11 INDUSTRIAL FIRE JOURNAL / April 2008

FIRE ENGINEERING REPORT

flow in a duct, there will be a development region
where the turbulent boundary layers grow and
eventually merge, leading to local equilibrium where
the flow no longer changes. Typically this should
occur around 25-40 ducts diameters from the inlet.
The research into this subject reviewed the results
of sampling experiments in exhaust duct stacks in
the nuclear industry where a passive tracer gas was
released from a single point in a duct. This showed
that even allowing for turbulence generated by
bends, the distance before well-mixed conditions are
developed can be significant and will be comparable
to the length of the development region for fully-
developed turbulent flow.
In a series of experiments, concentration
measurements were taken over the cross section of
the duct downstream from the release. To
characterise the degree of mixing a parameter called
the Coefficient of Variation (COV) was introduced.
This parameter is defined as the sample standard
deviation divided by the sample mean, for all points
sampled at a cross-section. It can be used to quantify
the uniformity of mixing in a duct, and can be readily
computed from CFD results.
Based on this principle the American standard
ANSI/HPS N13.1-1999 permits single point sampling
of gaseous contaminants in a duct, provided the COV
for both velocity and concentration of a tracer gas are
less than 0.2 over the central two-thirds of the duct.
The literature reviewed for this work reports on
experiments examining the COV and its dependence
on turbulence intensity, and obstructions or mixing
elements within ducts. The performance of static
mixing elements was very variable but the 0.2 COV
criterion was met for the examples tested within a
distance of nine duct diameters downstream. The
most effective mixing elements were able to meet
the criterion within three duct diameters of the
mixing element.
The effect of grilles on the turbulence in a duct is
relatively well understood, and either suppresses or
enhances turbulence depending on the grille
geometry. A fine mesh will tend to suppress
turbulence, and any mesh-induced turbulence will
decay quickly because of its small scale. A grid of
relatively large rods will enhance turbulence, but it
was found that even if turbulence was significantly
enhanced the COV is likely to remain high for long
distances downstream.
For typical HVAC duct equipment such as louvres,
grilles, and fire dampers, although these features
generate turbulence and the flow may well be
turbulent across the duct inlet, it would be wrong to
simply assume that mixing will be rapid and
contaminants will be quickly dispersed to give well
mixed uniform conditions. Even with sharp edged
duct entrances, it is concluded that up to 50 duct
diameters may be needed to achieve uniform mixing
of a passive tracer in a straight pipe.
Having reviewed the literature the research
proceeded to examine Computational Fluid
T
hat there is little information on siting of gas
detection for HVAC was highlighted by an
incident in late November 2004 on an
offshore platform in the North Sea. There
was a delay in confirmed detection of gas and
subsequent shutdown of HVAC systems, despite gas
being ingested in the HVAC intakes. Following this
incident the Health and Safety Executives Offshore
Safety Division funded research to examine
flammable gas releases around and into HVAC ducts.
In particular the circumstances that might lead to
non-uniform distributions of gas around and inside
HVAC duct inlets were examined, and the
suggestions regarding the gas detection
arrangements in place have been considered.
This article is a brief summary of a full research
project (2,3) and has highlighted some issues for this
topic, as well as reminding us of some points that
may have been overlooked with time.
Gas flow and mixing in a duct
A typical offshore HVAC duct will generally produce a
Reynolds number in the range from 105 to 106, and
this will ensure fully turbulent flow some distance
downstream from the entrance to a duct. However,
any gas flow outside the duct and in the vicinity of
the entrance may well be non-homogeneous. It is
important to know when gas can be considered to be
well-mixed across a duct, since this has implications
for selection and siting of gas detectors Normally, for

Arrangements for providing and siting gas detectors

for open areas and gas turbine enclosures (1) are
generally considered to be reasonably well
understood. But there is little information on the
provision and siting of gas detection systems for HVAC
(heating, ventilation and air conditioning) ducts
supplying air to accommodation modules, temporary
refuges or process areas, writes Kevin ODonnell,
Offshore Safety Division, Health and Safety Executive.
Figure 1, geometry and
location of HVAC inlets.
Top: figure 2,
Iso-surface of gas at 10
per cent LEL.
Flammable gas
mixing and detection
in HVAC ducts
www.industrialfirejournal.com
INDUSTRIAL FIRE JOURNAL / April 2008
FIRE ENGINEERING REPORT
12
HVAC inlet (approx. 6m x 4m) the gas distribution is
far from uniform, as illustrated in Figure 3.
