BRE 186_Design Principle for Smoke Ventilation in Enclosed Shopping Center

8/20/2019 BRE 186_Design Principle for Smoke Ventilation in Enclosed Shopping Center

http://slidepdf.com/reader/full/bre-186design-principle-for-smoke-ventilation-in-enclosed-shopping-center 1/39

Building Research Establishment Report

Design principles for smoke ventilationin enclosed shopping centres

H P Morgan, BSc, CPhys, MInstP, AIFireE, MSFSE and J P Gardner*, BSc

*Colt International Ltd

Fire Research StationBuilding Research EstablishmentGarston, WatfordWD2 7JR



Prices for all available

BRE publications can

be obtained from:

Construction Research

Communications Ltd

151 Rosebery Avenue

London, EC1R 4QX

Telephone

0171 505 6622

Facsimile

0171 505 6606

E-mail [email protected]

BR 186

ISBN 0 85125 462 4

© Crown copyright 1990

First published 1990

Reprinted with corrections 1991

Second reprint 1991

Republished on CD-ROM 1997

with permission of the Controller

of HMSO and the Building Research

Establishment, by Construction

Research Communications Ltd

Anyone wishing to use the

information given in this publicationshould satisfy themselves that it is not

out of date, for example with reference

to the Building Regulations

Applications to copy all or any part

of this publication should be made to:

Construction Research Communications Ltd,

PO Box 202, Watford, Herts, WD2 7QG



Contents

Page

Foreword v

Introduction 1

Chapter 1 General principles 3

Chapter 2 Design procedure for the mall smoke control system 6

Smoke reservoirs 6

D esign fire size 6

Single-storey malls 7

Two-storey malls with small voids 8

Two-storey malls with large voids 10

Multi-storey malls 14

Calcula ting smoke temperature 15

Mall sprinklers 17

Flowing layer depth 18

Local deepening 18

Inlet a ir 19

Minimum number of extract points 20

Natura l ventila tion - area required per reservoir 21

Powered ventilation 22

Chapter 3 Large shop opening onto a mall 23

Chapter 4 Some practical design considerations 25

Factors influencing the design fire 25

The effect of wind on the efficiency of a smoke ventilation system 25

False ceilings in the mall 25

The use o f a plenum cha mb er a bo ve a fa lse ceiling in a sho p 26

Stores with internal voids 27

Sloping malls 27

Assessment of effective layer depth 27

Smoke flow in low narrow malls 27

B asement service levels 27

E nclosed car parks 29

Smoke transfer ducts 29

E ntrances within the smoke layer 30

Other situations 30

Chapter 5 Some operational factors 31

References 32

iii



Foreword

This R eport has been prepared b y the Fire R esearch Sta tion of the B uilding Research

E stablishment (B R E ) and results from the latest scientific knowledge and practical experience in

smoke movement and control. It is an update o f Smoke control m ethods in enclosed shopp ing

complexes of one or m ore storeys: a design summary published in 1979, and is based upon

preliminary work carried out for B R E b y Colt Internat ional Limited.

The R eport ta kes into a ccount the comments of a Liaison Co mmittee consisting of industrial a nd

government representat ives. The Fire R esearch Station is also grat eful to the So ciety o f Fire Sa fety

E ngineers, the Institute of Fire Engineers and the U K Cha pter of the Society of Fire Protection

Engineers for providing detailed comments which have as far as possible been included.

The Fire R esearch Station a cknowledges the a ssistance given by C olt Interna tional L imited in the

preparation of the final Report.

v



This Repo rt is intended to a ssist designers of smoke

ventilat ion systems in enclosed shopping complexes.

Most of the methods advocat ed are the outcome of

research into smoke control by smoke ventilation at

the Fire Research Stat ion, but also take into account

the recommendations1 of the Working Part y on f ire

precautions in town centre redevelopment, as well as

experience gained and ideas developed whilst the

authors and their colleagues have discussed many

proposed schemes with interested part ies. The primary

purpose of this Repor t is to summarise the design

advice available from the Fire Research Station at t he

time of its preparation, in a readily usable form. As

such, the Report is neither a detailed engineering

manual nor is it a scientific review article. Perhaps

most importa nt of a ll, it is not a summary o f the

tota lity o f a pproaches possible. New methods such a s

those ba sed upon computational fluid dyna mics, will

be developed as time passes and there w ill alwa ys bespecial cases where existing alternat ive methods can be

adopted.

At peak times a shopping centre can be occupied by

thousands of people, and some larger centres by more

than a hundred thousand. A t ypical centre may

comprise many individua l shop units opening onto a

common mall. Although the individual units may be

separa ted from ea ch other by a dividing wall of fire-

resisting construction, usually t he shop is either open

fronted or only separat ed from the mall by a glass shop

front . This means that t he public areas of th e entirecentre can be effectively undivided.

Mea ns of escape from within ea ch shop unit will, in

genera l, be specified (eg B S 5588 Par t 22) in the same

wa y a s for shops which are not part of an enclosed

complex. This means that escape from w ithin the shops

is specified as if the mall were as much a place of sa fety

as the usual open-sky street. U nfortunat ely, the mall is

a street with a roo f, and so cannot be regarded a s being

as inherently safe as an open-topped street. People

escaping from a shop into a mall will still need to tra vel

along the ma ll before exiting to a true place of safety. It

follows that the mall constitutes an additional stage tothe escape route, which needs to be protected fro m the

effects of fire and smoke. Ideally, it should approach

the same level of safety a s a street for a s long as people

need to escape through it — even if smoke enters the

mall from the shop on fire. D etails of means of escape

provisions for the malls can be found elsewhere1.

A shopping complex is a public building and the

occupants w ill be a cro ss section of the community

including the elderly, children and the disab led. They

will not necessar ily be familiar w ith the building, or

perhaps more importantly, with all the escape routesthat might be provided for them. In many types of

building it is widely recognised tha t people will

commonly try to escape by the same route they had

used to enter the premises (see for example, Ca nter3).

It follows that escape via the malls must be assumed,

even where other exits are provided1. In this situa tion a

long evacuation period can b e expected. A detection

and a larm system is required to give early wa rning of a

fire, and sprinklers are needed to control its spread.

Sta tistics of fire deaths show that the majority of

fat alities are d ue to the effects of smoke.

The ideal option wo uld be to prevent a ny smoke from

a shop fire entering the ma ll at a ll. In the ma jority o f

cases, it would either be very difficult or extremely

expensive to fit a separate smoke extraction system to

each a nd every shop, however small1. Note that large

shops are, however, an exception to this rule and the

provision of smoke ventilatio n systems for such shops

is discussed in C hapter 3. Occasionally circumstances

dictate tha t smoke ventilation fitted to each shop unit,even small units, is the most viab le option for

protecting part or a ll of a ma ll. (This can a pply, for

example, when a n old complex is being redeveloped,

but the mall is too low and/or too narrow to a llow t he

installation of a viable smoke ventilation system in the

mall itself). There ha ve been severa l examples of this.

Nevertheless it remains generally true that this option

is rarely found to be appropriate for most ma lls.

Occasionally, one still hear s the suggestion tha t the

mall should be pressurised to prevent smoke moving

from a shop into the ma ll. This is not usually a via bleoption by itself where the o pening between shop a nd

mall is large (eg a n open-fronted shop, or a shop who se

glazing has fa llen awa y in whole or pa rt). This is

because the airspeed needed from ma ll into shop in

order to prevent the movement of smoky ga ses the

other wa y through the same opening, can va ry between1/2 and c. 2 ms-1 depending on fire size, gas tempera ture

etc. All of this air must be continuously removed from

within the shop unit in order to ma intain the flow. The

quantities of air-handling plant required will exceed

the size of smoke ventilat ion systems for most typical

shop-front openings.

Where smoke fro m a fire in a single shop unit could

spread rapidly via the malls through the entire centre,

a smoke vent ilation system in the ma lls is essential to

ensure that escape is unhindered, by ensuring that any

large q uantities of thermally buoya nt smoky gases can

be kept separa te from (ie ab ove) people who may still

be using escape routes thro ugh the ma lls. Therefore,

the role of a smoke vent ilation system is principally

one of life safety. It should also be remembered,

however, that firefighting becomes both difficult and

dangerous in a smoke-logged ma ll. It follows that to

assist the fire services, the smoke ventilation systemshould be capa ble of performing its design function for

a period of t ime longer than tha t req uired for the

Introduction

1



public to escape: so a n immediate att ack on the f ire

can be made a fter the arrival of the fire brigade.

G uidance on design principles for smoke ventilat ion

systems was summarised in a report by M organ4

published in 1979. This wa s ba sed on the knowledge

ava ilable at that time. Since then a great dea l of

relevant research ha s been carried out, which for t he

most part has confirmed the guidance given in the

original report, but has in some areas highlighted theneed to modify that guidance. A great d eal of practical

experience has been ga ined in designing such systems.

Also in the intervening yea rs, shopping centres

themselves have b ecome larger and more intricate with

man y open levels, interconnecting voids, sloping floors

and at rium features. These can result in a very complex

path o f smoke flow from the shop in which the fire

starts to the eventual point of extraction.

The purpose of this Report is to upda te the guidance

available in the earlier design summary4 to reflect

these changes, to a ssist designers of smoke controlsystems in enclosed shopping centres. As wa s the case

with the previous work, this Report only gives

guidance in line with current knowledge and generally

accepted practice. The guidance is based on results of

research where possible, but also on the cumulat ive

experience of design features, required for regulatory

purposes, of many individual smoke ventilation

proposals. Ma ny of these design features have been

evolved o ver a number of yea rs by consensus between

regulatory authorities, developers and fire scientists,

rather than by specific research. Such advice has been

included in this Repo rt with t he intention of giving the

fullest picture possible. It is therefo re likely tha t some

of this guidance will need to be mo dified in the future,

as the results of continued research become available.

A C ode of practice5 for enclosed shopping centres is

currently being prepared by t he B ritish Standa rds

Institution (B SI). The aim of this Report is to provide

guidance o nly on d esign principles for smokeventilation a nd it is hoped to support the Code ra ther

than pre-empt it. The Report cannot cover all the

infinite variat ions of shopping centre design. Instead it

gives general principles for the design of efficient

systems, with simplified design procedures for an idea l

model of a shopping mall and then further guidance on

frequently encountered practical problems. B ecause

the design procedures are of necessity simplified, the

R eport a lso gives their limitat ions so tha t, when

necessary, a mo re deta iled design by specialists can be

carried out.

A Co de of practice for at rium buildings is also being

prepared by the BSI . An atrium can be defined as any

space penetrating more tha n one storey of a building

where the space is fully or partially covered. Most a tria

within shopping centres may be considered as part of

the shopping mall and treated accordingly. Where atria

have mixed occupancies (eg shops and of fices) then

reference should be ma de to the ab ove document,

when available, or specialist advice sought.

2



Smoke fro m a fire in a shop rises in a plume to the

ceiling. As th e plume rises, air is entra ined into it,

increasing the volume of smoke and reducing its

temperature. The smoke spreads out underneath the

ceiling and f orms a layer which deepens as the shopbegins to fill with smoke. As the la yer deepens, there is

less height for t he plume of smoke to rise before it

reaches the smoke layer. Less air is being entra ined,

with the result that the temperature is higher on

reaching the smoke layer. As this continues, the

increasing smoke layer temperature at some point will

opera te the sprinklers. The fire may sha tter the shop

front glazing (if present) unless that glazing is fire-

resista nt. The smoke will then flow o ut of t he shop into

the ma ll (Figure 1). The change of d irection as the

smoke flows out o f the shop fro nt results in increased

mixing with the surrounding a ir.

There is so much mixing that , except close to t he fire

itself, the hot smoky gases can b e regarded a s

consisting entirely of warmed air, when calculating

flow. The quant ity of smoke par ticles (and hence the

optica l density a nd visibility thro ugh the smoky gases)

produced in a fire, depends strongly on the nat ure of

the burning material6.

