6 · THE SINGAPORE ENGINEER Feb 2009

Fire Protection

Smoke Control by Pressurisation in Stairwells

and Elevator ShaftsDr Richard S Miller and Dr Don

Beasley, Department of Mechanical

Engineering, Clemson University, USA,

explore the application of elevator

shaft and stairwell shaft pressurisation

systems, as a means of preventing smoke

migration through the ‘stack eff ect’ in

tall buildings.

� is article is based on a summary

of the technical results recorded from

a study funded by the Smoke Safety

Council, USA.

INTRODUCTION

� e stack eff ect is created in tall building

shafts when there is a temperature

diff erence between the building interior

and the ambient. For a cold ambient, the

lower fl oors have a net negative pressure

diff erence while the upper fl oors show

a net positive pressure diff erence. In

physical terms, air is being entrained into

the shaft on lower fl oors and forced out

into the building on the upper fl oors.

� e conventional calculation of the

stack eff ect pressure diff erence is actually

the diff erence in pressures between the

elevator (or stairwell) shaft and the

outside world. In the absence of other

interior pressure barriers, this total stack

eff ect pressure diff erence comprises the

sum of the pressure diff erences across

the elevator (or stairwell) doors plus that

across the building exterior. � e primary

problem associated with the stack eff ect

in tall buildings, in relation to the current

study, is its infl uence on smoke migration

during fi res. In this regard, it is the ‘across

door’ portion of the stack eff ect pressure

diff erence that is directly related to smoke

migration and control. A fi re located on a

lower fl oor can cause substantial damage,

injury, and even death on upper fl oors

due to the smoke migration through

the elevator or stairwell shaft. � e most

well-known example of the occurrence of

this eff ect, is the fi re at the MGM Grand

Hotel and Casino, in 1980. A fi re broke

out in a ground fl oor restaurant, killing

85 people, with the majority of the deaths

taking place on the upper fl oors, due to

smoke inhalation.

A variety of smoke control techniques

has been proposed for both stairwell

and elevator shafts, primarily involving

enclosed vestibules or lobbies surrounding

the doorways. However, the current study

sought to investigate the feasibility of

solely using shaft pressurisation as a means

of preventing smoke migration in elevator

shafts.

Pressurisation systems for stairwells

have been used for some time. � e

available literature has shown stairwell

pressurisation systems to be feasible,

although the system performance can be

quite sensitive to several design parameters,

including the opening of stairwell doors.

In contrast, elevator shaft pressurisation

has only been recently approved by the

International Building Code (IBC) for

smoke prevention in elevator shafts, and

relatively little research has been done to-

date, although experimental measurements

in a fi re tower and a limited number of

simulation results, have been reported.

� e primary objectives of the current work

are to both illustrate the fundamental

diff erences between stairwell and elevator

shaft pressurisation systems, and to provide

input for future code changes. Eff ects of

the elevator and exterior building doors,

ambient temperature, fan location, and

shaft venting, on the pressurisation system

performance, are also addressed.

MODELLING APPROACH

� e article presents the results from an

investigation of stairwell and elevator

shaft pressurization, on potential smoke

distribution through the shaft eff ect.

All results were obtained via computer

simulations using the CONTAM software

developed by the Indoor Air Quality

and Ventilation Group at the National

Institute of Standards and Technologies,

USA. � e CONTAM software has been

used extensively for similar simulations of

air fl ow and for both stairwell and elevator

shaft pressurisation.

� e results are presented for a

30-storey building model. A schematic of

the building model’s typical fl oor plan is

provided in Figure 1 (not to scale) along

with prescribed leakages. � e building is

specifi ed as a 30-storey building with a

fl oor height of 9.85 ft (3.0 m) and a fl oor

area of 10,000 ft2 (930 m2). On each fl oor,

there are two stairwells located at opposite

corners of the building. In the centre of

the building, are two (open) elevator shafts

having four sets of elevators and elevator

doors. All interior building leakage areas

are based on typical values reported in

the literature. Each of the closed elevator

doors (four per shaft) has a leakage area

of 75 in2 (0.0484 m2). However, the fi rst

fl oor elevator doors have a 854 in2 (0.558

m2) leakage area modelling the elevator

doors being open with the car on that

fl oor. Each stairwell has a single door with

leakage area 16 in2 (0.0103 m2). Each

fl oor of the building has exterior leakages

calibrated with experimental data for

either a ‘residential’ (37-storey building in

Korea) or a ‘commercial’ (data measured

in a 20-storey bank in Boise, Idaho)

building.

