6-1

6 Fire dynamics

This publication is supplied by CIBSE for the sole use of the person making the download. The content remains the copyright property of CIBSE
6.1 Introduction in the construction/building environment, it has been
found by experiment that the critical radiant heat flux for
ignition where there is already a flame present (i.e. pilot
Fire dynamics describes the complex subject of fire behav- ignition) is in the range 10–30 kW · m–2. For spontaneous
iour and encompasses chemistry, physics, heat transfer ignition, where there is no flame present, critical heat
and fluid dynamics. With knowledge of fire dynamics, a fluxes are higher, at about 40 kW · m–2. In both cases the
more fundamental approach to fire safety engineering can actual values depend on the fuel.
be applied at the design stage. It can also be used in
response to an incident that has highlighted a fire hazard
with a view to investigation and research.

Fire is a chemical reaction between combustible species

6.3 Fire growth
and oxygen from the air, which produces heat, the mode
of burning depending more upon the physical state and For sustained combustion to occur, oxygen, heat and a fuel
distribution of the fuel and its environment than on the source must all be present. The removal of any one of
chemistry. An example often quoted is that a wooden log these will terminate the reaction. The burning process in
is difficult to ignite but thin sticks can be ignited easily fires involves pyrolysis (i.e. thermal decomposition) of
and will burn fiercely when piled together. fresh fuel. This pyrolysis will produce volatiles from the
surface of the fuel and these gases will oxidise in the flam-
This section aims to present a basic understanding of the ing region, generating combustion products and releasing
processes which govern fire and smoke development and heat. If there are no control measures present, and both air
to guide the reader in the available techniques for calcu- and fuel are available, it must be assumed that the fire will
lating the important parameters. It is not exhaustive and continue to grow in a manner that may be predictable,
much use will be made of references to more detailed based on experimental or other evidence. However, the
publications, which should be consulted for further infor- calculation of flame spread or fire growth rates from first
mation. Important references include Drysdale’s An principles is not easy. Characteristic fire growth rates are
Introduction to Fire Dynamics (2011) and Karlson and Quin- given in section 6.5.3.1.
tiere’s Enclosure Fire Dynamics (2000).

It should be noted that most fire safety engineering calcu-

lations are based upon experiment and testing. Therefore, 6.4 Compartment fires
the validity of such calculations will be limited and extrap-
olation beyond these limits may not be appropriate. It may
be prudent to carry out further testing or modelling to 6.4.1 General
validate the design parameters used, if considered neces-
sary by designers or approvers. This can be in the form of A distinction may be made between fires arising in the
physical testing or computational simulation. open, where radiated heat is lost to the surroundings, and
fires which occur in confined spaces or compartments. In
the latter, heat is transferred to the compartment walls by
radiation from the fire and also by convection from the
6.2 Ignition hot gases that accumulate within the compartment. Re-
radiation from these hot boundaries can significantly
Ignition is the process whereby a material passes from a increase the heating of combustibles in the room.
relatively inert state to one where a reaction takes place that
can produce temperatures significantly in excess of ambi- If there are openings to the compartment to permit the
ent. Ignition of most materials requires the application of inflow of air, and if there is sufficient fuel, the fire will
an external source of heat, the incident heat flux causing continue to grow and the temperature of the hot gas layer
the surface temperature of the fuel to rise. In the case of at ceiling level will rise. Ultimately, the point may be
flammable liquids, this liberates vapour; solid materials reached where the downward radiation from this layer is
decompose to release flammable volatiles. Combustion so intense that all of the remaining fuel in the compart-
takes place in the gas phase above the fuel surface. ment becomes involved. This occurs at layer temperatures
of 500–600 °C (see section 6.8.4). The transition from
Whether or not ignition occurs, and whether the reaction growing to fully developed fire happens very rapidly, and
then becomes self-propagating, depends on a complex heat the event is often referred to as ‘flashover’.
balance between the incident heat flux, the convective and
radiative heat gains by the fuel, and the heat losses to the Following flashover, the rate of heat release of the fire
surroundings. For the types of materials commonly found increases rapidly and the oxygen content decreases.
6-2 Fire safety engineering

Flashover is unlikely to occur in large or tall compart-

ments with small fire loads, such as airport concourses,
multistorey malls and atria protected from fire in an adja-
cent enclosure. There is the potential for fire spread

This publication is supplied by CIBSE for the sole use of the person making the download. The content remains the copyright property of CIBSE
beyond the initial fire source by radiative heat transfer,
and therefore the risk of fire spread within large or tall
compartments cannot be discounted. Therefore, the siting
Temperature

of combustibles in such spaces should be considered as

Flashover part of the design process, and further guidance on this
issue is given in Annex B of BS 9999: 2017 (BSI, 2017).
However, given sufficiently high fire loads, such as in high
bay warehouses, fire development may reach flashover
conditions.

Fuel bed-controlled fires have excess air available and

Initiation Growth Steady-state Decay their combustion rate, heat output and growth are limited
by the fuel being burnt. All the burning takes place within
the fire compartment.
Time
Figure 6.1 Stages of development of a fire

Anyone remaining in a compartment which has under- 6.5 Calculation of fire

gone flashover is unlikely to survive. The risk of fire
spread from the compartment to adjacent areas increases
parameters
greatly and the structure becomes heated.

Because radiation from the smoke layer is the driving 6.5.1 General
force in initiating flashover, any factors that promote loss
of heat from the layer will tend to reduce the risk of its The expressions given in the following sections have
occurrence. In particular, in compartments that are high previously been published in the technical literature of
or wide and where there is limited material to burn, the the fire safety industry. They are the result of experiment
smoke will be unlikely to reach temperatures that would and observation and therefore each has its limitations.
result in flashover. Flashover is unlikely to occur where
sprinklers are operating.
6.5.2 Design fires
A useful way of showing the development of a compart-
ment fire is illustrated in Figure 6.1. The stages are: The design fire is characterised by the variation of heat
output with time. In the initial stages of fire growth it is
— Initiation: the fire will grow only slowly as a result assumed that the fire is well ventilated, its rate of burning
of flame spread over the item first ignited. being characterised by the type, amount and configuration
of the fuel. The fire is assumed to be confined initially to
— Growth: the fire will grow more quickly and begin to a single object or group of objects.
spread to other items, but remain effectively local.
— Fully developed steady-state or post-flashover: all the If unchecked, the fire may spread to adjacent objects and,
combustibles are involved and flames appear to fill once flames reach the ceiling, flashover may occur and the
the entire volume; the average temperature is very whole room or compartment becomes involved in a fully
high. developed fire. After flashover, the rate of smoke production
can be so great that smoke control becomes impracticable.
— Decay: at this stage, the average temperature of the However, if there is a post-flashover fire in a small room, it
fire has fallen considerably from its peak value. may be possible to design a smoke control system that
protects an adjacent large-volume space, such as an atrium,
6.4.2 Limiting fire development when smoke emerges from a window or doorway of the
room. Types of smoke control system and their practical
Once flashover has occurred, the development of the fire application are considered in chapter 10: Smoke ventilation.
in a compartment will be limited by the in-flow rate of air
(i.e. ventilation-controlled fires) or combustible material The parameter that governs most strongly the way in
(i.e. fuel bed-controlled fires), or by firefighting. which a fire and its products behave is its rate of heat
release, commonly termed ‘fire size’. In order to carry out
Ventilation-controlled fires have their combustion and a fire engineering design, it is essential to define at the
heat output controlled by the amount of air reaching the outset a series of design fires that represent the worst fire
fire, which is governed by the openings to the fire compart- situations likely to arise in the building under considera-
ment. A ventilation-controlled fire usually means that the tion. Information is available, both experimental and
whole compartment is involved and flashover has occurred. theoretical, that may be used by the designer in selecting
Flames may project from the openings of the compart- suitable design fires.
ment, and significant combustion of heated fuel gases may
take place outside, where they first come into contact with Pre-flashover fires are considered in section 6.5.3;
sufficient oxygen. post-flashover fires are dealt with in section 6.5.4.
Fire dynamics 6-3

6.5.3 Pre-flashover fires 2016). Many of the measurements relate to heat release
rates from goods such as those which might be stored in
The design fire size will depend on the characteristics of warehouses. There is also a significant body of data on
the type and arrangement of the fuel and may be catego- foam-filled furniture (Babrauskas, 1986).

