USING MODELS

TO SUPPORT
Smoke
Management
By James A. Milke, Ph.D., P.E.
System Design
A n engineering analysis is • analysis of the pre-venting, smoke- Froude modeling. Quintiere5 provided
needed to assess the ability of filling period or steady, equilibrium a review of scaling relationships to
a smoke management system conditions during venting. preserve the Froude number. The scal-
to satisfy stipulated performance crite- Algebraic equation methods cannot ing relationships seek to preserve the
ria. This analysis can be conducted to address the interaction between multi- following ratios:
verify acceptability, numerically test, or ple smoke management systems, such • fire energy/flow energy;
troubleshoot problems associated with as stair pressurization and atrium • fan flow/buoyant flow; and
the system. smoke exhaust, in the same building. • convection heat transfer/wall heat
The pressure difference between Several types of models are available transfer.
spaces is the focus of analyzing stair to assist design professionals either in
pressurization and zoned smoke con- lieu of or as a supplement to the alge- The scaling relationships are:
trol systems. For smoke management braic equations. These models include
systems in atria, the analysis consists small-scale physical models, computer- Temperature: Tm = TF (1)
of assessing the residual hazard posed based zone models, network flow
by smoke in terms of the extent of models, and CFD models. While this
l 
smoke spread, smoke layer depth, or article provides an overview of all of Position: xm = x F  m  (2)
smoke layer properties. In many cases, these models, the emphasis is on  lF 
a simplified analysis involving the small-scale models and network mod-
l 
application of algebraic equations1, 2 is els, given the extensive treatment of Pressure: ∆pm = ∆pF  m  (3)
suitable to assess the performance of zone and CFD models elsewhere.  lF 
smoke management systems. 1/ 2
SMALL-SCALE MODELS l 
However, in some cases, the Velocity: vm = vF  m  (4)
assumptions associated with the alge-  lF 
braic equations are unacceptable. For Small-scale models provide physical 1/ 2
representations of a space, though in a l 
stairwell pressurization systems, the Time: tm = t F  m  (5)
algebraic equations neglect vertical reduced scale. Scale models are espe-  lF 
leakage and wind, and require symme- cially useful in examining atria with Convective Heat Release:
try if more than one pressurized stair numerous projections or irregular 5/ 2
l 
is provided. Limitations of algebraic shapes. Milke and Klote review the Qc, m = Qc, F  m  (6)
equation methods for atrium smoke application of scale models as a design  lF 
management are: aid for smoke management systems3. Volumetric Flow Rate:
• steady or t2 fires only; Quintiere and Dillon developed a scale 5/ 2
model to assess the performance of a l 
• uniform horizontal cross-sectional V˙fan, m = V˙fan, F  m  (7)
area for all levels of the atrium; smoke management system in a fire  lF 
• uniform conditions throughout the incident in a covered mall.4
upper layer/zone, even in spaces A small-scale model may be The subscripts m and F correspond to
with large horizontal areas; and designed following the principles of model and full-scale, respectively.

S UMMER 2000 ©2000SFPE. All rights reserved. Fire Protection Engineering 17

 Many of the parameters in equations Thermal properties: COMPUTER-BASED ZONE
(1) to (7) are functions of time. Thus 0.9 MODELS
 lm 
the scaled parameters should also be
functions of time. Froude modeling
(kρc )
p w, m (
= kρc p ) w, F  
 lF 
(8)
Overviews of numerous zone mod-
has the advantage of conducting els are available.6, 7 Quintiere summa-
experiments in air at atmospheric pres- However, selection of enclosure rizes the assumptions of zone models8.
sure. While Froude modeling does not materials may be acceptably based on The principal advantage of computer-
preserve the Reynolds number, achiev- flow visualization needs, rather than based zone models is their ability to
ing fully developed flow by making scaling of thermal properties, given the address transient effects involving
the critical dimensions of the model at secondary effect of the thermal proper- smoke spread, delays in fan startup,
least 0.3 m minimizes this shortcom- ties on fluid flow. effects of environmental conditions,
ing. Fully developed flow only needs and a variety of fire-growth profiles.
to be achieved in those areas where EXAMPLE 1 Some computer-based zone models
the smoke behavior is of interest. The are applicable to spaces where the
critical dimension for a model of a A scale model is proposed to deter- ceiling is sloped or the horizontal
shopping center and atria could be the mine the equilibrium smoke layer cross-sectional area varies with height.
distance from the floor to the under- position for the atrium depicted in In addition, some of the computer-
side of a balcony. Figure 1. Because the horizontal cross- based zone models simulate conditions
In addition, Froude modeling does sectional area varies with height, alge- in multiple rooms or levels, where the
not preserve the dimensionless heat braic equation and computer-based algebraic equations are limited to a
transfer parameters. Often, this limita- zone models are of limited value. The single compartment.
tion has little effect because the tem- atrium height is 30.5 m, and the design
perature is the same for the scale fire is a 5 MW steady fire. The pro- Limitations of these models result
model and the full-scale facility. While posed exhaust fan capacity is 142 m3/s. from their assumptions. For example:
Froude modeling is inapplicable in By applying the scaling relationships, • the smoke layer forms immediate-
high-temperature locations, e.g., near the basic parameters for the scale ly, neglecting transport lag;
the flame, Froude modeling still pro- model are: • the plume is unaffected by wind
vides useful information about smoke • Height: 3.8 m tall model (1/8 scale) or mechanical ventilation;
transport away from the fire. • Fire size: 28 kW • the upper layer/zone is uniform,
Some surface effects can be pre- • Fan capacity: 0.78 m3/s independent of the area involved;
served by scaling the thermal proper- Given the 1/8 scale, the width of the and
ties of the construction materials for scaled spill plume would need to be • as smoke enters a tall room from
the model. The thermal properties can 1/8 of that at full scale. a short room, entrainment is
be scaled by: determined based on a new
axisymmetric plume rather than
from a balcony spill or line
plume.

