change and the bounding
surfaces of the compartment are heated. However, in the development of
the ASET models, the simplifying assumption was made that L, and L,
remain constant.
Values for L, which generally fall between 0.15 and 0.40, are available
in the literature (e.g., see Reference [lS], pages 3-78 to 3-8 1). A generally
accepted average value was proposed by Cooper [6] such that
Other values can be found in Table 4-3. Based on Equation (3.1 l),
approximately 35 percent of the heat released by a fire is radiated away
fiom the combustion region. Unless the user changes the source code, the
Comparison between ASET-QB and ASET-B 61
The derivation of these values will not be repeated here. There are
currently no simple accurate means for estimating the value of Lc, beyond
the simple rules offered by Cooper. A method to estimate L, will be
developed in Section 4.3.
FIGURE 3-2. ASET-QB input data for Walton's sample run of ASET-B [3]
The results of the comparison runs are provided in Table 3-1. The
ASET-QB results were obtained using mixed units for the input data, i.e.,
U.S. engineering units for all input data except heat release rate. The latter
is specified in kW instead of Btds. The small differences between the
upper layer temperature predictions from ASET-QB and ASET-B are
primarily due to the fact that
Note that the ASET-QB run terminates at 151 seconds, when the layer
interface drops below the fuel surface.
Limitations of ASET-QB 63
The plume model was presented in Section 2.6.2.1. The plume model
assumes a point source of heat release from the fuel surface. The model
does not account for virtual origin corrections or varying fuel surface areas.
A possible explanation for the lack of virtual source corrections in ASET
is that virtual source correctionswere just beginning to be studied when the
model was developed. Many of the plume experiments conducted were of
areas and heat release rates that correspond to small virtual origin
corrections, thus the corrections went unnoticed. When ASET-B was
released, it was based on the physics present in the original ASET model,
thus no virtual origin correction was included, although they were by then
available in the literature. According to Heskestad [46], the application of
the virtual origin correction to real fuels, such as upholstered fbmiture, is
untested, because the derivation of the expressions are based on pool-type
fires. Although the ASET models fail to account for virtual origin
corrections, they have been shown to agree with actual fire data in many
situations [37]. However, a virtual source correction may be important
when modeling some fire scenarios, and this is a limitation that should be
considered when applying and validating the models.
Equation (3.4) predicts the mass flow in the plume only, i.e., above the
flaming region. Several models that can be used to predict mass flow rates
in the flaming and intermittent regions are available in the literature [19].
An implicit assumption when using only the plume equation, is that the
model provides acceptable mass flow predictions in the flaming,
intermittent, and plume regions, not just hthe plume region alone. Relying
on a single plume equation, as in the ASET models, will ultimately
introduce some error into the model predictions.
Another assumption implicit in the derivation of the ASET models is
that stratification of the plume gases does not occur; that is, all plume gases
enter the hot layer that forms just below the ceiling. Stratification, which
in the present sense is defined as the layering of fue gases at some distance
below the ceiling due to aidgas density differences, occurs when the
buoyancy of the plume at the layer interface is too weak to penetrate into
the hot layer. Except for very low heat release rates in tall enclosures,
stratification should not be a concern in most compartment fire-modeling
problems. The convective heat release rate that results in stratification at
a height AZ, above the fuel surface, based on an ambient temperature rise
of AT,, over that height is given by (see Reference [l 51, page 4-255):
Limitations of ASET-QB 65
QC = 0.001 ~ ~ A z , Z . ~ A T ~ . ~ (3.13)
where
One of the most limiting features of the ASET model is their inability
to account for oxygen starvation. The models assume there is enough
oxygen present within the lower layer to allow fiee-burn conditions at all
times. This is obviously a gross oversimplification. The models use the
heat release given as input and do not consider if the heat release is even
physically possible. The he1 surface can be submerged in the descending
hot layer, but the model will continue to release fiee-bum heat into the
enclosure. This limitation can result in temperature predictions that are
physically impossible.
None of the ASET model programs, including ASET-QB, warn the
user when conditions occur that could result in predictions that are
physically impossible. However, ASET-QB terminates automatically when
the layer interface drops below the fuel surface, because the model
equations are defmitely invalid at that point.
Limitations of ASET-QB 67
The burning rate, and, more accurately, the heat release rate history of
a fire must be provided as input. The model cannot perform heat release
rate predictions. According to Cooper and Stroup [35], the model is not
considered accurate once the hot-layer temperature reaches approximately
350-440°C. This limit is due to the radiation augmentation of the hot layer
on the fbel-burning behavior. At these temperatures, the radiation emitted
by the hot, smoky layer that is incident on a burning fbel may substantially
alter the fi-ee-burn rates provided as input by the user. The concern of
compartment effects on burning rates will be discussed in greater detail in
Section 4.5.
CHAPTER 4
Modifications to ASET-QB
4.1.1 Introduction
where
It can be noted from inspection of Equation (4.1) that the flow will
approach zero when the layer approaches the midpoint of the vent. Quite
obviously, an expression such as this cannot be used in time-varying
enclosure f r e models.
A gas within an enclosure will move only if forced to. Overall, smoke
will move within buildings due to several forces, such as stack effect,
buoyancy, expansion, wind, and HVAC systems. These are covered in
detail by Klote and Fothergill [16]. Although stack effect is due to
buoyancy, the two are frequently considered separate forces in the
Venting of the Hot Layer 71
1 Cold outjlow only. This occurs early on, when the expansion of the
gases collecting below the ceiling acts like a piston pushing down and
expelling cool air.
2 Both hot and coldflow out. In this regime, the piston effect is still
present, but hot gases are also vented due to the hot layer dropping
below the vent soffit.
3 Hot gases venting out and coldjlow in.
4 ChokedJlow. This regime is typically associated with post-flashover
compartment fires that are oxygen-starved.
The most important regime, according to Mitler is No. 3 above, that of hot
flow out and cold (ambient) flow in. The f ~ s regime t always occurs,
provided the soffit of the vent is located at some distance below the ceiling.
The duration of the first regime is a firnction of this distance. The second
regime usually passes quickly. The inflow during the fourth regime is
proportional to the ventilation factor, A$H, [24]. The four regimes are
accounted for in FIRM-QB. The flow equations for the first three regimes
72 MODIFICATIONSTO ASET-QB
The neutrl plane is a simple but important variable in the vent flow
modeling process. Basically, the neutral plane height, by definition is the
vertical location within a vent at which the pressure difference across the
vent is zero. Thus, there is no flow at the neutral plane. Above the neutral
plane, hot gases flow out of the compartment, and below, ambient air
flows into the compartment.
The hot-layer interface plane and neutral plane are not the same,
although they may be numerically close. The layer interface height, Zi, is
the vertical elevation within the compartment, away fiom any vents, at
which the discontinuity between the hot and cold layer is located. The
neutral plane height, 2, is the vertical location within the vent at which the
pressure difference across the vent is zero.
Typical pressure profiles inside and outside the room during this
regime are shown in Figure 4-l(b). The inflow of ambient air, ki,is still
zero in this case. Equation (2.16a) indicates that the pressure difference
across the vent is constant at and below the layer interface, and, for
example, equal to the pressure difference at the interface height, AP(4).
Substitution of this constant into Equation (2.19a), and using the auxiliary
parameter defined by Equation (4.2), leads to the following expression for
the lower layer vent flow:
4.1.5.3 Transition between Flow Regimes 2 and 3 (Zi < 2,and Z,, = Zi)
Typical pressure profdes inside and outside the room during this
regime are shown in Figure 4-l(c). In this case, air inflow and lower gas
layer outflow are both equal to zero. An expression for the upper layer
Venting of the Hot Layer 75
outflow can be obtained fiom that derived in the previous section, but with
X, = 0:
Typical pressure profdes inside and outside the room during this
regime are shown in Figure 4-l(d). The integral in Equation (2.22a) has to
be split into two parts, because the expressions for the pressure difference
below and above the interface are different. Equation (2.22a), after
substitution of Equations (2.21a) and (2.21b), and using the auxiliary
parameters defined in Equations (4.2) and (4.3), can be written as
After integration, the following expression is obtained for the lower layer
vent flow:
FIGURE 4-1 (continued). Pressure profiles for different vent flow regimes
76 MODIFICATIONS TO ASET-QB
The upper layer vent flow follows fiom Equation (2.22b). Using Equation
(2.21a) for the pressure difference, and the auxiliary parameter X,defined
by Equation (4.5), this can be written as
The expression for the lower layer vent flow in regime 1 is identical to
that in ASET-QB, i.e., Equation (3.10). This expression is valid as long as
Z,2 G. Both rig and rid are equal to zero. Therefore, as far as the computer
solution is concerned, regime 1 does not present a problem because there
is a simple criterion for this regime (Zi 2 23, and it is clear which
equations must be used to determine the vent flows.
If the layer interface is located below the soffit, the situation is much
more complex There are three possibilities, corresponding to flow regime
2, the transition between regimes 2 and 3, and flow regime 3. The first task
of the vent flow routines is to determine which of the three cases is in
effect. A procedure to accomplish this task is developed below.
Equation (2.29), which is obtained by combining Equations (2.25) and
(2.28) to eliminate dZi/dt, can be written as
Venting of the Hot Layer 77
Replacing the lower layer vent flow on the left hand side of Equation
(4.19) with the expression on the right hand side of Equation (4.6), and
using the d e f ~ t i o nof X, fiom Equation (4.2), after rearranging, leads to
Therefore, for given values of Zi and T,, Equation (4.14) is again a non-
linear algebraic equation in @(&) that, with the aid of Equation (4.15),
can be written as follows:
This equation can also be solved with the bisection technique. In this
case it is much easier to fmd the limits of an interval that brackets the root.
dP(Z,) = 0 is again a suitable lower limit. An upper limit for the interval
follows fkom Equation (4.21), and the fact that Z, cannot exceed 2,:
The reader can easily verify that the slope of the h c t i o n in this case is
negative, becauseL(0) is positive.
Finally, in the transition between flow regimes 2 and 3, completing the
second task is straightforward because m, = = m:, and both m, and m,
are equal to zero.
A review of the vent flow routine will provide the reader with
additional understanding of the vent flow equations and their solution. This
can be found in the subroutine VentFlows of the QBasic source code of
FIRM-QB on the accompanying CD-ROM.
The vent flow model assumes that the lower layer remains at ambient
temperature, which affects the accuracy of the vent flow modeling
technique. This simplification has been discussed by Steckler et al. [49].
In some cases, the lower layer can reach significant temperatures.
However, the assumption of a f i e d lower layer temperature is acceptable
to predict the growing stage of most fues with reasonable accuracy.
4.2 OXYGEN-LIMITED BURNING
The vent flow equations for regime 3 developed in Section 4.1 allow
the interface height to descend to the bottom of the vent. In reality, when
the layer interface drops to a certain height, typically between one half and
one third of the vent height, the inflow of ambient air reaches a maximum
that corresponds to ventilation-controlled burning conditions. Based on
work by Kawagoe [24], Drysdale derived the following expression for the
maximum flow 1571:
where
Solving for rir, and renaming the variable (li2,)- to indicate it is the
maximum possible entrained air flow, yields
Table 4-1, Heats of Combustionaof Selected Fuels (@2S°C)
&,M &,ueJssbich &,neJrstoich
kJ/g fuel kJ/g air kW! 0 2
Carbon monoxide CO
Methane CH4
Ethane c,&
Ethene
Ethyne c2H2
Propane
n-Butane n-C4Hl0
n-~entane n-C5H12
n-octane n-C8H18
c-Hexane c-C6H,
Benzene C,&
Methanol CH30H
Ethanol C,H,OH
Acetone (CH3)2C0
D-Glucose C6H1206
Cellulose
Polyethylene
Polypropylene
Polystyrene
Polyvinylchloride
Polymethylmethacrylate
Polyacrylonitrile
Polyoxyrnethylene
Polyethyleneterephthalate
Polycarbonate
Nylon 6.6 29.6
" Apart from the solids glucose and the polymeric materials) the initial state of
the fuel and all the products are taken to be gaseous.
Heat Loss Fraction Calculation 83
The maximum possible heat release rate can be estimated fiom Equations
(4.26) and (4.28) such that
The heat release rate to be used for the governing equations is the smallest
value of that specified by the user, and that calculated fiom Equation
(4.29). The subprograms in FIRM-QB that calculate the right hand side of
the conservation equations (Derivatives), and the plume flow (PlumeFlow)
were modified to account for entrainment-controlled burning. The reader
is encouraged to review the source code of these subprograms.
The lower 0.6 value would relate to high aspect ratio spaces with smooth
ceilings, and fires positioned far away from walls. The intermediate values
and the high 0.9 value for L, would relate to low aspect ratio spaces, fire
scenarios where the fire position is within a room height or so from walls
and/or to spaces with highly irregular ceiling surfaces.
where
To allow for a more precise estimate of the radiative losses, the user
of FIRM-QB will be given the option to choose a value for L, that may be
different from the default value of 0.35. This option was not available in
the original ASET models without changing the source code. Some
values for L, are provided in Table 4-3. More data can be found in the
literature (e.g., see Reference [15], pages 3-78 to 3-81, where
L, = hHrad/AHCh).
Cooper calculated the heat losses to the ceiling on the basis of
experimental ceiling jet data, and plotted the ratio of ceiling heat losses
to convective fraction of the heat release rate as a h c t i o n of the ratio of
radial distance fkom the plume centerline to distance between the ceiling
and the fuel surface. Cooper's data are shown in Figure 4-2 as solid
circles. The following expression provides a good fit to Cooper's data:
where
0.4
0.3
0.2
Cooper [6]
0.1
0.0
The values recommended by Cooper are valid only if the ceiling span
exceeds four times the distance between the ceiling and the fuel surface.
In addition, heat transfer between the ceiling jet, after it hits a vertical
wall and is deflected downward, and the walls of the enclosure is not
accounted for. A new procedure will now be developed to address these
two problems.