� e two models take account of the

presence or absence of openable windows,

balcony doors, and other features

associated with the two specifi c buildings

used for the calibration. � e parameters

are not meant to imply that all residential

buildings will have more porous skins

than other buildings. It is noted that the

ground fl oors have larger leakage areas

than those on upper fl oors, due to the

presence of exterior doors and/or other

unique features. � e building temperature

is maintained at 70° F (21° C) on all fl oors.

Figure 1: Schematic representation of the 30-storey building fl oor plan. External leakages correspond to either a residential (R) or commercial (C) building model.

THE SINGAPORE ENGINEER Feb 2009 · 7

Fire Protection

No wind is present.

Each building model also has a roof

level with only the stairwells and elevator

shafts (where there are fans, they are

installed here). Elevator shaft and stairwell

shaft pressurisation are considered by

pressurising the shafts until a specifi ed

minimum pressure diff erence across any

elevator or stairwell door is achieved.

� is corresponds to a minimum pressure

diff erence of +0.05 inch water (+12.5

Pa) across elevator doors or +0.15 inch

water (37 Pa) across stairwell doors. � e

process is iterated with a model for the

average shaft air temperature based on

turbulent duct theory along with the

Dittus Boelter heat transfer correlation.

For elevator shaft pressurisation, the

elevator cars are on the fi rst fl oor with all

doors in the open positions and with all

stairwell doors closed, unless otherwise

specifi ed (where there is no pressurisation

system, all elevator doors are in the closed

position). For stairwell pressurisation,

all elevator doors and all stairwell doors

are in the closed positions. Simulations

are conducted for both pressurised

shafts and for non-pressurised shafts for

comparisons. Both ‘cold day’ (−12° C, 10°

F) and ‘hot day’ (38° C, 100° F) conditions

are considered.

RESULTS

One fi nding of the study is the importance

of the elevator or stairwell air temperature

on the simulation results. On very warm

or very cold days, the ambient air is at a

substantially diff erent temperature than

that of the building, and this diff erence

aff ects the pressure profi les within the

shaft. A detailed model for the average

shaft temperature was therefore derived

from turbulent duct fl ow theory and

incorporated into the simulations.

Stairwell pressurisation only

Results for stairwell shaft pressurisation

only, are presented in Figure 2, for the

residential and commercial building

models. Stairwell pressurisation is

observed to work well within the limits

allowed by the current code limits of

.

� e results of the simulations show that

the system calibration is highly sensitive

to the ambient temperature. � e location

of the minimum pressure diff erence across

Figure 2: Pressure diff erence across the stairwell doors as a function of the fl oor number for a stairwell only pressurisation system: (a) residential building with no pressurisation, (b) residential building with pressurisation, (c) commercial building with no pressurisation, and (d) commercial building with pressurisation.

a) b)

c) d)

8 · THE SINGAPORE ENGINEER Feb 2009

Fire Protection

stairwell doors is the ground fl oor when

the ambient temperature is less than

the building temperature. However, the

minimum pressure diff erence occurs on

the top fl oor when the ambient is warmer

than the building temperature. Fan

output is also dependent on the ambient

temperature.

Eff ects of the exterior building door position

While not particularly sensitive, stairwell

pressurisation systems can be aff ected

by the position of the building’s exterior

doors. Figure 3 illustrates the eff ects of

the ground fl oor exterior door position on

the system performance for the residential

building model, on a cold day. Results are

presented for systems that are calibrated

with the exterior building doors either

closed [Figure 3(a)] or open [Figure 3(b)].

! e eff ect of then opening (or closing)

the exterior doors, is also shown in each

fi gure. Required fan speeds also change

and are approximately 15% larger for

the open door cases. To understand the

results, the system should be considered

calibrated with the building doors closed.

As the doors are opened, the slightly

pressurised building loses pressure to the

ambient. ! e eff ect is felt throughout the

building as the elevator shafts provide a

relatively unrestricted fl ow of air between

fl oors. ! e resulting pressure diff erence

across the fi rst fl oor stairwell doors, is

slightly reduced below the code-specifi ed

minimum, but remains large enough to

inhibit smoke penetration. As discussed

below, a very diff erent behaviour results

with elevator shaft pressurisation systems,

as the stairwells are relatively well sealed

and do not provide an equivalent route

for air fl ow between fl oors.