This publication is supplied by CIBSE for the sole use of the person making the download. The content remains the copyright property of CIBSE
rised for design purposes as one of the following:
As a result of measurements, it has been found possible to
— a growing fire characterise fire growth rates in different ways:
— a fire having a fixed size for a finite time — t-squared fires (UK and USA)
— a steady-state fire. — t-cubed fires
Fixed size or steady-state fires will usually have grown to — standard fires, types 1, 2 and 3 (Japan)
some limit, further extension being restricted by one or — growing fires (Australia).
more of the following:

— fire control activities, such as automatic (or man- This Guide concentrates on the method of determining
ual) fire suppression fire growth rates used in the UK and USA. Fire growth
rates for various types of fire have been compared by
— sufficient space separation to neighbouring combust- Bukowski (1993) (see Figure 6.2).
ibles
— for hydrocarbon pool fires, the leakage versus 6.5.3.1 Fire growth rate
burning rate or, if bunded, the extent of the bund.
Much experimental work has been carried out in the USA
A fixed design fire size applicable to all situations is not on heat release rates in fires as a function of time. Some
feasible, especially when designing for means of escape or of the results are summarised in NFPA 92 (NFPA, 2015).
estimating the activation time of automatic detectors. Additional data on real fires are available from the National
Institute for Standards and Technology (NIST) (www.nist.
It is more realistic to design based on a growing fire, using gov) and BRE (www.bre.co.uk).
the widely accepted t-squared growth rates, and a maxi-
mum heat release rate. It is not possible to predict the These large-scale tests show fire growth and decay for a
length of the incubation period (see Figure 6.3), and there- series of objects and groups of objects. These data show
fore it is recommended that this period is ignored in this that fire curves are closer to spikes, with rapid growth and
approach. This provides inherent conservatism to the rapid decay. The fact that heat release rate peaks may be
design calculation. very high but last for a limited time should be taken into
account when designing fire systems and allowing for
A great deal of experimental work has been carried out on appropriate safety factors.
rates of heat release from different materials when burned
in fire tests. Much of this information is summarised in In many instances, building fires go through an initial
the SFPE Handbook of Fire Protection Engineering (SFPE, incubation period, when the growth rate is significantly

30 US slow

+ US medium
25 z z z z z z
> US fast
+
z
+ US ultra-fast
20
Heat release rate / MW

+
z Australia <1140 MJ·m–2

+
15 z Australia 1140 – 2280 MJ·m–2

+
Australia 1140 – 4560 MJ·m–2
z
10 + >
> Japan (t2)
z + >
> Japan (No. 1)
+ >
5 >
z
+ >
> z Japan (No. 2)
+ >
> >
+ >
+
z z
> z
>
0+
z
> z
+
> >
0 50 100 150 200 250 300 350 400 450
Time / s

Figure 6.2 Bukowski’s comparison of idealised fire growth curves (Bukowski, 1993)
6-4 Fire safety engineering

slower than the t-squared rates, such as the initial period Table 6.1 Characteristic growth time for various classes of fire
in small or smouldering fires. This period (see Figure 6.3) Fire class Characteristic growth time, tg / s Constant a / kW · s–2
is of indeterminate length and is ignored for design
purposes, although the fire may be detected during this Ultra-fast 75 0.1876

This publication is supplied by CIBSE for the sole use of the person making the download. The content remains the copyright property of CIBSE
period by the occupants or by an automatic detector, if Fast 150 0.0469
adjacent to the source.
Medium 300 0.0117
It has been found that, after this incubation period, the Slow 600 0.0029
heat release rate grows approximately as the square of the
time (NFPA, 2015), i.e.

Qt = at2 (6.1)
Table 6.2 Growth rates for growing fires

where Qt is the total heat release rate of the fire (kW), a is Building area providing fuel Growth rate
a constant (kW · s–2) and t is time (s).
Dwelling Medium

Figure 6.3 illustrates t-squared fire growth. The growth Office Medium
parameter for a t-squared fire is defined by the time taken Shop Fast
for the heat output to reach 1055 kW (i.e. approximately
Warehouse Ultra-fast*
1 MW). This is known as the characteristic growth time.
It has been suggested that fires may be conveniently clas- Hotel bedroom Medium
sified as ‘slow’, ‘medium’, ‘fast’ and ‘ultra-fast’, depending Hotel reception Medium
on the characteristic growth time.
Assembly hall seating Medium–fast
Table 6.1 gives the characteristic growth time, tg, and the Picture gallery Slow
corresponding values of constant a for the various classes of Display area Slow–medium
fire. The fastest burning upholstered sofas and plastic goods
stacked to a height of about 4.5 m give ‘ultra-fast’ growth * Depends on fire load
rates, while other upholstered furniture and lower piles of
plastic goods give ‘fast’ rates. Tightly rolled paper produces
a ‘slow’ growth rate. Experiments on burning computer
workstations suggest ‘medium’ to ‘fast’ growth rates.

Fire growth depends on the type of fuel and its arrange-

ment, but some growth rates are suggested in Table 6.2,
based on the experimental evidence available.

3 Figure 6.3 Illustration of

t-squared fire growth. (Reprinted
with permission from NFPA
92-2018 Standard for smoke control
systems, Copyright © 2017,
Continuously
National Fire Protection, Quincy,
growing fire
MA. This reprinted material is
not the complete and official
position of the NFPA on the
2
Heat release rate / MW

referenced subject, which is

represented only by the standard
in its entirety.) (Original in
Imperial units.)

Incubation period

0
Time Growth time

Effective ignition time

Fire dynamics 6-5

Research suggests that for a high rack warehouse (fire in However, it should be noted that the validity of this
the flue) the growth rate can be modelled as a ‘t-cubed’ assumption is the subject of some considerable debate in
fire, given by (Ingason, 1993) as fire engineering circles and may be potentially open to
some criticism. From the agreed value, it is assumed that

This publication is supplied by CIBSE for the sole use of the person making the download. The content remains the copyright property of CIBSE
Qt = 0.045 t3 (6.2) no further items of fuel ignite, and the value of mass flow
in the plume is calculated accordingly. After operation, it
For this rapid fire growth, the incipient stage is significant may be assumed that the sprinklers cool most of the smoke
and the curve is valid up to 10 MW for a 10 m high rack layer to a temperature less than the operating temperature
(there are no data for fires greater than 10 MW). For a fire of the sprinklers. For calculation purposes, an average
in the racking flue, the amount of entrainment of fresh air smoke layer temperature of 100 °C may be assumed with
in the rack plume is restricted, compared to that for a fire conventional sprinkler heads, while the sprinklers are
on the face of the rack. For a typical cellulosic fire, racking operating.
flue entrainment can be estimated as

mflue = Q1.08 # 10-4V t3

6.5.3.5 Transient fires
(6.3)

where mflue is the mass of smoke produced prior to sprin- To simplify calculations of smoke filling during the tran-
kler operation for a fire in a racking flue (kg · s–1) and t is sient phase (see section 6.8), an average value of Q may be
time (s) (CIBSE, 2010: appendix 6.A1). used:

Equations 6.2 and 6.3 are to be used together. #0 t Qt dt

Qt, ave = (6.5)
t
6.5.3.2 Unit heat release rate
where Qt, ave is the average total heat output of the fire
(kW) and t is time (s).
Estimates of heat release rates per unit floor area or per
unit fuel area for various commodities and materials can For t-squared fires:
be gained from the SFPE Handbook (SFPE, 2016) and
NFPA 92 (NFPA, 2015). Survey data from actual occupan- Qave = 333 Qt tgV2 (6.6)
cies in use have also been published (BSI, 2002a).

Measured survey loads, q, are given in MJ per square where tg is the characteristic growth time (s). Values of tg
metre of floor area. By assuming a conservative burn-out are given in Table 6.1.
time of 20 minutes (i.e. 1200 s), the unit heat release rate
can be estimated for well-ventilated compartment fires: 6.5.4 Post-flashover fires
Qu = q 1200 (6.4)
6.5.4.1 Condition for flashover
where QU is the unit heat release rate (kW·m–2) and q is the
measured survey load (kJ·m–2). Note that the measured For design purposes, it may be assumed that flashover
survey load is usually given in MJ·m–2 and must be converted does not occur if the smoke layer at ceiling level is at a
to kJ·m–2 for use with equation 6.4. Some commonly used temperature of less than 600 °C (McCaffrey et al., 1981). A
values of heat release rate are shown in Table 6.3. method for calculating this temperature is given in section
6.7.5. Note that in the plume above a fire the temperature
6.5.3.3 Steady-state fires: not sprinklered at the tip of intermittent flames is about 350 °C and at the
tip of sustained flames is about 550 °C.
Once the fire has spread from item to item until all the
If a correctly designed and maintained sprinkler system
available fuel is burning, the heat output will reach a
(or other approved fire suppression system designed to
steady value, eventually declining as the fuel decays. The
achieve fire control) operates, it may be assumed that flash-
estimation of the steady value is given in section 6.5.4.
over will not occur, since sprinklers are designed to operate
while the smoke layer is at a temperature much lower than
6.5.3.4 Steady-state fires: sprinklered the generally accepted flashover temperature of 600 °C.