Shortly after ignition, andjust before the upper layer starts to develop,
the ceiling jet covers the entire ceiling. R can be estimated as the radius
of a circle with the same area as the ceiling. In this early stage, the
temperature of the ceiling is close to the initial temperature. Cooper's
calculations of were based on the assumption that the ceiling
temperature is equal to ambient, and, therefore, are most accurate in this
stage. As time progresses, the upper layer volume increases, possibly until
86 MODIFICATIONSTO ASET-QB
One would expect that L, values calculated from Equation (4.3l), with
R = R,, would be excessive. However, the resulting heat loss fractions
actually would turn out to be consistently slightly smaller than the values
recommended by Cooper, and those used by Birk in the first edition of
this book. The lower values are probably more realistic for vented
enclosure fires that last more than a few minutes.
Note that L, is equal to 0.05 for rough ceilings, or 0.10 for very
rough ceilings, as suggested by Birk.
The method includes no correction for the heat release of the fie.
Obviously, the magnitude of the ceiling jet is proportional to the heat
release rate of the fire [19]. Thus, high heat release fires in small
compartments are likely to suffer fiom poor estimates.
The primary purpose of introducing the expression for the heat loss
fiaction is to provide novice users with a first guess of what L, should be.
The FIRM-QB model displays the value, and the user may select the
value or enter his~herown. This should not discourage those familiar with
selecting values for L, fiom choosing their own.
L, will have a profound effect on the results of the simulation, and,
therefore, these values must be selected with an understanding of their
importance. A discussion of the effects of L, and L, on the predictions of
the FIRM-QB model is presented in Chapter 7. This discussion reinforces
the need for careful selection of values for the heat loss fractions.
where
This method of predicting heat release rates has been widely reported and
used in the literature to determine spacings of detectors and sprinkler
heads. It is referenced in some of the National Fire Protection Association
(NFPA) codes and standards that deal with these issues [58,59].
However, this simple approach has significant limitations when used for
f i e hazard assessment. First, it requires that the user have knowledge of
the growth factor, a, which is fuel-specific, and is typically found
through curve fitting of full-scale fire test data of the particular fuel.
Some examples of growth factor values for upholstered fhxiture are
shown in Table 4-2. The expression cannot account for decay because
heat release rate increases with the square of time. Also, the heat release
is based on free-burn conditions. A more detailed discussion of the
limitations of using fires can be found in Appendix A of Reference
[601.
The data in Table 4-2 were obtained by fitting the power law in
Equation (4.33) to the results of 40 furniture calorimeter tests conducted
at NIST 1611. The first column in the table contains the test numbers used
in the original NIST reports. The second column is a brief description of
the item that was tested. In NFPA 72, fires are classified as being either
slow-, medium-, or fast-developing. A slow-developing f i e has a growth
factor of 0.0066 kW/s2or less. A fast-developing fire has a growth factor
greater than 0.0469 kw/s2.A mediumdeveloping fire has a growth factor
greater than 0.0066 kW/s2,but less than or equal to 0.0469 kw/s2. The
virtual time of origin, t,,, is the time at which the f ~ f ebegan to obey the
power-law f i e growth model. Prior to t,, the he1 might have smoldered,
but did not burn vigorously with an open flame. The last two columns
contain the time to reach and the value of the maximum heat release rate.
Table 4-2. Furniture Heat Release Data E581
Test a tv tm,
Q-
No. Item Description class* (kW/s2) (S) (S) (kW)
Metal Wardrobe (4 1.4 kg total) f 0.4220 10 42
Chair F33, Trial Loveseat (39.2 kg)
Chair F21, Initial (28.2 kg)
Chair F21, Later (28.2 kg)
Metal Wardrobe, Initial (40.8 kg total)
Metal Wardrobe, Average (40.8 kg total)
Metal Wardrobe, Later (40.8 kg total)
Chair F24 (28.3 kg)
Chair F23 (3 1.2 kg)
Chair F22 (3l .9 kg)
Chair F26 (19.2 kg)
Chair F27 (29.0 kg)
Chair F29 (14.0 kg)
Chair F28 (29.2 kg)
Chair F25, Later (27.8 kg)
Chair F25, Initial (27.8 kg)
Chair F30 (25.2 kg)
Chair F3 1, Loveseat (39.6 kg)
Chair F3 1, Loveseat (40.4 kg)
Chair F32, Sofa (5 1.5 kg)
W in. Plywood Wardrobe with Fabrics (68.5 kg)
?4in. Plywood Wardrobe with Fabrics (68.3 kg)
'/a in. Plywood Wardrobe with Fabrics (36.0 kg)
1 h in. Plywood Wardrobe with FR Interior Finish, Initial
42 1' 6 in. Plywood Wardrobe with FR Interior Finish, Later -** 1.1722 100 65 5000
Table 4-2 (continued). Furniture Heat Release Data [SS]
Qmaa
Test a tv tmax
No. Item Description class' (kW/s2) (S) (S) (kW)
Repeat of % in. Plywood Wardrobe (67.6 kg)
in. Plywood Wardrobe with F'R Latex Paint (37.3 kg)
Chair F2 1 (28.3 kg)
Chair F2 1 (28.3 kg)
Chair, Adj. Back Metal Frame, Foam Cushions (20.8 kg)
Easy Chair C07 (1 1.5 kg)
Easy Chair F34 (15.7 kg)
Chair, Metal Frame, Minimum Cushion (16.5 kg)
Chair, Molded Fiberglass, No Cushion (5.3 kg)
Molded Plastic Patient Chair ( l l .3 kg)
Chair, Metal Frame with Padded Seat and Back (15.5 kg)
Loveseat Metal Frame with Foam Cushions (27.3 kg)
Chair, Wood Frame and Latex Foam Cushions (1 1.2 kg)
Loveseat, Wood Frame and Foam Cushions (54.6 kg)
Wardrobe, % in. Particleboard (120.3 kg)
Bookcase, Plywood with Aluminurn Frame (30.4 kg)
Easy Chair Molded Flexible Urethane Frame (16.0 kg)
Easy Chair (23.0 kg)
Mattress and Boxspring, Later (62.4 kg)
Mattress and Boxspring, Initial (62.4 kg) S 0.0009 90 667 400
***s = slow, m = medium, f = fast
Fire growth exceeds design data
Reprinted with permission from NFPA 72-1996, National Fire Alarm Code, Copyright Q 1996, National Fire Protection
Association, Quincy, MA 02269. This reprinted material is not the complete and official position of the National Fire
Protection Association on the referenced subject, which is represented only by the standard in its entirety.
Heat Release Rate Predictions 91
I I 1 I I
100 200 300 400 500 600
TIME FROM 110 kW1 FIRE IGNlTlON (sec1
where
Equation (4.35) is valid for pools with a diameter greater than 0.2
meters that are burning in the open. Pool fires with a diameter greater
than 0.2 meters are characterized by burning that is governed by radiative
heating fkom the optically thick flame. This dependence is seen in the
expression by the presence of the kp parameter.
Equation (4.35) provides only the mass loss rate per unit area. The
heat release rate is obtained fkom:
Unless the user changes the pool diameter with respect to time (or any
other variable), the equation will yield a constant heat release rate.
Values to use in Equations (4.35) and (4.36) are shown in Table 4-3.
The expression can be applied to melting plastics, but data for these
materials are, however, quite limited at the present time.
Table 4-3. Data for Large Pool Burning Rate Estimates [91
Density Ahg Ahqne* tjt: kP
Material (kg/m3) (kJ/kg) J/kg) (kgfm2*s) (m) L,"
Cryogenics
Liquid H2 70 442 120.0 0.017 (fO.001) 6.1 (k0.4) 0.25
LNG (mostly CH,) 415 619 50.0 0.078 (k0.018) 1.1 (h0.8) 0.16-0.23
LPG (mostly c3&) 585 426 46.0 0.099 (f0.009) 1.4 (k0.5) 0.26
Alcohols
b
Methanol (CH,OH) 796 1195 20.0 0,017 (fO.001) 0.17-0.20
b
Ethanol (C2H,0H) 794 891 26.8 0.015 (kO.001) 0.20
Solids
Polymethylmethacrylate 1184 1611 24.9 0.020 (h0.002) 3.3 (k0.8) 0.40
(c5%02)
Polyoxymethylene 1425 2430 15.7 0.15
(CH20)n
Polypropylene (C,&), 905 2030 43.2 0.40
Polystyrene (C8H8), 1050 1720 39.7 0.44
a For diameters ca. 1 m. Decreases for small and for very large diameters
Value independent of diameter in turbulent regime
c Only two data points available
Reprinted with permission from Fire Protection Handbook, 16th Edition, Copyright O 1986. National Fire Protection
Association, Quincy. MA 02269
Heat Release Rate Predictions 95
f c THERMOCOUPLE
EXHAUST
'1 1 r mCi
BLOWER
THERMCA^' '-' - .- .- - - METER
SMOKE -
1 1
' -
U
PRESSURE
PARTICULATE FILTER U Ot ANALYZER
& COLD TRAP
TRANSDUCER
U CO, ANALYZER
U CO ANALYZER
PRESSURE
REGULATOR
fDuMp
I 1 I I I 1 I
2000 - SPECIMEN F2 1 -
1800 - -
curve does not include the "tail,"which is primarily the heat release rate
of the frame. Indeed, Babrauskas also found that the triangular part
accounted on average for 91%of the total heat released by furniture items
with a non-combustible fiame.
Babrauskas developed a simple model to predict peak heat release
rate (top of the triangle) and burning time (triangle base width) on the
basis of generic characteristics of the furniture item.
TIME IsL
According to the model, peak heat release rate can be estimated fiom
where
FF = fabric factor
1.O for thermoplastic fabrics (e.g., polyolefm)
0.4 for cellulosic fabrics (e.g., cotton)
0.25 for PVC or polyurethane film-type coverings
PF = padding factor
1.0 for polyurethane foam, latex foam, or mixed materials
0.4 for cotton batting or neoprene foam
CM = combustible mass (kg)
SF = style factor
1.5 for ornate convoluted shapes
1.2- 1.3 for intermediate shapes
1.0 for plain, primarily rectilinear construction
Heat Release Rate Predictions 99
where
The values for the different factors required in the equations presented
above were derived in Reference [71]. Babrauskas reported that this
method should not be used when the product of the fabric factor FF and
the padding factor PF is less than 0.225, because values this low are
indicative of extensively low burning rates that will not yield the
triangular-shaped heat release curve.
Table 4-4 provides some descriptive information and data regarding
13 items of fbmiture that have been tested in the furniture calorimeter at
NBS. Predicted [according to Equation (4.3 711 and measured peak heat
release rates for the 13 specimens are given in Table 4-5. The predictions
are within &15%of the measured values, except in two cases (F26 and
F28). However, the predictions in those two cases greatly exceed the
measured peak heat release rates, i.e., the error is on the conservative
side. Table 4-6 provides a comparison between burning times calculated
according to Equation (4.38), and the base width of triangles that were fit
to the heat release rate curves. The Ah,, values are those recommended
by Babrauskas et al. [64]. The calculated times are larger in most cases,
which again indicates that the model is generally consetvative.
100 MODIFICATIONS TO ASET-QB
Mass CM &m
, Q-
specimen (@) (kg) Style Frame Padding Fabric (gls) (kW)
traditional easy chair wood FR PU olefin
traditional easy chair wood FR cotton cotton
traditional easy chair wood FR cotton olefin
traditional easy chair wood FR PU cotton
traditional easy chair wood PU olefin
thinner easy chair wood FR PU olefin
traditional easy chair wood mixed cotton
traditional easy chair wood mixed cotton
traditional easy chair PP 1 PU olefin
traditional easy chair PU PU olefin
traditional loveseat wood FR PU olefin
traditional sofa wood FR PU olefin
traditional loveseat wood mixed cotton 75.0 940
Table 4-5. Predictions of Peak Heat Release Rate with Triangular Model
4.5.1 Introduction
FIGURE 4-7. Gas temperature vs. time for a typical compartment fire
The Prediction of Flashover 105
The first defmition (T, 2 600°C) has been widely used as an indication
of the onset of flashover, although many researchers prefer to consider a
change in temperature of 600°C as an indicator of flashover. The present
study will consider the onset of flashover to occur when the average
upper layer gas temperature reaches 600°C.
The FIRM-QB model developed here accounts for only a single fire
plume. Because multiple fires mean multiple plumes, it is clear that this
model is no longer valid once flashover is achieved. As such, the model
has a temperature check that will flag the user if the upper layer gas
temperature reaches 600°C. The user does have the option of continuing
the simulation, because he may be using another definition of flashover.
The first and most important application of the heat release data is to predict
room flashover. To make this determination, only the peak value of the heat
release is needed, supplemented by physical data characterizing the fire
room.
where
Ahg Is the heat of gasification. For a liquid fuel it is equal to the sum of
the enthalpy to raise the temperature of the he1 to its surface temperature
(slightly below the boiling point) and the latent heat of vaporization.
When a fuel is placed in an enclosure, Equation (4.39) changes to
where q: (in units of kW/m2)is the heat flux incident on the fuel surface
from extemal sources [57]. In enclosures, these sources can be from the
hot layer, heated walls, and other flames. It is clear that external heat
fluxes can affect the burning rate of fuels. Hence, the heat release rate
becomes
TIME ttscl
1 I 1 I I I I I 1
CHAIR F21
-Furniture calorimeter
-S--- Room fire t e s t no. 5
r Flashover
FIGURE 4-9. Free-burn vs. room-fire heat release rates for a chair
CHAIR F31
-F u r n i t u r e calorimeter
-- Room f i r e t e s t no. 1
.-. Room f i r e t e s t no. 2
Room fire t e s t no. 6
h
-.---a
r Flashover
L I I I I L I I I 1
FIGURE 4-10. Free-burn vs. room-fire heat release rates for a chair
1 10 MODIFICATIONS TO ASET-QB
It is clear from these two graphs that enclosure effects are in fact
negligible, even just after flashover. Complex he1 configurations, such
as wood cribs, can also explain, in part, the negligible difference between
free-burn and compartment-fire burning rates. Some chairs and other
similarly shaped fbmiture items may also have pronounced radiative
exchanges between different surfaces such that they behave like wood
cribs. The ability to use £tee-bum data of other fuels ($001 fires for
example) to predict flashover is unclear. No general guideline can be
offered, except that complex he1 configurations and optically thick
flames tend to reduce enclosure effects. Users of f ~ models e are
cautioned to determine the appropriateness of such data when predicting
flashover.