The qua ntity o f hot ga ses carry ing these par ticles is,

however, mainly dependent on the size and rate of

burning (related to the heat output per second) of the

fire. Increa sing the visibility t hrough these gases

req uires their dilution with clean a ir, but improving

visibility (and reducing toxicity) by dilution a lone in a

mall to sa fe levels af ter smoke has entered, is not a

pract ical proposition. One suggested safe level of

visibility7 (about 8 metres — rather a short d istance for

a ma ll) would require the hot smoky gases to be

diluted by a mass rat io possibly larger than 300, even

aft er the gases leave the shop of o rigin. It fo llows then

that hot smoky gases should alwa ys be kept spatially

separated from escaping people.

O ne can state as a general principle that a ir will mixinto a rising stream of h ot smoky ga ses in large

qua ntities, but w ill not mix appreciably into a

horizontally flowing stream of hot smoky gases except

under special (a nd usually local) conditions.

Smoke flow ing from the shop onto t he mall will rise tothe ma ll ceiling (Figures 1 and 2). Air w ill mix into t he

smoke as it rises. If no smoke control mea sures are

present, the gases will then flow along the mall as a

ceiling lay er (Figure 2a) at a speed8 typically b etween

1 and 2 m/s.

This is faster tha n the proba ble escape speed of

pedestrians in a crowded mall9. When the smoke

reaches the end of the closed mall it w ill dip down to a

low level and be dra wn ba ck toward s the fire10

(Figure 2b).

If the end of the ma ll is open, even light winds blow ing

into or a cross the opening will cause severe local

disturbance a nd mixing, and once a gain smoke at low

level will be draw n back tow ards the fire10 (Figure 2c).

H ence a single-storey ma ll could become smoke

logged w ithin minutes. An unsprinklered fire in a

single-storey shopping centre in Wolverham pton11, is

thought to h ave ca used a 100-metre-long ma ll to

become untenable within one minute. Similarly short

times can be expected to a pply to the upper floors of a

multi-storey ma ll. Since it is not usually pra ctical to

prevent smoke entering the mall, except for la rger

shops, a mall smoke contro l system is necessary to

control and remove heat a nd smoke.

The Fire Research Sta tion has extended the ideas

developed during its earlier work on t he fire venting of

factories12 to a pply to ma lls8 in order to use the

buoyancy inherent in fire gases to keep those gases

safely above the heads of people in the malls (Figure

3). There are three essential fea tures of a smoke

ventilation system, without a ny one o f w hich it w ill not

function effectively:

1 There must be some means of forming a smokereservoir to prevent the lat eral spread of smoke,

Chapter 1General principles

3

Figure 1 Smoke spread and ma in entrainment sitesin single and two -storey ma lls



which w ould result in excessive loss of buo ya ncy.

This reservoir must be designed to conta in the

smoke layer b ase well above head height. G enerally

speaking, malls with high ceilings or ro oflights will

allow a deeper smoke reservoir and hence a more

effective smoke ventilation system tha n in low,

narrow malls.

2 There must be extraction of smoke within the

reservoir, to prevent the smoke layer b uilding dow n

below t he design depth. This can be natural,

buoyancy driven ventilation or mechanical

extraction, depending on the circumstances. The

rate of t he exhaust must equa l the rate at w hich the

smoke enters the reservoir from below.

3 Since the gases being extracted consist almost

entirely of air tha t ha s mixed with the o riginal fire

gases, fresh air must enter the mall to ta ke its place.

It must enter at a rate eq ual to the rate of extraction

of smoke and a t a low enough height not to mix

prematurely with the smoke.

4

Figure 2a Crea tion of a moving smoke layer beneath the ceilingof a n unventilated ma ll, showing movement of thedisplaced a ir

Figure 2b R ecirculation of smoke in an unventilated and closed mall

Figure 2c Mixing of smoke into the air being drawn into anopen-ended ma ll, caused by wind

Figure 3 Principles of system needed to contain smoke in awell-defined layer (section a long ma ll)



In a fire most of the important fa ctors are time

dependent: the time for a fire to grow from ignition to

the design fire size, time for the mall to smokelog and

time needed for evacuation. C urrently there a re no

reliable da ta on fire growth rates in retail premises.

Smoke filling times are known to be rapid but are not

alwa ys easily q uantified. The time needed for

evacuation is unknown but it could be considerably

longer tha n the 2.5 minutes escape t ime used to size

exit w idths to cope with the peak flow rates during thisevacuation period. There is considerable anecdotal

evidence within the U K f ire services supporting this

view.

These uncertainties lead to ma ny prob lems in

designing a time dependent smoke ventilat ion system.

The design principles given in this R eport a re ba sed

instead o n steady sta te conditions. A design fire is used

which has a low probability of being exceeded, and the

smoke ventilation system is designed to remo ve the

mass flow rat e of smoke necessary to maintain clear

conditions in the ma ll for escape. Sprinklers in shopsare an essential part of the smoke ventilation design in

order to prevent the fire growing beyond t he design

fire size.

There are o ther essential factors tha t should be borne

in mind when designing a smoke control system. When

the ma ll has more than o ne level, it is also necessary to

restrict or channel smoke on the lower level in order to

restrict entra inment into the plume rising through the

upper level. Once the smoke has entered a ceiling

reservoir it will flow tow ards t he extraction points.

Since this horizontal flow is driven by the gases’ own

buoyancy, there will be a minimum depth for any

particular combinat ion of smoke temperature a ndmass flow ra te. This is discussed in great er deta il in

Cha pter 2, but this minimum depth will set a limit to

the design for the smoke reservoir, which cannot b e

shallower.

The capacity of the extraction system depends mainly

on the height the gases have risen to the la yer ba se,

and not at a ll on the area of the floor or the volume of

the space. Therefore a simple percentage rule fo r

natura l venting, such as 3% of the floor a rea, or a n air

change ra te, such as six air changes per hour for

powered ventilation, are misleading.

5



The following sections outline a genera l procedure

which can be followed when designing the mall smoke

contro l system. First plan the positions of the smoke

reservoirs, calculate the mass flow rate of smoke

entering the reservoir, and either the ventilation areaof a natural ventilation system or the extract ra te of a

powered system to remove the same a mount of smoke.

These procedures have been developed from ‘ idealised

models’ of shopping malls studied at the Fire Research

Sta tion. Commonly encountered variations are

discussed further in C hapter 4.

Smoke reservoirsNo smoke layer has a perfectly defined interface with

the colder, clearer air below; t here is always a small

amo unt o f cro ss mixing. This cross mixing has no

significant effect o n the q uantity o f smoke in a

horizonta l ceiling layer but can ca use progressive loss

of visibility in the air benea th. This loss of visibility

occurs more rapidly when:

• a ceiling layer is cool (indeed very coo l gases will

not persist as a layer)

• the hot smoke and the a ir beneath have a

high relative velocity

• turbulence disturbs the interface

H eat is lost from the smoke layer by radia tion

downwards and by radiation and conduction to the

surfaces of the reservoir. To prevent excessive heat loss

the size of an individual smoke reservoir within the

mall is usually taken to b e limited to 1000 m2 (or 1300

m2 if mechanica l extraction is used a s its efficiency is

less susceptible to heat loss). This recommenda tion

evolved d uring the ea rly 1970s and is perhaps best

understood by no ting that in many ‘industrial’ simple

undivided compartment applications of smoke

ventilat ion, it has long been the pract ice to limit

reservoir areas to betw een 2000 m2

and 3000 m2

.

In a mall, smoke from shops of up to 1000 m2 area (if

the ma ll is natura lly ventilated) or up to 1300 m2 (if the

mall has extract fans) can enter the mall reservoir,

giving a to tal a rea a ffected by smoke similar to t he

long-stand ing practice stated ab ove. The maximum

distance between screens forming the boundary of t he

reservoir should be 60 m. This distance

recommendat ion follows ea rlier considerations1

(confirmed in this form by the Liaison Pa nel consulted

in prepara tion of t he present document), and d erives

from concern over the distance people should be

expected to wa lk below a smoke layer while escaping.

For complex geometries of smoke reservoirs specialist

advice should be sought.

The smoke reservoir can of ten be fo rmed by t he

downsta nd fa scia of t he shops, combined with screens

at intervals along the mall (Figure 4). The fa scias a re

then necessary, not t o prevent the smoke flow ing into

the ma ll from a fire in the shop, but to prevent smokecontained in the mall reservoir flowing into the shops

adjoining the mall. If there a re raised rooflights abo ve

the ma lls, these can of ten be utilised as smoke

reservoirs.

Although the height through which the smoke may

have to rise to a ventilato r installed at the top of a high

rooflight ma y be grea ter tha n for lower ma lls, provided

that the smoke is allowed to build down to the same

level, the mass flow ra te of smoke entering the

reservoir is no greater, but the deeper reservoir of

smoke will produce a great er buoya ncy pressure a t theceiling.

The screens which form the smoke reservoir should be

arra nged so that smoke from a fire in any shop can

only flow into one reservoir.

E ven with a smoke reservoir limited t o 1000 m2,

excessive cooling and/or dow nwa rd mixing can occur

in stagna nt regions of the smoke layer. To prevent this

the extraction outlets, be they na tural or mechanical,

should be distributed over the reservoir so a s to

prevent sta gnant regions being formed. Pot entially

stagnant regions of the smoke reservoir can sometimes

be a voided by using smoke transfer ducts (see Chapter 4).

Design fire sizeB efore any smoke ventilation system can be designed

it is essential to d etermine a suita ble size of fire, for

design purposes. This fire size then forms t he ba sis of a

smoke ventilation system design.

Chapter 2Design procedures for the mall smoke control system

6

Figure 4 Smoke reservoir in mall with a ceiling heightgreater tha n that in shops (section a cross mall)



Ideally the d esign fire should show t he physical size

and heat o utput of the fire increasing with time,

allowing the growing threat to o ccupants to be

calculated as time increases. U nfortunately there is no

ava ilable research, at the t ime of preparing this

R eport, which allows assessment of the proba bility

distribution describing the va riation of fire growth

curves for retail areas. Clearly, one does not wa nt an

‘avera ge’ fire for safety design, since typically half o f

all fires would gro w fa ster. It is much simpler to a ssessthe ma ximum size a fire can reasonab ly be expected to

reach during the escape period, a nd to design the

system to cope with tha t. Note tha t even here, the

stat istical evidence is not strong (see for exa mple

Morgan and C handler13) for shopping malls. Further

research is currently in hand t o improve this sta tistical

basis.

It follows from the foregoing that there is a strongly

subjective element in a ssessing wha t fire size is

acceptably infrequent for safety design purposes.

A 12 m perimeter (3 m × 3 m) 5 MW sprinklercontrolled fire has become the accepted basis in the

U K for a smoke ventilation system in a sprinklered

shopping centre1,4,8.

Some fa ctors a ffecting the design fire size are discussed

in Cha pter 4. The design principles outlined in this

R eport a re ba sed upon a 12 m perimeter 5 MW fire.

Should a different design fire be considered for

whatever reasons, the equations, figures etc given in

this Report may no longer apply a nd ad vice should be

sought from experts. Ot her fire sizes have occasiona lly

been specified by designers, for both sprinklered and

unsprinklered shops. The pro blem o f unsprinkleredshops has been discussed in more detail by G ardner14,

who has shown t he importance of considering

‘flashover’ in such units and the consequent need to

consider potentially very large fires.

Single-storey mallsThe minimum height of t he smoke layer base must be

2.5 m from the ma ll floor to ensure saf ety, and

prefera bly at least 3 m in a single-storey ma ll (Figure 5).

It ma y be necessary to ra ise the smoke layer ba se

ab ove this for pra ctical reasons. This will result in an

increased mass flow rat e of smoke due to the

add itional entrainment, but as the smoke layer base is

higher, a greater degree of safety w ill have been

achieved. Shop fa scias should extend below this height

otherwise the layer ba se would be low enough to enter

unaffected shops. Indeed, fascias serve no other useful

smoke controlling function, except as part of a

separa te smoke extraction system within a shop, or tocontain smoke from a very small fire.

