Cirrus 7.0 Manual

CIRRUS version 7.
USER MANUAL
BP Cirrus Intranet Website
http://sunapps.bpweb.bp.com/cirrus/
to download a copy of the latest version
access the CIRRUS helpline
access latest news
give feedback
Alternatively: for CIRRUS HELPLINE enquiries contact:
Dr Steve Hall Tel +44 (0)1932 76 3574

Fax 3606
E-Mail - Hall, Steve M (HSE)
Internet - [email protected]
or send correspondence to:-
Group HSE Resource

BP plc
Chertsey Road
Sunbury-on-Thames
Middlesex TW16 7LN
SECTION 1. GENERAL INTRODUCTION
1. INTRODUCTION TO CIRRUS
1.1 PURPOSE AND DESCRIPTION
CIRRUS is a software package which was developed by BP International

Limited, London and others. The purpose of the package is to provide a standard
and validated set of consequence models which can be used to predict the effects
of a release of hydrocarbon or chemical liquid or vapour. We recommend that it
should be used to perform or cross-check all consequence modelling performed
for the BP Group in relation to the design of new facilities, the formulation of
emergency plans for existing facilities, or during the execution of Quantitative
Risk Assessment studies.
The individual models which make up the package have a variety of origins, and
have been selected for use in the code on the basis of their pedigree. In many
cases they can be obtained as a stand alone model from the originating
organisation; although not necessarily in the same form. Most of the models in
CIRRUS are those which are internationally recognised and have been developed
by government organisations or industry groups; a few of the models in CIRRUS
have been developed by the BP organisation.
CIRRUS aims to provide a consistent interactive shell around these individual

programs. This allows the user to enter data into a friendly and consistent
interface for the proposed simulation or analysis, and then to view the results
either graphically or as text. The data entered by the user is stored in a separate
directory or folder, and may be retrieved for re-analysis or slight modification
in an interactive what-if approach to modelling.
Each of the core models is written in FORTRAN. The shell is written within the
Microsoft Excel spreadsheet package, and operates under the Microsoft
TM
Windows graphical user interface.
Microsoft is a registered trademark and Windows is a trademark of Microsoft Corporation.
BP CIRRUS User Manual version 7.0 1

1.2 ABOUT THIS MANUAL
CIRRUS is a Microsoft Windows application, and in using CIRRUS you engage

many standard features of the Windows user interface. While it is not
necessary to be experienced with Microsoft Windows before using CIRRUS,
familiarity with the basic elements is obviously advantageous; the Microsoft
Windows User Guide provides a thorough introduction.
If you are new to CIRRUS and wish to use CIRRUS straight away, you should
work through the hands-on example in PART 1 of this manual before trying to
apply the models to your own problems, as this will introduce and explain the
essential concepts of using CIRRUS as they arise. It is assumed that a mouse is
connected; it is possible to use the package without a mouse but one is strongly
recommended, especially for new users.
If you are new to Consequence Modelling then you should read through PART 2
of this manual. This section of the manual describes how to use the outputs of
models in CIRRUS as inputs to other models in CIRRUS, and hence build up a
comprehensive picture of all the types of hazardous effects which might occur on
release of hydrocarbons and chemicals from hydrocarbon plant. PART 2 can be
used as a reference before modelling the scenarios for a particular type of
containment.
If you are new to using the various types of consequence model which are
contained in CIRRUS, then you should read PART 3 of this manual. This
section of the manual gives a general description of the types of model and their
development and background. It discusses how to use the particular models in
CIRRUS to estimate the effects of hazardous releases, and considers the
importance of various parameters on the model results. Similar to PART 1 of
this manual, PART 3 is written as a tutorial.
At the end of the manual we give some specific examples of the modelling of
accident scenarios in Refinery and E&P plant.
An INDEX is given at the end of the manual.