To summarize this section, it has been shown that an upper layer gas
temperature at or above 600°C can be used to predict the onset of
flashover. It has also been shown that the fiee-burn heat release data for
most upholstered fiuniture items can be used to predict flashover.
CHAPTER 5
The FIRM-Q6 Model
5.1 INTRODUCTION
1 Technical Documentation
2 User's Manual
3 Programmer's Guide
5.3.1.1 Assumptions
The three vent flows can be expressed as a function of the static pressure
difference at the layer interface height, AP(Zi). Hence, Equation (5.4) is a
non-linear algebraic equation, which is solved for AP(Zi) at every time
step. Once M(ZJ is known, the corresponding vent flows can be
calculated, and substituted into Equation (5.2). FIRM-QB does not allow
rir, to exceed Kawagoe's choked flow limit (see Section 2.6.2.2). The rate
at which hot layer gases are vented, mu, is also calculated in this process,
although it is not needed for the lower layer mass conservation equation.
A more detailed discussion of the vent flow equations in FIRM-QB can be
found in Section 4.1.
The conservation Equations (5.1) and (5.2) form a set of two ordinary
differential equations (ODES). This set is solved at every time step to
predict the upper layer temperature, T, and layer interface height, Zi, at the
next time step. A fourth-order Runge-Kutta method with stepsize control
is used for this purpose. The stepsize control algorithm reduces the time
step so that the estimated error of the solution vector is within certain
tolerances. The maximum errors permitted are 0.3 K and 1 mm for T, and
Zi respectively. These values are comparable to the tolerances used by
Walton in ASET-B [3]. The ODE solver and stepsize control algorithms
are discussed in more detail in Section B.5.3.
Equation (5.4) is a non-linear algebraic equation that is solved with the
bisection technique. First, upper and lower limits are found of an interval
that brackets the root of the equation. The root is then found by repeatedly
halving the interval, until the function value, within a certain tolerance, is
equal to zero. In FIRM-QB, this tolerance is set equal to 0.3, which
corresponds to a mass flow of less than 1 g/s. Further details concerning
the bisection method utilized in FIRM-QB can be found in section B.5.2.
Note that two different approaches are used to implement the
numerical techniques in FIRM-QB. The ODE solver was added to the
FIRM-QB Data Libraries 117
6.1 INTRODUCTION
This chapter contains the user's manual for the Fire Investigation and
Reconstructzon Model FIRM-QB. The information in this chapter is
structured according to the guidelines in ASTM E 1472, "Standard Guide
for Documenting Computer Software for Fire Models" [74f.
FIGURE 6-1. Typical sequence of questions and answers for FIRM-QB run
1 Path to the directory where the data andfire files are located. It is
assumed that the two types of files (output and fire are located in
the same directory. After the path has been specified, it remains the
same for all runs in a session. If the user wants to change the path, he
must terminate and restart the program.
2 Output file name, without path and extension. This is a string of
maximum eight characters. The name is not case-sensitive, and can
contain only the letters A-Z, the numbers 0-9, and the special
characters ', -, !, @, #, S, %, ", &, (, ), -,,(, ), and '. The data fle
name is automatically given the extension ".FOF." FIRM-QB checks
whether the f l e already exists, and verifies that the user wants to
overwrite the file if this is the case.
122 FIRM-QB USER'S MANUAL
6.6INPUT DATA
vent, and the dimensions of the fuel are usually relatively well known. The
main problem is to characterize the fire itself. This type of information can
be found in handbooks, journals, reports of standard tests, reports of
custom experiments, etc. For example, burning rates for a wide range of
fuels can be found in Chapter 3-1 of Reference [15]. An extensive number
of heat release rate curves are compiled in Reference [76]. The radative
loss fraction, L,, is generally in the range of 25 to 35%, and values for a
wide range of fuels are reported in the literature (e.g., see Reference [15],
pages 3-78 to 3-81, where L, = AHradAHCh). FIRM-QB can assist the user
in estimating the total heat loss fraction, L, (see Section 4.3.1).
All input values, except heat release rate data, are typed on a single line
in the appropriate format (numeric or text) followed by ENTER. Heat
release rate data (except at time zero) are entered in pairs. Each data pair
consists of the time and the corresponding heat release rate, separated by
a comma.
All variables, except the data and fire file path, are reset in between
runs. AU text variables are reset to null strings, except that the default units
are S.I. All numeric variables are reset to 0, except L,, which has a default
value of 0.6.
The input variables that describe the geometry of the compartment, and
that characterize the intensity of the fire, are listed below.
Floor area, A. It is assumed that floor and ceiling area are identical,
and that the floor plan is rectangular in shape. The model will provide
reasonable predictions for compartments with irregular floor plans, but
the estimate of L, might not be reliable. The units for this variable are
m2(S.I.) and ft2 (U.S. engineering and mixed).
Room height, H. This is the distance fiom the floor to the ceiling. Units
are m (S.I.) or ft (U.S. engineering and mixed).
Vent width, W,. This is the width of the vent. If multiple vents are
present, use the sum of the widths. Units are m (S.I.) and ft (U.S.
engineering and mixed).
Vent sill height, 2,. This is the height of the bottom of the vent. If
multiple vents with different sill heights are present, use the average
height as a starting point, and check the effect of sdl height (within the
124 FIRM-QB USER'S MANUAL
range for the different openings) on the predictions. Units are m (S.I.)
and ft (U.S. engineering and mixed).
Vent sofit height, Zt. This is the height of the top of the vent. If
multiple vents with different soffit heights are present, use the average
height as a starting point, and check the effect of soffit height (within
the range for the different openings) on the predictions. Units are m
(S.I.) and ft (U.S. engineering and mixed).
Fire base height, ZfThis is the height of the base of the flame. It is not
trivial to specify this variable for irregular objects, such as chairs or
sofas. If the height of the fuel surface is not very well defined (or if it
varies with time), it is recommended to repeat the simulations for a
range of he1 height values, and to retain the most conservative results.
Units are m (S.I.) and ft (U.S. engineering and mixed).
Radiative heat lossfiaction, L, This is the fraction of the heat released
by the fire that is lost in the form of thermal radiation. This variable is
dimensionless.
Total heat lossfraction, L,. This is the fraction of the heat released by
the fm that is lost to the room boundaries. FIRM-QB has an algorithm
to estimate the value of L,. The algorithm requires the user to specify
the length to width ratio of the floor, and whether the ceiling is smooth,
moderate, or rough. This variable is dimensionless.
Maximum simulation time, t,,. The simulated time will never exceed
this value. However, the program will terminate earlier if the end of the
fire file is reached, flashover occurs and the user instructs the program
to stop, or the layer interface drops below the fuel surface or the sill.
The unit for this variable is seconds.
Heat release rate, Q. If the user decides not to use an existing fire file,
heat release rate data must be entered. Except at time zero, the heat
release rate data are entered in pairs. Each data pair consists of the time
and the corresponding heat release rate. The unit for time is seconds,
and heat release rate is in kW (S.I. and mixed) or Btu/s (US.
engineering). The user must also specify the heat of combustion of the
fuel, in kJkg (S.L) or Btunb (U.S. engineering and mixed).
FIRM-QB (and ASET-QB) requires the user to specify the heat release
rate of the f i e . Heat release rate information is stored in external ASCII
Output Information 1 25
data files, i.e., f i e files. These files have the extension ".FIR," and consist
of numerical data in three parallel columns, followed by two lines of text.
The first c o l m is the time in seconds, the second column is the heat
release rate in kW at the corresponding time in the first column, and the
third column is the mass loss rate in g/s. The latter is equal to the heat
release rate divided by the heat of combustion, which is supplied by the
user. The mass loss rate is not used by FIRM-QB (and ASET-QB), but it
is needed for compatibility with FPETool. The values in the last row of the
three columns are equal to -9, to designate the end of the numerical data.
The first line of text contains the name of the fde, and the date it was
created. The second line is a description of the fire file entered by the user.
Fire fdes can be created directly fiom FIRM-QB. However, it is
recommended to use HRR-QB, because it was written specifically for
e and offers more options (see Section 4.4.5).
creating f ~ files,
It is recommended that the user press the CAPS LOCK key before
starting a FIRM-QB session. Thus, all character strings will be in
uppercase letters, resulting in a consistent and uniform output format.
Instructions for running FIRM-QB can be found in Appendix C.
The results of a run are displayed on the screen, saved to a file, and, if
requested by the user, sent to the printer. The format is nearly identical for
the three output media. A header is shown at the top of the screen, the start
of the data file, and the top of each printed page. The results are printed in
parallel columns below the header. Figure 6-2 shows the output from a m
with input data in Figure 6-1. The fust nine columns contain time in
seconds, upper layer temperature in "F and "C, layer interface height in ft
and m, heat release rate fiom the fire in Btds and kW, and upper layer
vent flow in lb/s and kg/s. The last two columns indicate which vent flow
regime is in effect (see Section 4. l), and whether the fire is oxygen-starved
(see Section 4.2).
FIRM-QB VERSION 1.00 - JULY 1999
ACETONE DIP TANK
SIMULATION OUTPUT DATA FILE : c:\FIRM\DATA\DIPTANK.FOF
FLOOR AREA : 139.35 m A 2 ( 1500.00 ftA2)
ROOM HEIGHT : 4.57 m (15.00 ft)
WIDTH OF VENT : 3.66m (12.00ft)
VENT SILL HEIGHT : 0.00 m ( 0.00 ft)
VENT SOFFIT HEIGHT : 3.05 m (10.00 ft)
FIRE BASE HEIGHT : 1.22 m ( 4.00 ft)
RADIATIVE FRACTION : 0.250
HEAT LOSS FRACTION : 0.550
MAXIMUM RUN TIME 900 S
HEAT RELEASE RATE DATA FILE : C:\FIRM\DATA\DIPTANK.FIR
l26
Personnel and Program Requirements 127
At the end of a run, the user has the option to view screen plots of
upper layer temperature, layer interface height, heat release rate and upper
layer vent flow vs. time. The plotted variables are in units that depend on
the system of units chosen by the user on input.
FIGURE 63. Westchase Hilton Hotel floorplan of the floor of fire origin (reprinted
with permission fiom NFPA Investigation Report Number LS-7, Westchase Hilton
Hotel Fire, copyright 01982, National Fire Protection Association, Quincy, MA,
00269)
Thinking the rescued occupant's date was still in the room, the second
occupant returned to the room in an attempt to search for the date. (Note,
occupant one is identified as the guest who was in the room of origin at the
time of the fire's inception, and who was subsequentlyrescued. Occupant
two is identified as the guest who returned to the hotel to find the fire and
rescue his roommate.) At this time, occupant two was unaware that his
roommate's date had left the hotel. Regardless, due to the rapidly
deteriorating conditions within the room of origin, guest two could not
reenter the room. The two guests then exited the hotel.
The two guests who occupied the room of origin survived the fire with
minor injuries. At postfire interviews, these guests reported that during the
course of the day, when the door leading to the hallway was open, the door
failed to completely close by itself due to interference by the carpeting in
the room. This corroborates the fire investigator's belief that, after the
rescue of occupant number one, the door was left open during the course
of the fire. Estimates suggest that the door may have been open between
6 and 18 inches.
At approximately 2:31 A.M.,the district fire chief arrived on the fire
scene and observed fire projecting out of the windows fiom the room of
origin, which indicated that room flashover had occurred some time after
2:20 A.M. and before 2:31 A.M. Fire department representatives estimate
first water was applied to the fire at about 2:38 A.M., and resulted in fire
extinguishment at Z:4 1 A.M.
130 FIRM-QB USER'S MANUAL
Summary time lines are commonly used for fire investigations. Time
lines can also prove beneficial for comparing model predictions with
eyewitness accounts and other reference data. Table 6-2 contains the time
line that was constructed ftom the information obtained in the
investigation. It is clear ftom the time line that the fire had smoldered for
over twenty minutes. Based on the rapid deterioration of the conditions
within the room of origin after the returning guest rescued his roommate,
it appears that the f ~ maye have been limited due to the depleted oxygen
within the room, or perhaps the chair was just attaining a flaming state.
After the overcome guest was removed ftom the room, the door was left
partially open, which provided a new source of oxygen for the fire. From
this point on, the fue grew rapidly resulting in flashover. From information
provided by the occupants of the room of origin, it is known that at about
2:20 AM., one guest rescued his roommate, assisted him down the hallway,
and returned to search for a second occupant who was believed to be in the
room. However, when the guest attempted to reenter the room, conditions
had deteriorated to the point where he could no longer enter.
The FIRM-QB model will be used to predict the f i e development
within the room of origin. It will be assumed that the first he1 ignited was
an upholstered chair, as shown in Figure 6-4. This assumption was based
on fire investigators attributing the cause of the fire to be a smoldering
cigarette in the chair. As a result of the fire, the Consumer Products Safety
Commission and the National Institute of Standards and Technology tested
hrniture removed from the hotel that was similar to the f.umiture located
Table 6-2. Time Line of Important Events for Westchase Hilton Hotel Fire
2:OOA.M. Odor of smoke on 10th floor.
2110 A.M. Light smoke noticed in room 804.
220 A.M. Room 404 (room of fire origin) occupant returns. Smoke in
4th floor corridor. Fire discovered in room 404. Unconscious
guest rescued by roommate.
2:25 A.M. Telephone alarm to Houston Fire Department.
2:27 A.M. Fire apparatus dispatched.
2:31 A.M. District fire chief on scene. Fire showing from 4th floor
window, indicating flashover occurred between 2:20 A.M. and
2:31 A.M.
to ble
ORIGIN
bed
bed
FIGURE 6-5. Example of FIRM-QB input for Westchase Hilton fire reconstruction
Figures 6-7 and 6-8 show the predicted location and temperature of the hot
layer in the room, after established burning has commenced, for various vent
widths. The various vent widths reflect the estimated door opening range of
6 to 18 inches based on information documented during the fire investigation.