FIELD MODELS

Figure 1. Small-Scale Model of Atrium Computational fluid dynamics (CFD)

models simulate fluid flow at a level of
detail impossible with other methods
of computer modeling.9, 10, 11 CFD mod-
els divide the fluid flow field into
numerous small cells and numerically
solve the conservation equations of
mass, momentum, and energy for each
cell. Boundary conditions are estab-
lished at the room boundaries, open-
ings to the outside, and exhaust inlets
by specifying velocities.
While generalizations concerning the
number and size of cells are difficult
to make, given the wide range of fea-
tures and capabilities of CFD models,
generally the smallest cells are near
the fire and at the ceiling. The govern-
ing equations cannot account for tur-

18 Fire Protection Engineering N UMBER 7

bulence on a scale smaller than the • interacting smoke management ρV˙ = .071Q˙ c1/ 3 z 5 / 3 + 0.0013Q˙ c (9)
cells. Further, it is important that the systems;
cell size and time step be coordinated • buildings with complex geome- where:
so that cells are not “skipped” from tries; and V̇ = volumetric entrainment rate (m3/s)
one time step to the next by the mov- • leakage paths between building ρ = density of smoke (kg/m3)
ing fluid. spaces. Q̇ = convective portion of heat release
rate (kW)
NETWORK FLOW MODELS Limitations of network flow models z = clear height (m)
such as CONTAM96 are:
Network flow models such as • uniform conditions (temperature The entrainment for a particular
CONTAM9612 can be applied to evalu- and concentration of contami- level needs to be determined based on
ate pressure differences between com- nants) are assumed throughout the amount of air entrained only with-
partments, direction of mass flows, each “zone”, where a “zone” is at in that increment of height. As such,
and movement of contaminants. least one room. the amount of air entrained for a par-
CONTAM96 is often described as the • transient fire conditions (e.g., tem- ticular level of the building is the dif-
successor to ASCOS, an early, widely perature or mass flow in the ference in the amount of air entrained
used network flow model13. Network smoke plume or temperature in up to the top of the level with that
models simulate a building as a net- the fire zone) resulting from a entrained up to the bottom of that
work of airflow paths comprised of growing fire are not considered. level. Buoyancy effects can be includ-
doorways, windows, vents, and leaks ed by setting the temperature at each
in building assemblies. However, fire conditions can be level of the “shaft” using Heskestad’s
incorporated into the model by analo- plume centerline correlation.14
The principle advantage of network gy. The volumetric flow in a smoke
flow models is their ability to consider: plume can be simulated as a shaft, with ∆Tc = 25Q˙ c2 / 3 z −5 / 3 (10)
• mechanical and natural ventilation; a fan supplying air at each level of the
• environmental conditions (includ- shaft. The air entrained at each level where:
ing wind); can be estimated using Heskestad’s Tc = plume centerline temperature (°C)
plume entrainment equation.14 Q̇ = convective portion of heat release
rate of fire (kW)
z = height above top of fuel (m)

While the centerline temperature of

the plume overestimates the buoyancy
of the overall plume, generally this
approach is adequate for design pur-
poses.
The mass release rate of contami-
nant can be estimated as:
Q˙
m˙ c = fc (11)
∆Hc
where:
ṁc = generation rate of contaminant
(kg/s)
fc = yield fraction (kg contaminant/kg
Atrium fuel) [the yield fraction depends
on the fuel, burning mode (flam-
ing, smoldering) and the avail-
able oxygen concentration]
Q̇ = heat release rate of fire (kW)
∆Hc= heat of combustion (kJ/kg)