1.3 INSTALLATION
To make full use of CIRRUS the minimum hardware you will need is a
TM
Pentium IBM-compatible PC with at least 16MB of random access memory
(RAM), 16MB of free disk space and a mouse. CIRRUS version 7.0 will work
with Microsoft Windows 3.1, Windows 95, Windows 2000, NT 4..0 and
Windows XP. Microsoft EXCEL 5.0c or later must be installed on the hard disk.
(Note that EXCEL 5.0a will not do). The program is compatible with the BP
COE.
The installation instructions are given on the program disk. You will be asked to
enter a password which would have been issued with your program disk. If you
have problems with the installation you should contact the CIRRUS helpline.
After successful installation in Windows 3.1, the screen will show the Program
Manager window in which there will be a number of smaller Program Groups.
One of these will be the CIRRUS group containing the CIRRUS icon. After
successful installation in Windows 95, Windows 2000 or Windows XP,
CIRRUS will be loaded as a Program in the start Menu. At the end of the
installation stage in Windows 95, when the program group first appears it is
often convenient to right click on the icon displayed, drag it onto the desktop
and select copy to copy the CIRRUS icon onto the desktop.
When installation is complete, CIRRUS is started by double clicking on the

icon. Note: At this stage you should ensure that you do not already have a copy
of EXCEL open. If you do you must close this down before loading CIRRUS.
After loading CIRRUS you can open another copy of EXCEL by double-clicking
on the EXCEL icon if you wish to use an EXCEL spreadsheet and the CIRRUS
package simultaneously. Open other spreadsheets from within this second copy
of EXCEL. Remember when using CIRRUS that you should not open another
spreadsheet by double clicking on the spreadsheet icon, as this will not open
another copy of EXCEL.
You must always leave CIRRUS using the Quit instruction from the Menu, do
not quit CIRRUS using the window spacebar (top left hand corner of Windows
3.1) or the close button (top right hand corner of Windows 95/2000). This will
quit EXCEL, and you will lose any previously defined Toolbar options.

This page is left intentionally blank
4
PART 1 TUTORIAL
FOR
NEW USERS
5
PART 1. TUTORIAL AND FEATURES SECTION 2. TUTORIAL
2. PART 1 - TUTORIAL FOR NEW USERS
On start-up the program will fill the screen with a single window showing the BP logo.
Following this you will see the initial screen as shown below:
This screen shows the disclaimer and copyright notices, and asks you to create a new
folder or select an existing folder where the work you will do will be stored. At a
later stage these folders can be copied to floppy or CD disk if you do not wish to leave
the data files on your local hard disk.
At this stage it is not recommended that you select the Expert user mode. This mode
shows you the results of some of the intermediate calculations required on the data
before input to a particular model. Also, if you hit Cancel at this stage you will exit
CIRRUS.

On selecting Continue you will arrive at a blank screen with the following menu:
The Source menu contains the models for estimating the release rate from various
types of containment and to model the flash fraction of a release. The Dispersion
menu contains the models used to estimate the distance before a vapour released into
the atmosphere will disperse to specified concentration. The Fire menu contains
models to estimate the thermal radiation from pool fires, torch fires and flares and
from fireballs. The Explosion menu contains models to estimate the overpressure
developed by a vapour cloud explosion, a pressure vessel burst and a model to estimate
the distance fragments of a vessel would travel following vessel burst or BLEVE.
If you have accidentally accessed one of the models on browsing the menus, then you
can return to the main menu by selecting previous.
The Options menu contains utilities to modify the CIRRUS databases, select and
create new folders and access a unit conversion program. Experienced users of older
versions of CIRRUS should note that this is the only main change to the architecture of
the package to have occurred following the conversion of CIRRUS to use EXCEL 5
and later.
All models in CIRRUS work in a very similar way, the following worked example will
help new users to discover the facilities offered by the package.