Additional opening widths are included for comparison purposes. The
simulation for a closed door was performed with ASET-QB, and the
corresponding data file on the accompanying CD-ROM is CH6-AO1.AOF.
FIRM-QB was used for the other simulations,with door widths between 3 and
36 inches. The corresponding data fdes are CH6-A02.FOF to CH6-A07.FOF.
FIRM-QB VERSION 1.00 - JULY 1999
WESTCHASE HILTON FIRE RECONSTRUCTION (1.5 FT DOOR WIDTH)
SIMULATION OUTPUT DATA FILE : C:\FIRM\DATA\CH6-A05.FOF
FLOOR AREA : 24.51 m A 2 ( 263.80 ftA2)
ROOM HEIGHT : 2.44 m ( 8.00 ft)
WIDTH OF VENT : 0.46 m ( 1.50 ft)
VENT SILL HEIGHT : 0.00m (0.00ft)
VENT SOFFIT HEIGHT : 2.00m (6.56ft)
FIRE BASE HEIGHT : 0.91 m ( 3.00 ft)
W I A T I V E FRACTION : 0.350
HEAT LOSS FRACTION : 0.660
MAXIMUM RUN TIME : 1200 S
HEAT RELEASE RATE DATA FILE : C:\FIRM\DATA\CH6-AO1.FIR
closed door
0.25 ft door width
m - - - - - - 0.50 ft door width 1
- - - - - m - 1 .OO ft door width :
.....,....... 1.50 ft door width 1
2.00 ft door width :
3.00 ft door width I
I I I I
Time (S)
FIGURE 6-7. Layer interface height vs. time for various door widths
,.pH- P
closed door
----- 0.25 ft door width
------- 0.50 f t door width
-----. - 1.OO ft door width
............. 1.50 ft door width
2.00 ft door width
v
-.-.-.m
-,.-..-... 3.00 ft door width
Time (S)
FIGURE 6-8. Upper layer temperature vs. time for various door widths
136 FIRM-QB USER'S MANUAL
From the time versus temperature plots (Figure 6-8) it can be noted that
vent widths from 1 to 3 feet result in flashover being achieved in a rather
narrow time frame of 122 seconds (2.03 min) to 142 seconds (2.37 min).
It is interesting to see that the time to flashover decreases from 142 seconds
(2.37 min) to 125 seconds (2.08 min) when the door width is changed
from 1 to 2 ft, and increases back to 140 seconds (2.33 min) for a door
width of 3 ft. This can be explained by the fact that the fire is not oxygen-
starved prior to flashover for door widths of 2 ft or greater. In that case, the
heat losses associated with the upper layer vent flow are greater for larger
widths, resulting in longer flashover times. Below 2 ft, the fire becomes
oxygen-starved prior to flashover, and the resulting heat release rate
reduction compensates for the decreasing vent flow heat losses. The reader
is encouraged to review the data files to verify this.
Flashover, for the minimal reported door opening width of 6 inches
(0.5 feet), occurs at 199 seconds (3.32 min), and at 299 seconds (4.98 min)
for a door opening of only 0.25 feet. Thus, the time to flashover begins to
differ appreciably for door openings greater than approximately 1 foot.
With the door closed, the simulation was terminated before flashover
occurred, i.e., when the layer interface dropped below the fuel height. If it
is assumed that the door was left open at least one foot, then flashover can
be predicted to have occurred approximately at 2-2.5 min past established
burning. As noted, the prediction of the time to flashover is extremely
dependent on the true opening of the door.
The interface prediction (see Figure 6-7) is less affected by the door
opening width. The interface height is between 1.2 and 1.5 m (4 and 5 ft)
when flashover occurs, and decreases with increasing door width.
Based on values presented by Cooper and Stroup, untenability in
enclosure fres is assumed to occur when the hot layer reaches 183°C
(361°F) when the hot layer is above eye level, or when the hot layer
reaches l OO°C (2 12°F) when the hot layer descends below eye level [35].
The selection of 183°C is based on the fact that the radiant flux fiom a
black body (i.e., a perfect radiator, see Section B.3.3.1) at this temperatwe
is 2.5 kw/m2,which is near the threshold of human tenability. The 100°C
(i.e., the boiling point of water) is based on the damages suffered by the
respiratory system due to the inhalation of hot gases and due to irritation
of the skin and eyes. Using 5 feet (1.5 m) as an estimation of eye level, it
can be noted fiom Figures 6-7 and 6-8 that the eye level criterion is not
important since the hot layer drops below that level when the upper layer
temperature is already much higher than 183°C. Between 50 and 55
Sample Problems 137
seconds into the fire, regardless of the door width, the hot layer is above
the 183°C untenability limit. Therefore, there is less than one minute for
occupants to safely exit the hotel room after the onset of flaming
combustion. If it took approximately I minute for the rescuing guest to
assist his roommate to the west exit stairwell and then return to the room
of origin, the conditions within the room would have deteriorated to the
point where reentry to the room was impossible. In another minute and a
half, flashover in the room would occur based on the model predictions.
This would correspond to a real time of about 2:23 A.M. This is about 8
minutes prior to the arrival of the district chief, who reported observing
flames h m the window of the room of origin. Flashover predictions based
on the minimal door openings of one-quarter and one-half of a foot would
also result in flashover times prior to the arrival of the district chief.
The f i e development predicted by the model appears to be reasonable.
In particular, the time to flashover is predicted within the allowable range
based on eyewitness accounts. The time elapsed between flashover and the
time of the district chief S arrival is unknown. Also, the error associated
with the times provided by the eyewitnesses is unknown. Within these
limitations, the true time to flashover cannot be more adequately
quantified As such, an evaluation of the model predictions versus the true
event time line cannot be more specific. However, the predictions do
appear to be acceptable. The results could be used to further study the fue
and to evaluate the effect of varying input data, such as fuel loading, and
first fuel ignited, i.e., bed versus chair. An indication of the effects of
varying door widths has already been provided.
Several limitations must be considered with studies such as the actual
fire reconstruction just provided. The difficulty associated with the
incomplete time lines for comparison has already been mentioned. Also,
the effect and importance of varying the heat loss fractions will be
demonstrated in Chapter 7. This was not included in this study. An
assumption of the study was that when the door was opened to the room
of origin, the smoke present quickly cleared, and the fxe then began as a
small flame on the chair. The effect of varying the input data to account for
varying initial conditions would also be beneficial. Also, it was assumed
that the first fuel ignited (the chair) was the only fuel appreciably
contributing to the fue up to flashover. In the same regard, it is not known
for sure that the fire actually originated in the chair; it could have
originated in the bed near the chair.
138 FIRM-QB USER'S MANUAL
Table 6-6. Heat Release Rates and Times to Flashover of Various Sofa Designs
Computed by FIRM-QB
Sofa Peak Heat Release Rate Time to Flashover (S)
Full Polyolefin
%-Polyolefin
%-Polyolefin
Full Cotton
%-Cotton
%-Cotton
Sample Problems 14 1
FURNITURE WIMETER
TME (S)
FIGURE 6-9. Effect of varying fuel composition on heat release rate [72]
The dip tank is a square tank with Cfoot sides, and the top of the acetone
is approximately 4 feet above the floor. The dip tank is located in a large
room that measures 30 by 50 feet, with a 15-foot ceiling. On each end of
the room are 10-foot high by 6-foot wide openings, where the fiuniture
passes through. Since the model allows for a single vent only, it will be
assumed that a single vent, tvvice the width of the two equal-area vents,
will result in vent flows equivalent to that of the two single vents. Thus, the
vent will be modeled as being 10 feet high by 12 feet wide.
The authority having jurisdiction (AHJ) is concerned that the
unprotected acetone tank poses an unacceptable hazard to the life safety of
the workers, due to the threat of spreading fire caused by flashover.
Predictive modeling can be used to evaluate the threat posed by the acetone
tank. For the present analysis, the primary concerns will be, "Can the
acetone tank cause flashover?'and "How much time is available for safe
egress?"
From measurements of the facility, the geometric data required for
input is available. Based on the information on page 3-78 of Reference
[H], the radiative heat loss fiaction for acetone can be estimated to be
27%. With L, = 0.27, LIW = 1.67, and a moderately rough ceiling,
FIRM-QB suggests a value of 0.62 for L,. Again, since life safety is a
primary concern, low heat loss fractions will be used to afford a factor of
safety, i.e., L, = 0.25, and L, = 0.55. An estimation of the rate of heat
release fiom the burning acetone tank is required. The acetone tank can be
adequately modeled as a pool fire. From Equations (4.35) and (4.36), and
Table 4-3, the heat release rate can be predicted by
The heat release rate, which is considered to be a constant until the fuel is
consumed, is found to be 1457.8 kW. Note that the diameter of the pool,
D, must be specified in meters. A data input screen similar to that for the
acetone dip tank example is shown in Figure 6-1. For this example,
different names were used for the data and fire files, i.e., CH6-Col.*
instead of DIPTANK.", and a maximum simulation time of 900 seconds
was specified. The results for the first 150 seconds of the simulation are
identical to those in Figure 6-2. To review the complete results of the
simulation, the reader can browse through the data file, CH6-CO1.FOF.
Figures 6-10 and 6- 11 show the predicted layer interface height and
hot-layer temperature as a function of time for the acetone fire. Based on
a hot-layer temperature of 600°C, the FIRM-QB model shows that
flashover does not occur (see Figure 6-11). The available safe egress time
can be determined on the basis of untenability criteria, such as the ones
proposed by Cooper and Stroup (see Section 6.1 1.1). Using 5 feet (1.5 m)
as an estimation of eye level, it can be noted fiom Figure 6-10 that the
interface never drops below eye level, since the hot layer becomes steady
at 8.5 fi (2.6 m) above floor level. However, at 35 seconds into the fire, the
hot layer temperature is equal to 183OC. Therefore, there is less than one
1 ' " ' I " "
ceiling
soffit
FIGURE 6-10. Layer interface height for acetone dip tank hazard analysis
FIGURE 6-11. Hot layer temperature for acetone dip tank hazard analysis
144 FIRM-QB USER'S MANUAL
minute for all occupants to safely exit the room containing the acetone dip
tank. If someone is injured and requires assistance, the time available for
safe egress may be insufficient. If occupants delay egress, for whatever
reason, they could be subjected to untenable conditions. Furthermore, since
acetone has a boiling point of only 56S°C, ignition of other acetone tanks,
if present, is quite likely. Based on this analysis, the AHJ may have
sufficient cause to require some form of suppression system protecting the
dip tank (if not already required), or may prevent the continuation of
dipping operations. In other situations, the analysis may suggest that a
variance in the code requirements is permissible, although this would
require a more extensive analysis to verify the effect of uncertainties in the
input data on the results of the predictions.
FIRM-QB will run on virtually any IMB compatible PC. The source
code can be loaded and executed fiom the QBasic environment. The
QBasic interpreter is included with DOS (version 5.0 and higher) and
Windows 95. The executable can be run directly fiom DOS or Windows.
Therefore, there are no significant hardware or software restrictions.
FIRM-QB predicts the consequences of a user-specified fire in a
compartment with a single vent in one of the vertical walls of the
compartment. Therefore, the application of FIRM-QB is limited to certain
types of problems. The predctive capability of FIRM-QB has been verified
for compartments with a floor area of 100 m2 (or 1000 fl?) or less and a
height of the order of 3 m (or 10 ft), and fires of a few Megawatts (see
Chapter 7). If FIRM-QB is used to simulate a much bigger fire and/or a
fue in a much larger compartment, it is strongly recommended that the user
evaluate the predictive capability of the model according to the guidelines
in ASTM E 1355.
7.1 INTRODUCTION
7.2.1 Documentation
7.2.2 Validation
correctness of the assumptions that are made and the approaches that are
used.
7.2.3 Verification
7.2.4 Evaluation
1 Standard tests. Standard test data are useful for the evaluation of
models that predict how a material or assembly performs in the test.
Only a few standard test procedures involve a room, and most standard
test data are therefore not applicable for the evaluation of compartment
fire models. ASTM E 603, "Standard Guide for Room Fire Tests" [87]
provides general guidelines for conducting full-scale fire experiments,
and is perhaps the most useful standard test procedure in terms of
generating data suitable for compartment fire model validation.
2 Tests conducted speclJicallyfor thispurpose. Due to the high cost, it
is very unusual that full-scale tests are conducted specifically to
provide data for evaluation of a particular model. If experiments are
conducted, they should be designed judiciously to assure the data
produced by the tests affords the best data for comparison. For
example, a model that does not calculate layer species concentrations
certainly would not require any experiments where these data are
measured.
3 Test data in the literature. For obvious reasons, the open literature is
by far the most common source of data for model evaluation.
4 Fire experience. Fire risk assessment involves a very large number of
deterministic computer fire model runs, and can be used to evaluate the
model by comparing the results of the risk assessment to fire statistics.
Compartment fire models are usefd tools in the reconstruction of fires,
and can be evaluated by checking whether model predictions are
consistent with the time line and other pieces of information in the fire
investigation report (e.g., see Section 6.11.1).
Predictive Capability of Fire Models 151
There are many problems in comparing the results from fire model
simulations to data fiom full-scale fire experiments. Some of the problems
are due to the differences between the form of the recorded experimental
data and the form needed for comparison with model predictions. For zone
models, the compartment is divided into two distinct zones, a lower cool
layer and a hot upper layer. In reality, there is no such clear and sharp
change distinguishing the lower and upper layers. A typical temperature
profile inside a f i e compartment is shown in Figure 7-1. Also shown in
Figure 7-1 is the corresponding idealized temperature distribution for a
two-zone situation. The difference between actual and ideal is obvious. To
use experimental data for comparison with zone model results requires that
the experimental data be cast into an idealized form, i.e., isothermal upper
and lower layers separated by a sharp interface. This is commonly
accomplished by identifying the ideal interface level, and then simply
averaging the temperatures within the hot layer based on thermocouple
data. Thus the problem is one of identifying the ideal interface level. A
common procedure to determine the location of the layer interface on the
basis of vertical temperature profile measurements was developed by
Cooper et al., and is referred to as the NO/o rule [93]. According to this rule,
the interface is located at a height at which the gas temperature rise above
ambient is some percent of the temperature rise (10, 15, or 20 percent, for
example) of the top-most thermocouple in the test room. The dependence
on the selection of the interface on the data used for comparison is evident.