For design purposes, the value of Q may be assumed as Because flashover is such a serious event, a great deal of
steady after operation of the first sprinkler (see section research effort has been invested in methods to predict the
6.6.3 regarding sprinkler response and activation). conditions which give rise to it. A presentation and
comparison of the different correlations which are availa-
ble are given in the SFPE Handbook (SFPE, 2016). The
Table 6.3 Commonly used values of heat release rate (NFPA, 2015) simplest of these relates the heat release rate required for
Occupancy Unit heat release rate, QU / flashover, Qf (kW), to — what has become known as the
kW · m–2 ventilation factor (Avo √ ho), such that
Offices 290
Qf = 600 Avo ho (6.7)
Shops 550
Industrial 260 where Qf is the heat release rate required for flashover
(kW), Avo is the area of the opening to the compartment
Hotel rooms 249
(m2) and ho is the height of the opening (m).
6-6 Fire safety engineering

The total heat release rate is given by 6.6 Effect of sprinklers

Qt = Hc # R (6.8)
The following paragraphs deal with the effect of sprinklers

This publication is supplied by CIBSE for the sole use of the person making the download. The content remains the copyright property of CIBSE
where Hc is the heat of combustion (kJ · kg–1) and R is the on fire growth. Sprinkler design is considered in detail in
mass rate of burning (kg · s–1). (The heat of combustion, chapter 11: Fire suppression.
Hc, is discussed in chapter 10: Smoke ventilation, and
values for various materials are given in PD 7974-1: 2003
(BSI, 2003a) or the SFPE Handbook (SFPE, 2016). The 6.6.1 General principles
rate of burning, R, is considered in sections 6.5.4.2 and
6.5.4.3.) The plume of hot smoky gases from a fire rises as a result
of its buoyancy. When it hits the ceiling the plume turns
6.5.4.2 Ventilation-controlled fires and spreads laterally, where it may interact with sprin-
klers, eventually causing them to operate. The time to
sprinkler operation depends on:
The mass rate of burning is given by

R = 0.02 !Ao h1o/2 QAt - AoVQw / dV$1/2 (6.9) — fire growth rate

— sprinkler location
where R is the mass rate of burning (kg · s–1), Ao is the area
of ventilation opening (door, window etc.) (m2), ho is the — sprinkler sensitivity.
height of ventilation opening (m), At is the area of room
surface (wall, floor, ceiling) (m2), w is the width of wall
containing the opening (m) and d is the depth of room 6.6.2 Sprinkler location
behind the opening (m) (Thomas et al., 1963).
As the smoke plume rises from a fire, it draws in air from
Effective values for these parameters for rooms with more the surroundings, which causes it to cool. Therefore, the
than one opening are beyond the scope of this document higher the ceiling, the lower will be the temperature of the
but can be derived using the procedures given in the smoke that reaches the sprinklers. Additional cooling then
Appendix to this chapter. occurs as the smoke spreads laterally. Clearly, the hotter
the smoke, the more rapidly the sprinkler will operate.
Equation 6.9 has been derived from experiments with
wood cribs and can be used for most types of fire load
found in houses, offices and shops. Conventionally, fire 6.6.3 Sprinkler and smoke detector
load may be expressed in terms of the equivalent weight sensitivity
of wood. If expressed in MJ or MJ · m–2, the fire load may
be converted to kg or kg · m–2 of wood by dividing by
18 MJ · kg–1; for example, 360 MJ · m–2 is equivalent to In order to operate and release water onto a fire, a sprin-
20 kg · m–2 of wood. kler must be heated to its operating temperature, usually
about 70 °C, at which point a temperature-sensitive
element is designed to fail, e.g. a solder link melts or a
6.5.4.3 Fuel bed-controlled fires glass bulb breaks. The rate at which the element heats up
when exposed to hot smoke depends on its shape and
With low values of fire load, equation 6.9 overestimates mass. A heavy, short bulb will take longer to reach a given
the mass rate of burning by a factor of 2 or 3. temperature than a light, slim bulb. The parameter used
to describe sprinkler sensitivity is known as the response
An effective fire duration of 20 minutes may be assumed, time index (RTI), see chapter 11: Fire suppression. Sprin-
with R then being given by klers with RTI values below 50 m1/2 · s1/2 are described as
having a quick response, while those with values up to
R = L 1200 (6.10) 350 m1/2 · s1/2 are regarded as having a standard response
(BSI, 1999).
where R is the mass rate of burning (kg · s–1) and L is the
total fire load (kg), or Concealed sprinkler heads are not designated a thermal
L = Q L Af V Af
sensitivity response rating by manufacturers due to the
(6.11) nature of the sprinkler assembly. However, work by Anna-
ble (2006) determined RTI values for the overall assembly
where Af is the floor area (m2) (Law, 1978). arrangement for a limited number of concealed sprinkler
head types. This work demonstrated that for a concealed
Values of L / Af (kg · m–2) are derived from surveys or design sprinkler with a temperature-sensing element with a quick
data. Where such data are expressed in MJ · m–2, they may response, the RTI of the overall arrangement was not quick
be converted by dividing by 18 (see section 6.5.4.2). response, but was within the expected range of a standard
response head. Common practice assumed in design is that
For design purposes, R should be calculated from equa- the RTI of the concealed sprinkler head temperature-
tions 6.9 and 6.10 and the lower value adopted. sensing element should be quick response, but with the
overall assembly arrangement assumed to be a standard
If the heat release rate is needed for the approving author- response head, and the value of RTI used for any concealed
ity, this can easily be calculated by multiplying R by the sprinkler activation calculation should be discussed and
heat of combustion of the fuel. agreed with the approving authorities.
Fire dynamics 6-7

Note that a smoke detector can be considered as an equiv- to grow but at a slower rate. Whether or not such an
alent heat detector having an RTI of 0.5 m1/2 · s1/2 and a assumption is made, the fire may continue to burn until
fixed temperature rise of 13 °C (Evans and Stroup, 1985). all the fuel is consumed.

This publication is supplied by CIBSE for the sole use of the person making the download. The content remains the copyright property of CIBSE
It is possible to use computer software to assess the effect
of fire growth, sprinkler location and sprinkler sensitivity
on sprinkler activation as an alternative to hand calcula- 6.7 Smoke plumes
tion. One such computer zone model is B-RISK (Wade et
al., 2013), which can be used to predict sprinkler activa-
tion times and corresponding fire sizes at sprinkler 6.7.1 Introduction to smoke
activation. B-RISK utilises the differential equation
describing the temperature of the sprinkler sensing ventilation calculations
element based on the RTI concept of Heskestad and Bill
(1988), incorporating the gas temperature and velocity in For most smoke ventilation calculations (see chapter 10:
the vicinity of the sprinkler sensing element, using the Smoke ventilation), it is essential to know:
unconfined ceiling jet correlation by Alpert in chapter 14
of the SFPE Handbook (SFPE, 2016). (a) the plume type; for example, an axisymmetric
plume (see section 6.7.3) or a plume flow from an
opening (see section 6.7.4) — the entrainment
6.6.4 Effect on fire size equations associated with different plume types cal-
culate the mass flow rate of smoke produced (msmoke)
Real fire tests are rarely performed. However, should a
series of tests be carried out on the intended typical layout (b) the temperature of the hot gases (see section 6.7.5)
(i.e. room dimensions, fuel type etc.) and if these tests show and
that a fire will be quickly suppressed with the installed (c) the volume flow rate of smoke (see section 6.7.6).
sprinkler system, then it seems reasonable to assume that
combustion effectively ceases when the sprinklers operate. By using this methodology, the effectiveness or perfor-
mance of a smoke ventilation system can be determined.
In a room equipped with sprinklers, a fire may grow until
the heat in the plume sets off the first sprinkler heads. The Section 6.7.7 provides information on the flowing layer
effect of the sprinklers on the design fire size can be taken depth, which can be used in conjunction with the guid-
into account by assuming that the fire stops growing when ance given in chapter 10: Smoke ventilation to calculate the
the sprinklers are activated. The design fire is then esti- depth of channelling screens etc.
mated as the size the fire has grown to at the moment of
sprinkler actuation unless there is reason to suspect that The proportion of the total heat release rate in the plume
the fire will continue to spread after the sprinklers have varies with the type of combustible material and the char-
been actuated. Since the sprinklers will cool most of the acteristics of the compartment (for flow out of an opening).
smoke layer to below 100 °C, flashover is not likely to For the purposes of design, the convective portion (Qp)
occur where they are installed. It can then be assumed can be assumed to be 66% of the total heat output of the
conservatively that the fire will have a constant rate of fire (Qt) (SFPE, 2016).
heat release (see Figure 6.4).