Elevator shaft pressurisation only

Results for elevator shaft pressurisation are

presented in Figure 4, for the residential

and commercial building models. Several

of the major potential problems with

elevator shaft pressurisation systems are

illustrated. Elevator shaft pressurisation

is markedly diff erent from stairwell

shaft pressurisation. Required fan

fl ow rates are approximately 50 times

larger for elevator shaft pressurisation

systems than for stairwell pressurisation

systems. ! e current code limits of

Figure 3: Pressure diff erences across stairwell doors as a function of the fl oor number for the residential building model. Data are for a stairwell only pressurisation system calibrated with the exterior building doors in either the (a) closed or (b) open position, and show the eff ects of opening or closing

the exterior building doors. All data are for ‘cold day’ conditions (10°F).

a) b)

a) b)

c) d)

Figure 4: Pressure diff erence across the elevator doors as a function of the fl oor number for an‘elevator shaft only’ pressurisation system: (a) residential building with no pressurisation, (b)residential building with pressurisation, (c) commercial building with no pressurisation, and (d)commercial building with pressurisation.

are also impossible to meet. Furthermore,

pressure diff erences across upper fl oor

elevator doors far exceed any reasonable

limits for proper door functioning.

! e reasons for the resulting ‘across

elevator’ door pressure diff erences can

be explained. Air is forced into the shaft

from the roof and some is ‘lost’ along the

way through the closed elevator doors

and into the building interior. However,

a relatively large fl ow rate is needed to

achieve the +0.05 inch water (+12.5 Pa)

pressure diff erence across the fi rst fl oor

open elevator doors, due to their much

larger leakage areas. As the ground fl oor

elevator doors are open and have relatively

large leakage areas, this required fl ow rate

can be considerable.

! e air fl owing into the fi rst fl oor

from the shaft then pressurises the fi rst

fl oor until the fl ow rate out of the fi rst

fl oor (through exterior and stairwell

THE SINGAPORE ENGINEER Feb 2009 · 9

Fire Protection

leakages) equilibrates with the fl ow rate

entering through the elevator shafts. � e

second fl oor interior building pressure is

much less than that on the fi rst fl oor as

the closed stairwell doors have a relatively

small leakage area (in cases where there is

a coupled stairwell pressurisation system,

no air would be allowed to fl ow into the

stairwell shaft).

However, the pressure within the

shaft only varies hydrostatically, and

so is only slightly lower at the second

fl oor. � erefore, the ‘across elevator’

door pressure diff erence is increased

substantially on the second fl oor (as well

as all remaining fl oors). � is pressurisation

of the ground fl oor is due to the large

open door leakage areas and is the primary

eff ect distinguishing stairwell and elevator

shaft pressurisation systems. � e eff ect

is enhanced as the fi rst fl oor leakage

becomes smaller for the commercial

building model (and vanishes if the fi rst

fl oor exterior door is open). � e outside

temperature has relatively little infl uence

on the fi nal system pressure diff erences.

However, signifi cantly diff erent fan fl ow

rates are required based on the exterior

temperature (fl ow rates are approximately

10% larger on hot days than on cold days).

� erefore, a system calibrated and tested

during one season, may have signifi cantly

diff erent behaviour during other seasons.

Eff ects of the exterior building door position

One possibly tempting means of

overcoming the large pressure diff erences

across elevator doors, is to calibrate

Figure 5: Pressure diff erences across elevator doors as a function of the fl oor number for the residential building model. Data are for an ‘elevator shaft only’ pressurisation system calibrated with the exterior building doors in either the (a) closed or (b) open position and show the eff ects of opening or closing the

exterior building doors. All data are for ‘cold day’ conditions (10° F).

the system with the exterior ground

fl oor building doors open. � is would

eliminate the overpressure on the

ground fl oor. However, in contrast to

stairwell pressurisation systems, elevator

shaft pressurisation systems are much

more sensitive to the ground fl oor door

position. Figure 5 shows the eff ects of

the exterior building door positions for

the residential building model on a cold

day, with elevator shaft pressurisation.

System performance is considered for

systems calibrated with the exterior doors

in either the open or closed positions. A

set of double doors propped wide open, is

modelled with a 42 ft2 (3.90 m2) leakage

area on the ground fl oor. � e eff ects of

then changing the exterior door position

are also included in the fi gure.