A lower value for the interface will likely lead to a lower hot-layer
l ' l - l ~ l - l ~ l - i
Two-loyer profile
2.0 - P Measured profile - 2.0
-
C
0
- U-
-
1.5- /--U- 1.5
E 0
W -0'
3
N
3
- 1 .o
...........................................................
.-
0,
0,
C*
0
t
Neutral plane height a
I
0.5 -
P
0 - 0.5
0
0.0 I . I . I . I . I . I .
0.0
0 50 100 150 200 250 300 350
Temperature (OC)
7.3.1 Documentation
7.3.2 Validation
7.3.3 Verification
The numerical algorithms that are used to solve the non-linear algebraic
and ordinary differential equations are discussed and verified in
Appendices B and D.
7.3.4 Evaluation
The first data set was reported in a paper by Peacock et al. in the
Journal of Research of the National Institute of Stamkwds and Technology
[95].This paper provides an extensive discussion on the subject of room
Predictive Capability of FIRM-QB 157
fire testing and accuracy assessment of room fire models. Five sets of
experimental data, which can be used to evaluate the predictive capability
of zone-based fire models, are described in detail. The data are available
fiom MST on a CD-ROM as "Fire Data Management System" (FDMS)
ASCII data files [96]. The first set, referred to as "Single Room with
Furniture" was chosen to evaluate FIRM-QB, primarily because the test
scenario is identical to that simulated by the model, i.e., a single furniture
item in a room of fixed size but with varying vent sizes and shapes. The
upholstered furniture items were tested previously in the furnitwe
calorimeter. The description of the tests in the following paragraphs is
largely taken fiom the aforementioned journal article.
Test Room. An experimental room, 2.26 X 3.94 X 2.3 l-m, with a window
opening in a narrow wall was constructed inside the large-scale fire test
facility. The dimensions of the window openings for the various tests are
given in Table 7-2. The sofit depth of the window opening was the same
in all cases. For tests 1 and 2, the opening height (and therefore the
ventilation parameter AV=) only was varied. For test 6, the same A ~ J H ,
was retained as in test 2, but the shape of the opening was changed. Test
5 resembled test 6, except that the fuel was an armchair instead of a
loveseat. Thus, for specimen type, ventilation factor, and opening aspect
ratio, a pair of tests each was provided where these variables were singly
varied, the other two being held constant. The walls and ceiling materials
in the room were 16 mm thick Type X gypsum wallboard, fiured out on
steel studs and joists. Floor construction was normal weight concrete. The
test room was conditioned before testing by gas burner fires, where the
paper facing was burned off the gypsum wallboard, and the surface
moisture driven off The room was allowed to cool overnight and between
tests. The room was equipped with an instrumented exhaust collection
system outside the window opening that can handle fires up to 7 MW size.
Table 7-2. Chairs and Vent Sizes for NIST Single Room Tests with Furniture
Test # Chair 5 (m) Zt (m) Hv(m) W , (m) AVJH(mm)
Experimental Conditions. The test furniture for four of the six tests is
listed in Table 7-2, and included a 28.3 kg armchair (F21) and a similar
40.0 kg loveseat (F3 1). A description of these items is provided in Table
4-4. A single piece of test furniture and the igniting wastebasket were the
only combustibles in the test room. The furniture items were tested
previously in the furniture calorimeter (see Section 4.4.4). The tests in the
furniture calorimeter [64] made use of a gas burner simulating a
wastebasket fire as the ignition source. Because of practical difficulties in
installing that burner in the test room, actual wastebasket ignition was
used. This involved a small polyethylene wastebasket filled with 12
polyethylene-coated paper milk cartons. Six cartons were placed upright
in the wastebasket, while six were tom into six pieces and dropped inside.
The total mass of the wastebasket was 285 g, while the 12 cartons together
weighed 390 g, for a total mass of 675 g. The gross heat of combustion
was measured to be 46,320 kT/kg for the wastebasket and 20,260 W/kg for
the cartons, representing 2 1.1 MJ in all. Using an estimated correction, this
gives a heat content of 19.7 MJ, based on the net heat of combustion. To
characterize this ignition source, a constant mass loss rate of 1.8 g/s
(equivalent to 52.5 kW) was assumed for the fxst 200 S.The mass loss rate
of the ignition source was considered negligible thereafter.
Predictive Capability of FIRM-QB 159
Table 7-3. Fire and Output Files for Single Room with Furniture Tests
Measured Heat Release Rate
--
Triangular Heat Release Rate
Test # Fire Filea Outnut Fileb Fire File" Output Fileb
-----
............. FIRM-Q (Triangle HRR)
FIGURE 7-2. Upper layer temperature measurements and predictions for test 1
FIGURE 7-3. Layer interface height measurements and predictions for test 1
Single Room with Furniture: Test #l(CH7-BO1)
FIGURE 7-5, Upper layer vent flow measurements and predictions for test 1
Single Room with Furniture: Test #2 (CH~-B02)
1- 11
Experiment
----- FIRM-Q (Measured HRR)
.........
.,,.
FIRM-Q (Triangle HRR)
FIGURE 7-6. Upper layer temperature measurements and predictions for test 2
FIGURE 7-7. Layer interface height measurements and predictions for test 2
162
Single Room w i t h Furniture: Test #2 (CH7-602)
FIGURE 7-9. Upper layer vent flow measurements and predictions for test 2
163
Single Room with Furniture: Test #5 (CH7-B03)
F'IGtTRE 7-10. Upper layer temperature measurements and predictions for test 5
FIGURE 7-11. Layer interface height measurements and predictions for test 5
Single Room with Furniture: Test #5 (CH7-€303)
Time (S)
FIGURE 7-13. Upper layer vent flow measurements and predictions for test 5
Single Room with Furniture: Test #6 (CH7-004)
FIGURE 7-14. Upper layer temperature measurements and predictions for test 6
Experiment
FIRM-Q (Measured HRR)
,,,.......... FIRM-Q (Triangle HRR)
FIGURE 7-15. Layer interface height measurements and predictions for test 6
Single Room with Furniture: Test #6 (CH7-B04)
1.0 -
0.8 - /,?%L
......-11 %+.,
\'..
. I \
\ ...
"
*\-L..
:..-.-..---
: I
i l
0.6 - i l 1.
i #
:
l
I ::
: I -
.-
-d
.-
c
.
i I --C--_
: l L.-
i lI
*.-
0.4 M : 1
i lI
I
Y
- :i
C:
0.2 I :
I
I
:i
II II
FIGURE 7-17. Upper layer vent flow measurements and predictions for test 6
well with the measurements, except during the peak burning period, when
the measured interface height drops down temporarily to 0.5 m (or less)
from the floor. Figures 7-4, 7-8, 7-12, and 7-16 provide a comparison
between the measured heat release rates, and those used by the model.
There is a slight reduction of the heat release rate due to oxygen starvation
around the peak burning rate in tests 2 and 6. The triangular approximation
is quite good in test 1, but appears to be rather crude for the remaining
tests. However, the model does not seem to be very sensitive to the exact
shape of the heat release rate curve, and the triangular model seems to be
adequate for engineering analyses. The calculated upper layer vent flows
are shown in Figures 7-5,7-9,7- 13, and 7- 17. The measured flows are not
shown because Peacock et al. concluded on the basis of the tests' data that
the arrangement of velocity probes spaced along the centerline of the
window opening leads to serious errors in computed mass flows.
Test Room. The experiments were conducted in the test room depicted in
Figure 7- 18. The dimensions of the test room were 2.8 X 2.8 X 2.13-m. The
vent (opening) configurations that were used for the various tests are
shown in Figure 7-19. The ceiling and walls of the test room were lined
with ceramic fiber insulation. The thickness of the linings were not
provided in the report documenting the experiments, nor was a description
of the floor material included.
Predictive Capability of FIRM-QB 169
MOVABLE B I
m c
T I
oNA 1
,VUOCRY
PROBES AND THERMOCOUPLES
ASPIRATED THERMOCOUPLE
DOOR OPENINGS
WINDOW OPENINGS
FIGURE 7-19. Vent configurations for steady vent flow experiments
170
Predictive Capability of FIRM-QB 171
Table 7-4. Data Files and Test Conditions for Steady Vent Flow Tests
Q
Test P Output Fileb Fire Filec Vent Configuration (kW)
216 Door
316 Door
416 Door
416 Door
516 Door
616 Door
616 Door
616 Door
616 Door
7/6 Door
816 Door
Full Window
213 Window
213 Window
113 Window
616 Door
616 Door
616 Door
" Numbers are for reference purposes, and are not in original report
Fire file names have extension .FIR
Output files have extension .FOF
it was expected that the total heat losses were very low. Therefore, instead
of using a generic value, L, was estimated for each test on the basis of the
steady-state energy conservation equation for the compartment (Equation
7-l),and measured gas temperatures and vent flows reported in Reference
[49]. The data and resulting total heat loss fractions are given in Table
7-5.The steady-state energy conservation equation is as follows
Table 7-6. Comparison of Vent Flow Test Data and FIRM-QB Predictions
Measurements FIRM-QB Predictions
Test # & (m) T,- T,('C) m,,(Ws) Zi (m) T,- T, ('C) % (k@)
174 EVALUATION OF THE PREDICTIVE CAPABILITY OF FIRM-QB
Steckler et al. observed that the flame was pushed over by the flow of
air entering the room through the lower part of the vent. This phenomenon
was discussed in more detail by Quintiere et al. in Reference [97]. In this
paper, average flame angle measurements based on video recordings are
presented for some of the tests conducted by Steckler et al. Quintiere et al.
suggested increasing the vertical distance between the fuel surface and the
layer interface in Zukoski's equation to bring the calculated entrainment
rates in better agreement with the vent flow measurements.
The PlumeFlow subprogrm in FIRM-QB was modified to account for
the flame and plume angle. The modified source code is shown in Figure
7-20. The distance between the kel surface and the layer interface is
simply divided by the sine of the flame angle. If the centerline of the tilted
flame hits a wall before it reaches the layer interface, the increased
entrainment height is set equal to the distance between the fuel surface and
the w d . The variable Theta is the flame angle, which was added to the list
of global variables that are shared between the main program and all
subprograms through a COMMON SHARED statement in the main
program. A few lines were added to the InputData subprogram to allow the
user to input the flame angle fiom the keyboard.
The measured flame angles reported in Reference [97], the flame angle
values used for the calculations with the modified FIRM-QB program, and
the results of the calculations are given in Table 7-7. The layer interface
quickly dropped below the sill in tests 13, 14, and 15, leading to
termination of the modified FIRM-QB program. A comparison of the
calculated results in Table 7-7 and the measurements in Table 7-6 shows
that the inclusion of the flame angle effect greatly improved agreement
between FIRM-QB predictions and the experimental data. Figures 7-2 1,
7-22, and 7-23 provide the same comparison in graphical form.
7.4 CONCLUSIONS
END I F
l IFme
ELSE
' ( ( 1 - Lr)
Zi= KZf* THEN * Q * DeltaZi A 5) A (1 / 3 )
me = 0
END IF
I PlumeFlow
END FUNCTION
= me
FIGURE 7-20. Modified version of Zukoski's plume flow model for a tilted plume
One of the limitations of FIRM-QB is due to the fact that the lower
layer is assumed to be at ambient temperature. However, this results in
higher, and therefore conservative predictions of the hot layer temperature.
A major strength of FIRM-QB is that the source code is provided and can
easily be modified to address a particular circumstance or problem, as was
illustrated in Section 7.3.5.2.
FIGURE 7-21. Calculated vs. measured Zi for steady vent flow tests
h
200 - 0 .,. . .
.."a
P
W 8 O ..:'
o&00 ,,$..,'
'
l-
3
150
8 . ,./
-U 8-
Q, &.....a
. '
ZI
100- m,.
:..'
-
0 0 ,......"
0 ....."'
0
50 -
0 Without flame angle correction -
With flame angle correction
O ; . . ~ . ; . . . l s . . . l . . . l m . m m l . , ~ m
0 50 100 150 200 250
Meosured T,,-To ("C)
FIGURE 7-22. Calculated vs. measured T, - T,for steady vent flow tests
............. Line of perfect agreement
- ;.m"
.a
,.""
,..."' 0,
.. "'
.aa. m
* "'
0
.. .'i 0
@
0 ,go
0
,,. 0'
...' 0 -
Measured mu (kg/s)
FIGURE 7-23. Calculated vs. measured m, for steady vent flow tests
CHAPTER 8
Conclusion
FAST and FASTLITE programs, and older fire modeling software such as
FIRST, FPETool, and LAVENT Alternatively, the same programs (as well
as data, publications, and other bits of information that may be of interest
to the fire modeler) can also be downloaded fiom the MST BFRL web site
OJRL: http ://fiie.nist.gov/).
NIST also developed a CFD model to predict smoke and air flow
movement caused by fire, wind, ventilation systems, etc. This model is
referred to as the FIRE DYNAMICS SIMULATOR (FDS). A separate
program, called SMOKEVIEW, visualizes the predictions generated by
FDS. These programs (as well as a user's guide and example data files)
can be downloaded fiom the aforementioned MST BFRL web site.
SMARTFIRE is another popular CFD-based fire model. The
SMARTFIRE program was developed by the Fire Safety Engineering
Group at the University of Greenwich in the UK. SMARTFIRE is user
friendly, and appears to be an excellent educational tool for the novice
CFD fire modeler. It is continuously being improved and expanded. For the
latest information on SMARTFIRE, it is recommended the reader consult
the FSEG web site OJRL: http://fseg.gre.ac.uk/).
It is hoped that the reader, who might have avoided using models in the
past because of the potentially intimidating mathematical equations and
computer programs, will start to appreciate and enjoy mathematical fire
modeling as a result of this book.