2.1 A WORKED EXAMPLE
In this example you will model a Non-Momentum Release (Spillage) of Propane

which is a simulation of the Maplin Sands Experimental Test Case No. 46 (a gas
dispersion trial conducted by Shell at Maplin Sands, England, 1980).
2.1.1 Step 1 - Creating a folder in which to store your data

From the main menu select the Options menu and then select the Folders option and
New Folder from the strip-off menu.
You should be asked for the name of the new folder. Type any name in the dialog box,
for example Test Example. CIRRUS will create the folder Test Example and
change to the folder. All subsequent work will be stored in the folder Test Example
until you create or select a new folder.
2.1.2 Step 2 - Choosing the model to use

CIRRUS should return to the main menu. From this menu select the Dispersion
option and then the click on the model D2 - Heavy Vapour Dispersion as shown
below:
If you accidentally select the wrong model, return to the main menu by selecting
Previous and try again, (this time holding down the left mouse button and moving the
arrow to model D2 before releasing the left mouse button).

2.1.3 Step 3 - Creating a new Dataset for input to the model
Now create a new Dataset. Select New from the menu:
In the following screen type in a dataset name (up to eight alphanumeric characters)
Type MAP46 as the dataset, and then press TAB, then type Maplin Sands Test Case
46 as the description name. Your screen should now look like this
Now click the CONTINUE button (or press RETURN). A dialog box will appear as
shown overleaf.

Notice that the only input required from you in this box is a choice from the six
mutually exclusive radio buttons. A single mouse click on one of the options will
turn that one on, and turn the others off. Select in this case the radio button which
refers to a Continuous Steady State Release with output showing Concentration
contours, and the radio button for dispersion over water, then click the CONTINUE
button:
Use the mouse or the arrow keys to select the required material.
The list will scroll up or down as required automatically when, as you move the cursor,
it hits the bottom or top of the box. You can also use the scroll bar, either by clicking
on the up and down arrows, or by dragging the scroll box up or down the bar. (Point to
the scroll box, which is the white square in the grey bar between the two arrows, and
press and hold the left mouse button. While you hold the button down, you can drag
the box up and down, and you will see the list scroll with it).

Alternatively you can move quickly through the material list by typing the first letter of
the material you wish to select.
Select 'Propane' as shown, then click the continue button.
To recap, you have so far created a folder, chosen the Non-Momentum Release
Dispersion Model, named a new dataset, chosen to investigate Concentrations during a
Steady-State Continuous Release onto water, and selected Propane as the material
being released. CIRRUS will now display an input form appropriate for these choices,
as below.
Notice that the required physical properties of the chosen material have already been
entered into the form for you. The default values of concentration are the upper and
lower flammable limits for the material.
The majority of the input forms within CIRRUS are of a similar format to this one. In
general you must enter data into all of the blank boxes, and then tell the program to
process the data by using the Run menu. The cursor appears as a coloured border
around one of the cells, into which you may type. The cursor can be moved between
the data entry cells using the TAB key (forwards) and Shift TAB (backwards). You can
also of course move the cursor with the arrow keys or the mouse, but notice that the
cursor may be moved onto parts of the form which are NOT data entry cells.
(Anything typed while the cursor is not over a data entry cell will be rejected).

For the example, enter the data as shown below.
Note that you do not need to enter the decimal points and trailing zeros for values
which are whole numbers; as you move the cursor away from each box (or press
RETURN), the data will be formatted automatically as shown here. To reproduce this
particular example, you will overwrite the default values offered by CIRRUS for the
Concentrations of interest and for the Source Diameter associated with the release.
The latter is an example of a dynamic default - notice that when you first enter the
Vapour Release Rate and Initial Cloud Temperature, a Source Diameter of 16.6 metres
was suggested, which you must overwrite when, as in this example, you wish to
specify the source diameter directly.
N.B. there is a deliberate, fundamental mistake in the data shown; if you notice it,
please DON'T correct it at this stage!
2.1.4 Step 4 - Checking data and trying to run the model

With all the data entered, pause to check it, and then pull down the Run menu, and
select the Validate and Run option to process this case:
It is quite likely, as a first-time user, that you will find this menu Item Greyed, and
that nothing happens when you try to select it:

This will be because you have inadvertently failed to complete the entry of data into
the last cell you selected (notice how the 18 for the Source Diameter is still at the
left hand side of its box, as the automatic formatting only occurs when entry into the
cell is complete). To enter 18 successfully, leave the menu, and either click the
mouse button anywhere on the form, or press RETURN before trying the Run menu
again. (Excel users will recognise what is happening here). With the value entered
correctly, you are now able to successfully choose Run, and a data validation
message will appear:
Read the message, then click OK. The cursor will be left on the offending cell, in
this case the roughness parameter. This is a good time to explore the on-line help for
input forms.