System performance and fan

requirements change dramatically based

on the exterior doors. If the system is

calibrated with the exterior doors in the

closed position, a relatively large fan

speed is required (larger by in excess of

300% than if the doors are open). If the

exterior doors are then propped open,

the minimum pressure diff erence is still

maintained across all elevator doors

(although prohibitively large pressure

diff erences exist). In contrast, if the

system is calibrated with the exterior

doors open and then operated with the

doors closed, there will be essentially no

pressure diff erence across the ground fl oor

elevator doors. In this situation, there is

a high probability of smoke entering the

shaft if a fi re is present on the ground

fl oor. In summary, the only way to achieve

a potentially code-compliant system

performance is with the ground fl oor

doors open. However, the system will fail

if it is operated with the doors closed as

may very well be the case in an actual fi re

situation.

Eff ects of other parameters

Further studies have also examined the

eff ects of the fan location, secondary

pressurisation systems, multiple injection

points, and the eff ects of various louvre/

vent systems to alleviate overpressures.

� e results clearly show that each of these

approaches is incapable of alleviating the

above problems. Since the elevator shaft

is relatively wide, it experiences negligible

frictional resistance and the shaft pressure

simply equilibrates to pressure changes

as would occur in a large tank. � e shaft

pressure is therefore independent of the

fan location or to multiple injection

points. Louvres and vents are similarly

incapable of properly controlling the shaft

pressure distribution because they are

only capable of uniformly changing the

pressure in the entire shaft. � erefore, any

reduction in the maximum shaft pressure

due to a roof (or otherwise located) vent

or louvre, simply shifts the entire pressure

distribution within the shaft evenly.

� is results in the minimum +0.05 inch

(+12.5 Pa) being violated as the fi rst fl oor

pressure diff erence drops. For example,

if a louvre system that allows 2,500 ft3/

min (70 m3/min) of air to fl ow from the

top of the shaft, is installed, the fan speed

a) b)

10 · THE SINGAPORE ENGINEER Feb 2009

Fire Protection

would need to be increased by the same

2,500 ft3/min (70 m3/min) to compensate

and to re-acquire the minimum pressure

diff erence across the ground fl oor elevator

doors. � e net eff ect is to re-acquire the

original pressure profi le but with a larger

fan requirement. Relying on transients is

also ineff ective. An analysis shows that the

system response time to changes in door

positions, fan fl ow rates, etc. is ~5 sec for

the current model building.

Coupled stairwell and elevator shaft

pressurisation

Results for simulations of the building

models with coupled stairwell and

elevator shaft pressurisation systems

(that is, operating simultaneously and

jointly aff ecting the building fl oor

pressures) are presented in Figure 6 for

the residential and commercial building

models. � e simulation results illustrate

an additional and very serious problem

for stairwell pressurisation systems if

used in conjunction with an elevator

shaft pressurisation system. � e addition

of the elevator shaft system results in an

additional fl ow of air into the building

on all fl oors. � is raises the pressure of

the building interior and would result in

negative pressure diff erences across the

stairwell doors, if the stairwell-only fan

speeds were used. � erefore, substantial

modifi cation of existing stairwell

pressurisation would be required if an

elevator system were later installed.

Furthermore, another problem occurs

after the stairwell system is re-calibrated

to acquire a minimum +0.15 inch water

(+37 Pa) pressure diff erence across any

stairwell doors. In this case, a phenomenon

occurs, as was observed previously for the

elevator shaft pressurisation systems. � e

overpressure on the fi rst fl oor as compared

to the second fl oor of the building, also

creates very large pressure diff erences

across all upper fl oor stairwell doors.

� ese pressure diff erences are far too large

for proper stairwell door functioning. For

example, if a 7 ft x 3 ft (approx 1 m x

2 m) stairwell door has a 1.5 inch water

Figure 6: Pressure diff erences across either stairwell or elevator doors as a function of the fl oor number for coupled stairwell and elevator shaft pressurisation systems: (a) residential building, stairwell doors, (b) residential building, elevator doors, (c) commercial building, stairwell doors, and (d) commercial building, elevator doors.

a) b)

c) d)

THE SINGAPORE ENGINEER Feb 2009 · 11

Fire Protection

(approx 375 Pa) pressure diff erence, this

would require a force of approximately

165 lbf (approx 750 N) to open the

door! � ese results show that in addition

to the problems described previously for

stand-alone systems, an elevator shaft

pressurisation system will also make the

standard stairwell pressurisation system

fail.

Eff ects of the elevator door position

Simulations were also conducted

assuming that the system is calibrated

with all the elevator doors in the closed

position. � e results are similar to those

described above, for the exterior building

door position. Both the stairwell and the

elevator pressure diff erences are within

reasonable limits when the elevator

doors are closed. However, if the elevator

doors are opened on the ground fl oor,

the minimum pressure diff erence nearly

vanishes on the ground fl oor. � is occurs

also for the stairwell doors on the ground

fl oor. In this case, there is a strong potential

for smoke to enter either shaft, if present

on the ground fl oor.