APPENDIX A
Conversion Factors and Constants
The following properties describe the state of a fluid, and may vary
from point to point.
where
where
Although a fluid at rest is unable to sustain shear forces, this is not the
case with a fluid in motion. Furthermore, it has been observed
experimentally that a moving fluid is at rest immediately adjacent to a
stationary solid boundary. Therefore, there must be a region close to the
solid boundary where the velocity changes from zero to the main stream
value. Figure B-l shows a typical velocity profile in this transition region,
which is referred to as the boundary layer.
The shear stress at a point for most fluids of practical interest is
proportional to the rate of deformation due to shear forces at that point:
I Velocity v /
-I Free S t r e a m
Layer
Solid Boundary
where
Fluids for which Equation (B.3) is valid are called Newtonian. The
viscosity, p, is a physical property of the fluid. It is related to the transverse
transfer of flow momentum due to molecular motion. Viscosity is therefore
a function of temperature, composition, and pressure of the fluid. The ratio
of viscosity to density occurs often in engineering problems. This ratio,
pip, is referred to as the kinematic viscosity, v. To distinguish p from v,
the former is called the dynamic viscosity. The units of p are Pa-S(S.I.
unit) and lb/ft=s(U.S. engineering unit). The kinematic viscosity, v, is
expressed in m2/s(S.1. unit) and ft2/s (U.S. engineering unit).
It has been observed experimentally that there are two very different
types of fluid motion. Under certain conditions, the flow will be regular or
laminar. Equation p . 3 ) is valid for this type of flow. Under other
conditions, however, the flow appears to be irregular with transverse
eddies. When this type of flow prevails, Equation (E3.3) has to be modified
to account for the transverse transfer of momentum due to eddy movement.
The modified equation can be written as follows
where his the turbulent viscosity, associated with the momentum transfer
due to eddy movement, which is usually much greater than the molecular
viscosity, p. Unfortunately, p, is not a physical property of the fluid, but a
function of the flow itself. Therefore, in order to characterize a turbulent
flow field, the conservation equations need to be complemented by one or
more additional equations that describe the dissipation of turbulent energy.
With a few exceptions (very small fires), the flows associated with
enclosure fires are turbulent.
1
P + -pv2 + pgh = constant (B4
2
where h is the height above a fuced datum level (m). Bernouilli's equation
can be used to describe the flow through a sharp-edged orifice plate,
commonly used to measure fluid flow through a pipe (see Figure B3).The
stream emerges fiom the orifice as a jet. The jet converges to a vena
contracta just downstream of the orifice plate, and then breaks up into a
turbulent flow region. Application of the continuity equation between
sections 1 and 2, assuming the fluid is incompressible, leads to
This equation is used in Section 2.6.2.2 to determine the mass flow through
a vent in a vertical wall of an enclosure that contains a fie.
where
The Froude number is the ratio of inertial to gravity forces. Fr is, for
example, used to correlate experimental data fiom gas burner jet fires.
8.2THERMODYNAMICS
The density, pressure, and temperature of an ideal gas are related by the
equation of state, (B.l), which can also be written as
where
c = compressibility factor
APPENDIX B
Temperature ("C)
FIGURE B-4. Compressibility factor of some gases
(B.l4b)
where
(B.16)
Equation (B. 16) shows that the equation of state can also be applied to a
mixture of gases, whereby the gas constant of the mixture is related to the
gas constants of its components by the following expression
B.2.2.1 Work
Subscript 1 refers to the original state (prior to the increase of F), and
subscript 2 refers to the fmal state after compression to the new
equilibrium. The value of W,, is a h c t i o n of the process, i.e., the path of
successive states passed through between the initial and fmal states. For
example, if the cylinder is water-cooled to keep its temperature constant,
the work done by the system can be calculated fiom
Wl, = i PdV =
im~mix'
v
1 1
Note that W,, is negative for a compression process (V, is smaller than V,),
and positive for an expansion process.
The type of work described in this section is referred to as
displacement work. Many other types of work are important in engineering
thermodynamics, for example, work associated with fluid friction, stirring,
electrical effects, etc.
B.2.2.2 Heat
where c,, is the specific heat at constant volume of the gas, which is a
material property that varies with temperature. The specific heat at constant
volume of a gas mixture can be calculated ffom the specific heat of its
components using the following relationship.
where c, is the specific heat at constant pressure of the gas (the reason for
this name will be explained in Section B.2.2.6). As with c,, the specific
heat at constant pressure is a material property that varies with temperature.
Equation (B.23) is valid only for very small temperature changes. To
calculate the enthalpy change for a large temperature difference, the
temperature-dependency of c, must be accounted for. Denoting the
temperature difference by AT = T2 - Tl, the corresponding specific
enthalpy change follows fiom
where 5 is the average specific heat between Tl and T2. The average
specific heat of most gases does not vary strongly with temperature, and a
cubic temperature function fits the data quite well:
The polynomial constants for gases that are of primary interest to the fire
modeler are provided in Table B- l. These constants are for T in "C, with
0°C as the reference temperature (Tl = O°C). The resulting equations are
valid for temperatures between 0 and 3000°C.
Table RI. Polynomial Constants to Calculate c,(T) of Some Gases
M A B C D
Species (kglkmol) (kJ/kg°C) (kJ/kg.oC2) (kJ/kg°C3) (kJ/kg°C4)
Ar 39.95 0.5207 0 0 0
where the subscripts and superscripts are the lower and upper limits
respectively, of the temperature interval over which the average specific
heat is calculated.
The specific heat at constant pressure of a gas mixture is equal to the
mass-weighted average of the specific heats of the components.
For example, dry air consists of 1.28% Ar by mass, 0.05% CO,, 75.53%
NZ,and 23.14% 0,. The reader can verify that the constants in Table B-l
are identical to the mass-weighted averages of the component constants.
Review of Fundamentals of Engineering for Fire Modeling 199
In the case of an open system, the subscript 1 refers to the state of the fluid
when it enters the system, and 2 refers to the state of the fluid when it
leaves the system. Since an open system control volume is (usually) fued,
work is not associated with a change in volume, but with a change in
pressure. If the process is isobaric, i.e., the pressure in the open system is
constant and P, = P,, no work is performed on or by the system. In this
case, Equation (B.29) indicates that the heat transferred to the system is
equal to the enthalpy rise of the fluid flowing through the system. This is
the reason why c,, as defmed by Equation (B.23), is referred to as the
specific heat at constant pressure.
B.3.1 Conduction
where q" is the rate of heat transfer per unit area in the X direction. There
is a minus sign on the right hand side of the equation, because heat is
transferred fiom a point at higher temperature to a point at lower
temperature, i.e., in the opposite direction of the temperature gradient. The
proportionality constant is called the thermal conductivity, and is a material
property. Insulating solids have a thermal conductivity of the order of 0.1
WImK, while metals have a thermal conductivity that is 100 to 1000 times
greater. Thermal conductivity generally increases as a function of
temperature.
If conduction is the only mode of heat transfer, as, for example, in an
opaque solid, the conservation of energy leads to the following general heat
conduction equation
where p is the density (in kg/m3),and c, is the specific heat (in Jkg-K), T
is temperature (in 'C or K), t is time (in S), and k is the thermal
conductivity (in W/mK). In a Cartesian coordinate system, Equation
(B.31) has the following form
Review of Fundamentals of Engineering for Fire Modeling 201
The left hand side of this equation is equal to the amount of energy that is
stored in the solid, per unit time and volume. The right hand side is a
summation of the gradients of conduction heat flux in the three directions,
X , y, and z. If the thermal conductivity is constant, i.e., does not vary with
temperature (and, therefore, with location), Equation (B.32) simplifies to
B.3.2 Convection
where q" is the rate of heat transfer per unit area fiom the solid to the fluid,
T, is the solid surface temperature, and Tf is the fluid's fiee stream or bulk
temperature. If T, is higher than Tf, heat is transferred from the solid to the
liquid and Q" is positive (by convention). If T, is higher than T,, q" is
negative. Equation (B.34) is referred to as Newton 'S Law of Cooling.
As in the case of pure fluid flow without heat transfer, the systematic
analysis of convection problems is greatly facilitated by the use of
dimensionless parameters.
where
where
laminar and turbulent flow). For example, the Nusselt number for natural
convection over a vertical flat plate at constant temperature is given by [29]
The characteristic length, I, in this case is the height of the plate. Once the
Nusselt number is known for the geometry and flow conditions at hand, the
heat transfer coefficient (Equation B.36), and rate of convective heat
transfer pquation (l3.34)] can be calculated.
where c is the speed of propagation of the wave. Matter is not required for
radiant heat transfer, and c in a vacuum is equal to the speed of light
(approximately 300,000 kmh). A body at a temperature greater than
absolute zero emits thermal radiation over a range of wavelengths fiom 0.1
to 100 pm. This includes the visible region, which extends from 0.38 to
0.76 pm. Thermal radiation at longer wavelengths than the upper limit of
the visible range is called infared. In ffires, radiation heat transfer occurs
primarily at wavelengths in the hfkared region.
Thermal radiation that falls upon a body is partly absorbed, partly
reflected, and partly transmitted through the body, or in equation form
where
Review of Fundamentals of Engineering for Fire Modeling 205
Most solid materials absorb all radiation within a very thin surface
layer. These materials are called opaque, and have a transmissivity equal
to zero ('G = 0). In fire modeling applications, solid surfaces can generally
be considered diffuse, i.e., they reflect and emit radiation uniformly in all
directions. Gases do not reflect (p = O), but absorb andor transmit thermal
radiation.
where
0 1 2 3 4 5
Wavelength, h (pm)
FIGURE B-6. Blackbody monochromatic emissive power
where
Wien's displacement law explains why the color of a hot metal surface
changes to bright red as its temperature increases above 600°C.
The total ernissive power (in w/m2) fiom a blackbody, 4,at a given
temperature is equal to the area under the spectral distribution at that
temperature, and is given by the following remarkably simple relationship
where
o = Boltzmann constant (5.67010-*w/m2K4)
A real body generally does not absorb all incident radiation, i.e., part
of the incident radiation is reflected at the surface. A real surface also emits
less radiation than a blackbody at the same temperature. The ratio of the
monochromatic emissive power of a real body to that of a blackbody at the
same temperature is referred to as the emissivity, E*. The emisivity of a real
body varies with wavelength. For engineering purposes, it is often assumed
that the emissivity is independent of wavelength. A surface with this
property is referred to as a grey surface. The emissive power of a grey
surface is related to its temperature by
AU gases, except those with diatomic molecules, absorb (and also emit)
thermal radiation. However, they do not absorb radiation at all
wavelengths, but only over specific bands. A graph of the absorptivity of
a gas as a function of wavelength is referred to as its absorption spectrum.
The absorption spectrum of a gas mixture is equal to the sum of the
absorption spectra of its components (each weighed on the basis of the
component's concentration in the mixture), except at wavelengths where
component spectra overlap. Rather complex corrections are necessary to
determine total absorptivity at overlapping wavelengths. Smoke from fres
nearly always contains significant amounts of soot particles. The
absorption spectrum of a soot cloud is continuous, and must be combined
with the discrete spectrum of the carrying gas mixture to determine the
radiation characteristics of the smoke. Despite the fact that the resulting
spectrum has discrete peaks at wavelengths where one or several gas
components absorb, an approximate grey gas assumption is often used for
engineering calculations of radiation heat transfer through smoky fire
gases. The absorptivity of a grey gas volume is a fhction of the length of
the path that a beam has to travel to cross the volume. If the path length is
denoted as L, the absorptivity is given by
Review of Fundamentals of Engineering for Fire Modeling 209
B.4 COMBUSTION
One molecule of methane reacts with two molecules of oxygen to form one
molecule of carbon dioxide, and two molecules of water. The total number
of carbon (l), hydrogen (4), and oxygen (4) atoms is the same on both
sides of the equation.
Most combustion reactions take place in air, instead of pure oxygen.
The mole (or volume) fi-action of oxygen in dry air is 20.95%. The balance
consists of nitrogen (78.09%) and small amounts of other non-reacting
gases, primarily argon (0.93%) and carbon dioxide (0.03%). For
engineering calculations, dry air is assumed to consist of 2 1% by volume
oxygen, and 79% by volume nitrogen. Hence, complete combustion of
methane in air is described by the following equation, which can be
derived from Equation (B.49)
where the numerator is the ratio of the mass fraction of air to the mass
fraction of fuel in a stoichiometric mixture, and the denominator is the
same ratio for the actual mixture. Q, of a fuel-rich mixture is greater than
one. is less than one when the mixture is fuel-lean. A stoichiometric
fuellair mixture, by d e f ~ t i o nhas
, an equivalence ratio equal to one.
This energy balance can be expressed on a fuel mass basis by dividing each
side of the equation by m, Denoting the ratio of the rate of heat extracted
to the rate of fuel supplied as q, this leads to
If fuel and air were supplied at a reference temperature, To, and if heat
were extracted at the exact rate to cool the products down to the same
temperature, Equation (E3.60)would take the following form
CHAMBER
2600
1
Equivalence Ratio, @
where
Even the simplest fire models (such as the ones presented in this book),
consist of a collection of relatively complex equations, which cannot be
solved analytically. Numerical methods must be used to obtain approximate
solutions. Two such methods are discussed in this section. First, however,
it is important to provide some discussion concerning computer accuracy.
Computers store numbers in the form of binary digits, or bits. A bit can
have a value of 0 or 1.Bits are combined into groups of eight, called bytes.
Numbers are represented using a binary, or base 2, numbering system. An
integer can be represented exacly, as long as its value does not exceed the
range that can be represented. The highest number that can be represented
is a function of the number of bits (or bytes) that are used. For example,
the maximum value that can be represented in one byte is 1111 1 111 in
binary. The corresponding decimal number is given by
Negative integers can be represented by using the first bit as a sign bit (0
for positive numbers, 1 for negative numbers). In this case, numbers
between - 128 and 127 can be represented in one byte.
More sophisticated methods are needed to represent real (or floating-
point) numbers. A single precision representation of real numbers consists
of a group of four bytes, divided into portions. The first portion, the
mantissa, holds the base value of the number. The second portion, the
exponent, indicates to what power the mantissa must be raised to obtain the
final value of the number. As with integers, as sign bit is used to
distinguish between positive and negative numbers. The most common
Review of Fundamentals of Engineering for Fire Modeling 221
in FIRM-QB (and FIRM-VB) to solve the vent flow equations (see Section
4.1.6).
