1. Introduction
Mechanical stair pressurization is a preferred system for staircases in high-rise buildings to
control the smoke movement in case of fire though the survey conducted by Cowlard et al. [1]
implies that pressurization systems fail in 9 out of 10 cases. Lay [2] stated that there is suspiciousness about the functionality among fire professionals.
Requirements published in codes worldwide address the pressure difference to the adjacent
floors or a minimum air inflow velocity into the fire floor by upper and lower bounds.
The factors that influence the performance of the pressurization system in the stairwell are
closely linked to the pressure profile along the flow path caused by the flow through the
stairwell that pushes the smoke into the outside. Flow path resistances along this streamline
and room pressures can be calculated by formulas that are derived from one dimensional fluid
dynamical formulas like Bernoullis principle, the ideal gas law and using electrical resistance
calculations, all of them are extended with empirical correlations. Those are the fundamentals
for multizonal software like CONTAM [3] and state of the art of the performance based design method of pressurized stairwells. The CFD code Fire Dynamics Simulator (NIST) [4] is used
to highlight the limitations of the existing design methods by multizonal software [3].
Scenario
Design criteria
Design methodology
NFPA 92 2012
USA
vestibules permitted
>8 stories: multiple injection
requires hBuilding>15m / 3stories
pmin= 12.5 Pa
Fmax,opening= 133N
Opposed airflow dependent
on fire size
MHRRL
Germany
UAE
Fire and Life
Safety Code of
Practice
TRVBS 112
Austria
DIN EN
12101-6
udoor = 1 - 2 m/s
pmax = 60 Pa
pmin = 30 Pa
Qmin = 7.500 15.000 mh-1
pmin = 12.5 / 25 Pa
pmax = 60 Pa
Fmax,opening = 133N
udoor = 0.75 - 2 m/s
2
p1
u1 g h1 p2
u 2 g h2 ploss 1 2
(1)
2
2
velocity
static
pressure P1 pressure P1
gravitational
static
velocity
pressure P1 pressure P1 pressure P2
gravitational
pressure P2
pressure
loss
With known conditions at point 1, the pressure loss has an influence on the velocity pressure
and therefore on the mass flow rate at point 2 and has to be considered carefully in design of
the pressurization system because supply fans perform dependent on the summated pressure
losses on the flow path. In contrast, because pressure loss is caused by friction, it is dependent
on the velocity or the Reynolds number. Consequently there is a feedback between supplied
volume flow in the staircase and created pressure loss on the flow path that is governed by
local velocities.
In engineering, different approaches exist to consider the pressure loss, for the wall friction
this is for example illustrated in the Moody chart, local resistances or head loss coefficients
are experimentally determined by the measured ratio of pressure loss between two points to
velocity pressure of a reference point: C [12], K [11]or [13] (eq. 2) The related velocity is
closely linked to the considered flow area and can be determined with mass conservation by
ureference=Q/Areference. For common duct fittings loss coefficients are listed in fundamental literature [13-15] and are tabularized, printed in curves or can be calculated by equations dependent on velocity or Reynolds number and geometrical parameters of the fitting.
ploss 1 2
ploss 1 2
C K
2
(2)
2
u2
u
2
5. Pressure Losses in Stairwells
Internal pressure losses that occur during the flow through the stairwell have a great influence
on the design of the pressurization system and are strongly influenced by the geometry.
1976 Shaw and Tamura [16] related the overall pressure to the Darcy-Weibach-equation for
friction in a rectangular duct where the pressure loss depends on an empirically determined
friction factor K SW (eq. 3), an equivalent hydraulic diameter Dh= 4A/P=4ab/2/(a+b) and the
height of the floor. Additionally, an approach was introduced to relate the pressure loss to an
empirically determined equivalent orifice area for the stairwell Ao (eq. 4) based on the Bernoulli equation, which represents the flow resistance by an orifice at each floor level. Here the
coefficient Cd for turbulent discharge of an orifice completes the equation. This approach was
also used by Achakij [17] whose experiments were the basis for todays design methods and
his findings are a part of the contemporary design procedure [5].