H aving established the clear lay er height in the mall,

the mass flow rate of smoke can then be ca lculated.

R ecent work by Hinkley15 has confirmed the rat e of

entrainment of a ir into a plume of smoke rising ab ove

a fire as:

M = 0.19 P Y 3/2 (1)

where M = mass flow ra te o f smoke enter ing the

smoke la yer w ithin th e shop (kg/s)P = the perimeter of the fire (m)

Y = the height f rom the base of the f ire to

the smoke layer (m)

There is no information available to show how

E quation 1 (or any current alternatives) should be

modified to allow for the eff ects of sprinkler-spray

interactions. Co nsequent ly it is used here unmodified.

E xperiments have shown16 that the smoke flowing

from the shop onto the mall becomes turbulent with

increasing mixing of a ir. The mass flow ra te of smoke

entering the reservoir is approximately double theamount given by Eq uation 1, where Y is now the

height from t he fire to the ba se of the mall smoke layer

(Figure 5). Figure 6 shows the mass flow rate of smoky

gases entering the layer at different layer heights in a

single-storey ma ll. This can be ca lculat ed fro m:

M = 0.38 P Y 3/2 (2)

The Fire Resea rch Sta tion is currently studying

entrainment into smoke flow from compartments. The

purpose of this work is to determine more accurately

the influence of such factors as compartment openinggeometry, the presence of a downsta nd fa scia a nd

ba lcony/do wnstand combinations. It follows that

E quation 2 may be superseded in due course.

If the clear layer in the ma ll is much higher than the

shopfront o pening, then the plume of smoke rising to

the smoke layer w ill be similar t o a t wo -storey

shopping centre (Figure 7). Where this height

difference (hr) between the shopfront opening and the

smoke layer ba se is more tha n 2 m (Figure 7) the mass

flow ra te should be o bta ined using the procedure for

tw o-storey shopping centres given later in this chapter

on page 10.

7

Figure 5 Smoke ventilation in a single-storey mall



Two-storey malls with small voidsWhen a fire occurs on the lower level of a t wo -storey

mall, a plume of smoke has further to rise before

entering the smoke reservoir. This results in grea ter

entrainment of a ir, hence a larger q uantity of smoke.

The smoke layer on the upper level will then be cooler

and less well defined. In this case the smoke layer

should be not less than 3 m ab ove the upper floor level

and preferably more tha n 3.5 m. Again, shop fasciasshould extend below this height otherwise the layer

base wo uld be low enough to enter una ffected stores.

In a shopping centre w hich has small voids connecting

levels, smoke from a fire in a low er level shop will flow

out of the shop and spread in a complicated horizontal

circulation pattern beneath the ceiling (ie beneath the

upper deck) (Figure 8). Where smoke rea ches the edge

of a void linking the tw o levels, some will flow o ver the

edge producing an extensive plume ab ove each void,

rising thro ugh the upper level.

Air mixes into these plumes, resulting in extremelylarge q uant ities of very cool gas collecting in the upper

level ceiling reservoir. This in turn reduces the

efficiency of buoyancy-driven venting as well as

increasing dow nwa rd mixing from the ceiling layer. To

minimise this mixing of a ir into t he plume, smoke

screens of at least 1.5 m depth17 (actuated by smoke

detectors or a s permanent features) should be hung

below the low er level ceiling (ie below the upper deck)

in order to restrict the lateral spread of smoke and

ensure that a ll the smoke from a fire passes through

only one void (Figures 9 and 10). The appro ach

outlined in this section w ill only apply when the void issmall enough in relat ion to the ma ll, and the screens

are a rranged so that smoke can flow freely into the

void a round its perimeter in a ‘swirling’ pattern

(Figure 9).

The perimeter o f the rising plume, which stro ngly

affects the rat e of the mixing of a ir into t he plume,

then depends only on the size of the void. Experiments

with scale models17 suggest that the largest void

consistent w ith this type of smoke control system

wo uld have a perimeter of between 35 and 45m.

There are no a dequa te theories to relate the q uantities

of smoke entering the reservoirs, to the height from

the upper floor to the smoke base. Figure 11 shows the

values of the mass flow rate (M) entering the reservoir

for d ifferent heights of ceiling reservoir smoke ba se

above the upper floor, and for three different sizes of

void (indicated by th eir respective perimeters), ba sed

on t he voids used in the scale-model experiments.

Other void perimeters must be inferred by

interpolation.

Figure 11 applies to a 5 MW fire occurring in a shop o n

the low er levels, and is derived from experiments on aone-tenth scale mod el17 representing a ma ll of 5 m

floor-to-floor height. Note tha t extra polating too far

8

Figure 6 R ate of production of hot smoky gases ina single-storey mall from a 5MW fire

Figure 7 Additiona l entrainment with the smoke basewell above the shop front opening



9

Figure 8 Plan view of smoke circulation pat tern below upper deck.Without smoke-restraining screens

Figure 9 Plan view of smoke circulation pat tern below upper deck.With smoke-restra ining screens

Figure 10 Smoke and a ir movements in two-storey ma ll with smallconnecting voids



will risk an increasing (but unknow n) error, since these

curves are purely empirical.

The fire on the upper level can be t reat ed for designpurposes as a fire in a single-storey ma ll. If the smoke

ventilat ion system will cope with the smoke from a fire

on the low er level, it should usually be a ble to cope

with fires on the upper level. Note, how ever, tha t

upper level fires can result in higher tempera ture

smoke in the reservoir.

Two-storey malls with large voidsWhen a shopping centre has large voids through the

mall connecting levels, the smoke flow is somew hat

different from tha t given earlier for small voids. Anidealised model of a two-storey mall with large voids is

shown in Figure 12, where the upper pedestrian

wa lkways take the form of a balcony on either side of a

large void.

Aga in, the ‘worst ca se’ is when a fire occurs in a shop

on the lower level. Smoke will flow from the shopfront

to the underside of the balcony. It will then flow

forwa rd to the ba lcony edge, as well as sidewa ys

beneath the balcony (Figure 12). B ecause of this

sidewa ys flow beneath the ba lcony, there may be a n

extensive line plume flowing upwa rds from the balcony

edge like an ‘inverted wa terfall’.

U nless the length of this line plume flowing over the

edge of the ba lcony is limited, the extensive

entrainment of a ir that takes place may result in very

large q uantities of cool smoke.

Screens installed beneath the ba lcony to channel thesmoke from the shopfront to the edge of the b alcony

will restrict lateral spread of smoke (Figure 13),

thereby producing a mo re compact line plume. This

results in less entrainment of air and therefore a more

mana geable q uantity of hot ter smoke. These screens

can be permanent structures or, alterna tively, screens

activated a utomatically on smoke detection.

Recent, as yet unpublished, research18 suggests tha t

channelling screens may be unnecessary if the ba lcony

projects no more than 1.5 m beyon d the shop front .

The minimum depth req uired for a pa ir of these

screens to cha nnel all the smoke is dependent on theirsepara tion at the void edge (L). Some values for a

5 MW fire in a t ypical ma ll are given in Table 1a for a

balcony with a d ownstand a t the void edge, which is

deep in relation to the a pproaching layer, and Table 1b

for a ba lcony without a downsta nd at the void edge

(based on Morgan19). If sidew ay s spillage occurs in

quant ity, the visibility on t he upper level can be

significantly worse.

The mass flow ra te entering the reservoir can, in

principle, be related to the height from the b alcony to

the reservoir layer b ase a nd to the plume length (ie to

the screen separat ion) by the theory of M organ a nd

Marshall20. U nfortunately, this theory applies to a

plume rising through free space, w here the a ir outside

the plume is uniformly a t a mbient temperature.

10

Figure 11 Mass of smoky gas entering ceiling reservoir per secondfrom a 5 MW fire — small voids



The smoke layer in a mall ceiling reservoir do es not

have a well defined base (especially in a two-storey

mall where there might be a deep layer of cool smoke).

E ven below the nominal layer base (ie d1 below the

ceiling, Figure 13, which corresponds closely to thevisible layer base17), there is a tempera ture excess

relative to the temperature of t he incoming air w hich

increases with height (Figure 14).

To a pply the theo ry20 to a calculation of the ma ss flow

ra tes of smoke entering the reservoir, one must

introduce a correction factor for t he smoke layer d epth

in a reservoir. Experiments with flat roofed models21

have shown that for ca lculating plume entrainment, the

effective layer depth (d2) is 1.26 times the nominal and

visible layer d epth (d1) which has been chosen for

reasons of visibility and safety. In practice many

shopping malls have roofs which are not flat, a nd it is

necessary to a ssess the result of this on the effect ive

layer depth. This is examined more closely in

Cha pter 4.

R esults from such calculations for a number of va lues

of channelling screen separation, are shown graphically

in Figure 15. These results include an a llowance for

entra inment of a ir into the ends of the plume. This

correction method follows that of Mo rgan andMarshall21, which supersedes their ear lier approa ch20.

Once the height of the nominal layer base (h-d1) has

been chosen on safety grounds, and the channelling

screens separation L (and hence also channelling

screen depth using Tables 1a or 1b) has been chosen on

practicability grounds, (eg such that the screens contact

the wa lls separa ting the shop) then Figure 15 can be

used to find the mass flow rat e of smoke entering the

reservoir.

It should be noted tha t once the height of the lay er

base (h-d1) has been selected, d1 can be used instead of

d2 for grea ter simplicity in interpreting Figure 15 and

in designing the consequent smoke extract ion. This

results in an ‘overd esign’ of the extract ca pacity, which

11

Figure 12 Smoke spread beneath a balcony producinga long line plume



thus errs on the side of saf ety. In experiments values of

(h-d2) as low as a full-scale equivalent 0.75 m were

obta ined. For a very deep layer, one sometimes finds

that (h-d2

) can somet imes be negat ive. This

corresponds to a plume moving downwa rds which is

impossible in this context and shows that the method

breaks down under these conditions. It follows that

wherever (h-d2) is less than 0.75 m, it w ould be safe t o

use (h-d1) instead for estimating entrainment.

The results given in Figure 15 are representative of

typical shop/ba lcony geomet ries. In practice the

shopfront geometry, presence or absence of a deep

downsta nd fascia and a ba lcony will affect the mass

flow ra te of smoke. For example, many malls will have

the upper walkway set back ab ove the shop units on

the storey below w ith no ba lcony projecting beyondthe low er shop front. In such designs the shop wa lls

themselves act as channelling screens. Where such a

shop has a dow nstand fa scia, the plume’s rise (h-d2)

should be measured from the bott om of the

downstand. Figure 15 can a gain be used to estimate the

entra inment into the plume, but a more precise

calculation fo r this case is feasible22.

Similarly, results given in Figure 15 are for a line plume

rising through a n upper level where air is entrained o n

bot h sides. If the plume is rising aga inst a vertical

surface (such as a wall or shopfront on the level

abo ve), then air will only be entra ined into o ne side.

R ecent research work22,23 has enabled a more deta iled

ana lysis of the fire compartment conditions and

subsequent plume entrainment to b e carried out taking

into a ccount these fa ctors. This fire engineering

approach is of necessity more complicated and needs

individual consideration.