An equation can always be written in the following form
X,,-X mm
.
S accuracy @.W
2"
(B.69a)
-
To obtain a numerical solution, the solution domain is first subdivided
into N steps. The step size is denoted as h, and the upper limit of each
interval is given by xi X, + i h, with i = 1 ... N. Instead of a continuous
solution, a numerical method frnds a discrete solution in the form of
estimates of y at X,,X,, ..., X,.
The simplest approach to numerically solve an initial value problem is
the Euler or tangent method. Since X, and y, are known, the slope of the
tangent line to the solution at x = X, can be calculated fiom the differential
Equation (B.69a). If h is small, the slope can be used to obtain a reasonable
estimate of y, as illustrated in Figure B-9. This process can be repeated at
the estimated (X,, y,) to obtain an estimate of y,, subsequently at (X,,y,) to
224 APPENDIX B
If we denote the exact solution of the initial value problem as @(X), then,
a Taylor series expansion to express @(xi+ h) as a function of @(xi)leads
to
Comparison of Equations (B.70) and (B.72) leads to the conclusion that the
local truncation error of the Euler method, i.e., the error of the numerical
solution when advancing one step h in the solution domain, is proportional
to h2. It can be shown that the cumulative error over the entire solution
domain is proportional to h. Therefore, to reduce the cumulative truncation
error by a factor of 2, h must be halved. However, besides doubling the
computation time, this also increases the cumulative roundoff error, often
to a point that it defeats the purpose of using a smaller stepsize. Because
Review of Fundamentals of Engineering for Fire Modeling 225
its accuracy is limited by the roundoff error, the Euler method is generally
not recommended for the numerical solution of initial value problems.
Around the turn of the 19th century, the German mathematicians
Runge and (later) Kutta developed some improvements to the Euler
method. The basic idea was to estimate the slope at an intermediate point
inside the step interval, instead of using the slope at the start of the interval
to advance the solution to the end of the interval. The most frequently used
Runge-Kutta method is the fourth-order method described below.
Derivation of these equations is beyond the scope of this book, and can be
found in many textbooks on numerical analysis (see, for example, Chapter
8 in Reference [102]).
where
C.1 INTRODUCTION
The QBasic program and data files on the CD-ROM are not
compressed. The software is therefore installed by copying all fdes to the
227
hard disk. It is recommended that the program files be copied to C:\FIRM\,
and the data fdes be kept in C:WIRMUIATA\. The latter is the default
directory where the Visual Basic programs expect to fmd all fire and
output files. In Windows 95/98, the installation can be performed as
follows:
There are several other ways to run DOS programs from within Windows.
The reader is referred to the Windows user's guide for more details.
To run the source programs, the QBasic interpreter must first be
loaded. Follow the aforementioned instructions for starting a DOS program
to run the interpreter. In the QBasic environment, choose the Open
command from the File menu. Then switch to the directory where HRR-
QB.BAS, ASET-QB-BAS, and FIRM-QB reside, and double-click on the
desired program name. Alternatively, scroll down until the name is
highhghted, and press Enter or click OK. The source code is loaded in the
QBasic environment, and can be edited and executed fi-om this
environment.
To run the Visual Basic programs, the reader must have a system that
meets the following requirements:
The setup routine on the CD-ROM must be used to install the Visual
Basic software. The installation can be performed as follows:
The setup routine decompresses the files on the CD-ROM, creates the
necessary directories on the user's hard drive (if needed), copies the
programs and Windows DLL files fiom the CD-ROM, updates the
Windows registry, and creates a FIRM-VB program group.
If the reader intends to customize the Visual Basic programs, it is
recommended that the source code be copied fkom the \FIRM\ directory on
the CD-ROM to the hard disk. This is not necessary if the QBasic software
already has been installed (see Section C.2.2).
There are two ways to run the Visual Basic software. The fust way
consists of running the compiled Visual Basic programs. Click the Start
button, then choose Programs and FIRM-VB, and click on the desired
program name, i.e., HRR-VB, ASET-W, or FIRM-W. The QBasic
executables can also be started from the FIRM-VB program group. A
special icon was created for each of the three Visual Basic programs. It is
recommended that the same icons be associated with fire files (*.FIR
extension), ASET output files (*.AOF extension), and FIRM output files
(*.FOF extension) respectively. This can be done by clicking the Start
button, then choosing Settings and Folder Options, and selecting the File
Types tab. After pressing the New Type button, the user can specify an
extension and select the corresponding icon. Select HRR-VB.EXE,
ASET-W.EXE, or FIRM-VB.EXE as the source file for the icon.
The second way must be used if the reader wants to modify the Visual
Basic source code. In this case, the program (*.WP extension) must be
loaded into the Visual Basic programming environment. There are several
books on the market that come with a working model of Visual Basic 6.0
(e.g., Perry, G. and S. Hettihewa, 1998. Teach Yourself Visual Basic 6 in
24 Hours, Sams, Indianapolis, IN), which provides the reader with an
inexpensive way for exploring the source programs (retail prices are as low
as $20). Working models do not allow the creation of stand-alone compiled
programs. Readers who wish to compile and perhaps distribute their
modified programs should consider the Visual Basic 6.0 Deluxe Learning
Edition, published by Microsoft Press. With a list price of $130, the
Deluxe Learning Edition is much more expensive than any of the working
models. However, it comes with two books, and a multimedia tutorial on
CD-ROM. If the reader requires more advanced capabilities, he may need
to obtain the Professional or Enterprise Edition of Visual Basic 6.0.
APPENDIX D
QBASIC Programmer's Notes
D.l INTRODUCTION
l Feldman, P., and T. Rug, 1993. Using Basic (2nd Edition), Que
Corporation, Cannel, IN.
2 Dyakonov, V., V. Munerman, E. Yemelchenkov, and T. Sarnoylova,
1996. The Revolutionary Guide to QBasic, Wrox Press, Birmingham,
m.
The first book is a tutorial and reference for beginning to intermediate
QBasic programmers. The second book addresses more advanced topics
and discusses the development of professional QBasic applications.
A quick search through the world wide web shows that there still is an
extensive following of die-hard QBasic afficionados, despite the fact that
QBasic is a somewhat obsolete DOS-based programming tool. The
following two web sites are among the most frequently visited:
1 http://neozones.quickbasic.com
2 http://www.qbasic.com
Both sites offer book reviews, tutorials, tons of sample programs, links to
other QBasic web sites, and much more.
The QBasic interpreter is needed if the reader wants to modify the
source code provided with this book. It can be copied from the
\OTHER\OLDMSDOS folder on the Windows 95 CD, or can be
downloaded fiom the neozones web site.
A compatible compiler is much harder to find. The executable QBasic
programs on the accompanying CD-ROM were generated with Microsoft
PDS 7.1. However, Microsoft discontinued its support of the Professional
Development System, and legal copies of QuickBasic 4.5 are no longer
available either. There are a few BASIC compilers that can be downloaded
from the neozones web site, but none of these compilers are fdly
compatible with QBasic. The efforts to convert the source code are
probably not worth the trouble, since execution in the QBasic environment
is quite fast (in particular on Pentium class PCs, which are very common
nowadays, see Section 6.1O), and the interpreter can handle much more
complex programs than the ones presented in this book.
D.4 HRR-QB
HRR-QB allows the user to create fire files that are read by ASET-QB
and FIRM-QB. The f i e name extension (*.FIR) and file format are
identical to those of the fire files in FPETool3.2 [4 l]. Therefore, if the
user already has a database of FPETool fire fdes, there is no need to re-
enter the data to conduct simulations with ASET-QB or FIRM-QB.
The fire files consist of numerical data in three parallel columns,
followed by two lines of text. The first column is the time in seconds, the
second column is the heat release rate in kW at the corresponding time in
the first column, and the third column is the mass loss rate in g/s. The latter
is equal to the heat release rate divided by the heat of combustion, which
is supplied by the user. The mass loss rate is not used by ASET-QB and
FIRM-QB, but it is needed for compatibility with FPETool. The values in
the last row of the three columns are equal to -9, to designate the end of
the numerical data. The first line of text contains the name of the fde, and
the date it was created. The second line is a description of the fire file
entered by the user.
HRR-QB allows the user to create fire fdes for 3 fires as described in
Section 4.4.1, a semi-universal fire as described in Section 4.4.2, pool fires
as discussed in Section 4.4.3, and upholstered firrnitwre fires based on
Babrauskas' triangle model described in Section 4.4.4.1. The program
QBASIC Programmer's Notes 235
offers the user the option to enter a series of (t, 0 )data points on a curve,
which can also be done directly fkom within ASET-QB and FIRM-QB.
The main program code starts with a declaration section where
functions, subroutines, and global variables are declared; constants are
defied; and dimensions of m y s are specified. Next, an error trapping
routine is enabled to handle disk and file errors. This is followed by the
screen display of the main menu, which gives the user six choices:
1. create a ? fire fie, 2. create a semi-universal f i e file, 3. create a pool
fire me, 4. create a fiuniture f i e fie, 5. create a custom fire file, or 6. exit.
If the user selects the last option, the program terminates. If the user selects
any of the fist five choices, he is prompted to do the following:
The subroutine XYPlot is called to display a graph of the heat release rate
curve on the screen. XYPlot in turn calls the DrawAxes and LinePZot
subroutines. DrawAxes draws the x-axis and y-axis on the screen; with tick
marks,labels, and titles. The description of the f i e is printed at the top of
the graph. The Side Write subroutine writes the title of the y-axis vertically.
The range for each of the axes is determined by a call to the MaxValue
function, which finds the highest value in a one-dimensional array (also
called a vector). Lineplot finally draws a line on the screen that connects
all the data points. When the user is done viewing the graph, he is taken
back to the main menu.
236 APPENDIX D
D.5 ASET-QB
l ~ n a b l eerror trapping 1
e et input data (lnputData subroutine)/
6 Initialize variables
t
l solve ODES ( O ~ ~ S o l vsubroutine)
e /
Update screen/file/printer
+
Z E n d of s i m u l a t i o n ' ? ?
Close files
D.6 FIRM-QB
E.l INTRODUCTION
In this section the reader will be shown how to use HRR-VB to create
a fire file. Walton's ASET-B example case f r e (see Section 3-5) will be
used. To start HRR-VB, follow the instructions in Section C.3.3. A
welcome screen is displayed with the graphic image, the name of the
program (KRR-VB), the version number, and some copyright information.
This is commonly referred to as a splash screen (see Figure E-l). If the
user clicks on the Quit button, the program terminates. Clicking on the
Continue button (or pressing Enter) leads to the main data input screen.
Figure E-2 shows the input screen, with the description field already
completed. Because HRR-VB is not a very complex program, there is no
extensive help feature. Some assistance is provided in the form of
ToolTips, i.e., informative text boxes that pop up when the user points at
a control on the screen. Figure E-2 shows the ToolTip that appears when
the user points at the U.S. Engineering Units option button. The U.S.
engineering units for length, area, mass, temperature, and energy are
displayed in the text box.
When the user clicks on, or moves the cursor to the Fire File Name
Field, a dialog box pops up that allows the user to select a file name (see
Figure E-3). A separate dialog box minimizes the risk that an existing file
name is specified and overwritten by mistake, because the user can verify
whether the fde already exists and will be asked to confm overwriting any
existing file. Figure E-3 shows that the specified file, APPE-A0 1, does not
exist. This will not be the case if the user copied the data files fiom the
CD-ROM as specified in Section C.2.2. The extension .FIR is
automatically appended to the file name. Note that C:\FIRMU>ATA\is the
default directory for fire files. To go back to the main input data menu,
click on the Save button.
Since we will enter heat release rates in kW, there is no need to change
the units. The Heat of Combustion field can also be left unchanged,
because it is only needed to calculate mass loss rates, and does not affect
the fire data that are used by ASET-VB and FIRM-VB. The user can enter
a delay time in the text box in the lower left corner of the main data input
screen. This may be useful toaccount for delayed ignition of the fuel. For
example, for the furniture fire simulations discussed in Section 7.3S.1, an
ignition delay was added to synchronize the triangular heat release curves
244 APPENDIX E
based on Babrauskas' model with the experimental data. We are now ready
to enter the heat release rate curve, so click on the Custom Fire command
button (or press the Tab key several times until the Custom Fire button is
highlighted and press Enter, or press Alt + C). The program verifies that
the delay time is indeed zero, and returns to the main data input screen if
a delay time has to be specified. If no delay time is needed, the program
displays the custom fue parameter screen. Figure E-4 shows the screen
after the data points of Walton's ASET-B example case fue curve have
been entered. If there are more than 10 data points on the curve, the user
can click the Next 10 Points command button when the first screen is
completed. A second blank screen is then presented. This process can be
repeated until 1800 data pairs have been entered. In our case we only have
four data pairs, so we can click the OK button.
The next screen is a graph of the heat release curve (see Figure E-5).
If the user is satisfied with the curve, he can click the Save button to write
the data to disk. A File Save dialog box is displayed to give the user a fmal
opportunity to change the fde name, and avoid possible data loss by
overwriting an existing file. After the data are saved, the program returns
to the graph screen. The Save button is disabled to indicate that the data
were indeed saved. Clicking the Close button takes the user back to the
main input data screen. If the Close button is clicked before the Save
button, the user is asked to confirm that he does not want to save the data.
FIGURE E-4.HRR-VB custom fire parameter screen
245
246 APPENDIX E
FIGURE E-6.Main HRR-VB input data screen for acetone dip tank fire
The completed input data screen is shown in Figure E-8. Clicking Run
initiates the calculations. An output data screen is displayed, as shown in
Figure E-9. Only seven lines of results fit inside the window, and the
vertical scroll bar allows the user to go through and inspect the entire data
set. The results on this screen are identical to the ASET-QB data presented
in Table 3- 1, and those stored in CH3-A0 1.AOF.