Other approaches where introduced by Poreh et al. [18] and [19], who relate their experiments
to the initial loss coefficient SW , floor perfloor(eq. 5)while [20-23] use the effective area with
Aeff Cd Ao (eq. 6).
ploss 1 2 K SW
ploss 1 2
hhloor ( a b ) 2
Q
3
2 ASW
2
SW , floor 2
Q
2
ASW
ploss 1 2
(3)
(5)
p loss 1 2
2
1
Q
2
Cd2 Ao 2
A
2
ASW eff
A
SW
(4)
2
1 2
A 2 2Q
eff
(6)
Besides the flow rate or the mean velocity related to the stairwell area, the complexity of geometrical dependencies of the pressure loss varies significantly. While the use of the friction
factor K SW requires the input of width, length and height of the stairwell, for the approach
referring to the equivalent orifice area and the flow coefficient SW , floor it is sufficient to specify the stairwell area. The empirical coefficients are additionally related to open and closed
treads [17], shape (conventional with 2 landings vs. helical with 4 landings [18] ), the existence of railing, existence and size of well hole or running length [20-23].
In 2003 Poreh et al.[18] figured out by experimental studies, that the geometry of the stairwell
(conventional vs. helical) has a great influence on the flow resistance.
The fact that persons inside the stairwell have a significant effect on the pressure drop was
observed in 1988 by Achakij [17], though further tests with additional parameters were not
carried out since then. Nevertheless the effect of pressure losses is not considered in performance based codes at all.
In table 2 loss coefficients that are determined by different approaches are compared in terms
of the scenario of a stairwell 5m length, 3m width and 3m height with the volume flow of
4m/s related to a critical velocity of 2m/s through a door (1m x 2m).
Table2Determinationofpressurelosses[16],[17],[18],[23]&[22],[21],[20],[19]
7
6
5
4
3
2
1
0
max
pr EN 12101 -13
min
max
model 1:6
min
no well hole
max
model
min
realscale
Achakij
1988
low resistance
persons
high resistance
realscale
Tamura
& Shaw
1976
max
no
persons
min
max
min
max
min
model 1:6
model
1:10
Ostertag
2001
Gerhardt
&
Konrath
2001
(13)
Here the loss coefficient founds input into the velocity pressure (dynamic pressure) that is
related to the mean velocity of the opening and calculated by mass balance and can be related
to the orifice equation as well as to the mass flow between two nodes of different pressure:
1
2
m Cd , mass Aopening uopening
Aopening p1 2
(14)
Cd ,local
Hence it is obvious that according to equation (15) there is a relation between the local loss
coefficient and the mass discharge coefficient which is used to relate the geometrical opening
area to the real flow area due to contraction.
For simplification the velocity inside the rooms is assumed to be
1
C
d
,
mass
(15)
negligible small. This does not meet the physics in terms of
Cd ,local
pressurization systems, e.g. the critical velocity of 2 m/s in a door
(1 m x 2 m) corresponds to a velocity of 0.10 m/s in a common
room width of 10 m and a height of 2.2 m.
Because the empirical models are mainly tailored for thermal or wind induced flow and not
for mechanical induced flow in rooms with high mass exchange rates, their applicability for
pressurization systems is not yet validated. One point are missing experiments on the determination of the local or mass discharge coefficient for the flow through doors. Literature values state Cd,mass to be typically close to 0.6 for a sharp-edged orifice and slightly higher for
other openings in buildings [3]. Klote [5] suggest for open doors in stairwells Cd,mass = 0.35
though it is recognized that stationary vortices may reduce the flow significantly.
According to recent studies, it is very difficult to determine loss coefficients for external
openings in wind driven ventilation [24].
Additionally Chu et al. [24] pointed out that internal openings have a significant effect on the
ventilation rate as they trim down flow velocity and an internal coefficient of Cd,loca l= 2.58
respectively Cd,mass = 0.62 for wind driven ventilation was derived though their findings are
tied to a fixed flow path.