An a lternative method of calculating the entrainment

into the line plume is due to Thomas24. This treat s the

plume in a 'far plume' approximation apparently rising

12

Figure 13 The use of smo ke-channe lling screens to produce acompact rising plume



from a line source of zero thickness some distance

below t he void edge. The relevant fo rmula is:

where M = mass flow of smoky gases passing height

z (kg s-1)

ρ = density of warm gases at height z (kg m-3)

Q = hea t flux in ga ses (kW)

L = length of void edge past which gases spill

(m)

C p = specific heat of air (kJ kg-1 K -1)

T1 = absolute ambient temperature (K)∆ = empirical height of virtual source below

void edge (m)

z = height above void edge (m)

Morga n’s re-ana lysis25 of L aw ’s earlier paper26,

concluding that the effective depth d 2 of the reservoir

layer should b e used when the plume rises within a

closed space such as a ma ll, should apply eq ually to

Thoma s’s work. This means tha t z should be ta ken to

be (h-d2), as earlier in this section. Aga in from

Morgan25, one can take ∆ = 0.3 times the height of the

compartment (ie shop unit) opening, although

comparisons between E quation 3 and t he modified

form22 of Morga n and M arshall’s method20 suggest that

the value of ∆ may be sensitive to t he parameters of

the horizontal flow a pproaching the void. In this

context it is noteworthy tha t La w 26 derived a value of

∆ = 0.67 times the height of the compa rtment ba sed on

different experimental dat a.

It should be realised that the derivation of E quation 3

limits its application to scenarios where smoky gases

issue directly from the compa rtment o n fire, with abalcony projecting beyond. For two-storey malls

E quation 3 and Figure 15 should give broadly similar

13

Table 1a Minimum depth of channelling screens— downstand at void edge

Note: The minimum depth is that below the

lowest tra nsverse obstacle

Screen separation Minimum screen

at edge L depth(m) (m)

4 2.7

6 2.4

8 2.2

10 2.1

14 1.8

Table 1 b Minimum depth of channelling screens— no downstand at void edge

Note: The minimum depth is that below the

lowest tra nsverse obsta cle

Screen separation Minimum screen

at edge L depth(m) (m)

4 1.6

6 1.4

8 1.3

10 1.2

14 1.1

gQ L2 1/3

(z + ∆) (3)M = 0.58ρ ρC pT1

0.22 (z + 2∆) 2/3

× 1 + L



results, but under some circumsta nces significant

discrepancies can occur because of the apparentvariability of ∆. Nevert heless, E quation 3 can be a very

useful way of estimating entrainment for geometries

departing significantly from Figure 15.

Multi-storey mallsFrom the previous section it can be seen that w hen a

fire occurs in a ground floor shop and rises through a n

upper level, a very large quantity of smoke is

produced. I f this is extended to a t hree-storey ma ll the

result is an impracticably la rge qua ntity of very cool

smoke. Therefore it is not usually possible to design a

practical smoke ventilation system which allows smokefrom more t han t he top tw o levels of a multi-storey

mall to rise up through the mall and maintain a clear

layer for escape o n the upper level. The limiting fa ctor

here is not the height to the to p of the smoke reservoir,

but the height of rise to the smoke layer ba se.

A multi-storey mall can instead be treat ed a s a sta ck of

single-storey ma lls, with each level having a separa te

smoke ventilation system. Clearly this technique can

also be used in a tw o-storey ma ll if so desired. Figures

16 to 18 illustrate in schematic form a mall whose

upper floor (tw o levels only are show n in the figures) is

penetrated by voids which leave a considerable a rea

for pedestrians. On the lower level there is a large area

situated below this upper deck. If screens (activated by

smoke detectors or as permanent fea tures) are hung

down from the void edges, the region below the upper

deck can be turned into a ceiling reservoir similar tothat of a single-storey mall, albeit a more complicated

geometry.

This reservoir can t hen be provided with its ow n

smoke extraction system. Other screens can be

positioned a cross the mall to limit the size of this

reservoir, a s for a single-storey mall.

The screens around the voids w ill, in general, be fa irly

close to potentia l fire compar tments (ie shops). Being

close, smoke issuing from such a compart ment w ill

deepen locally on meet ing a tra nsverse barrier. The

depth of these screens should take into account local

deepening27 — see page 18. Smoke removed from

these lower level reservoirs should usually be ducted to

outside the building but can be ducted into the ceiling

reservoir of the top floor ( Figure 18). The mass flow

rate o f smoke exhaust for the top floor can b e

calculated as if it were a single-storey ma ll.

There w ill often be some small smoke spillages under

the void screens, which will contribute to a ‘fogging’ of

the upper levels. This can be cont rolled by permitting

some ventilation o f these upper levels to operat e, to

flush out this stray smoke. To minimise such spillagesand limit the a mount of smoke below the ceiling layer

on the a ffected level, it is desirab le to simplify the

geometry of t he ceiling reservoir where possible.The

14

Figure 14 A ty pical temperature profile for a reservoir layer



void screens tend to split the smoke flow into separate

streams within the reservoir (Figure 16).

These streams can meet further a long the reservoir, as

shown. Such opposed smoke flows produce turbulence

and downw ard mixing of smoke into the air below. It

follows that it is an ad vanta ge to keep a simple

reservoir geometry. It is also important to provide the

full ‘flushing’ clean a ir inflow below the ceiling

reservoir at the a ffected level.

A lesser amount o f ‘flushing air flow ’ is desirab le on

the top level when lower reservoirs are vented to a

common top level reservoir (Figure 18). This can easily

be provided by increasing the smoke extraction from

the top reservoir by 10% above tha t needed to remove

all the smoke arriving from belo w.

Calculating smoke temperatureThe tempera ture rise of the smoke layer, above

amb ient, is given in Table 2, for a 5 MW fire (ignoring

any further cooling) and ca n be calculated from:

where Q = hea t flow ra te (kW)

θ = temperature rise of the smoke layer

abo ve ambient (°C)

C p = specific heat capacity of the gases

(kJ /kgK )

M = mass f low rate of smoky gases (kg/s)

H igh smoke lay er tempera ture will result in intense

heat radiation causing difficulties for people escaping

15

Figure 15 Mass of smoke entering ceiling reservoir per second froma 5MW fire — large voids

θ =Q

(4)MC p



16

Figure 16 Schematic plan of multi-storey ma ll with a smoke reservoiron ea ch level

Figure 17 Schematic section of a two -storey mall with a smoke reservoiron the low er level

Figure 18 Schematic section of a two -storey mall with low er smokereservoir venting into the upper reservoir



benea th the smoke la yer in the ma ll. To reduce the

intensity of heat radia tion the smoke layer

tempera ture in the ma lls should be less tha n 200°C. In

general a higher clear lay er beneath t he smoke layer

will lead to more air being entrained into the rising

smoke plume and therefore lower smoke

temperatures. If the height of the mall is restricted,

then it may not be pra ctical to increase the clear layer

height in order to reduce the smoke layer t emperature,

in which case considera tion ma y be given to installingsprinklers into the malls specifically to reduce the

smoke layer tempera ture by sprinkler cooling. This is

discussed further in the follow ing section.

Mall sprinklersSprinklers in shops are a n essential pa rt of t he smoke

ventilation d esign in order to prevent a fire growing

beyo nd the design fire size. Sprinkler opera tion in the

malls will lead to increased hea t loss reducing the

buoya ncy of smoke, which in turn can contribute to a

progressive loss of visibility under the smoky layer.

However, gases sufficiently hot enough to set offsprinklers will remain initially as a thermally buoyant

layer under the ceiling.

When the fire occurs in a shop, opera tion of sprinklers

in the mall will not assist in controlling it. If too many

sprinklers opera ted in the ma lls sprinklers in the shops

could b ecome less effective as the a vailable wa ter

supply a pproaches its limits.

Ma lls should be sprinklered if they conta in sufficient

combustibles to support a fire larger than the design

fire size o f 5 MW, 12 m perimet er, during their

operational lifetime. Note how ever tha t sprinklersinstalled a t high level in a multi-storey ma ll are

unlikely to opera te unless the fire size reached is much

larger than this.

Sprinkler cooling can be used in the ma lls to red uce

the smoke layer temperature to below 200°C, above

which heat ra diation from t he layer is likely to impede

escape benea th.

A natura l ventilation system relies on the buoya ncy of

the smoke for extra ction, therefo re if sprinkler cooling

is underestimated, the use of unrealistically high smoke

temperatures could lead to the system being

underdesigned. C onversely a powered extra ct system,to a reasonable a pproximation, removes a fixed

volume of smoke irrespective of tempera ture.

Therefore if t he extent o f sprinkler cooling is

overestimat ed the system could be und erdesigned. The

heat lost from smoky ga ses to sprinklers in the ma ll is

currently the subject of research a lthough data suitab le

for design application is not yet ava ilable.

An a pproximate calculation a pproach can be used, as

follows:

If the smoke passing a sprinkler is hotter tha n the

sprinkler operating temperature that sprinkler willeventua lly be set off. The sprinkler spray will then cool

the smoke. If t he smoke is still hot eno ugh, the next

sprinkler will opera te, cooling the smoke f urther. A

stage will be reached w hen the smoke temperature is

insufficient to set off further sprinklers. The smoke

layer temperature can thereafter simply be assumed to

be a pproximately equa l to the sprinkler operating

temperature if natura l vents are used.

The cooling effect o f sprinklers in the ma lls can be

ignored in determining the volume extract rate

req uired for fa ns. This will err on the side of safet y.

Alternatively this cooling and the consequent

contraction of t he smoky ga ses can be a pproximately

estimated o n the ba sis of an a verage value between the

sprinkler operat ing temperature a nd the calculated

17

Table 2 Volume flow rate and temperature of gases from a 5 MW fire(ignoring cooling)

Mass flow rate Temperature of gases Volume rate of extraction

(Mass rate of extraction) above ambient (at maximum temperature)

kg/s °C m3/s

9 556 21.5

12 417 24

15 333 26

18 278 2924 208 34

30 167 39

36 139 43

50 100 55

65 77 67

80 63 79

95 53 91.5

110 45 103.5

130 38 120

150 33 136



initial smoke temperat ure, since a fire ma y occur close

to a t least one of the extract openings, whereas the

other openings may be well outside a ny proba ble zone

of opera ting sprinklers. The number of potent ial ‘hot’

and ‘cool’ intakes must b e assessed separa tely for ea ch

shopping complex when calculating the average

temperature of extra cted gases.

If the sprinkler operat ing temperature is set high

enough and is abo ve the calculated smoke layertempera ture, then sprinkler cooling in the ma lls can be

ignored.

Note that the effect o f sprinkler cooling is to reduce

the heat flux Q w ithout significantly changing the mass

flux. It follows that once a new va lue of θ has been

estimated, t he new heat flux can b e found using

E quation 4.

Note that sprink lers must not be installed close to

ventilator extract positions.

Flowing layer depthSmoke enter ing the ceiling reservoir will flow from t he

point of entry to wa rds the vent or fa ns. This flow is

driven by the buoya ncy of the smoke. Even if there is a

very large ventilation area downstream (eg if the

downstream roo f were to be removed) this flowing

layer wo uld still have a depth related to t he width of

the mall, the temperature of the smoke and the mass

flow ra te of smoke. Work by M organ19 has shown tha t

this depth can be calculated fo r unidirectional flow

under a flat ceiling, as follows:

where d1 = flowing smoke layer depth (m)

Tc = absolute temperature of the smoke

layer (K)

θ = temperature rise of the smoke layer

abo ve ambient (°C)

W = channel width (m)

γ = downstand factor 36 if deep downstand

is present at right angles to the flow;

downstand factor 78 if no downstand is

present at right angles to t he flow

M = mass flow rate of smoky gases (kg/s)

The resulting values of layer dept h for d ifferent

reservoir widths and ma ss flow rates of smoke are

shown in Table 3a . This ignores the effect s of coo ling in

the layer, therefore if sprinklers are insta lled in the

mall E quation 5 should be used after estimating the

effects of such cooling (see previous section). Ea ch

depth show n in this ta ble is the minimum possible

regardless of the smoke extraction method employed

downstream: consequently it represents the minimum

depth for that reservoir.

The depth must be measured below the low est

transverse downsta nd obstacle to the flow (eg

structural beams or ductwork) ra ther than the true

ceiling. Where such structures exist and are an

appreciable fra ction of the overall layer depth, the

depth below the obsta cle should be found using Tab le

3b rather than 3a.

It is also necessary in some malls to determine the

flowing layer depth between channelling screens

benea th ba lconies. This flowing lay er depth w ill be

altered by t he presence or absence of a downsta ndwhich is deep in comparison to th e approa ching lay er.