FIGURE E-8. Completed ASET-VB input data screen
248
Visual Basic Programs 249
FIGURE E-10. FIRM-VB input data screen for acetone dip tank example
250 APPENDIX E
FIGURE E-11. FIRM-VB output data screen for acetone dip tank example
If the reader wants to modifi the Visual Basic source code, he will
need to load the source program (*.VBP extension) into the Visual Basic
programmitlg environment. As a minitnum, the reader will need a working
model of Visual Basic 6.0 to run the programs. There are several books on
the market that come with a working model of Visual Basic 6.0 (see
Section C.3.3). If the reader wants to compile his programs, he will need
more than a working model. Microsoft released different versions of
Visual Basic 6.0, to suit the needs of any programmer, ranging fiom the
novice to the professional.
All changes to the Visual Basic programs must be clearly documented
by adding clarifying comment statements to the source code, and by
providing an extensive discussion of the changes in any report or
publication of the model calculations. This discussion must be detailed
enough so that others can reproduce the results. If the changes are
extensive, the predictive capability of the modified model may have to be
re-evaluated (see Chapter 7).
REFERENCES
16. Klote, J. H., and J. W. Fothergill. 1983. "Design of Smoke Control Systems for
Buildings," NBS Handbook 141, National Bureau of Standards, Washington,
DC.
17. Butcher, E. G. and A. C. Parnell. 1979. Smoke Control in Fire Safety Design.
London: E. & F. N. Spon Ltd.
18. 1986. "Measured Air Leakage in Buildings," STP 904, ASTM, West
Conshohocken, PA.
19. Beyler, C. L. 1986. "Fire Plumes and Ceiling Jets," Fire Safety Journal,
11:53-75.
20. Forney, G. P. 1994. "Analyzing and Exploiting Numerical Characteristics of
Zone Fire Models," Fire Science & Technology, 14:49-60.
2 1. Zukoski, E. E., T. Kubota, and H. Cetegen. l98OA98 1. "Entrainment in Fire
Plumes," Fire Safety Journal, 3: 107- 121.
22. Morton, B. R., G. F. Taylor, and J. S. Turner. 1956. "Turbulent Gravitational
Convection from Maintained and Instantaneous Sources," in Proceedings of the
Royal Society, A234: 1- 23.
23. Yokoi, S. 1960. Building Research Report No. 34, Ministry of Construction,
Tokyo, Japan.
24. Kawagoe, K. 1958. "Fire Behavior in Rooms," Report No. 27, Building
Research Institute, Ministry of Construction, Tokyo, Japan.
25. Thornton, W. 1917. "The Relation of Oxygen to the Heat of Combustion of
Organic Compounds," Philosophical Magazine and J. of Science, 33.
26. Huggett, C. 1980. "Estimation of the Rate of Heat Release by Means of Oxygen
Consumption," J. of Fire and Flammability, l2:6 1- 65.
27. Siegel, R. and J. Howell. 1981. Xhermal Radiation Heat Transfer.New York,
NY: Hemisphere Publishing Company.
28. Modak, A. T. 1979. "Mation from Products of Combustion," Fire Research,
1:339-361.
29. Oziqk, M. N. 1985. Heat Transfer.A Basic Approach. New York, NY: McGraw
Hill Book Company.
30. Cooper, L. Y. 1993. "Fire-Plume-Generated Ceiling Jet Characteristics and
Convective Heat Transfer to Ceiling and Wall Surfaces in a Two-Layer Fire
Environment: Uniform Temperature Walls and Ceiling," Fire Science &
Technology, 13:1- 17.
31. Beller, D. K. 1987. "Alternate Computer Models of Fire Convection Phenomena
for the Harvard Computer Fire Code," M.S. Thesis, Worcester Polytechruc
Institute, Worcester, MA.
32. Satterfield, D; B. and J. R. Barnett. 1990. "User's Guide for WPUFire Version
2 Compartment Fire Model," Worcester Polytechnic Institute, Worcester, MA.
33. Sarnarskii, A. A., and P. N. Vabishchevich. 1995. Computational Heat Transfer
(2 Volumes).Chichester, England: John Wiley & Sons Ltd.
34. Petzold, L. R. 1982. "A Description of DASSL: A DifferentiaVAlgebraic System
Solver," Technical Report 8637, Sandia National Laboratories, Albuquerque,
NM.
References 253
35. Cooper, L. Y., and D. W. Stroup. 1982. "Calculating Safe Egress Time (ASET)
- A Computer Program and User's Guide," NBSIR 82-2578, National Bureau
of Standards, Washington, DC.
36. Cooper, L. Y. 1983. "A Concept for Estimating Available Safe Egress Time in
Fires," Fire Safely Journal, 5:135- 144.
37. Cooper, L. Y. 1982. "A Mathematical Model for Estimating Available Sale
Egress Time in Fires," Fire and Materials, 6:135- 144.
38. Cooper, L. Y. and D. W. Stroup. 1985. "ASET - A Computer Program for
Calculating Safe Egress Time," Fire Safety Journal, 9:29- 45.
39. DiNenno, P. J. 1985. "ASET-B: A Room Fire Program for Personal Computers,"
Software Review, Fire Technology, 2 1:322- 323.
40. Walton, W. D. 1985. 'ASET-B: A Room Fire Program for Personal Computers,"
Fire Technology, 2 11293-309.
41. Deal, S. 1995. "Technical Reference Guide for FPETool Version 3.2,"NISTIR
5486-1, National Institute of Standards and Technology, Gaithersburg, MD.
42. Cooper, L. Y. 1983. "The Development of Hazardous Conditions in Enclosures
with Growing Fires," Combustion, Science and Technology, 33:279-297.
43. Boyce, W. E. and R. C. DiPrima. 1977. Elementary D~flerentialEquations and
Boundary Value Problems. New York, NY: John Wiley & Sons.
44. Press, W. H., S. A. Teukolsky, W. T. Vetterling, and B. P. Flannery. 1992.
Numerical Recipes in Fortran - The Art of Scientrfic Computing. Second
Edition. Cambridge, England: Cambridge Universty Press.
45. 1997. "Standard Guide for Use and Limitations of Deterministic Fire Models,"
ASTM, West Conshohocken, PA.
46. Birk, D. Personal Communication with G. Heskestad.
47. Milke, J. A. 1989. "Smoke Management Design Considerations for Covered
Malls and Atria," International Fire Protection Engineering Institute, Ottawa
Ontario, Canada.
48. Babrauskas, V. 1983. "Upholstered Furniture Heat Release Rates:
Measurements and Estimations," Journal of Fire Sciences, 1 :9-32.
49. Steckler, K. D., H. R. Baurn and J. G. Quintiere. 1984. "Fire Induced Flows
through Room Openings - Flow Coefficients," NBSIR 83-2801. National
Bureau of Standards, Washington, DC.
50. Nakaya, I., T. Tanaka, M. Yoshida, and K. Steckler. 1986. "Doorway Flow
Induced by a Propane Burner," Fire Safety Journal, 10:185- 195.
5 1. Peacock, R. D., S. Davis, and B. T. Lee. 1988. "An Experimental Data Set for
the Accuracy Assessment of Room Fire Models," NBSIR 88-3752. National
Bureau of Standards, Washington, DC.
52. Lawson, J. R. and J. G. Quintiere. 1985/1986. "Slide Rule Estimates of Fire
Growth," Fire Technology, 2 1 :267-292, and 22:45- 53.
53. Nelson, H. E. 1986. "FIREFORM- A Computerized Collection of Convenient
Fire Saftety Computations," NBSIR 86-3388, National Bureau of Standards,
Washington, DC.
54. Rockett, J. A. 1976. "Fire Induced Flow in an Enclosure." Combustion Science
and Technology, 12:l65 - 175.
254 References
55. Mitler, H. E. 1978. "The Physical Basis for the Harvard Computer Fire Code,"
Home Fire Project Technical Report No. 34, Harvard University, Cambridge,
MA.
56. Babrauskas, V. 1984. "Upholstered Furniture Room Fire Measurements.
Comparison with Furniture Calorimeter Data and Flashover Predictions,"
Journal of Fire Sciences, 2:2- 19.
57. Drysdale, D. 1985. An Introduction to Fire Dynamics. New York, NY: John
Wiley & Sons.
58. 1996. "National Fire Alarm Code," NFPA 72, National Fire Protection
Association, Quincy, MA.
59. 1998. "Guide for Smoke and Heat Venting," NFPA 204, National Fire
Protection Association, Quincy, MA.
60. Babrauskas, V. 1996. "Fire Modeling Tools for FSE: Are They Good Enough?"
Journal of Fire Protection Engineering, 8:97-96.
61. Schifiliti, R. P. 1986. "Alternate Computer Models of Fire Convection
Phenomena for the Harvard Computer Fire Code," M.S. Thesis, Worcester
Polytechnic Institute, Worcester, MA.
62. Babrauskas, V. 1986. "Free Burning Fires," Fire Safety Journal, 11.
63. Babrauskas, V. 1983. "Estimating Large Pool Fire Burning Rates," Fire
Technology, 19:251- 26 1.
64. Babrauskas, V., J. R. Lawson, W. D. Walton, and W. H. Twilley. 1982.
"Upholstered Furniture Burning Rates as Measured with a Furniture
Calorimeter," NBSIR 82-2604, National Bureau of Standards, Washington, DC.
65. Parker, W. J. 1982. "Calculations of the Heat Release Rate by Oxygen
Consumption for Various Applications," NBSIR 8 1-2427, National Bureau of
Standards, Washington, DC.
66. Janssens, M. L. 1991. "Measuring Rate of Heat Release by Oxygen
Consumption," Fire Technology, 27:234-249.
67. 1996. "Standard Test Method for Fire Testing of Real Scale Upholstered
Furniture Items," ASTM E 1537, ASTM, West Conshohocken, PA.
68. 1987. "Upholstered Furniture Fire Behavior - Full-scale Test," NORDTEST
NT Fire 032, NORDTEST, Helsinki, Finland.
69. Krasny, J. F., V. Babrauskas, and W. J. Parker. 1998. "Fire Behavior of
Upholstered Furniture Items," in press.
70. 1997. "Standard Test Method for Heat and Visible Smoke Release Rates for
Materials and Products Using an Oxygen Consumption Calorimeter," ASTM E
1354, ASTM, West Conshohocken, PA.
71. Babrauskas, V. and W. D. Walton. 1986. "A Simplified Characterization of
Upholstered Furniture Heat Release Rates," Fire Safety Journal, 11:18 1- 192.
72. Babrauskas, V. and J. Krasny. 1985. "Fire Behavior of Upholstered Furniture,"
NBS Monograph 173, National Bureau of Standards, Washington, DC.
73. Peacock, R. B. and J. N. Breese. 1982. "Computer Fire Modeling for the
Prediction of Flashover," NBSIR 82-2516, National Bureau of Standards,
Washington, DC.
74. 1992. "Standard Guide for Documenting Computer Software for Fire Models,"
ASTM E 1472, ASTM, West Conshohocken, PA.
References 255
75. 1997. "Standard Guide for Evaluating the Predictive Capability of Deterministic
Fire Models," ASTM E 1355, ASTM, West Conshohocken, PA.
76. Siirdqvist, S. 1993. "Initial Fires: HRR, Smoke and CO Generation from Single
Items and Room Fire Tests," LUTVDGtTVBB-3070-SE, Lund University,
Department of Fire Technology, Lund, Sweden.
77. Bell, J. R., T. J. Klem, and A. W Willey. 1982. "Investigation Report -
Westchase Hilton Hotel Fire," NFPA No. LS-7, National Fire Protection
Association, Quincy, MA.
78. Mitler, H. E. and H. W. Ernrnons. l98 1. "Documentation for CFC V, The Fifth
Harvard Computer Fire Code," NBS-GCR-81-244, National Bureau of
Standards, Washington, DC.
79. 1997. "Code for Safety to Life from Fire in Buildings and Structures," NFPA
101, National Fire Protection Association, Quincy, MA.
80. Babrauskas, V. 198111982. "Will the Second Item Ignite?" Fire Safety Journal,
4:281-292.
81. Khoudja, N. 1988. "Procedures for Quantitative Sensitivity and Performance
Validation Studies of a Deterministic Fire Safety Model," NBS-GCR-88-544,
National Bureau of Standards, Washington, DC.
82. Davies, A. D. 1985. "Applied Model Validation," NBSIR 85-3154-1, National
Bureau of Standards, Washington, DC.
83. Bukowski, R. W. 1985. "The Application of Models to the Assessment of Fire
Hazards from Consumer Products," NBSIR 85-3219, National Bureau of
Standards, Washington, DC.
84. Watts, J. M., Jr. 1987. "Validating Fire Models," Fire Technology, 23:93-94.
85. 1999. "Standard Terminology of Fire Standards," ASTM E 176, ASTM, West
Conshohocken, PA.
86. Keski-Rahkonenen, 0 . 1996. "CIB W l 4 Round Robin of Code Assessment: A
Comparison of Fire Simulation Tools," in Proceedings of the Conference on
Fire Safety Design of Buildings and Fire Safety Engineering, Oslo, Norway,
August 19-20,l- 14.
87. 1999. "Standard Guide for Room Fire Experiments," ASTM E 603, ASTM,
West Conshohocken, PA.
88. Peacock, R. B. and J. N. Breese. 1982. "Computer Fire Modeling for the
Prediction of Flashover," NBSIR 82-25 16, National Bureau of Standards,
Washington, DC.
89. Iman, R. L. and Shortencarier. 1984. "A FORTRAN 77 Program and User's
Guide for the Generation of Latin Hypercube and Random Samples for Use with
Computer Models," NUREGICR-3624, SAND83-2365, Sandia National
Laboratories, Albuquerque, NM.
90. 1997. "Standard Practice for Conducting an Interlaboratory Study to Determine
the Precision of a Test Method," ASTM E691, ASTM Standards on Precision
and Biasfor Various Applications, 5" Edition, ASTM, West Conshohocken, PA.
91. 1986. "Precision of Test Methods - Determination of Repeatability and
Reproducibility for a Standard Test Method by Inter-Laboratory Tests," IS0
5725, International Origanization for Standardization,Geneva, Switzerland.
256 References