CONTAM96 can conduct a steady or

unsteady analysis of the flow of conta-
minants (smoke). Input for CONTAM96
Hotel rooms includes:
• Area and height of spaces
• Shaft characteristics
Figure 2. Typical floor of five-story building. • HVAC system

S UMMER 2000 Fire Protection Engineering 21

 7 Walton, W.D., and Budnick, E.K.,
Table 1. “Deterministic Computer Fire Models,”
Fire Protection Handbook, 18th Ed., J.L.
Pressure Difference at Stairwell (Pa) from CONTAM96 Analysis Linville (ed.), NFPA, Quincy, MA, 1997.
8 Quintiere, J.G., “Compartment Fire
No Atrium 160 m3/s exhaust 160 m3/s exhaust Modeling,” SFPE Handbook of Fire
Protection Engineering, 2nd Ed., P.J.
Floor Exhaust 76 m3/s supply 94 m3/s supply DiNenno (ed.), NFPA, Quincy, MA, 1995.
9 Chow, W.K., “A Comparison of the Use
1 54.8 65.2 61.0 of Fire Zone and Field Models for
Simulating Atrium Smoke-Filling
2 55.0 65.7 61.5 Processes,” Fire Safety Journal, 25, 4,
1995, pp. 337-354.
3 55.3 66.2 62.0 10 McGrattan, K.B., Baum, H.R., and Rehm,
R.G., “Large Eddy Simulations of Smoke
4 55.8 67.0 62.7 Movement,” Fire Safety Journal, 30, 2,
1998, pp. 161-178.
5 56.8 68.0 64.0
11 Rho, J.S., and Ryou, H.S., “A Numerical
Study of Atrium Fries Using
6 58.0 69.5 65.2 Deterministic Models,” Fire Safety
Journal, 33, 3, 1999, pp. 213-230.
• Fans: constant volume, mass, or as an aid in establishing or testing the 12 Walton, G., CONTAM96, National
curve design of a smoke management sys- Institute of Standards and Technology,
• Environmental conditions: wind, tem. The appropriateness of assump- Gaithersburg, MD, 1997.
temperature tions should be confirmed, either by 13 Klote, J.H., “A Computer Program for
• Connection of spaces via leakage comparing predictions to data or con- Analysis of Pressurized Stairwells and
paths ducting a sensitivity analysis. Pressurized Elevator Shafts,” NBSIR 82-
• Release rate of contaminant 2512, National Bureau of Standards,
(kg/s): unsteady or steady James Milke is with the University of Gaithersburg, 1982.
Maryland. 14 Heskestad, G., “Engineering Relations
Output from CONTAM96 includes: for Fire Plumes,” SFPE TR 82-8, SFPE,
• Pressure difference between zones Boston, MA, 1982.
• Airflow between zones
• Contaminant concentration in REFERENCES
zone
1 Klote, J.H., and Milke, J.A., Design of
An example application of Smoke Management Systems, American
CONTAM96 is provided for a five-story Society of Heating, Refrigerating and Air-
building (see Figure 2). Using Conditioning Engineers, Atlanta, GA.
1992.
CONTAM96, the interaction between
the atrium smoke management and 2 NFPA 92B, Guide for Smoke Management
stairwell pressurization systems is Systems in Malls, Atria, and Large Areas,
investigated. The capacity of the stair- Quincy: NFPA, 1995.
well pressurization fans is 2.83 m3/s. 3 Milke, J.A., and Klote, J.H., “Smoke
The exhaust fan capacity in the atrium Management in Large Spaces in
is 160 m3/s, with two simulations con- Buildings,” Building Control Commission
ducted with the capacity of the make- of Victoria, Victoria, Australia, 1998.
up air fans being either 76 or 94 m3/s. 4 Quintiere, J.G., and Dillon, M. E., “Scale
The resulting pressure differences are Model Reconstruction of Fire in an
provided in Table 1. The pressure dif- Atrium,” Proceedings of the Intl. Conf.
ferences for the stairwell pressurization On Fire Research and Engineering, BFRL
systems acting alone or together with and SFPE, 1995, pp. 397-402.
the atrium smoke management system 5 Quintiere, J.G., “Scaling Applications in
are appreciably different. Fire Research,” Fire Safety Journal, 15,
Computer-based and physical mod- 1989, pp. 3-29.
els are applicable as aids for smoke 6 Friedman, R., “An International Survey of
management design. Because each of Computer Models for Fire and Smoke,”
the models provides simplifications of Journal of Fire Protection Engineering, 4,
actual behavior, models can be used 3, 1992.

22 Fire Protection Engineering N UMBER 7