If there is a deep dow nstand at the balcony edge use

Table 1a .

Local deepeningWhen a buoya nt layer of hot smoke flows along

beneath the ceiling and meets a tra nsverse barrier it

deepens locally a gainst that barrier27; the kinetic

energy of the approaching layer is converted to

buoyant potential energy aga inst the ba rrier as the

gases are brought to a halt.

Table 3b Minimum reservoir depth — deep downstandacross the flow

* For b i-directional flow of smoky ga ses this should be tw ice the actual

reservoir width

Mass flow rate Width of reservoir (m)*entering reservoir

(kg/s) 4 6 8 10 12 15

10 1.8 1.4 1.2 1.0 0.9 0.8

15 2.3 1.8 1.5 1.3 1.1 1.0

20 2.8 2.2 1.8 1.5 1.4 1.2

25 3.3 2.5 2.1 1.8 1.6 1.4

30 3.8 2.9 2.4 2.1 1.8 1.6

40 4.7 3.6 3.0 2.6 2.3 2.0

50 5.7 4.3 3.6 3.1 2.7 2.4

70 7.5 5.8 4.8 4.1 3.6 3.1

90 9.4 7.2 6.0 5.1 4.5 3.9

110 11.2 8.6 7.1 6.1 5.4 4.7

130 13.1 10.0 8.2 7.1 6.3 5.4

150 15 11.5 9.5 8.2 7.2 6.2

18

Table 3a Minimum reservoir depth (m)

* For bi-directional flow o f smoky gases this should be twice the a ctual

reservoir width

Mass flow rate Width of reservoir (m)*

entering reservoir(kg/s) 4 6 8 10 12 15

10 1.1 0.8 0.7 0.6 0.5 0.5

15 1.4 1.1 0.9 0.8 0.7 0.6

20 1.7 1.3 1.1 0.9 0.8 0.7

25 2.0 1.5 1.2 1.1 1.0 0.8

30 2.3 1.7 1.4 1.2 1.1 0.9

40 2.8 2.2 1.8 1.5 1.4 1.2

50 3.4 2.6 2.1 1.8 1.6 1.470 4.5 3.4 2.8 2.4 2.2 1.9

90 5.6 4.3 3.5 3.1 2.7 2.3

110 6.7 5.1 4.2 3.6 3.3 2.8

130 7.8 6.0 4.9 4.2 3.8 3.2

150 9.0 6.8 5.6 4.9 4.3 3.7

MTc 2/3(5)

d1

= γ θ1/2W



When designing a smoke ventilat ion system for

shopping centres of more tha n one level, it is oft en

necessary to control the path of smoke flow using

dow nstand smoke curtains. These are typically

installed a round the edge o f voids to prevent smoke

flow ing up through the vo ids. If the void edge is close

to t he shop this local deepening could cause smoke to

under-spill the smoke curtain a nd flow up through the

void, possibly affecting escape from other storeys.

C learly, the void edge screens must be d eep enough tocontain not only the established lay er, but a lso t he

additional local deepening outside the shop unit on

fire.

The extent of local deepening can b e found from

Figure 19. The depth of t he established layer (d m in

Figure 19) in the mall immediately down stream o f the

local deepening must first be found using the design

procedure given in the preceding sections. U sually this

means in the channel formed between the void edge

screen and the shop front. The additional depth ∆ dw

can then be f ound by inspection of Figure 19, allowingthe necessary minimum overall depth (dm + ∆ dw ) of

the void edge screen to b e found. Alternatively it has

been shown28 that the formula given below can be used

to determine approximate values for local deepening:

where ∆ dw = additional deepening below the

established smoke layer a t t he

transverse barrier (m)

dm = established layer depth (m)

w = d ist ance bet ween the sho p f ro nt a ndthe tra nsverse barrier (m)

Note tha t extrapolating too far will risk an increasing

(but unknow n) error, since equa tion 6 is a purely

empirical fit to the da ta a nd has no theoretical

derivation. Note also that E quation 6 only applies

strictly to a 5 m high ceiling. Figure 19, on the other

hand, can rea dily be scaled.

Inlet airThere must be a dequa te replacement a ir for the

efficient operat ion of a smoke ventilation system. Theratio of inlet area to extract a rea is used in the

calculation of natural ventilation area required to

account for the effect of inlet restrictions on the

19

Figure 19 Local deepening at a transverse barrier

∆dw =2(1 – logedm)

(6)log

ew



efficiency of the system (see page 21). If doors are

required to o pen for replacement air, an appropriate

entry coef ficient should be used. Natural vents in

unaffected smoke zones can often be opened

automatically, and simultaneously with the main smoke

ventilation system extracting smoke, contribute to the

tota l air inlet required. If a powered smoke extract

system is used, smaller a reas for inlet a ir may be

sufficient.

If the a rea a vailable for inlet becomes too restricted,

incoming air flow through escape doors may b e at too

high a velocity for ea sy escape. Studies at the B uilding

R esearch Estab lishment29 have shown that winds

ab ove 5 m/s can cause discomfort to pedestrians. Such

air inflows through doors w ould hinder escapees and

could b e da ngerous since they might a lready b e

predisposed t o pa nic. It would perhaps be w ise to

design a smoke ventilation system such that air

velocit ies through doors a re less than 3 m/s.

A high relative velocity b etween the smoke layer andincoming air occurs when air is drawn in through an

inlet of limited area and the resulting air stream (or

jet) passes below t he smoke lay er immediately. Smoke

will be drawn do wn into such a jet by t he venturi

effect17, causing a significant loss of visibility in the

lower cold air regions; this can occur when doors are

used for inlet air (Figure 20). It can be minimised by

placing screens defining the end o f the reservoir a t

least 3 m back from the a ir inlet, giving the inflow a n

increased cross-section and a drop in velocity. This

measure a lso permits turbulence in the entering a ir

stream (caused by external winds) to da mp out before

it can d isturb the smoke/air interfa ce and cause

excessive loss of visibility. If the la yer base is designed

at least 2 m abo ve the top of the doo rs (or air inlets)

there is no need to set ba ck the reservoir screen17.

Any sta gnant region in the cold clearer a ir beneath the

smoke layer w ould suffer from a steady a ccumulation

of smoke. The fire and o ther ma jor entrainment sites

will act as air pumps and ca use air to flow from the

inlets tow ards themselves. The a ir inlets shouldtherefore be chosen to ensure that these flows of cold

air w ill flush through a ll areas of the malls below the

ceiling reservoir. Any smoke wisps tha t enter t he lower

clearer a ir will thus be swept back into the ma in body

of the hot smoke.

A f an-driven inlet a ir supply can give problems when

mechanica l extraction is used (the b uilding will usually

be f airly well sealed in such circumstances). This is

because the warmed air taken out will have a greater

volume than the inlet air. As the fire grows a nd

declines, the mismatch in volume betw een the inlet airand the extracted fire-wa rmed air w ill also change.

This can result in significant pressure d ifferences

appearing across any doo rs on the escape routes. For

this reason simple ‘push-pull’ systems should be

avoided.

Minimum number of extract pointsThe beginning of this chapter outlined the importa nce

of d istributing extraction points ab out a smoke

reservoir to prevent the formation o f stagna nt cooling

regions. The number of extr actio n points within the

reservoir is also importa nt since, for a ny specified layerdepth, there is a maximum rat e at which smoky gases

can enter a ny individual extract vent (be it na tural,

chimney, or mechanical). Any further attempt to

increase the extraction through that vent merely serves

to d raw air into the orifice from below the smoke layer.

This is sometimes know n a s ‘plug-holing’. It follows

that , for efficient extraction, the number of extract

points must be chosen to ensure tha t no a ir is dra wn up

in this way. Table 4, which is ba sed on experimental

work30 subsequently modified by H eselden31, lists the

minimum numbers needed fo r different reservoir

conditions and for a variety of ma ss flow rat es being

extracted from t he vents in the reservoir. Table 4strictly applies to vent s which are small compared to

the layer depth below the vents.

If ca lculat ion is preferred to using Table 4, the

following apply a t the critical point where a ir is abo ut

to be d raw n into the openings31:

At a n opening,

m = α (gd 5 T1 θ/Tc2)1/2 (7)

where m = crit ical extract rate for eff icient ventingat one vent (kg s-1)

α = 1.3 for a vent near a wall (kg m-3)

20

Figure 20a Smoke from a buoyant ceiling layer mixing into ahigh velocity air inflow

Figure 20b Mixing is reduced by allowing the incoming a ir flow to slow before conta cting the smoke layer



or α = 1.8 for a vent distant from a wa ll (kg m-3)

g = acceleration due to gravity (ms-2)

d = depth of smoke layer below the vent (m)

T1 = ambient absolute temperature (K)

θ = excess temperature of smoke layer (°C )

Tc = T1 + θ.

The req uired number of extra ct vents (N) is then given

by:

where M is the total extract rate required from the

reservoir

Where very large o r physically extensive vents are used

(eg a long inta ke grill in the side of a horizonta l duct)

an a lterna tive method is possible. For this case, Tab le

3a can be used with the ‘w idth of reservoir’ being taken

as the tota l horizontal accessible perimeter of all the

vents within the reservoir (eg the tota l length of intake

grilles in the example abo ve) and the ‘minimum

reservoir depth’ corresponds to the depth of the smokelayer beneath the top edge of the intake orifice. In

practice for a given mass flow rat e and layer depth one

can use Table 3a or E quation 5 to find the minimum

value o f a ccessible perimeter.

Intermedia te size inta kes (ie where the vent size is

comparable to the layer depth) cannot be treat ed so

simply and it is recommended that Table 4 be used

since it errs on t he side of sa fety.

Natural ventilation — area required

per reservoirA natura l ventilation system uses the buoyancy of the

smoke to provide the d riving force for extract ion. The

rate of extraction is largely dependent upon the depth

and t emperature of smoke. The adva ntage of a natura l

ventilat ion system is tha t it is very simple and reliable,

and ca n cope with a w ide range of fire conditions.

Should for any rea son the fire grow la rger than the

design fire size, a grea ter depth a nd temperature of

smoke leads to a n increased extraction rate, so to an

extent a na tural ventilation system has a self-

compensating mechanism.

Ca re must be ta ken to ensure that nat ural ventilators

are not sited in a position subject to positive wind

pressures (see Cha pter 4). If smoke ventilation is

req uired in such positions powered extr act ion must be

used instead of na tural vents. Note tha t nat ural vents

and powered extracts should never be used t ogether in

the same reservoir.

Table 5 gives the minimum aerodyna mic free area of

ventilation required, ignoring the effect of any inlet

restriction. To a llow for the effect o f limited fresh air

inlets the following guide can b e used:

If the inlet area for the whole mall is the same as the

vent area for the reservoir given by Table 5, thisindicated vent area should be increased by

appro ximately 35%.

If the to ta l inlet area is twice the reservoir vent area ,

the indicated vent a rea should be increased by 10%.

