Gas Detector Coverage Using Gaussian Dispersion Modeling PDF
Gas Detector Coverage Using Gaussian Dispersion Modeling PDF
Gas Detector Coverage Using Gaussian Dispersion Modeling PDF
Edward M. Marszal
President and CEO
Kenexis
Edward.marszal@kenexis.com
+1 (614) 451-7031
Abstract
In the past few years, the use of quantitative risk analysis tools to determine the required number of
location of gas detectors has gotten more attention in the literature and also in real world project
execution. While there is great interest in these performance based methods, often referred to as fire
and gas mapping, there is little information available on how these types of studies are performed. This
paper will provide a discussion of how gas detector coverage, as defined in the ISA 84.00.07 technical
report Guidance on the Evaluation of Fire and Gas System Effectiveness is calculated. Specifically, the
paper will address the definition of scenario coverage and how it is more accurate and effective than
more common, but less sophisticated geographic coverage, because it takes into consideration the
actual leak sources and nature of the leak scenarios.
The paper will begin with a discussion of the creation of Gaussian dispersion models from a sufficient set
of leak sources and leak conditions. These modeling results will then be manipulated to consider
multiple release orientations and their relative frequency, along with consideration of wind directions by
modifying the relative directional release frequency based on wind direction frequency. Next a
discussion of integration of the risk is presented, discussing accumulation release frequency at all
geographic positions based release scenarios present in those locations. Coverage calculation is then
described as the determination of which scenarios are detectable given the detector locations relative
to locations of released gas clouds. Resulting the calculation of coverage as a fraction of detected
scenario frequency over total scenario frequency. The paper will then go on to provide an example of
how results are typically presented both graphically and in a tabular format. The paper will give an
overview of the entire process using a case study of a well bay in an offshore oil production platform.
Introduction
All safety instrumentation needs a basis of safety, or a set of rules or procedure that allows the designer
to make systematic and repeatable decisions about the two key factors of safety instrumentation
where? and how much? For the problem of gas detector placement, the most basis of safety is expert
judgment. That means that a single person or a handful of people in a plant will basically use their gutfeel, supplemented with some rules of thumb to place detectors. Even worse, sometimes these experts
are not even operating company stuff, but subcontractors from engineering companies and equipment
vendors. This informal design process resulted in designs that were inconsistent and not repeatable.
In 2005, the International Society for Automations SP 84 committee undertook the development of a
technical report that would define the techniques and processes that could be used to convert the
process of defining how many detectors are required and where to place them into a systematic and
repeatable process based on quantitative risk analysis. Working group 7 was formed, and released the
technical report ISA TR 84.00.07 technical report Guidance on the Evaluation of Fire and Gas System
Effectiveness. This technical report defined a new quantitative concept that represents how well a fire
or gas detection system functions called Coverage. The technical report defined two types coverage
geographic coverage and scenario coverage.
Geographic coverage was quickly adopted based on its ease of calculation. Geographic coverage is the
fraction of a given area where if a gas cloud of design basis size or larger were to exist, the gas cloud
would be detected by gas detector array. This process only requires the determination of a design basis
or critical gas cloud size. For combustible gases, many operating companies used a 5m diameter
spherical cloud based on work from the United Kingdoms Health and Safety Executive. An example of
the result of geographic mapping is shown in Figures 1 and 2.
Figure 2 Geographic Coverage Map Superimposed into 3D Model with Gas Detector Coverage
Bubbles Shown in Wire Frame
While geographic coverage was quickly adopted, it was not long before risk practitioners began to
question its accuracy and effectiveness. The same attributes that made geographic coverage
calculations easy to perform were also viewed as limits on the processes accuracy. Specifically,
geographic coverage suffers from the following limitations.
Only performs geometric analysis that is based on geometry of Design Basis gas cloud
Does not consider origin of the leaks
Does not consider how wind direction and speed affect gas cloud locations
Does not consider the relative frequencies of leaks from different sources
In order to address the limitations of geographic coverage, many risk practitioners moved to scenario
coverage in order to improve the quality of their analysis over the geometry only results generated by
scenario coverage.
Scenario Coverage
Unlike geographic coverage, which focuses on the geometry of a volume of detection (called a cone
of vision for fire detectors) that represents the what the detector see, scenario coverage delves much
deeper into the analysis by considering the size and location of all the potential gas clouds that could
occur as the result of a leak in a piece of equipment in the covered area intersecting the location of the
hazard with the location of the detection equipment. It can easily be seen that this type of approach is
significantly more powerful the geographic coverage approaches, but it is also very apparent that the
level of effort to perform the analysis has increased by orders of magnitude, not only in the engineering
effort required to developed all of the new scenarios, but also in terms of sheer computing power as the
quantitative analysis that must be performed for this type of analysis must be repeated thousands of
times in different directions and weather conditions.
Scenario coverage can be summarized in the following steps:
For the balance of this paper, the steps will be discussed in more detail and implemented in an example
problem. The example problem is combustible gas detection of for well bay of an offshore gas
production platform, but the concepts can be directly applied to virtually any facility in the process
industries. A simplified plot plan of the facility is shown in Figure 3.
conditions are close enough together that a single modeling run will be representative for all of the
selected equipment. A stream is a set of process conditions that includes the following parameters:
Temperature
Pressure
Composition (Fractions of each component chemical)
Development of a stream list is typically a fairly straightforward exercise. Many process flow diagrams
from chemical process industry plants already contain this information. Not only is the stream defining
information included in heat and weight balance tables, the equipment items where those conditions
apply are also generally clearly shown on the process flow diagram itself with flag that relate a heat and
weight balance table to the related equipment section.