Alternatively, it could be assumed that, after sprinkler 6.7.2 Entrainment

activation, the heat output will slowly decrease. Experi-
ments in small compartments have suggested that fire It is assumed that the volume of smoke produced is equal
heat release rates will fall by 50% over a period of a few to the amount of air entrained into the plume. The concen-
minutes (Madrzykowski and Vettori, 1991). In some tration of smoke particles and toxic products depends on
circumstances, it may be assumed that the fire continues the type of fuel and ventilation rate.

Figure 6.4 Typical fire model

Steady-state fire

t2 growth period Decay period

Fire growth

Design fire

Natural fire

It may be assumed that decay occurs

after sprinkler activation or after a period
of steady-state burning; conservatively,
it may be assumed that there is no decay
Time
6-8 Fire safety engineering

At a given height, entrainment depends on the heat output where msmoke is the mass flow of smoke in the plume
and, at small plume heights, on the geometry of the source. (kg · s–1), p is the perimeter of the source (m) and z is the
At large plume heights, entrainment is equivalent to that height of the plume above the fuel surface (m) (Thomas et
above a point source. The plume itself may be in the room al., 1963).

This publication is supplied by CIBSE for the sole use of the person making the download. The content remains the copyright property of CIBSE
of fire origin (e.g. axisymmetric plume directly above the
source) or it may be outside the room, having emerged
from an open door or window (e.g. a spill plume). 6.7.4 Flow from an opening

For the purposes of smoke ventilation design, the zone of This Guide recommends that the following equations be
interest is above the flame tip height. used to calculate flow from an opening. It is, however,
acknowledged that there are alternative calculation meth-
ods (Morgan et al., 1999; Kumar et al., 2008; NFPA, 2015).
6.7.3 Axisymmetric plume
Spill plumes occur when the flow of gases leaving a fire
An axisymmetric plume is expected for a fire originating compartment opening rotate around and vertically rise
on the floor away from the walls. It has a virtual point from the opening or balcony edge into the adjacent space.
source. Air is entrained from all sides and along the entire The ascending spill plume entrains air as it rises. The total
height of the plume until the plume becomes submerged flow in the spill plume depends on the size of the opening,
in the smoke layer beneath the ceiling. the convective enthalpy of the gases flowing through the
opening, the presence of a balcony, downstand or other
The choice of calculation depends on the height of the construction element that affects either the flow through
plume above the fuel surface (z) and the linear dimension the opening or the subsequent entrainment into the rising
of the source (Ds) (BSI, 2002b). Typically, the fuel surface spill plume, and the vertical height between the spill edge
is taken as the finished floor level and the linear dimen- and the smoke layer interface in the adjacent space.
sion is taken as the diameter of the fire, unless otherwise
specified by the designer. 6.7.4.1 Flow from an opening of a room

The interface between the ‘far field’ and the ‘near field’ is The horizontal mass flow from an opening of a room
often assumed to be at the flame tip height, with the ‘far containing a fire (see Figure 6.5) is given by
field’ above the flame height and the ‘near field’ below it.
See section 6.9.3 for further guidance on calculating the mo = 0.09 Qp1/3 Wo2/3 ho (6.16)
flame height of the plume.
where mo is the horizontal mass flow of smoke from the
6.7.3.1 Far field room opening (kg · s–1), Qp is the convective heat output of
fire (kW), Wo is the width of the opening (m) and ho is the
In the ‘far field’, height of the room opening (m).

msmoke = 0.071 Q1p/3 Qz - z0V5/3 (6.12)

The smoke layer depth below the compartment opening,
do (m) (see Figure 6.5), is given by
where msmoke is the mass flow of smoke in the plume
# &
1 mo 2/3
(kg · s–1), Qp is the convective heat output of the fire (kW), do = (6.17)
Cd 2Wo
z is the height of the plume above the fuel surface (m) (see
Figure 6.8) and z0 is the height of the virtual source above where mo is the horizontal mass flow of smoke from the
the fuel surface (m), room opening (kg · s–1), Wo is the width of the room open-
ing (m) and Cd is 1.0 for a flat ceiling below the spill edge
z0 =- 1.02 Ds + 1.38 Q* 5/3 D s (6.13) or 0.6 for a downstand below the spill edge (BSI, 2003b).
where Ds is the linear dimension of the source (typically
diameter) (m), and 6.7.4.2 The adhered spill plume

Qt Qt Adhered spill plumes are applicable where there is a wall

Q* = 5/2
= (6.14) or solid construction directly above the spill edge, prevent-
1/2
t0 T0 cp g D s 1110 D s5/2
ing entrainment from one side and leading to the plume
where Q* is a dimensionless heat release rate, Q is the total adhering to the wall. Adhered plumes are also known as
heat output of the fire (kW), g is acceleration due to grav- single-sided plumes, since entrainment occurring above the
ity (typically 9.81) (m · s–2), t0 is the ambient air density spill edge occurs only on one face (see Figure 6.6).
(typically 1.204 at 20 °C) (kg · m–3), T0 is the ambient air
temperature (K) and cp is the specific heat of air at constant Calculating total flow (Ws / ds”
pressure (typically 1.006) (kJ · kg–1 · K–1) (Zukoski et al.,
1981). Where Ws / ds ≤ 13, the entrainment for the adhered plume
is given by
6.7.3.2 Near field msmoke = 0.3 Q1p/3 Ws1/6 d1s /2 zs + 1.34 ms (6.18)

In the ‘near field’, where msmoke is the mass flow of smoke in the plume
(kg · s–1), Qp is the convective heat output of fire (kW), Ws
msmoke = 0.19 p z3/2 (6.15) is the width of the flow at the spill edge (m), ds is the
Fire dynamics 6-9

This publication is supplied by CIBSE for the sole use of the person making the download. The content remains the copyright property of CIBSE
Rising plume

mo, do
Wo Qp

Compartment ho
Fire compartment
tment opening Figure 6.5 Plume from a room
opening
Front view Section

msmoke

Adhered
spill plume

zs

gp
Rising plume
Wall

Spill edge ms, ds, hs

Ws Qp

Fire compartment
parrtment
Figure 6.6 Adhered spill plume
Front view Section

smoke layer depth below the spill edge (see equation 6.19) spill edge (see equation 6.20) (kg · s–1) (Harrison and Spear-
(m), zs is the height of rise from the spill edge to the point, 2010a).
underside of the smoke layer over which entrainment can
occur (m) and ms is the mass flow of smoke in the smoke Large entrainment heights
layer below the spill edge (see equation 6.20) (kg · s–1),
For large entrainment heights, where zs > ztrans, the plume
# &
1 ms 2/3 flow becomes axisymmetric in nature, and the total mass
ds = (6.19)
Cd 2Ws flow in the adhered spill plume is given by equation 6.12,
with z0 taken to be zero.
where Cd is 1.0 for a flat ceiling below the spill edge or 0.6
for a downstand below the spill edge, and The transition height ztrans is calculated as (Harrison and
Spearpoint, 2010b)
ms = 0.09 Q1p/3 Ws2/3 hs
ztrans = 3.4 QWs2/3 + 1.56 ds2/3V
(6.20) 3/2
(6.22)
where hs is the height of the spill edge above the level of
the fire (m) (Harrison and Spearpoint, 2010a). 6.7.4.3 The balcony spill plume