The precise relationship betw een the mass flow ra te

extracted, the vent area, the inlet area a nd the smoke

layer is12:

21

Table 4 Minimum number of extraction points needed in asmoke reservoir

Note: In reading the abo ve table, the first number is for extraction points well

aw ay fro m the wa lls, the second is for extraction points close to the wa lls

Total mass Depth of layer below extraction point (m)

rate of extraction

(kg/s) 1 1.5 2 3 4 5 7 10

9 4–5 2–2 1–1 1–1 1–1 1–1 1–1 1–1

12 5–6 2–3 1–1 1–1 1–1 1–1 1–1 1–1

15 6–8 2–3 1–2 1–1 1–1 1–1 1–1 1–1

18 7–9 3–4 2–2 1–1 1–1 1–1 1–1 1–1

20 8–10 3–4 2–2 1–1 1–1 1–1 1–1 1–125 9–13 4–5 2–3 1–1 1–1 1–1 1–1 1–1

30 11–16 4–6 2–3 1–1 1–1 1–1 1–1 1–1

40 6–8 3–4 1–2 1–1 1–1 1–1 1–1

50 8–11 4–5 2–2 1–1 1–1 1–1 1–1

70 6–8 2–3 1–2 1–1 1–1 1–1

90 3–4 2–2 1–1 1–1 1–1

110 4–5 2–3 1–2 1–1 1–1

130 3–3 2–2 1–1 1–1

150 3–4 2–3 1–1 1–1

N ≥ M(8)

m

M Tc2 + (A v C v/A iC i)

2ToTc 1/2

(9)AvC v = ρo 2g db

θc

To



where

A vC v = aerodynamic free area of natural

ventilation (m2)

A v = measured throat area of ventilators for the

reservoir being considered (m2)

A i = total area of all inlets (m2)

C v = coefficient of discharge (usually between

0.5 and 0.7)

C i = entry coefficients for inlets (typically about

0.6)M = mass f lo w ra te o f sm oke to be ext racted (kg/s)

ρ° = ambient air density (kg/m3)

g = a ccelera tion due to gra vity (m/s2)

db = depth of smoke beneath ventilator (m)

θc = temperature rise of smoke layer above

ambient (°C )

Tc = absolute temperature of smoke layer (K)

To = absolute temperature of ambient air (K)

Natura l ventilat ion can sometimes be enhanced by

using chimneys to increase the buoya nt head of hot

smoke. H ow ever, a system of chimney vents can bedifficult to design. The flow resistan ce must be taken

into a ccount. Typical values may be found in the

CIBSE G uide32, but it should be noted tha t

unpublished experiments at FR S ha ve shown the entry

coefficient to t ake on a value of a pproximately 0.7 (as

compared w ith the more usua l value of 0.5 used in

H E VAC calculations) for a sharp-edged opening in

draw ing gases from a relatively shallow buoya nt layer.

A powered extract system should be used where

positive wind pressures are likely to b e a pro blem, or

where it is necessary to extra ct smoke via an extensive

ductwork system.

Powered ventilationA powered smoke extract system consists of f ans a nd

associated ductwork designed to remove the mass flow

ra te of smoke entering the smoke reservoir, and to be

capable of withstanding the anticipated smoke

temperatures.

The controls and w iring should of course be protected,

to maintain the electrical supply to the fans during a

fire.

The mass flow ra te of smoke determined from t he

previous section can be converted to the corresponding

volume flow rate and temperature, using Table 2 (orthe following equation) for selection of the a ppropriate

fans:

where V = volume flow rate of gases to be extracted

from the smoke layer (m3/s)

22

Table 5 The minimum total vent area (m2) needed in one

ceiling reservoir (from Equation 9with Cv = 0.6)

Note: Add on 10% if the tota l inlet area in the mall is twice the vent

area.

Add o n 35% if the total a ir inlet area in the mall is equal to

the vent area

Mass rate Smoke depth beneath vents (m)

of extraction

(kg/s) 1.5 2 3 4 5 7 0

9 4.8 4.1 3.4 2.9 2.6 2.2 1.8

12 6.1 5.3 4.3 3.7 3.3 2.8 2.4

15 7.5 6.5 5.3 4.6 4.1 3.5 2.9

18 9.0 7.8 6.4 5.5 4.9 4.2 3.5

24 12.2 10.5 8.6 7.5 6.7 5.6 4.7

30 15.6 13.5 11.0 9.5 8.5 7.2 6.0

36 19.2 16.6 13.6 11.8 10.5 8.7 7.4

50 25 20 17.5 15.7 13.2 11.1

60 25 22.0 19.7 16.7 13.9

75 34 29 26 22 18.690 43 37 34 28 24

110 57 49 44 37 31

130 71 62 55 47 39

150 87 75 67 57 48

V = MTc (10)

ρoTo



As outlined in C hapter 2, the preferred option for the

majority of shops is to provide a common smoke

ventilat ion system in the malls unless the shops are

grea ter than 1000 m2 (if the mall is naturally vented) or

1300 m2 (if the mall has powered extraction). Shopslarger than these sizes need additional measures to

protect the ma ll.

A smoke layer w ithin a large shop will lose heat as

smoke spreads throughout t he store. This heat loss is

caused by the cooling action of the sprinkler spray and

heat loss by conduction and ra diation to the building

structure. Thus the cooling is related to t he size of the

shop. If there is too much heat loss, the effect in the

mall will be as if the mall reservoir itself were to o

large, since there will again be excessive downward

mixing and loss of visibility in t he ma ll. To ensure saf econditions in the mall therefore, smoke from large

stores must be prevented f rom flow ing onto it. To

achieve this the store must be either isola ted fro m the

mall or ha ve its own smoke ventilation system. Fire

shutters should only be used to isolate the store from

the ma ll if there is no other practicable choice — and if

the implicat ions for means of escape a nd the possible

psychological effect on escapees have been fully taken

into account.

This excessive cooling of smoke is thought only t o be

serious in stores over 1000 m2 in area (or 1300 m2 if

powered extraction is used in the mall). If they a re notisolated f rom the ma ll, such large stores should have

ceiling reservoirs formed by similar met hods to those

for ma lls. Since no smoke is allowed to enter the ma ll,

each reservoir can be limited t o 2000 m2 (or 2600 m2 if

powered extra ction is used) to prevent excessive

cooling, ie to the same maximum area as the combined

maxima for the ma ll plus a small shop unit. Note t hat

any area less tha n 1000 m2 ‘left over’ from t his internal

sub-division, and adjacent to the mall, can then be

ventilated via t he mall in the same way t hat a small

shop would be. The quantity of smoke entering the

ceiling reservoir within the shop is given by E quat ion1, and results for a 5 MW fire are shown gra phically in

Figure 21. H aving determined the mass flow rat e of

smoke, the design procedure given for the ma ll

ventilation system in Cha pter 2 can b e used to

determine the extract capacity req uired within the

store. It should be noted tha t, provided smoke cannot

enter the ma ll, the height of the smoke base in the

shop need not be the same a s it would be in the ma ll.

When calculating the overall smoke layer temperature

the procedure f or ma ll sprinklers given on page 17

should be used (but note t hat any individual intake or

duct might be located immediately abo ve any fire and

this should be ta ken into account w hen specifying the

temperature requirements for fans a nd ducts).

An a lternative, particularly a ppropriate where the

tota l width of openings between the store and the mall

are restricted, is to provide large capacity ‘slit’extraction33 in the ceiling over the w hole widt h of such

openings, including doors, but not including fixed glass

windows (Figure 22). Such a system is likely to w ork

best with further extraction distributed w ithin the

store, which may possibly be provided by the normal

ventilation extraction system, with the norma l

ventilation input and recirculation of air being stopped.

Whilst this system is designed to prevent coo l smoke

entering the ma ll, it will not necessar ily mainta in a

clear lay er within the sto re itself. The extract ion should

be provided very close to the opening from a

continuous slit which may b e situa ted in the plane of

23

Chapter 3Large shop opening onto a mall

Figure 21 R ate of production of hot smoky gases in astore from a 5MW fire



the fa lse ceiling. The extraction ra te (V) can be fo und

from:

w here W = to ta l w idth of stores opening (m)

Wm = width of stores opening onto the mall(m)

H = height of st ore opening (m)

Vs = volume extract rate f rom store

ventilation system (m3/s)

24

Figure 22 Slit extraction

3.45 0.74Vs (11)V =

W3.24 – Wm m3/s–1

H W2/3



Factors influencing the design fireUnspri nkl ered shops

If sprinklers are not installed in the shop, it is likely

tha t the fire w ill grow larger than t he 5 MW, 12 m

perimeter, design fire14. An a nalysis of fires in publicarea s of reta il premises13 has suggested that fewer t han

4% of fires exceed the design fire area . (Note tha t the

‘sprinklered’ curve in Figure 23 is drawn to conform to

Morgan a nd Cha ndler’s13 interpretation of the

statistical da ta a nd to have a pproximately the same

form a s the 'unsprinklered' curve; it can be seen that

the lat ter is far more reliab le). The same stat istical

probability of being exceeded when sprinklers are not

present gives a fire size in excess of 100 m2 (Figure 23)

— potent ially 10 times larger.

This could lead t o ‘fla shover’, when the fire w illdevelop very ra pidly to involve the entire shop.

E vidence from a ctual fires suggests that flashover can

happen quickly34.

A fully involved shop fire in a ‘t ypical’ shop unit may

not ha ve a large enough opening onto the mall for

sufficient oxygen to be available to support complete

combustion w ithin the shop. This will result in unburnt

gases flowing from t he shop, mixing with air and

further combustion taking place in the mall itself,

resulting in considerable dow nwa rd heat radia tion.

U nless the ma ll roof is high it is unlikely to be pra ctical

to design a smoke ventilat ion system that will ensure

safe escape, for people in normal clothing, below t he

mall ceiling smoke reservoir.

C learly if sprinklers are no t installed in the shops a

5 MW design fire size is inappropriat e, and the

formula, equa tions and much of the guidance given in

this R eport a re invalid. This is discussed in more deta il

in a paper by G ardner14.

M ult i -occupancy mal ls

As shopping centres are becoming increasingly multi-

use developments, it is importa nt to consider thepotent ial fire risk in occupancies other tha n shops, if

they ad join the mall without any fire separation.

Info rmation on fire sizes in some other classes of

occupancies has been developed during research into

smoke control in atr ium buildings35,36.

It is importa nt to assess the likely fire size in any o f

these occupancies, and a check should be made to test

whether a system designed on the ba sis of a 12 m

perimeter 5 MW fire will be ad equa te to deal w ith a

fire from any a lternative occupancy opening onto tha t

section of the ma ll. The mall smoke ventilat ion system

should be designed to cope with the most severeconsequences of any of the selected fire sizes in any o f

these occupancies.

The effect of wind on the efficiency of a smoke ventilation systemWhen natural ventilators are used for smoke

extraction, it is important that they a re positioned

where they will not be adversely affected by external

wind cond itions. A positive wind pressure can be much

greater tha n the pressure head developed by a smoke

layer. Should this occur the ventilator ma y a ct as a n

inlet rather tha n as an extra ct. How ever, if sited in an

area of negative wind pressure, the resultant suction

force on a natura l ventilator w ould assist smoke

extractions.

Tall buildings or ta ller area s of the same building (such

as roof top plant rooms etc) can create a positive wind

pressure on lower nearby roofs. Steeply pitched roofs,

ie those with o ver 30° pitch, may also have a positivewind pressure on the w indward slope.

A suggestion sometimes advanced for offsetting wind

over-pressures is to increase the total area of natural

ventilat ion per reservoir. Since the over-pressure is, by

definition, force per unit area , this will usually not

work a nd indeed could exacerbate the problem by

allowing even greater q uantities of a ir to be driven

through the vent to mix into the smoke. A powered

smoke extraction system should be designed to

overcome the anticipated wind pressures.

In some cases it ma y be possible to reta in natural

ventilation openings in a vertical plane by arra nging

them to fa ce inwards to either a region sheltered from

wind action, or where the wind w ill alwa ys produce a

suction. In other cases the erection of suitably designed

screens or wind baffles (outside the vertical wall or

window holding the vents) can o vercome wind

interference and ma y even be a ble to convert a n over-

pressure into a suction. There is also t he possibility o f

selectively opening vents in response to signals from a

wind direction sensor. E xpert advice should be sought

for such designs.

D ue to the complexity of w ind induced air flow over a

shopping centre and surround ing buildings, it may

sometimes be desirable to carry o ut boundary layer

wind t unnel studies to esta blish the wind pressure over

the b uilding’s envelope. Once areas o f o ver-pressure

and suction have been identified fo r all possible wind

directions, design of vents or fans can proceed as

before.