7. Spatial Resolution: Electrical Resistance Concept
Spatial distribution and local pressure losses due to flow patterns in a room govern the distribution of mass flow via different openings in one room. This has an influence on the real flow
path and is essential to determine pressure loss resistances for the design of the fan, because in
multizonal models the pressure loss depends on the mass flow through an opening. According
to the program documentation of the predecessor to CONTAM, AIRNET [29] as well as according to Klote [5] the distribution of flow between multiple openings can be calculated similar to the electrical resistance concept (Kirchhoffs law) of resistance in parallel and series.
While in CONTAM laminar flow is considered with the laminar flow coefficient K, in turbu-
m
LAMINAR
K p (18)
parallel:
K e K i (19)
series:
1
1
(20)
Ke
Ki
TURBULENT
C
(21)
1
1
i
e
sure loss does not occur within the fitting
2 (22)
2
Ce
Ci
under observation but as well upstream and
downstream. For flow through rooms inAnalytical Approach - Klote [5]
duced by a pressurization system this effect
is reinforced due to more complex flow
parallel:
series:
phenomenon.
1/ 2
Ae Ai (23)
1
Ae 2 (24)
An isothermal FDS simulation was run to
Ai
visualize the influence of the outflow and
inflow conditions on flow path resistance.
Table3electricalresistanceconcept
While the multizonal and analytical approach integrates pressure losses due to dissipation in one loss coefficient, the CFD tool FDS
simulates pressure losses for every point in the flow path by Navier-Stokes equations and dynamic Smagorinsky turbulence model in detail. Figure 1 shows the flow through one door
(1m x 2m) between two rooms in a 3m height compartment. The inflow conditions (4m/s 2m/s - 1m x 2m) are established by a door positioned a) parallel (direct flow) and b) rectangular to the door under observation (crossflow). Additionally the inflow and outflow conditions
are varied with room size of 10m in flow direction to 10m width (room) versus 4m width
(corridor). Cell sizes of xyz=0.05m0.05m0.05m were determined as optimum mesh
size by a sensitivity study.
Figure 1 visualizes the flow pattern through the different inflow conditions by velocity magnitude; red color indicates high velocities of >3m/s while blue indicates 0m/s. It can be observed that local velocities are much higher (>3m/s) than calculated mean velocities from
mass balance (2m/s).
Consequently there exist different flow patterns for the scenarios that lead to different distributions of pressure drop. Everywhere in the room, where a current is set up, a pressure drop is
created and intensified in regions with steep velocity gradients. For instance in the scenarios
with rectangular inflow conditions pressure drop is also caused by the flow that occurs within
the room far away from inflow or outflow door. The magnitude of pressure drop via the door
between the rooms / corridors shows a difference of 3.5Pa for the considered scenarios (simulated in the middle of the door on the longitudinal axis of the room in a height of 1m). The
resulting pressure drops of approx. 3Pa and 6.5Pa correspond to a flow coefficient of 0.89
respectively 0.60 according to equation (2) and cover a wide range. For smaller flow rates e.g.
1 m/s (corresponding to a velocity of 0.5 m/s through a door 1m x 2m) the pressure drop is
about half an order of magnitude and varies between 0.28 and 0.33 Pa. For the determination
of a flow coefficient in the examined scenarios, it should be considered that the flow coefficient of the jet flow from the inflow door interacts with the flow coefficient of the observed
door between the rooms. Hence smooth inflow conditions via the full room width, not through
a door are additionally visualized for the direct flow scenario (dashed line). According to figure 1, the inflow condition through a door that causes a highly turbulent flow behavior within
the room can be able to mitigate the pressure loss of a door in comparison to smooth inflow
condition.
Derived flow coefficients are in the range of 1.25-2.71 for Cd, local and 0.89-0.60 for
Cd,mass referring to a flow rate of 4m/s and to the flow rate of 1m/s in the range of 1.67-2.2
for Cd, local and 0.77-0.67 for Cd,mas .The latter correspond very well to the loss coefficients
of 0.68-0.73 suggested by Steckler in 1985 for fire induced mass loss coefficients through
windows and doors [32].