Instead of calibrating the pressurisation

system with either the exterior building

doors propped open or with the elevator

doors closed, it may be proposed that the

system be calibrated by simply ignoring

the pressure diff erences across the open

ground fl oor elevator doors. Aside

from the ignored pressure diff erences,

Figure 7: Pressure diff erences across elevator doors as a function of the fl oor number for theresidential building model with coupled stairwell and elevator shaft pressurisation on a cold day. Pressure diff erences across the open elevator doors on the ground fl oor are ignored for system calibration. ! e results show the eff ects of having two elevator cars move to the 15th fl oor with open elevator doors.

all systems are essentially able to meet

the specifi cations. Both the stairwell

doors and the elevator doors experience

reasonable pressure diff erences. However,

the open elevator doors may still be

problematic. In all cases, these pressure

diff erences are essentially null. In the

event that smoke was present on the

ground fl oor and the elevator doors were

opened, it could be forced into the lower

level fl oors either just above the ground

fl oor or into the lower (basement) levels.

Occupants may be forced to evacuate

towards the fi re containing fl oor. In

contrast, a pressurisation fan mounted

on or below the ground fl oor would

prove catastrophic as the smoke would

be blown throughout the entire building.

Another potential problem with a system

calibrated, ignoring the open elevator

door pressure diff erences, is illustrated

in Figure 7. In this case, the calibrated

system for the residential building model

on the cold day conditions is examined.

� e calibrated building model is altered as

follows: two of the elevator doors from a

single shaft are now closed on the ground

fl oor and the same two doors are opened

on the 15th fl oor (mimicking the eff ects

of two cars in use by either fi re fi ghters

or building occupants). In this case,

the pressure diff erence across all of the

elevator doors is lost on the 15th fl oor,

as air from the shaft pressurises the fl oor.

� e results show that if the elevators are

brought to a smoke containing fl oor, there

is a high probability of smoke entering the

shaft. In this case, the fan pressurisation

system would actively distribute the

smoke throughout the building (and at a

higher rate than the stack eff ect the system

was originally designed to overcome).

� erefore it is recommended that pressure

diff erences across open elevator doors,

should not be ignored, if there is any

potential for elevator usage during a fi re

situation.

Eff ects of the building height and

number of elevator cars

� e results to this point have shown

that a robustly operating elevator shaft

pressurisation system with reasonable

pressure diff erences across elevator doors

is nearly impossible to design in the

30-storey building model if (1) these

pressure diff erences apply to both open

and closed elevator doors, and (2) if the

system must function properly when the

ground fl oor exterior building doors are

closed. Additional simulations have shown

that these results are not directly aff ected

by the building height but are directly

aff ected by the number of elevator cars and

shafts. � erefore, while tall buildings may

have the characteristics that produce large

‘across elevator’ door pressure diff erences

(due to larger numbers of elevator cars),

it is not the building height directly that

causes the behaviour observed in the

study.

CONCLUSIONS

� e operation of stairwell shaft

pressurisation systems was found to be much

simpler than elevator shaft pressurisation

systems (and quite feasible). In contrast,

elevator shaft pressurisation was found to

require substantially larger fan fl ow rates

to achieve the required minimum pressure

diff erences. Prohibitively large pressure

diff erences across upper fl oor elevator

doors were found for all cases where the

exterior building doors are kept closed,

and open elevator doors contribute to the

minimum pressure diff erence.

� e elevator shaft system also

catastrophically interferes with the

stairwell pressurisation system in these

cases. In contrast, systems calibrated

with either the exterior building doors

open, all elevator doors in the closed

position, or ignoring the open elevator

door pressure diff erences, were all

found to maintain reasonable ‘across

door’ pressure diff erences on all fl oors

(stairwell and elevator). However, each

of these will lead to situations in which

nearly null ‘across elevator’ door pressure

diff erences on some fl oors, could allow

smoke to enter the shaft and be actively

distributed throughout the building. Fan

location, vents and louvres were all found

to be ineff ective as a means of controlling

the shaft pressures. Little eff ect of the

ambient temperature was observed on the

fi nal elevator door pressure diff erences.

However, signifi cantly diff erent fan speeds

are required.

(More information on the study, and on

the references cited in the summary of the

technical results, may be obtained from Dr

Richard S Miller at rm@clemson.edu)