False ceilings in the mallWhere there is an unbro ken fa lse ceiling in the mall it

must be trea ted a s the top of the smoke layer. If the

fa lse ceiling is porous to smoke, ie if it has a n

appreciable free area, any smoke screens forming the

25

Chapter 4Some practical design considerations



smoke reservoir must be continued above t he ceiling.

If the proportion of free a rea is large enough the

reservoir and its screens may be to ta lly a bove the fa lse

ceiling. The permeab le ceiling ought no t to interfere

appreciably w ith the flow of smoke from the fire to the

smoke ventilation openings above the fa lse ceiling.

It has been shown experimentally37

that a minimumfree area of 25% can be used as a ‘rule of thumb’ value

for a llowing safe escape. Coo l smoke can sometimes be

expected to a ffect nearby shops but w ould not

significantly hinder sa fe escape. Free area s of less than

25% are possible in some circumsta nces; expert advice

should be sought where t his possibility is felt desirable.

The use of a plenum chamber above afalse ceiling in a shopSome designs have been seen in which the space above

a ma inly solid fa lse ceiling in a la rge store is used fo r

the extraction of air for normal ventilation purposes. Afan extracting air from this space (effectively a plenum

chamber) reduces its pressure and so draws air from

the store through a number of openings in the false

ceiling. In the event of a fire a fa n of suitably larger

capacity starts up and dra ws smoky gases into the

chamber in a similar wa y. This system can o f course

also be used for ma lls.

A potentially valuable bonus of such a system is that

the sprinklers which are norma lly required in the space

ab ove the fa lse ceiling will cool the smoky ga ses befo rethey reach the fan.

The plenum chamber should not be larger in area than

its associated reservoir. Larger chambers should be

subdivided by smoke screens tha t a re the full height of

the chamber and w hich extend dow nwa rds to form a

complete smoke reservoir below the fa lse ceiling. The

minimum number of o penings through the f alse ceiling

required within a single subdivision can be found from

Table 4. The tot al a rea o f such openings per reservoir

should be decided by consideration of the design

pressure differences between chamber and smoke

layer, and of t he flow impedance of the openingsconcerned. A system of reasonably w ide (perhaps one

or two metres) slots surrounding a region of false

ceiling could perhaps be used instead of screens below

26

Figure 23 R etail premises — horizontal firedamaged area



the f alse ceiling.

It can be d esirable to leave the false ceiling below the

extraction points ‘solid’ (ie not a ble to pass smoke) to

prevent air being drawn up through the smoke layer. A

sufficiently extensive area of ‘solid’ fa lse ceiling will

ensure tha t the smoke passes through at least one

sprinkler spray en route to the extract.

The principles described in this section could be usedwith extraction through a shaft vent (or chimney) from

the space above the false ceiling provided that

sprinklers w ere not installed in this space, since they

would rob the gases of buoya ncy.

Stores with internal voidsWhen a store o f more tha n one level, with internal

voids, is open to a multi-level mall, smoke flow o nto

the ma ll at more tha n one level simultaneously should

be avoided. In some circumstances this may mean

isolating the store fro m the mall on one or mo re levels.

This can b e achieved in a number of w ay s, includingthe use of fire shutters on detection of smoke to isola te

the store. Whenever these are used a number of fa ctors

need to b e assessed, such as the implications for means

of escape and the psychological effect on people of fire

shutters operating.

Sloping mallsIf t he mall floor slopes significant ly, a 3 m clear layer a t

one end o f the smoke reservoir may be considerably

less at the ot her end. The smoke ventilation system

should be designed to ma intain an a cceptably clear

layer a t the high end for a fire in a shop at t he low end,resulting in a grea ter height of rise to the smoke layer

base and ma ss flow rat e of smoke.

There may b e a diff erence in floor level betw een the

mall and the shops either side. If so, the height o f rise

to t he smoke layer b ase should be mea sured from t he

lowest shop level and the clear la yer in the malls

should be measured fro m the highest mall level.

Assessment of effective layer depthAs explained in C hapter 2 the effective layer depth

introduces a correction in the procedure fordetermining entrainment, to account for t he layer of

warmed air beneath the visible smoke layer base. The

experimental w ork wa s carried out on a scale model

with a fla t roof a s shown in Figure 13. The depth o f

smoke beneath b oth the ventilator and the ceiling wa s

found to be the same.

In practice pitched roofs, pyramids, domes or

roof lights are commonly used a s smoke reservoirs. No

experimental data exist on these other reservoir forms

and using the depth beneath t he apex of a pitched roof,

for example, will increase the effective layer depth,

possibly resulting in an underestima te of the mass flow

rate of smoke (Figures 24a and 24b). Therefore it is

desirable to determine an equivalent flat ro of position

for a ny given reservoir. U ntil more reliab le

information becomes ava ilable, it seems reasonable to

assume that the equivalent flat roof is one where the

layer ba se is at the same height and the lay er has the

same cross-sectional a rea. The effective lay er depth

can then be d etermined with d 1, now the depth of

smoke beneath the eq uivalent flat roof position.

Smoke flow in low narrow mallsThe problems of smoke flow in low na rrow m alls are

more often encountered when dealing with the

refurbishment o f older existing shopping centres.

Table 3 gives the depth o f flowing layer in malls of

different widths and varying mass flow rates. Narrow

malls less tha n 5 metres floor-to-floor or with d eep

beams across the mall beneath which the smoke must

flow, could have insufficient clear height for escape a nd

correspondingly high smoke t emperat ures. These high

smoke layer temperatures can be reduced by installing

sprinklers in the ma lls, specifically to cool do wn t he

smoke layer.

The smoke layer ba se cannot be ra ised by increasing

the rate of extraction unless there is sufficient floo r-to-

ceiling height to a chieve this. From Table 3 it can be

seen that if t he smoke flows in two directions in the

smoke reservoir, the flow ing layer depth is smaller tha n

for a single direction flow.

This can be a chieved by ensuring tha t 50% of the

extract capacity is installed at either end of the

reservoir such tha t bi-directiona l flow will occur. Aga in

from Table 3 it can be seen that if the mall is made

wider the flow ing layer depth is reduced. O ne methodof achieving this without moving the shop fronts back,

which has been used in practice, is shown in Figure 25.

The mall is widened a t high level witho ut any change

at lower levels by ‘stepping back’ the shop fronts. For

aesthetic reasons this could be abo ve a permeable false

ceiling, allowing the flowing layer to exist wholly above

the f alse ceiling line.

Basement service levelsWhere regulato ry a uthorities feel that a smoke

ventilation system is required for the basement service

level of a shopping centre, there ha s previously beenno published guida nce.

A fire size of 7 MW, 15 m perimeter cor responding to a

fully burning tractor (of a n a rticulated vehicle) or a

partly burning trailer or a va n, has become widely used

as a maximum practical fire size for smoke ventilation

design. Note however that larger fires are possible for

large d elivery vehicles or t railers. The design

procedure is similar to tha t for a large store given in

Cha pter 3.

Figure 26 gives mass flow rat es for varying clear layer

heights, for the ba sement service level design fire size.The temperatures and volume extract ra tes for fa ns

can then be d etermined from Table 6. If na tural

27



28

Figure 24a Assessment of effective layer depth from a pex ofpitched roof

Figure 24b Assessment of effective layer depth from the equivalentflat roof position

Figure 25 R educing the flowing smoke layer depth by w ideningthe mall at high level



ventilation is used the a erodynamic free area required

can then be found from E quation 9.

Enclosed car parksShould regulato ry a uthorities require a smoke

ventilation system for a n enclosed car park, there has

previously been no published guida nce. This section

outlines an a pproach that has been used for some time

by the Fire Research Station.

A design f ire size of 2.5 MW, 12 m perimet er

corresponding to a single burning car has become

widely used a s the design fire size based o n a

recalculation of experimental results38. Ho wever,

concern has been expressed about t he increasing use of

plastics in car bodies, and of plastic petrol tanks and

the implications that these would have on the design

fire size. Further research is needed to eva luate thepract ical significance of such concern. Where the

appro ving autho rities feel tha t such aspects might

significantly influence their opinion of any ind ividual

building design, it can b e noted t hat a fire in one car is

unlikely to spread to a neighbouring parked ca r, even

one with a ‘plastic’ body shell, when sprinklers have

been fitt ed in the ca r park. This comment should not

be ta ken as a general recommendation tha t sprinklers

are regarded as essential in all enclosed car parks.

The design procedure is similar to that for large stores

given in Cha pter 3 since the perimeter is the same a s

for a shop fire. Figure 21 can be used to ob tain themass flow rat e for va rying clear lay er heights.

H ow ever, since the hea t f lux is 2.5 MW the

temperature and volume extract ra tes must be

determined f rom Table 7. If na tural ventilation is to be

used the a erodynamic free area ca n be determined

from E quation 9.

Smoke transfer ductsStagnant regions of a smoke reservoir will suffer from

continued hea t loss resulting in downward mixing into

the a ir below a s discussed in C hapter 2. Go o d

distribution of ventilator extract positions can preventthis being significant. Where th is solution is

impracticable smoke transfer ducts can be installed to

move smoke from the stagnant region to a nother part

29

Table 6 Volume flow rate and temperature of smoke from a 7 MW 15 m perimeterfire (ignoring cooling)

Temperature of Volume extract

Mass flow smoke above ambient ratekg/s °C m3/s

9 778 28

12 583 30

15 467 33

18 389 35

24 292 40

30 233 45

40 175 53

50 140 62

70 100 79

Table 7 Volume flow rate and temperature of smoke from a 2.5 MW 12 m perimeterfire (ignoring cooling)

Temperature of Volume extractMass flow smoke above ambient rate

kg/s °C m3/s

9 278 15

12 208 17

15 167 20

18 139 22

24 104 27

30 83 3240 63 41

50 50 49

70 36 66

Figure 26 Mass flow ra te of smoke entering the ceilingreservoir from a 7MW fire, 15m perimeter



of the smoke reservoir to rise with a n existing flow

towards a vent or extract fan (Figure 27).

As a ‘ rule of thumb’, if the reservoir continues more

than 3 times as deep beyond a n extract opening as the

reservoir is wide, then a smoke tra nsfer duct ma y benecessary. A currently recommended value of the

minimum extract ra te is 4% of the smoke la yer’s net

flow o r 1 m3/s, whichever is the greater.

Entrances within the smoke layerThere have been shopping centres designed with

entrances within the smoke layer (eg from car parks

above the shopping levels). Any pedestrian area inside

these entrances and open to the ma ll needs to be small

enough to be evacuated quickly through smoke sealed

doors.

Other situationsThe possibilities for special fea tures are b y no mea ns

exhausted b y t he a bove. Whenever another special

case occurs that is not explicitly covered by this

Report, advice should be sought from experts.

30

Figure 27 U se of smoke transfer ducts in otherwise stagna ntregions



All reservoir screens, smoke channelling screens and

natura l vents that are not permanent features should

operate a utomatically, actuated b y smoke detectors, as

should mechanical extract systems. Other types of

detector a re less useful since, at b est, they have longerresponse times and ma ny (eg fusible links) may no t

operate a t a ll. It is much easier to prevent smoke

spreading to escape routes than it is to clear such

routes subsequently. If doors a re to be relied on a s air

inlets, devices to open t hem auto matically are needed.

When non-permanent a utomatic drop curtains are

used as reservoir screens the gaps betw een curtains

should be minimised. If a powered smoke extract

system is installed, care should be ta ken to ensure that

smoke leakage through gaps between curtains does not

inadvertently activate a djacent powered extract zones.It can be a n adva ntage to open all the natura l vents

installed in a shopping centre, since those in zones

unaffected by smoke can contribute to the fresh air

inflow. Wherever it is feared t hat wind pressures on

vents to unaffected reservoirs may be much more

negat ive than on vent s in a smoke-filled reservoir, it is

recommended tha t w ind-tunnel studies and/or a

deta iled pressure ana lysis be carried out to evaluate

the eff ect on vent ing efficiency. When mechanical

extraction is used for a number of separate smoke

reservoirs, it wo uld be better to o perate the smoke

contro l system individually by zo nes. Facilities for

manual override should be installed to a llow greaterflexibility in firefighting, testing an d ma intenance.