Calculating total flow (Ws / ds! Balcony spill plumes are applicable where a balcony
projects beyond the compartment opening and there is no
Where Ws / ds > 13, the entrainment for the adhered wall or solid construction directly above the spill edge,
plume is given by therefore allowing entrainment to occur from both sides
of the rising plume. Balcony plumes are also known as
msmoke = 0.08 Qp1/3 Ws2/3 zs + 1.34 ms (6.21) double-sided plumes, as entrainment occurring above the
spill edge occurs on both faces (see Figure 6.7).
where msmoke is the mass flow of smoke in the plume
(kg · s–1), Qp is the convective heat output of fire (kW), Ws Channelling screens (or side walls) below the level of the
is the width of the flow at the spill edge (m), zs is the spill edge and extending from the compartment opening
height of rise from the spill edge to the underside of the can be used to reduce the lateral spread of the spill plume,
smoke layer over which entrainment can occur (m) and ms thereby reducing the amount of entrainment above the
is the mass flow of smoke in the smoke layer below the spill edge.
6-10 Fire safety engineering

msmoke
Balcony spill plume

This publication is supplied by CIBSE for the sole use of the person making the download. The content remains the copyright property of CIBSE
b

zs
Rising plu
plume
me
Balcony Balcony
Room opening

Spill edge ms, ds, hs

Ws Qp
Channelling Channelling
elling
screen screen
en
Figure 6.7 Channelled
balcony spill plume
Front view Section

Channelled balcony spill plume 6.7.5 Smoke temperature

The total flow in a channelled balcony spill plume is given The excess average temperature of the hot gases can be
by calculated using the following equation:

msmoke = 0.16 Q1p/3 QWs2/3 + 1.56 d2s /3V zs + 1.34 ms (6.23) Qp

i=
msmoke cp (6.25)
where msmoke is the mass flow of smoke in the plume
(kg · s–1), Qp is the convective heat output of the fire (kW),
Ws is the width of the flow at the spill edge (i.e. separation where msmoke is the mass flow of the smoke (kg · s–1), Qp is
between channelling screens) (m), ds is the smoke layer the convective heat output of the fire (kW), cp is the
depth below the spill edge (see equation 6.19) (m), zs is the specific heat of air at constant pressure (kJ · kg–1 · K–1) and
height of rise from the spill edge to the underside of the i is the excess average temperature (°C).
smoke layer over which entrainment can occur (m) and ms
is the mass flow of smoke in the smoke layer below the Smoke temperature can be found by adding the ambient
spill edge (see equation 6.20) (kg · s–1) (Harrison and Spear- temperature of the air to the excess temperature (i).
point, 2008).
The axial plume temperature is given by
Note that for large entrainment heights, where zs> ztrans,
(see equation 6.22) the plume flow becomes axisymmetric Tc = 2i + T0 (6.26)
in nature and the total mass flow in the adhered spill
plume is given by equation 6.12, with zo taken to be zero. where Tc is (absolute) axial temperature (K) and T0 is
ambient air temperature (K).
No channelling screens below balcony
6.7.6 Volume flow rate of smoke
Where there are no channelling screens below the balcony,
the entrainment in the vertical plume is given by
The volume flow rate of smoke is
msmoke = 0.16 Q1p/3 Q!Wo + b$2/3 + 1.56 d2o/3V zs + 1.34 mo (6.24)
T
V = msmoke
where msmoke is the mass flow of smoke in the plume t0 T0 (6.27)
(kg · s–1), Qp is the convective heat output of fire (kW), Wo
is the width of the flow at the compartment opening (m), where msmoke is the mass flow of smoke (kg · s–1), t0 is the
b is the breadth of the balcony (m), do is the smoke layer ambient air density (typically 1.204 at 20 °C) (kg · m–3), T0
depth below the compartment opening (see equation 6.17) is ambient air temperature (K) and T is smoke temperature
(m), zs is the height of rise from the spill edge to the (K).
underside of the smoke layer over which entrainment can
occur (m) and mo is the mass flow of smoke in the smoke
layer below the compartment opening (see equation 6.16) 6.7.7 Ceiling flow
(kg · s–1) (Harrison and Spearpoint, 2010c).
Smoke will flow along the ceiling towards the vents or
This equation is limited to cases where Wo ≥ 2b. It has not fans. This flow is driven by the buoyancy of the smoke.
been verified for cases where a downstand exists prior to Irrespective of the reservoir or ventilation area, this flow-
the spill edge, so should not be applied in that situation. ing layer would still have a depth related to:
Fire dynamics 6-11

— the width of the reservoir

— the temperature of the smoke, and
— the mass flow rate of smoke. h

This publication is supplied by CIBSE for the sole use of the person making the download. The content remains the copyright property of CIBSE
z

This depth (dl) can be calculated as zf

dl = T Y
Ml Tl 2/3
(6.28) Figure 6.8 Axisymmetric smoke filling a room with a low-level opening
ci l0.5 Wl

where dl is the depth of the flowing layer (m), Ml is the Calculation routines for simple smoke filling can easily be
mass flow of smoke entering the layer (kg · s–1), Tl is the written. A specified growth curve (e.g. fast, medium or
smoke layer temperature, c is the downstand factor, i is slow) is subdivided into time elements and the entrain-
the excess temperature (e.g. rise of the smoke layer above ment equations are applied to each successive element.
ambient) (°C) and Wl is the width of the reservoir or the The layer depth in the reservoir at the end of each time
distance between channelling screens etc. (m) (BSI, element can then be taken as the starting point for the
2003b). next element. The smoke layer will therefore consist of a
number of elemental thin layers. In addition to adding
The downstand factor (c) is equal to 36 if a deep down- elemental layers, elemental smoke extract may be
stand is present at right angles to the flow, or 78 if no subtracted, depending on what type of smoke control (if
downstand is present at right angles to the flow. any) is applied. The output of the program can show, as a
function of time, the following:
Ml would usually be taken to be the mass of smoke
(msmoke), while Tl is taken to be the average temperature — clear layer position
of the smoke plume as it enters the layer. — average temperatures
— average visibilities.

6.8 Accumulated ceiling

6.8.3 Smoke filling: rooms with
layer low-level ventilation openings
In such rooms, there is no smoke flow out of the low-level
6.8.1 General opening in the wall (see Figure 6.8). Heat loss to the room
surfaces, which would result in slightly smaller fire devel-
The simplest zone model postulates that smoke rises to opment, is neglected.
form a smoke layer of uniform depth and temperature
with a substantially smoke-free layer below it. Smoke
control systems are frequently required to maintain a 6.8.3.1 Axisymmetric plume
minimum height for the smoke-free layer for a specified
time (see chapter 10: Smoke ventilation). The elapsed time at which the smoke-free layer is at a
height z (m) is obtained by solving the differential
equation
6.8.2 Smoke filling times
dz Qp
t0 Af + msmoke + =0 (6.29)
For steady-state smoke control design, the entrainment dt T0 cp
equations may be used to calculate the smoke exhaust
required. where Af is the floor area of the room (m2), msmoke is the
mass flow of smoke (kg · s–1), Qp is the convective heat
However, in some large spaces the volume of the smoke output of the fire (kW), cp is the specific heat of air at
reservoir is so large that the size itself is a form of smoke constant pressure (kJ · kg–1 · K–1), t0 is the ambient air
control, since any smoke reservoir will take a finite time density (typically 1.204 at 20 °C) (kg · m–3), T0 is the ambi-
to become full. This time may be calculated by a number ent air temperature (K), z is the height of the plume above
of methods, as follows: the fuel surface (m) and t is time (s).

— by using a computer program to integrate calcu- The variation of msmoke with z is described in section 6.7.
lated smoke volumes produced at small time
intervals (e.g. the ‘available safe egress time’ (ASET) Solutions to equation 6.29 are given in Figure 6.9 for an
model) axisymmetric plume (equation 6.12) and constant Qp,
using dimensionless parameters as follows:
— by integrating various relationships mathemati-
cally, using simplifying assumptions, to derive a Z=z h
formula (see below).
Q* = Qp "t0 T0 cp QghV1/2 h2% = Qp Q1100h5/2V
The latter method, being more approximate in nature, will
usually produce a conservative figure. x = t Qg hV1/2 Qh2 AfV = Q3.13th3/2V Af
6-12 Fire safety engineering

where h is the floor-to-ceiling height of the room or the where Hc is the heat of combustion (kJ · kg–1), YCO is the
height of the ceiling above the base of the fire (m) (see mass yield of carbon monoxide per unit mass of fuel
Figure 6.8) and g is acceleration due to gravity (m · s–2). decomposed (kg · kg–1), R is the rate of burning (kg · s–1)
and Q is the total heat output of the fire (kW).