However, it is not just large ducts which can exhibit
such non-uniformity in gas concentration. Figure 4
shows the gas distribution inside the smallest of the
three ducts, with a 1.5 m square cross-section. The gas
concentration is again very non-uniform; the COV is
0.43 at 2 m inside the duct, and gas concentration
varies from approximately 3% LEL to well over 20%
LEL at this location. Although the average gas
concentration in this smaller duct is approximately
16 % LEL, it falls below 10% over a significant part of
the cross-section. Post-processing of these results
shows that beam detectors located 2 m inside the
duct and oriented across the width of the duct would
indicate a gas concentration of approximately 16%
LEL per m which is comparable to the average
concentration.
Discussion
In the CFD simulations performed for this research,
the distribution of gas at HVAC inlets is found to be
non-uniform. This is consistent with empirical data
on unimpeded high pressure gas releases. Thus for a
release of pure methane at a stagnation pressure of
100 bar from a hole of 12 mm diameter, the
concentration at approximately 10 m downstream
from the release would be 100% LEL on the jet
centreline but just 10% LEL at a radius of 1.9 m. This
distance, over which the concentration varies by a
factor of ten, is broadly comparable to the
dimensions of typical offshore HVAC inlets. If such a
release were ingested into an HVAC inlet then
significant non-uniformity in gas concentration
could be expected outside and immediately inside
the HVAC duct: large variations in gas concentration
can be expected to be present over the cross-section
of the modelled inlets, and within the HVAC ducts.
This implies a notable potential for gas releases to be
missed by detection systems unless this non-
uniformity in gas concentration is anticipated in the
selection and siting of gas detectors at HVAC inlets.
The CFD results also show that a variation in gas
concentration over a duct cross-section only reduces
slowly with distance along a straight duct. These
findings are consistent with theoretical
considerations for gas distribution in a high pressure
jet or low pressure buoyant plume, and from the
literature covering the sampling of gas distributions
in the exhaust ducts of nuclear stacks.
The literature highlights that purpose-designed
mixing elements and bends in a duct can be effective
in creating well-mixed conditions but at the cost of
increased pressure drop. It also points out that
relatively small-scale obstructions, such as louvres
and fire dampers, are unlikely to significantly
enhance mixing. This is borne out by CFD modelling
of such obstructions in this study.
The implications of the modelling work,
substantiated by the literature, are that in the
absence of purpose-designed mixing elements or a
series of bends upstream from gas detectors, no
significant benefit would be gained from siting
detectors a significant distance downstream from an
HVAC inlet. Also, no significant benefit can be
expected to be gained from siting detectors inside an
HVAC duct compared to locating them immediately
outside the HVAC inlet.
Therefore based on the results of this work so far,
the recommendations made in the report for
flammable gas detection strategies for offshore
HVAC ducts are as follows:
Detector alarm levels should be set as low as
reasonably practical: 10% LEL or less.
Point catalytic, point infra-red, extended path
point infra-red, cross-duct beam infra-red and
Dynamics (CFD) modelling of some representative
scenarios.
CFD modelling of gas ingestion and
distribution inside HVAC ducts
A number of scenarios were modelled, but here well
concentrate on one which draws upon some
elements of an incident on Brae Alpha in 2004. A
high pressure gas riser failed, the failure location
being at the top of the riser underneath the main
body of the platform. The diagrams at Figure 1
illustrate the simplified representation used for the
CFD modelling. The domain includes a large region
of atmosphere surrounding the platform, with a
wind speed of 12.3 m/s, and a gas release rate of 2 kg/s
modelled as a high pressure release using a pseudo-
source approach in conjunction with empirical data.
CFD simulations were undertaken using ANSYS CFX
10 software, in time-dependent mode.
Figure 2 illustrates one aspect of the simulations, an
iso-surface of gas concentration at 10% LEL. Inside
this volume the concentration will be greater than
10% LEL. The gas is shown as spreading throughout
and underneath the lower elements of the platform
and is clearly illustrated being ingested into the
smallest of the three HVAC inlets. It is not so clear in
the diagram, but inside the very large Hazardous
Figure 3, gas
concentration
distribution 3.5 m
inside the hazardous
HVAC inlet duct.
With the new Altair 4, MSA reckons that it has
developed one of the most rugged four-gas detectors
available.
It measures O2, H2S, CO and combustible gas
simultaneously, plus it features the MotionAlert.
When enabled, the man-down alarm option alerts
with LEDs and a piercing audible alarm anyone in
the vicinity if the user is in distress .
The Altair 4 comes with a high performance triple alarm system, 24-hour bump
test check-mark and comprehensive data-logging facility as standard. The
rechargeable battery provides upto 20 hours of continuous operation.