The norma l ventilation systems fitted to ma ny, if not

most, malls and shops blow a ir into the mall or shop at

a high level, frequently where smoke would be flowing.

This air w ould simply increase the q uant ity of smoke

that has to be dealt w ith. Hence the normal ventilation

inlet system should be a utomatically shut down when

smoke is detected .

If t he smoke to be removed from a reservoir is too hot

for the ducting or for the fans, sprinklers or water jets

could be installed in a suitable part of the extract

ducting to cool the smoke. A t emperature rating

quot ed for ma ny (but not a ll) smoke extract fans is to

survive a tempera ture of 300°C fo r half an hour. Many

aut horities regard this time as sufficient for escape

from most malls — but note that this should not be

regarded a s being universally true. A ll such criteria

should be agreed w ith the a pproving authorities. All

electrical apparatus and power supply cables used inthese smoke ventilation systems should, of course, be

protected to ensure sustained operation in a fire.

Cold smoke tests are sometimes used for the

acceptance testing of smoke ventilation systems.

Whilst this cold smoke can b e used to opera te the

smoke detection system and therefore activate a ll the

components of the smoke ventilation system, it should

be noted that since the smoke is cold it would not have

the buoya ncy that smoke in a true fire condition would

have, and cannot therefore adeq uately test the

ventilation efficiency of the system.

Smoke ventilat ion systems should be regularly testedand adequa tely ma intained.

31

Chapter 5Some operational factors



References

1 Home Office and Scottish Home and HealthDepartment.Fire precautions in town centre

redevelopment. Fire Prevention G uide 1. Londo n,

HMSO, 1972.

2 British Standards Institution.Fire precautions in

the design and construction of buildings. Code ofpract ice for shops. Bri tish Standard B S 5588: Part

2: 1985. Londo n, B SI , 1985.

3 Canter D.Studies of human behaviour i n f ir e:

empi ri cal r esults and their impli cations for

education and design. B uilding Research

E stablishment Report. G arston, B R E , 1985.

4 Morgan H P.Smoke control methods in enclosed

shopping complexes of one or more storeys: a

design summary. B uilding R esearch E stablishment

R eport. London, H MSO , 1979.

5 British Standards Institution.Fire precautions in

the design and construction o f buildings. Par t 10:

C ode of pra ctice for enclosed shopping complexes.

Bri tish Standard B S 5588: Pa rt 10. Lo ndon, B SI. To

be published.

6 Heselden A J M.Fire problems of pedestrian

precincts. Part 1. The smoke production of var ious

materials. Fi re Research Station Fi re Research N ote

856, B orehamwo od, B R E , 1971.

7 McGuire J H, Tamura G T and Wilson A G.Considerations in the design of smoke control

systems in ta ll buildings. Proceedings of A SHRA E

Symposium on Fire H azards in Buil dings,

San Francisco, January 1970.

8 Hinkley P L.Work by the F ire Research Sta tion

on the cont rol of smoke in covered shopping

centres. Bu ilding Research E stablishment Curr ent

Paper CP 83/75, B oreha mwo od, B R E , 1975.

9 London Transport Board.Second Report of the

Opera tional R esearch Team on the capa city offootways. Research Report 95.

10 Heselden A J M and Hinkley P L.Smoke travel in

shopping malls. E xperiments in co-opera tion w ith

G lasgow Fire Brigade — Part 1. Fi re Research

Station Fi re Research Note 832, B orehamwoo d,

B R E , 1970.

11 Silcock A and Hinkley P L. Fire a t Wulfrun

shopping centre, Wolverhampt on. Fi re Research

Station Fi re Research Note 878, B orehamwoo d,

B R E , 1971.

12 Thomas P H et al. Investigations into the flow o f

hot ga ses in roof venting. Fi re Research Techni cal

Paper No 7, Londo n, H MSO , 1963.

13 Morgan H P and Chandler S E.Fire sizes and

sprinkler effect iveness in shopping complexes and

reta il premises. Fi re Surveyor, 1981, 10(5) 23–28.

14 Gardner J P.U nsprinklered shopping centres.

D esign fire sizes for smoke ventilation. Fire

Surveyor , 1988, 17(6) 41-47.

15 Hinkley P L. R ates of production of hot ga ses in

roof venting experiments. Fi re Safety Jour nal ,

1986, 1057–65.

16 Heselden A J M, Wraight H G H and Watts P R.Fire problems of pedestrian precincts. Pa rt 2.

Large-scale experiments with a shaft vent. Fire Research Station Fi re Research N ote 954,

B orehamwood, B R E , 1971.

17 Morgan H P, Marshall N R and Goldstone B M.Smoke ha zard s in covered multi-level shopping

malls: some studies using a model tw o-storey mall.

Buil ding Research E stablishment Curr ent Paper

C P45/76, B oreha mwo od, B R E , 1976.

18 Hansell G O, Marshall N R and Morgan H P.Privat e communication. Fire R esearch Sta tion

1988.

19 Morgan H P.The horizontal flow of buoyant gases

towa rd an opening. Fi re Safety Jour nal, 1986, 11,

193–200.

20 Morgan H P and Marshall N R. Smoke hazard s in

covered, multi-level shopping malls: an

experimentally-based theory for smoke

production. Buil ding Research E stablishment

Cur rent Paper CP 48/75, B orehamwoo d, B R E ,

1975.

21 Morgan H P and Marshall N R.Smoke controlmeasures in covered tw o-storey shopping mall

having balconies as pedestrian w alkwa ys. Building

Research E stabli shment Cur rent Paper CP 11/79,


22 Morgan H P and Hansell G O.Atrium buildings:

calculating smoke flows in atria for smoke control

design. Fi re Safety Jour nal, 1987, 12(1) 9–35.

23 Hansell G O, Morgan H P and Marshall N.Smoke

flow experiments in a model a trium: Pa rt 2 plume

entra inment in the a trium. To be published.

32



24 Thomas P H. On t he upwa rd movement of smoke

and related shopping mall problems. Fi re Safety

Journal , 1987, 12(3) 191–203.

25 Morgan H P.Co mments on ‘A note on smoke

plumes from fires in multi-level shopping malls’.

Fi re Safety Jour nal, 1987, 12(1) 83–84.

26 Law M.A no te on smoke plumes from fires in

multi-level shopping malls. Fi re Safety Jour nal,1986, 10197–202.

27 Morgan H P and Marshall N R.The depth o f void-

edge screens in shopping ma lls. Fir e Engineers

Journal, 1989, 48(152) 7–9.

28 Hansell G O. Fir e Engineers Journal, 1989, 48(152) 9.

29 Penwarden A D.Acceptable wind speeds in tow n.

Buil ding Research E stablishment Curr ent Paper

C P1/74, G arsto n, B R E , 1974.

30 Spratt D and Heselden A J M. E fficient extraction

of smoke from a thin layer under a ceiling. Fire

Research Station Fi re Research N ote 1001,


31 Heselden A J M.P rivate communication. Fire

R esearch St at ion, 1976.

32 Chartered Institution of Building ServicesEngineers.CIB SE G uide , Volume C, 1986.

33 Marshall N R and Heselden A J M.Smoke control

in large stores opening onto enclosed shopping

malls. Fi re Surveyor, 1986, 15(1) 18–22.

34 Morgan H P and Savage N P.A study of a large

fire in a covered shopping complex: St J ohn’s

C entre 1977. Bu ilding Research Establi shment Cur rent Paper C P10/80, B oreha mwo od, B R E 1980.

35 Hansell G O and Morgan H P.Fire sizes in hotel

bedro oms — implicatio ns for smoke control

design. Fi re Safety Jour nal, 1985, 8(3) 177–186.

36 Morgan H P and Hansell G O.Fire sizes and

sprinkler effectiveness in off ices — implications fo r

smoke control d esign. Fi re Safety Jour nal, 1985, 8(3) 187–198.

37 Marshall N R, Feng S Q and Morgan H P.Theinfluence of a perforated false ceiling on the

performance of smoke ventilation systems. Fire

Safety Journal , 1985, 8(3) 227–237.

38 Butcher E G, Langdon-Thomas G J andBedford G K. Fire and car park buildings. Fire

Research Station Fi re Note 10, Bo rehamwood ,

B R E , 1968.

33



Liaison Panel

The preparation of the final report wa s undertaken with the a ssistance of a Liaison Pa nel which

consisted of the follow ing members, representing both industry and G overnment interests:

D epartment of the E nvironment, Construction D irectora te

H ome Office, Fire Inspectora te

Scottish D evelopment D epartment

Co lt International Limited

G radwood LimitedNuaire L imited

34



Smoke control in buildings: design principlesBRE Digest 396 £4.50Outlines the design principles for systems which will providesafe escape routes from buildings.

Fire spread in buildings

BRE Information Paper IP21/84 £3.50This paper is concerned with the spread of fire in occupiedbuildings and its implications for safety of life. It makessuggestions on how problems are avoided.

Smoke spread within buildingsBRE Information Paper IP22/84 £3.50Complementary to Fire spread in buildings , IP21/84, thispaper is concerned with smoke spread to areas beyond thesource, its implications for life safety and for propertydamage.

Fire safety in buildingsRef BR96 £16 BRE Report 1986The author’s view of how Building Regulations and fireprotection measures relate to actual fires. The Reportmakes a number of suggestions for the improvement ofprotection measures and its legislation.

Aspects of fire precautions in buildingsRef BR137 £16 BRE Report 1988This book introduces students and others to the passiveaspects of fire precautions—the measures taken andprotection provided in a building, for example, to minimisethe risk to occupants, contents and structure from anoutbreak of fire. It describes the nature of fire, structural fireprotection techniques and means of escape.

Selection of sprinklers for high rack storage in

warehousesBRE Information Paper IP5/88 £3.50Demonstrates the likely benefits of a ‘fast-response’ solder-link sprinkler head for use in a warehouse. A ‘fast-response’solder-link sprinkler has been shown to provide effective in-rack protection for goods stored in high racks up to about10 m. The information will be useful to manufacturers, thoseconcerned with high rack storage, insurers and the fireauthorities.

Guidelines for the construction of fire-resistingstructural elementsRef BR128 £13 BRE Report 1988This Report establishes guidelines for the fire resistance ofelements of structure. It includes tables of notional periodsof fire resistance based on current test data; it also makes

revisions to tables and text concerning concrete, masonryand timber included in the earlier 1982 edition.

ASKFRSPart 1 The scientific contentPart 2 A programmer’s guideBRE Computer package AP46 £100.00 + £17.50 VATAn interactive computer program comprising a manual and511 / 4 disk, to help fire engineers quantify various firescenarios using common algebraic expressions and amaterials database. Part 1 provides ten types ofcalculation—Part 2 describes ASKFRS. The program iswritten in FORTRAN-77 and can be run on various VAXcomputers and IBM PCs.

Experimental programme to investigateinformative fire warning characteristics formotivating fast evacuationRef BR172 £30 BRE Report 1990Gives details of an experimental programme to evaluate therelative effectiveness of alarm bells, text messages,computer generated voice and graphics displays to motivateescape. The quantitative values of each, in promoting adecision to escape, are given together with the subjects’message assimilation times.

Psychological aspects of informative fire warningsystemsRef BR127 £9 BRE Report 1988

A summary of research using simulations of informative firewarning systems in a number of buildings, and how theinformation they carried, and training procedures, can mosteffectively promote rapid and safe evacuation in the eventof fire.

Studies of human behaviour in fire: empiricalresults and their implications for education anddesignRef BR61 £11 BRE Report 1985A model of human behaviour in fire is presented. Theimplications of the model and the findings of the variousstudies are discussed in relation to education, training,publicity, design and legislation. Essential reading for allthose involved in decision making in these fields.

Also available from BRE

BRE 186_Design Principle for Smoke Ventilation in Enclosed Shopping Center

Documents