This publication is supplied by CIBSE for the sole use of the person making the download. The content remains the copyright property of CIBSE
Figure 6.9 solves the following integral:
6.8.3.2 Spill plume
1 dZ
#z
0.195 QQ*V1/3 Z 5/3 + Q*
x= (6.30)
For a spill plume, such as given by equation 6.18, 6.21,
6.23 or 6.24, i.e.
Where the ceiling area and the smoke base area are both
equal to Af , the average density of the smoke layer (ts) is msmoke = aQp1/3 Ws2/3 zs (6.36)
given by

ts t0 = 1 - Q* x Q1 - ZV
where a is a numerical factor, Ws is the width of the flow
(6.31) at the spill edge (m), msmoke is the mass flow of smoke in
the plume (kg · s–1), Qp is the convective heat output of the
The average temperature of the smoke layer Ts (K) is fire (kW), zs is the height of the rise from the spill edge to
given by the underside of the smoke layer over which entrainment
QTs - T0V T0 = 1 !1 - Q* x Q1 - ZV$
can occur (m), the differential equation is
(6.32)
dzs Qp
Where an impurity such as carbon monoxide can be t0 Af + aQ1p/3 Ws2/3 zs + =0 (6.37)
dt T0 cp
related to Qp by the expression

mCO = C QQp cp T0V

where Af is the floor area of the room (m2), cp is the
(6.33) specific heat of air at constant pressure (kJ · kg–1 · K–1), t0
is the ambient air density (typically 1.204 at 20 °C)
where mCO is the mass rate of generation of carbon (kg · m–3), T0 is ambient air temperature (K) and t is time
monoxide (kg · s–1), the mass fraction in the ceiling layer (s).
(fm) is given by

fm = CQ* x Q1 - ZV
The solution to this equation, with constant Qp, is
(6.34)
a + QQ*V2/3
QQ*V1/3 x = ln U 2 Z
1
a2 Z + QQ*V2/3
where C is given by equation 6.35b below. Note that (6.38)
a2
mCO = YCO R = YCO Qt Hc (6.35a) where

C = YCO cp T0 Hc (6.35b) Q* = Qp "t0 T0 cp QghV1/2 QhWsV% = Q Q1100 h3/2 WsV

1.0 Figure 6.9 Solutions of equation

6.29 for an axisymmetric plume

0.8

0.6
Value of Z

0.4

0.2

0.002

Q* = 0.5 0.2 0.1 0.05 0.02 0.01

0.0
0 2 4 6 8 10 12
Value of [(Q*)1/3 o ]
Fire dynamics 6-13

x = t Qg hV1/2 QhWs AfV = 3.13 th1/2 Ws Af the negative pressure zone, and the extract air is taken
from the positive pressure zone. The height of the neutral
a2 = a QT0 cp t20V1/3 = 2.72 a plane, therefore, must be known. If h1 and h2 are the
distances from the neutral plane to the lower and upper

This publication is supplied by CIBSE for the sole use of the person making the download. The content remains the copyright property of CIBSE
openings, respectively, then
Z = zs h
A2 T
QQ*V1/3 x = 0.303 QQWs2V1/3 t Af
h1
= 22 1 (6.43)
h2 A1 T0

where A1 and A2 are the areas of the lower and upper

where h is the floor-to-ceiling height of the room or the openings, respectively, and T0 and T1 are the lower and
height of the ceiling above the base of the fire (m) and g upper temperatures, respectively (Thomas et al., 1963).
is acceleration due to gravity (m · s–2). The sum of h1 and h2 must always equal the total distance
between the upper and lower openings, and therefore the
Equations 6.31, 6.32 and 6.34 can be used to calculate the location of the neutral plane can be determined.
average temperature, density and mass fraction by insert-
ing the above values.
6.8.4 Smoke filling: open rooms
This solution can be used where smoke flows from a
communicating space into a large volume space, such as a
approaching flashover
shopping mall or atrium, by entering equation 6.18, 6.21,
6.23 or 6.24 and the dimensions Af and h of the large The calculations given in section 6.8.3 are not suitable
volume. where flames are approaching ceiling height or where
smoke flows out of the wall opening. Under these circum-
stances, the following equation may be used:

Ts - T0 = 9.15 !QQ2p QAo h1o/2 ak AtVV$

6.8.3.3 Room filling with smoke extract 1/3
from layer (6.44)

where Qp is the convective heat output of the fire (kW),

A critical height of the smoke layer may be dictated by the
Ao is the area of ventilation opening (door, window etc.)
need to keep it above eye level, inside a reservoir or, if
(m2), ho is the height of the ventilation opening (m), ak is
otherwise too hot, well above head level. This is covered
the effective heat transfer coefficient (kW · m–2 · K–1) and
in further detail in chapter 10: Smoke ventilation.
At is the area of the room surface (wall, floor, ceiling) (m2)
(McCaffrey et al., 1981).
If the critical ‘clear layer’ height zc (m) would be reached
before the occupants have escaped, then extract from the
smoke layer can be provided, under steady-state condi- Equation 6.44 was derived for At / (Ao ho1/2) values between
tions, as follows: 16 and 530 m–1/2.

Mout = Mc (6.39) By substituting (Ts – T0) = 580 K in equation 6.44, the

value of Q at flashover is given by

Qf = 505 QAo h1o/2 ak AtV

where Mout is the mass flow rate of the vented smoke
(kg · s–1) and Mc is the mass flow rate in the plume (kg · s–1)
1/2
(6.45)
at height zc (m).
where Qf is the convective heat output of the fire at flash-
The temperature of the vented smoke, Ts (K), under over (kW).
steady-state conditions will be given by
For [Anet / (Avo ho1/2)] < 10 m–1/2, the following value for Qf
Ts - T0 = Qp QMout cpV (6.40) is recommended (Thomas, 1981):

and the volume flow rate, v (m3 · s–1), by Qf = 5.2 Anet + 252 Ao h1o/2 (6.46)

v = QMout t0V + Qp Qt0 T0 cpV (6.41) where

With natural ventilation, the mass flow rate of the vented Anet = At + Ao (6.47)
smoke is given by
The effective heat transfer coefficient is derived from
Cd Avo t0 !2gQh - zVQTs - T0V T0$1/2
ak = Qmw tw cw tcV
Ts1/2 "Ts + QAvo AviV T0%
Mout = 2 1/2 (6.42) 1/2
(6.48)

where Avo is the outlet ventilation area (m2), Avi is the where mw is the thermal conductivity of the wall material
ventilation inlet area (m2) and Cd is the discharge coeffi- (kW · m–1 · K–1), tw is the density of the wall material
cient (BSI, 2002b). Values for discharge coefficients (kg · m–3), cw is the specific heat capacity of the wall mate-
(between 0.6 and 0.9) are provided by the vent rial (kJ · kg–1 · K–1) and tc is the characteristic burn time (s)
manufacturer. (Drysdale, 2011).

The location of the extract points is determined by the Table 6.4 gives values of ak for a characteristic burn time
location of the neutral plane, so that inlet air enters within of 900 s.
6-14 Fire safety engineering

Table 6.4 Effective heat transfer coefficient to 6.8.7 Stratification

the surfaces of a room or compartment
Material of surface ak / When the ambient temperature at ceiling level is signifi-
kW · m–2 · K–1 cantly higher than at the level where the fire starts, the

This publication is supplied by CIBSE for the sole use of the person making the download. The content remains the copyright property of CIBSE
Concrete 55 × 10–3 upward movement of the smoke plume may cease, due to
lack of buoyancy, and stratification may occur.
Brick 36 × 10–3
Plaster 21 × 10–3 The maximum height of rise of an axisymmetric plume is
Plasterboard 13 × 10–3
given by
Fibre insulating board 5.2 × 10–3 zm = 5.54 Q1p/4 QdT/dzV-3/8 (6.53)