OUT
OUT
IN
IN
a)
corridor
direct
flow
b) room crossflow
b) corridor
crossflow
Figure1FlowpatternthroughadoorQ=4m/s,A=2m
FDS was used to demonstrate the behavior of flow via multiple resistances. As it can be seen
in figure (2) flow does not necessarily distribute equally via 3 resistances of the same size as it
is calculated with CONTAM. Distances and positions between inflow and different outflow
elements have a significant influence on the distribution of mass flow within the 3 rooms,
especially the set-up of a free jet flow determines the mass distribution. Figure 3 illustrates the
flow distribution in terms of fraction mass inflow in room to mass outflow through 3
openings.
Regarding the distribution and distance of inflow and outflow elements, figures (4) and (6)
confirm this behavior. In this simulation an eccentric set-up was chosen. However due to
other flow pattern the difference between mass flows via the openings decreases slightly. Additionally the influence of different opening sizes (1.5m x 1m vs. 0.3m x 0.5m) that leads to
different flow resistances is investigated (figure 5).
It is figured out that the eccentric inflow conditions in the room and hence the formation of a
free jet have a great influence on the mass distribution via the both openings, this is not reproduced with CONTAM modeling. However, the change of resistance due to changed opening
size within the flow path and centric flow distribution lead to the same results (figure 6). On
the other hand for suction out of the room scenario the comparison of FDS and CONTAM
show nearly no differences except in the modeling of the change in opening size (figure 7).
9. Pressure Losses and Flow Paths within a Stairwell
The discrepancies between empirically determined pressure losses inside the stairwell with
O1
O2
O3
Figure2Flowvisualizationforflowthrough3
rooms
O1
Figure3Massdistributionforflowthrough3
rooms
O1
O2
O2
Figure4Flowvisualizationforflowthrough2 Figure5Flowvisualizationforflowthrough
roomseccentric
2roomscentricwithchangedresistance
Figure6Flowthrough2doorsoverpressure Figure7Flowthrough2doorsunder
pressure
respect to their importance obey the need for further investigation. For this purpose, a simulation study with the Fire Dynamics Simulator was carried out. Here, two types of stairwells
where examined: one conventional stairwell with two landings, and one helical stairwell with
four landings as it was suggested by Poreh [18] (figures 8-11). Geometrical variation regarding well hole is taken into account. Furthermore a closed railing and a simple handrail are
compared. Different flow paths through a 30 m high stairwell (hfloor = 3.75m) with one door
open on the top floor are examined. Additionally a scenario with two open doors is considered, one on the top and one on the middle floor with varying position (a-e in Fig. 12, 13).
d c b
IN
Figure10planviewstairwell2
landings
IN
Figure8FDSmodel Figure9FDSmodel
2landingsclosedwell 4landingsclosedrailing
hole
Figure11planviewstairwell4
landings
During the investigation the pressure distributions on the landing are compared. All simulations have a volume flow of 4m/s, which corresponds to a (mean) velocity of 2m/s in a door
of 2m. Regarding figure 12 it can be observed that there is a great discrepancy between the
pressure losses caused by different geometries. Here the helical stairwell spans the widest
range, confirming Porehs observation of the smallest pressure loss for an open well hole.
Both stairwells have the lowest pressure drop when the flow is guided through the well hole
via a closed railing. The influence of leakages was studied likewise. Leakages through tight
and loose walls as well as through gaps of one door were summarized in the gap below a door
with Atigh = 0.003 m vs. Aloose = 0.02m (located at position d according to figure 10). Figure
13 shows that the leakage area has a great influence on the pressure relief with height, nevertheless the influence of mass leakage is comparatively small. For tight walls a mass loss of
3% was observed, for loose walls a value of 10% results.
Figure12Influenceofstairwellgeometryon
pressuredrop
Figure14Influenceofdoorpositiononflow
distributioninsidethestairwell
Figure13Influenceofwallleakagefor
stairwellwith2landingsandclosedwellhole
The influence of door position inside
the stairwell with multiple open doors
has also a great influence. Figure 14
shows that the mass flow rate through
the top door, that should be 2.4 kg/s
(20C, u = 1m/s) for a uniform distribution, is reduced up to 60 % depending on the door position (a-e according to figures (10-11). Hence also the
mean velocity through the top door
decreases to 0.6 m/s while the mean
velocity through the middle door increases to 1.4 m/s. In summery the
internal flow path and internal pressure losses have a great influence on
flow distribution and accordingly the
attained inflow velocity into the fire
floor.