It also features a durable rubberised housing for water and dust ingress
protection [IP 67]. A large numeric display allows users to view gas
concentration, alarm conditions, battery status, set points and peak reading with
accuracy and confidence. An intuitive three-button operation ensures the Altair 4
is extremely easy to use, even when wearing gloves.
The triple alarm system includes a distinctive 95 dB alarm, strong vibrating
alarm and super-bright LEDs on the top and bottom to ensure visibility from all
angles. A separate LED clearly shows the unit is operating correctly by flashing
every 15 seconds.
Simple testing of the detectors alarms along with response to a known gas
concentration can be carried out by the QuickCheck test station. Alternatively,
if a stand-alone automated record keeping, calibration, testing and charging
system is required, Galaxy is the Altair 4s perfect companion.
The toughest kid on the block?
No significant
benefit can be
expected to be
gained from siting
detectors inside an
HVAC duct
compared to
locating them
immediately outside
the HVAC inlet.
www.industrialfirejournal.com
INDUSTRIAL FIRE JOURNAL / April 2008
FIRE ENGINEERING REPORT
14
aspirated point detector systems all have the
potential to be effective in detecting non-
uniform distributions of flammable gas in and
around HVAC ducts, provided their sensitivity is
sufficiently high (low detection limit), and it is
recognised that gas could be distributed non-
uniformly.
Extended path point infra-red detector systems
currently appear to offer the greatest sensitivity
for this situation, but multiple detectors should
be used and sited so as to anticipate non-uniform
mixing.
Cross-duct beam infra-red, extended path or
aspirated point detector systems should be
based on two approximately orthogonal beams
or lines of aspirated point probes.
No significant benefit can be expected to be
gained from siting detectors inside an HVAC duct
compared to locating them immediately outside
the HVAC inlet.
In the absence of purpose-designed mixing
elements or a series of bends upstream from gas
detectors no significant benefit is to be gained
from siting detectors a significant distance
downstream from an HVAC inlet.
Mixing elements have the potential to reduce
any non-uniformity in the distribution of gas in a
duct but their effectiveness should be proven by
physical tests. Now it should be noted that these
recommendations are based on a limited
amount of work from CFD modelling and the
published literature. CFD modelling has inherent
uncertainties and it is recognised that the
findings from the literature are not based on
offshore HVAC systems as such. However the
CFD modelling demonstrates that there is a
possibility of significant non-uniformity in the
distribution of gas inside and around an HVAC
inlet.
It is difficult to provide firm guidance on how many
point or extended path detectors should be used
since this depends on the size and shape of a duct.
However, it is recommended that there should be
good coverage of the cross-section of the duct. This
can be achieved by two infra-red beams arranged
approximately orthogonally, either as open-path
cross duct or extended path point infra-red, or lines
of aspirated point probes.
For the time being it is recommended that the
findings of this work should be taken into account by
those who have responsibility for HVAC system
design, manufacture and maintenance, where
hazardous atmospheres are likely to be encountered.
HSE are considering substantiating these results
through physical trials using real detectors. S
References:
1. Guidance note PM 84: Control of safety risks at gas
turbines used for power generation. HSE, 2000.
2. Assessment of gas detection strategies for offshore
HVAC ducts based on CFD modelling. Health and Safety
Laboratory. Report RR602,
http://www.hse.gov.uk/RESEARCH/rrpdf/rr602.pdf3.
Assessment of flammable gas ingestion and mixing in
offshore HVAC ducts: Implications for gas detection
strategies. C J Lea, M Deevy and K ODonnell, HAZARDS
XX, University of Manchester, April 2008.
The MX6 iBrid is the new full featured multi-gas
monitor from Industrial Scientific.
The user-friendly design of the unit features a full-
color liquid crystal display and five way navigation
button to scroll through the menu-based
operating system.
Designed to meet a variety of monitoring
applications, the MX6 iBrid can be set up to detect
from one to six gases, including options for PID
and infrared sensors for combustible gases and
Carbon Dioxide.
Configured with an internal sampling pump or for
diffusion operation, it is ideal for
applications
from confined
space entry to
personal
survey
monitoring.
The MX6 iBrid
is compatible
with the DS2
Docking
Station and
iNet
Instrument
Network to
further
simplify and
automate
calibration,
function
(bump) testing
and data
management.
Figure 4, gas
concentration
distribution inside the 1.5
m2 PLQ duct, at 2m
inside the duct (right),
and at 6 m inside the
duct (bottom).
New iBrid MX6 from InduSci
Figure 4 shows the
gas distribution
inside the smallest
of the three ducts...
The gas
concentration is
again very non-
uniform; the COV is
0.43 at 2 m inside
the duct, and gas
concentration varies
from approximately
3% LEL to well over
20% LEL at this
location.
www.industrialfirejournal.com