Flashover is not expected until there are sustained flames where zm is the maximum height of smoke rise above the
at ceiling level. For axisymmetric sources of base dimen- base of the fire (m), Qp is the convective heat release rate
sion less than the ceiling height, the minimum condition (kW) and dT / dz is the rate of change of ambient temper-
for flashover is given by ature with respect to height (assumed to be linear) (K · m–1)
(NFPA, 2015).
hf 1 0.094 Q2p/5 (6.49)

where hf is the height of the ceiling above the base of fire

for flashover (m) (Cox and Chitty, 1980). 6.9 Flame calculations
For extended area sources, the minimum condition for
flashover is given by 6.9.1 General
hf 1 0.035 Q2p/3 Qd1 + 0.074 Q2p/5V
2/3
(6.50) Various methods are available to calculate flame heights
for both hydrocarbon and cellulosic fires, and for
where dl is the longer dimension of the source (m). post-flashover fires. These are mainly used to estimate
radiant heating or radiant and convective heating of
combustible materials and elements of the structure,
6.8.5 Heat transfer to building although it may be necessary to assess radiant effects on
surfaces personnel, such as firefighters. Calculating flame height
can show where flame impingement is likely to occur. For
In the simple room filling model considered in 6.8.3, heat example, if it can be shown that a steel member is not
transfer to the ceiling and wall surfaces is neglected. This engulfed in flame, it may be possible to use materials
is a conservative assumption in that the volume of smoke which that shorter fire resistance periods.
is overestimated. However, if low-temperature smoke is
filling a large reservoir, then cooling may lead to loss of
buoyancy, which should be taken into account. In the 6.9.2 Heat flux calculation
absence of experimental data, it is suggested that cooling
effects should be allowed for where the area of the reser- By assuming flame heights and areas of burning, it is
voir is greater than 2000 m2, and/or the average layer possible to calculate the radiation due to a fire which
temperature is less than 10 K above ambient when calcu- impinges on a separate fuel package.
lated by neglecting cooling. Further guidance is given in
chapter 10: Smoke ventilation. In areas not equipped with sprinklers, fires will tend to
grow until limited by lack of fuel or air. In compartments
where items of fuel are very widely spaced, it is possible
6.8.6 Heat transfer from smoke to predict whether fire spread will occur from item to
layer by radiation item. This is done by calculating the radiative heat flux
originating from the fire and which falls on the target
The radiation emitted from a hot smoke layer is given by item:

Ir = fs v Ts4 (6.51) Ir = zff v Tf4 (6.54)

where Ir is the intensity of the emitted radiation (kW · m–2), where Ir is the radiative heat flux (kW·m–2), z is a configu-
fs is the emissivity of the smoke layer, v is the Stefan– ration factor (see below), ff is the flame emissivity, v is the
Boltzmann constant (5.67 × 10–11) (kW · m–2 · K–4) and Ts is Stefan–Boltzmann constant (5.67×10–11) (kW·m–2 ·K–4)
the average (absolute) smoke layer temperature (K). and Tf is the flame temperature (K) (SFPE, 2016).

As a conservative assumption, fs may be taken as unity. The configuration factor, z, represents the geometrical
Alternatively, it may be estimated for a ceiling layer from relationship between the source and target. The above is a
very general method for calculating radiative heat flux. A
fs = 1 - exp !- Q0.33 + 470 mfVQh - zV$ (6.52) more detailed analysis, including techniques for calculat-
ing z, is given in the SFPE Handbook (SFPE, 2016).
where h is the height of the ceiling (m), z is the height of
the layer interface (m) and mf is the mass concentration of The heat flux impinging on combustible material will
smoke aerosol (kg · m–3) (SFPE, 2016). cause it to heat up. Whether this heating results in
Fire dynamics 6-15

ignition depends on the intensity of the incident flux. drawn to the SFPE Handbook (SFPE, 2016). Hydrocarbon
Experimental work by Babrauskas (1981) suggests that for fires in the open are likely to be influenced strongly by the
very thin materials, such as curtains, the heat flux required wind, and this should be taken into account.
for ignition could be relatively low, at around 10 kW · m–2.

This publication is supplied by CIBSE for the sole use of the person making the download. The content remains the copyright property of CIBSE
For thick materials, the value may be higher, i.e. about
40 kW · m–2. BR187 (Chitty, 2014) gives a conservative 6.9.4 Flame projection (post-
value of 12.6 kW · m–2 to be used in design, based on the flashover)
piloted ignition of unprotected dry wood; although, spon-
taneous ignition and the effects of paint and moisture can On occasions it may be necessary to calculate the flame
further increase the critical radiation intensity required to projection from openings in a compartment that is
cause ignition. The Ignition Handbook (Babrauskas, 2003) involved in a fully developed fire (see Figure 6.10). Flame
provides ignition conditions for a range of materials. It is projection from the windows or doors in a compartment
suggested that a value of 20 kW · m–2 be taken as appropri- can be estimated from the work of Law and O’Brien, as
ate for most materials. This figure is the same as that contained in Eurocode 1 (BSI, 2002a: annex B). The
found by Thomas and Bullen (1979) as the critical heat height of the flame above the top of the opening, zfo , is
flux for flashover in a room. given by

zfo + ho = 12.8 QR wV2/3 (6.57)

6.9.3 Calculation of flame height
where zfo is the flame height above the top of the opening
The larger the flame or the surface that is radiating heat, (m), ho is the height of the opening (m), R is the rate of
the larger will be the total heat that is emitted. This fuel combustion (kg · s–1) and w is the width of compart-
implies that larger flames give larger values of z. There- ment openings (m).
fore, the estimation of flame heights is a crucial part of the
calculation process. For cellulosic fires, the ventilation-controlled rate of burn-
ing, R, may be calculated from Thomas’s correlation
For most fires away from walls, the plume can be consid- (Thomas, 1973) as follows:
ered to be axisymmetric. The mean flame height of
luminous flames for fires is given by R = 0.02 "QAt - AoVQAo ho VQw dV%
1/2
(6.58)
zf = 0.2 Q2t /5 (6.55) where R is the rate of fuel combustion (kg · s–1), At is the
area of enclosing walls (m2), Ao is the area of the opening
where Qt is the total heat output of the fire (kW) and zf is (m2), ho is the height of the opening (m), w is the width
the mean flame height of the luminous flame (m) (Cox and of the wall containing an opening (m) and d is the depth
Chitty, 1980) (see Figure 6.8). of the room behind an opening (m).
As an alternative to equation 6.55, the mean flame height The heat output of the fire is given earlier by equation 6.8.
is also given by

zf = 0.235 Qt2/5 - 1.02 D f (6.56) 6.9.5 Fire resistance assessment

where Df is the fire diameter (m) (SFPE, 2016). If The fire resistance value is based, for example, on a furnace
unknown, the fire diameter may be estimated from the test specified in ISO 834-1: 1999 (ISO, 1999), BS 476-20:
heat output by assuming an average fire load density and 1987 (BSI, 1987) or BS EN 1363-1: 2012 (BSI, 2012) (or
then calculating the area of burning. from conditions for fire resistance testing for specific
applications). A real fire may be shown to be less or more
As equations 6.55 and 6.56 do not perfectly agree, the more severe (see Figure 6.11), in which case the fire resistance
conservative choice should be made if there is any doubt. period may be reduced (Butcher and Parnell, 1983) or may
need to be increased. In Figure 6.11, curves 60(1/4) and
The above relationships do not apply to hydrocarbon fires. 60 (1/2) are typical of shop fires (60 kg · m–2), and curves
The calculation of such fires is complex and attention is 30(1/4) and 30 (1/2) are typical of office fires (20–30 kg · m–2).

Flame projection Figure 6.10 Flame projection

Flame projection
Fla from an opening
zfo zfo d

Ao = h o w
ho ho
Co
Compartment
op
opening
w Compartment
opening

Front view Section

6-16 Fire safety engineering

1200 (a) Simple case

1000 Furnace curve

(BS 476)

This publication is supplied by CIBSE for the sole use of the person making the download. The content remains the copyright property of CIBSE
Average temperature / °C

800

600
60(1/4)

60(1/2)
400 30(1/4)
30(1/2)
15(1/4)
200
15(1/2) Af = w1 w2 Ao = wo ho
7.5(1/4)
7.5(1/2)
0 Anet = 2 Af + 2 h Qw1 + w2V - Ao
0 10 20 30 40 50 60
Time / minutes
d w = w2 w1
Note: 60(1/2) means fire load 60 kg·m–2 of floor area and ventilation
50% of one wall

Figure 6.11 Effect on fire temperature of fire load and ventilation.

(b) More than one window
(Reproduced from Designing for Fire Safety by EG Butcher and AC
Parnell, by permission of David Fulton Publishers Ltd.)

Methods for calculating compartment temperatures are

beyond the scope of this section. However, detailed calcu- ho3
lation procedures are given by Law and O’Brien in
Eurocode 1 (BSI, 2002a: annex B), in the SFPE Handbook ho2
wo3
(SFPE, 2016) and by Thomas (1986). ho1 wo2

wo1

Appendix: Dimensions of a room

or compartment
Ao1 = wo1 ho1 Ao2 = wo2 ho2 etc.