For the sample facility, all of equipment shown in the module, whether well head or piping, would be
included in a single stream because the temperature, pressure, and composition contained in the
process equipment is very similar.
Dispersion Modeling
Once the streams have been defined, dispersion modeling is performed. While this paper specifically,
and the state-of-the-art in industry in general, is to use Gaussian dispersion models, the same concepts
apply to computational fluid dynamics models, which the author expects to replace Gaussian dispersion
modeling for this task in the future. A series of dispersion models will be run for each stream. The
number and types of models will generally follow the customs of quantitative risk analysis studies with a
few modifications to suit the problem of gas detection mapping.
While a discussion of dispersion modelling is out of the scope of this discussion, a few points that make
dispersion modeling for gas detector mapping are important. When performing dispersion models for
gas detection modeling, the number of hole sizes selected for releases is smaller, and focuses on the
smallest traditionally 5mm diameter. While large hole sizes may contribute to higher risk levels
because of the consequences they can generate, due to their large size, they are also very easy to detect
with a small number of detectors. As a result, including these scenarios into the risk profile generally
does not assist in the task of detector placement. Generally, only one wind direction and release
orientation are considered in the dispersion modeling, the full range of these variables are accounted
for later in the analysis through probabilistic methods. While traditional dispersion modeling calculates
distances out to end points that represent various levels of consequence to humans or equipment, the
dispersion modeling for gas detection mapping is only concerned with the alarm set point of the gas
detection equipment.
The final consideration for dispersion modeling for gas detector mapping is elevation. When modeling is
done in three dimension, the only elevation of interest with respect to results are the two detection
elevation planes. Most processes have one or two gas detection elevation planes depending on the gas
hazards that are present. For dense gases, a detection plane is typically set at 0.5 meters, low enough to
detect dense gases but high enough to prevent kicking, splashing, and minor flooding from causing
damage to the detectors. For buoyant gases, a detection plane is typically set at 2.5 meters, high
enough so that personnel at grade cannot reach up and interfere with them, but low enough that they
can be maintained from a step ladder.
For the sample problem, the gas released is primarily methane, so the detection plane and associated
dispersion modeling results were calculated at 2.5 meters for the 10,000 PMM endpoint (shown in blue)
that corresponds to the set point of the combustible gas detectors (i.e., 20% of the lower flammability
limit [LFL]). Figure 4 presents the dispersion modeling results for the single scenario analyzed in this
example.
in accordance with the legend shown in the drawing. The frequency of each individual cloud is equal to
the overall release frequency divided by the number of release directions, assuming that a release is
equally likely in any direction.
Figure 5 Gas Release Scenarios (Modified) Plotted on Plan Drawing 32 Release Directions
Figure 6 is the same concept as Figure 5, but in this case the size of the gas cloud is increased to 20 m in
length and 3 meters in width (which is still smaller than the calculated dimensions. It is quickly apparent
that the frequency of existence of a gas cloud in the overlap areas of the clouds is higher than in the
main part of the cloud. The reason for this increase in frequency is that all of the frequencies of all of
the clouds are summed for each location to develop a composite frequency of existence of a gas cloud in
a given location.
Figure 6 Gas Release Scenarios (Modified, Larger) Plotted on Plan Drawing 32 Release Directions
In order to make the assessment even more accurate, the number of direction of the release is then
increased from 32 to 720. As shown in Figure 7, this removes the jaggedness from the geographic risk
profile, and also clearly demonstrates that the frequency of the existence of a gas cloud is higher nearer
to the release source.
Figure 7 Gas Release Scenarios (Modified, Larger) Plotted on Plan Drawing 720 Release Directions
Considering Wind Directions
The geographic risk profile that is represented in Figure 8 did not directly take into consideration wind
directions. In order to address this issue, an analyst could run a separate dispersion model for each
combination of release direction and wind direction, but even if only 16 wind direction and 32 release
direction are considered, this would require 512 dispersion models to be run for each wind speed /
atmospheric stability combination, all of which would need to be separately entered into the mapping
tool. At this time, this degree of effort is not feasible for a typical industrial project. Instead, wind
directions are accounted for through probabilistic methods.
Current best practices utilize probabilistic approaches to address different wind directions by weighting
the probability of different release directions based on wind direction probability. Reconsider Figure 5.
All of the gas clouds are the same frequency (thus their colors are all the same). Addressing the wind
direction effect of gas cloud location can be estimated by modifying the condition probability of a
release direction by the conditional probability that the wind is blowing in the direction of the release.
For the sample problem, the probability of wind direction is entered based on location weather station
data as shown in Figure 8.
Figure 9 Gas Release Scenarios (Modified, Larger) Plotted on Plan Drawing 720 Release Directions
Directional Frequency Weighted by Wind Direction
After demonstration of the effect of weather conditions on the reduced gas cloud size, the dimensions
of the gas cloud are returned to the actual calculated gas cloud dimensions of 56 meters in length and
23 meters in width. The resultant geographic risk profile is shown in Figure 10.
equipment. Then a graphical coverage map of residual risk can be created, along with a calculation of
the frequencies of the detected gas clouds divided by the total frequency of all releases, the scenario
coverage. This coverage is then compared against the target to determine whether or not the gas
detection array is sufficient.