Leaving the cursor where it is, press F1. After a few seconds your screen should look
like this:
The help text is shown in a separate Window, which has been posted in front of the
entry form. You can alter the size of this help window by dragging the sides of the
box, and you can move the whole help window by clicking and holding down the
mouse key on the window top bar, then moving the mouse. If you need to, you can
scroll the text in this Window with the scroll bar - in much the same way as for
material selection described earlier.
When you have finished reading the help text, you can close the help window by a
double-click (2 clicks in rapid succession) on the control menu box (box with
picture of a book) at the top left corner of the help window - or just click back onto the
form i.e. just above the help window, so that it comes to the front and hides the help
window for the time being.
Obviously we must reduce the roughness parameter. Change it to 0.00038, the correct
value for the Maplin Sands example, and try to run the model again.

2.1.5 Step 5 - Running the model

If you have entered all the model parameters correctly then there will be some activity
from your computer's hard disk and various changes on screen, resulting after a short
while in this display.
This particular example should take no more than 10 seconds or so to execute on most
computers, but some of the CIRRUS models can run for many minutes, depending on
the input conditions. Unless you are fully experienced in using Windows as a true
multi-tasking environment, it is strongly recommended that you wait for the model to
complete, without striking any keys or making any mouse clicks.
Note that the clock is displayed to give you an idea of the progress of the simulation,
and is also a useful confirmation that everything is still working. Should the clock
ever stop, it is an indication of serious technical problems with your computer system,
from which you will probably not be able to recover without re-booting. Should you
get bored with waiting for the model to finish running, you can quit from the program
by clicking on the QUIT CIRRUS bar in the centre of the timer.
When the simulation finishes, there will be further activity from your hard-disk, and
then there will be a period in which CIRRUS post-processes the output from the
simulation, reporting progress in the status bar at the bottom of the screen.

Soon the display will be as follows:
2.1.6 Step 6 - Reviewing the results
As with most of the CIRRUS models, after the simulation calculations are completed,
the display usually shows one of the more useful graphical displays; in this case the
cloud centreline concentration as a function of distance. Ready in the left of the
status bar is an indication that control has been passed back to you.
(Screen Colours should make it easy to distinguish between the contours. If not there
is something wrong with your Windows set-up which will need further investigation).
This display currently shows the axes with Floating Scales, in which the
concentration scales and downwind distance scales are allowed to adjust automatically
so that the graph fills the screen. Often it is more appropriate to view the data using
equal scales or log scales. To switch between the type of scale, pull down the Axes
menu, and make the appropriate selection.

In this particular case select the log/lin option:
A dialog box will appear:
Click on Log/Lin, and then click CONTINUE; your screen should look like the one
below. You can now view the plot of ground centreline concentrations on a log/lin
scale:
In some cases you may get an alert box saying that zero or negative values cannot be
plotted on a log plot. When this occurs Click OK, or press RETURN, to get rid of the
alert box. You will often see this box when choosing log. scales - it is a warning
from the underlying software because there are zeros in the output data. Unfortunately
this message cannot be suppressed, but it is not of relevance in this context.

By pulling down the Graph menu, three other plots may be displayed:
Select the next graph Horizontal contours of 2 concentrations which shows you the
plan view of the two concentrations which you input; in this case the UFL and the LFL
of the Propane cloud. Your screen should look like the following:
This graph of the horizontal contours is shown with 'Floating Scales' as a default, it is
sometimes times helpful to view dispersion results with 'Equal Scales', again this is
adjusted by selecting the Axes menu and selecting the appropriate option as shown
below:

The result is a representation of the cloud in equal proportions:
Again from the Graph menu we can select some other plots. Firstly a vertical section
of the cloud at the centreline showing the height of the cloud:

and by selecting the cumulative mass in cloud, we can view the total mass of propane
within the upper and lower concentration contours:
To see the tabular output from which these charts were derived, pull down the
Results menu:
Selecting the Concentration option will display the table of ground centreline gas
concentration as a function of distance which is used to plot the graphs,

alternatively selecting Output will display the following data, which is the original
output from the Fortran code.
You can scroll up and down this tabular window with the arrow keys and/or the Page
Up/Page Down keys.
Provided a suitable printer (e.g. an HP LaserJet) is connected and properly installed in

Windows, then any of the output screens (and also the input forms) may be printed by
pulling down the Print menu and selecting Print. Printer Setup is also provided on
the menu for convenience - this will invoke the normal Windows printer set-up dialog
boxes, and may be used to switch between installed printers, or to change the
characteristics (e.g. Portrait vs. Landscape) of the default printer, but cannot be used to
install a printer.

To recap, you created a dataset called MAP46 in a new folder called Worked
Examples, you entered data into the dataset and then ran the model. After this you
reviewed the results both graphically and as tables, and optionally you have printed
these.
2.1.7 Step 8 - After modelling

So far you have not used the Previous menu, which has been shown on most of the
menu bars.
This always has three options, Return to Input Dataset, Previous and Quit. The
Return to Input Dataset option takes you back to the input sheet where you can make
changes to the conditions you have simulated e.g. wind speed, release rate, air
temperature etc. for the case you have just modelled. The Previous option returns
you back to the model menu from which you can select another model to run. The
Quit option allows you to completely exit from the CIRRUS system and return to the
Windows environment. For this example we will quit from CIRRUS, and start again
to retrieve and modify the dataset. In this second session you will use some of the
alternative ways of interacting with the program, especially short-cuts. Now select
Quit. When you select this option you are always asked to verify it, to avoid
problems when you inadvertently select the Quit option.
Note: Do not Quit CIRRUS by using the normal EXCEL menu bar buttons (far left and far right on the top
menu bar). Quitting in this way will cause you to lose your normal EXCEL setup options.
When you shut down CIRRUS you will return you to the Desktop (Windows 95) or
Program Manager (win 3.1) unless you have been running other Windows applications
simultaneously with CIRRUS, when control will very probably return to one of those
applications):
Once you have returned to a state in which the CIRRUS icon is visible, then double
click on it to restart CIRRUS. From the initial dialog box select the folder which you
created previously (e.g. Test Examples) and you should see the model menu bar. As
we are going to retrieve the example which you have just run you should select
Dispersion and then D2 - Heavy Vapour Dispersion. CIRRUS will then move to
the program menu bar. From this bar select Retrieve as shown below:

you should then be able to see a list of all input cases which have been created for
model D2 - the heavy gas dispersion model within the folder Test Examples.
As there is only one dataset saved so far in this folder, you can immediately press
ENTER or click retrieve; the input form, complete with the values you entered
previously will be displayed.
You are then asked if you want to rename the dataset. This enables you to make any
changes you wish and then save the data under a new name, so as not to cause
confusion with the old dataset.

Suppose you want to run this set of input data model again, but for a higher wind
speed. Select YES. You will then be asked to enter a new dataset name and
description. Enter the new name and description you require, and press ENTER to
continue. In this case call the dataset name WIND20MS as shown below:
Click on Continue, and then on the model input sheet change the Wind Velocity to a
new value of 2 (m/s) and change the Stability Category to F - and press ENTER to
finish editing the cell.

When the run has completed, go to the Graph menu and select Horizontal Contours
of 2 Concentrations. You should see the following graph:
From here you can either choose the Return to Input Dataset option to re-run this case
or select Previous to select another model, or use the Quit option to close down
CIRRUS again. To continue the tutorial Select Previous to return you to the model
menu.
All the various models in CIRRUS work in a similar manner. If you have completed
this tutorial you should be able to run any of the models in CIRRUS without difficulty.
The rest of PART 1 of this guide describes the other features of CIRRUS.