The following dimensions and areas should be Ao = A1 + A2 + etc. wo = wo1 + wo2 + etc.
calculated:
Ao1 + Ao2 + etc.
h=
Ao
Af Floor area (m2)

Ao Area of opening (window or doorway) of room (c) Windows in more than one wall
(m2)
Wall 3 Wall 4
Anet Internal surface area of room minus area of open-
ings (m2)

c Core dimension (m)

d Depth of opening (m)

h Floor-to-ceiling height of room or height above

base of fire (m)

ho Height of opening (window or doorway)

w Width of wall containing an opening (m)

wo Width of an opening or doorway (m) (Wall 1 contains the greatest window area)
Fire dynamics 6-17

Aow1 = window area on wall 1 BSI (2003b) BS 7346-4: 2003 Components for smoke and heat control systems.
Functional recommendations and calculation methods for smoke and heat
exhaust ventilation systems, employing steady-state design fires. Code of practice
Aow2 = window area on wall 2 etc. (London: British Standards Institution) (2003)

This publication is supplied by CIBSE for the sole use of the person making the download. The content remains the copyright property of CIBSE
BSI (2012) BS EN 1363-1: 2012 Fire resistance tests. General requirements
Ao = Aow1 + Aow2 + etc.
(London: British Standards Institution)
w2 Aow1 BSI (2017) BS 9999: 2017 Fire safety in the design, management and use of
d w=
w1 Ao buildings. Code of practice (London: British Standards Institution)

Bukowski RW (1993) ‘A review of international fire risk prediction

(d) Compartment with core methods’ Proceedings of the 6th International Fire Conference (Interflam ’93)
(Oxford: Interscience Communications)

Butcher EG and Parnell AC (1983) Designing for Fire Safety (London:

David Fulton Publishers)

Chitty R (2014) External Fire Spread: Building separation and boundary

distances BR 187 (2nd edition) (Garston, Watford: IHS BRE Press)

CIBSE (2010) CIBSE Guide E (3rd edition) (London: Chartered

Institution of Building Services Engineers)

Cox G and Chitty R (1980) ‘A study of the deterministic properties of

unbound fire plumes’ Combustion and Flame 39 191–209
c2 c1 Drysdale DD (2011) An Introduction to Fire Dynamics (2nd edition)
(Chichester: Wiley)

Evans DD and Stroup DW (1985) Methods to Calculate the Response of

Heat and Smoke Detectors Installed Below Large Unobstructed Ceilings
Af = w1 w2 - c1 c2 NBSIR 85-3167 (Washington, DC: National Bureau of Standards)

Harrison R and Spearpoint M (2008) ‘Characterisation of balcony spill

Anet = 2 Af + 2 h Qw1 + w2 + c1 + c2V - Ao plume entrainment using physical scale modelling’ Proceedings of the 9th
Symposium of the International Association for Fire Safety Science (London:
Qw2 - c2V Aow1 IAFSS) 727–738
d w=
Qw1 - c1V Ao
Harrison R and Spearpoint M (2010a) ‘Physical scale modelling of
adhered spill plume entrainment’ Fire Safety Joural 45 (3) 149–158

Harrison R and Spearpoint M (2010b) ‘A simple approximation to predict

the transition from a balcony spill plume to an axisymmetric plume’
References Journal of Fire Protection Engineering 20 (4) 273–289

Harrison R and Spearpoint M (2010c) ‘A comparison of channelled and

Annable K (2006) Effectiveness of Sprinklers in Residential Premises – An
unchannelled balcony spill plumes’ Journal of Building Services Engineering
Evaluation of Concealed and Recessed Pattern Sprinkler Products. Section 5:
Research and Technology 31 (3) 265–277
Thermal sensitivity BRE Report 218113 (Watford: BRE Ltd)
Heskestad G and Bill RG (1988) ‘Quantification of thermal responsiveness
Babrauskas V (1981) Will the Second Item Ignite? Report NBSIR 81-2271
of automatic sprinklers including conduction effects’ Fire Safety Journal
(Gaithersburg, MD: National Institute for Standards and Technology)
14 113–125
Babrauskas V (1986) ‘Free burning fires’ Fire Safety Journal 11 33–51
Ingason H (1993) Fire Experiments in a Two-dimensional Rack Storage
Babrauskas V (2003) Ignition Handbook (Issaquah, WA: Fire Science (Brandforsk project 701-917) SP Report 1993:56 (Borås, Sweden: Swedish
Publishers/SFPE) National Testing and Research Institute (SP))

BSI (1987) BS 476-20: 1987 Fire tests on building materials and structures. ISO (1999) ISO 834-1:1999 Fire-resistance tests. Elements of building
Method for determination of the fire resistance of elements of construction construction. Part 1: General requirements (Geneva: International
(general principles) (London: British Standards Institution) Organization for Standardization)

BSI (1999) BS EN 12259-1: 1999 Fixed firefighting systems. Components for Karlson B and Quintiere JG (2000) Enclosure Fire Dynamics (Boca Raton,
sprinkler and water spray systems. Sprinklers (London: British Standards FL: CRC Press)
Institution)
Kumar S, Thomas PH and Cox G (2008) ‘A novel analytical approach for
BSI (2002a) BS EN 1991-1-2: 2002 Eurocode 1: Actions on structures Part characterising air entrainment into a balcony spill plume’ Proceedings of
1–2 General actions – Actions on structures exposed to fire (London: British the 9th Symposium of the International Association for Fire Safety Science
Standards Institution) (2002) (London: IAFSS) 739–750
BSI (2002b) PD 7974-2: 2002 The application of fire safety engineering Law M (1978) ‘Fire safety of external building elements – the design
principles to fire safety design of buildings - Part 2: Spread of smoke and toxic approach’ Engineering Journal 15 59–74
gases within and beyond the enclosure of origin (Sub-system 2). (London:
British Standards Institution) (Note: PD 7974-2: 2002 has been replaced McCaffrey BJ, Quintiere JG and Harkleroad MF (1981) ‘Estimating
by PD 7974-2: 2019) room temperatures and the likelihood of flashover using fire test data
correlations’ Fire Technology 17 (2) 98–119 and 18 (1) 122
BSI (2003a) PD 7974-1: 2003 The application of fire safety engineering principles
to fire safety design of buildings. Initiation and development of fire within the Madrzykowski D and Vettori RL (1991) A Sprinklered Fire Suppression
enclosure of origin (Sub-system 1) (London: British Standards Institution) Algorithm for the GSA Engineering Fire Assessment System (Gaithersburg,
(Note: PD 7974-1: 2003 has been replaced by PD 7974-1: 2019) MD: National Institute for Standards and Technology)
6-18 Fire safety engineering

Morgan HP, Ghosh BK, Garrad G, Pamlitschka R, De Smedt J-C and Thomas PH (1986) ‘Design guide – structural fire safety’ (CIB W14
Schoonbaert LR (1999) Design Methodologies for Smoke and Heat Exhaust Workshop) Fire Safety Journal 10 (2) 77–137
Systems BRE Report 368 (Garston, Watford: BRE Press)
Thomas PH and Bullen ML (1979) ‘On the role of KtC of room limiting
NFPA (2015) NFPA 92 Standard for smoke control systems (2015 edition) materials in the growth of room fires’ Fire and Materials 3 (2) 68–73

This publication is supplied by CIBSE for the sole use of the person making the download. The content remains the copyright property of CIBSE
(Quincy, MA: National Fire Protection Association) Thomas PH, Hinkley PL, Theobald CR and Simms DL (1963)
Investigations into the Flow of Hot Gases in Roof Venting Fire Research
SFPE (2016) SFPE Handbook of Fire Protection Engineering (5th edition)
Technical Paper No. 7 (London: The Stationery Office)
(Boston, MA: SFPE; Quincy, MA: NFPA)
Wade C, Baker G, Frank K, Robbins A, Harrison R, Spearpoint M and
Thomas PH (1973) Behavior of Fires in Enclosures – Some recent progress Fleischmann C (2013) B-RISK User Guide and Technical Manual BRANZ
(Pittsburgh, PA: Combustion Institute) Study Report 28 (Judgeford, Porirua City: BRANZ Ltd)

Thomas PH (1981) ‘Testing products and materials for their contribution Zukoski EE, Kubota T and Cetegen B (1981) ‘Entrainment in fire
to flashover in rooms’ Fire and Materials 5 (3) 103–111 plumes’ Fire Safety Journal 3 (3) 107–121