The design criteria that affect the pressurization systems, the thresholds for door opening
force versus minimum pressure difference to the floors to avoid smoke from floors entering
through leakages of the wall and the critical inflow velocity require further investigation: Because the stairwell ventilation state has a significant effect on the heat release rate in the floor
(open doors below and above the fire floor) [33], the interaction between fire and flow
through the stairwell should be part of the performance based design. Likewise the applicability of the critical ventilation velocity as a principal parameter in tunnel design with a similar
approach to stairwell pressurization was questioned recently by Vaitkevicius et al. [34]. It was
discovered, that local flow phenomena in tunnels also have a significant impact on the spread
of smoke that could not be captured by one critical design value. Hence the pressurization
concept could not be described with a simple balance of smoke flow (low pressure side) versus critical velocity (high pressure side).
11. Conclusion
For pressurization systems, airflow patterns inside the stairwell as well as across doors control
the quantities of interest: the pressure drop due to dissipation caused by internal friction which
increases with intensity of turbulence. Available coefficients are tied to fixed scenarios and
may be suitable for ventilation purpose but in pressurized case with higher flow rates, velocities and dissipation effects need to be expanded.
Additionally the electrical resistance analogy concept used for flow distribution does not take
into account the complex geometry of the flow. Especially the formation of free jets in rooms
can lead to strong deviations between model and reality, which does not appear when flow is
caused by suction. For modern open plan scenarios this effect could be captured by the implementation of additional analytical relations for free jets or the coupling of CFD with multizonal models [35]. It is essential that the designer knows about those limitations and is sensitive to the degree of accordance between multizonal modeling and real flow behavior.
The quantification of air leakage paths and leakage areas with common flow coefficients for
realistic calculation of the pressure profile due to stack effect is nearly impossible to meet by
designers: First of all measured and tabularized values have a wide range and engineering
judgement is very difficult for one certain building, secondly building leakage is determined
on the construction side and varies per building in every floor. Only with field measurements
under operating conditions the designer will be able to identify real leakage in detail.
Additionally, pressure fluctuations due to wind effects on the building as well as the uncertainties associated to the nature of the fire [1] lead to a highly transient flow behavior that
cannot be represented by steady-state correlations. Prtrel et al. [36] showed that within the
first 5 minutes of the fire in a confined and force ventilated enclosure there was a huge pressure variation measured that was affected by the fire size, the ventilation features and the enclosure characteristics like the conductivity possibility of the walls or the air flow resistance
of the ventilation system (flow inversions in the ventilation branches). This can only be included by transient CFD calculations depending on fire size.
Furthermore, factors that influence the pressure profile at doors to the fire room are not yet
completely understood e.g. air pumping effect of room fires which is related to the size of the
room openings in relation to the heat released [37].
Hence for the safe design of a pressurization system it is of crucial importance to define pressure drops on the flow path and furthermore ensure that flow paths are build-up and kept in a
fire scenario. This should also be implemented in fire management and the evacuation concept. Human influence in case of unintended closing and opening of doors can significantly
affect the pressurization system and need to be considered in design. Additionally the influence of human beings in the flow path resulting in greater pressure losses needs to be studied
further.
12. References
1. Cowlard, Adam, et al. Fire Safety Design for Tall
Buildings. 2013. pp. 169-181. 9th Asia-Oceania
Symposium on Fire Science and Technology
2. Lay, S. Pressurization systems do not work and present
a risk to life safety. Case Studies in Fire Safety. 2014,
2214-398x.
3. Walton, G.N and Dols, W.S. CONTAM 3.0 User Guide
and Program Documentation. Gaithersburg, MD : NIST,
2013.
4. McGrattan, K., et al. Fire Dynamics Simulator, User's
Guide. NIST, 2013. NIST Special Publication.
5. Klote, John H. Handbook of Smoke Control
Engineering. [Ed] Paul G. Turnbull James A. Milke.
sASHRAE, 2012.
6. Emmerich, S. J. und Persily, A. K. Analysis of U.S.