PART 1. TUTORIAL AND FEATURES SECTION 2. FEATURES
2.2 OTHER FEATURES
2.2.1 Definition and usage

As a particular copy of the CIRRUS program is likely to be used by several different
people over a period of time, CIRRUS supports the creation of 'folders' in which a
number of datasets (cases) may be stored for retrieval. Each folder is created as a
separate directory on the hard disk. For example, folders can be created for specific
projects where all the modelling for a project can be stored, or for specific people
where all the modelling of a specific person can be stored. In CIRRUS only the input
datasets are stored, and must be re-run in order to reproduce the model outputs.
Folders are manipulated from the Options menu. From the model menu click on
Options and then folders to look at the ways in which folders can be managed. (To
access the strip-off menu you will need to hold the cursor on the folders option).
As you can see, from this menu you can select a different folder from the one in which
you are currently located without leaving CIRRUS. You can also delete datasets from
the current folder. Also you can create a new folder, copy folders to or from a floppy
disk and delete folders using this menu.
2.2.2 User level

Currently CIRRUS has two user levels, Advanced and Expert. The default level is
the Advanced level. If you select the Expert model you will find that for the more
complex models, for example the dispersion models, in Expert mode the data which
is actually written to the program is shown as a secondary sheet after input of the
initial data. The second sheet shows you the result of the post-processing of the simple
data to make it more relevant to the case in question. This may initially be confusing
as parameters such as temperature, hole size etc. may not be the same as the physical
conditions which you specified. It is not recommended that you change any processed
data unless you fully understand why CIRRUS has made the changes. For details on
post-processing you should request a copy of the relevant Validation manual from the
CIRRUS Helpline.

2.2.3 Unit conversion facility
All CIRRUS input data is specified in metric units. The unit conversion facility is
provided to allow in-situ conversion of other units to those required by CIRRUS. It
can be accessed either as a stand alone utility from the options menu or from the input
screen when using a model.
From the model menu bar select Options and then Unit Conversion Facility. Select
the type of parameter which you wish to convert; e.g. Mass, volume etc., and then
select the units for conversion and input the known value. Conversion will occur
automatically. If you want to copy the converted value for later input you should click
on the Copy button, otherwise select Close. An example of the Unit Conversion
used from the Options menu is shown below:
From within a model itself you should select the input box containing the parameter
you want to input, and then either select Units from the menu or the keyboard
command ctrl-U. In this case a similar conversion to that shown above will appear,
except in this case there is an option to paste the result directly into the parameter box.

2.2.4 The Single-Component Database
In CIRRUS there are two different types of database; Single-component and Multi-
component. The method of adding or modifying data differs between the two types
and they are used in different models. The only models which use the Multi-
component database are the two-phase pipeline release model S7 and the Flash model
S9. All other models access the single component database.
The data held in the single component database is used to help set up input cases for
the various models. While some of the data is used almost everywhere, for example
Molecular Weight; other data is used only by certain models, for example flame
temperature for use with the Jet Fire model and molecular diffusivity for use with
the Evaporation model. Ideally therefore when adding a new material to the single
component database all data for the material for all models should be entered, even if
for your immediate needs only a subset is important. Understanding the coefficients
used in the various models requires a thorough comprehension of the underlying
model. Therefore we recommend that all database modifications should be discussed
with the CIRRUS Helpdesk. A password to allow user modifications can also be
obtained from the CIRRUS Helpdesk.
The following examples show how to view the data in the single-component database,
to add a new material to the single-component database, to modify that material and to
remove it from the database.
To view the database without the ability to make changes, select Options from the
model menu and then Single Component Database.

From the strip-off select View. This action will produce the following screen:
The arrow keys or the mouse will move the cursor, and the screen will scroll up, down,
left or right as the cursor hits the edges. Alternatively, if you set Scroll Lock (a setting
which toggles on and off when you press the Scroll Lock key), then the arrow keys will
cause the display to scroll directly. For help on the properties of a material, select the
column of the required property and either select Help and then Context Specific
from the menu, or hit the F1 function button, a 'hypertext' help facility on the material
property is then displayed.
You cannot modify the data in this tabular display; to do so you must choose
Continue to return to the model menu, and then choose Options, single Component
Database and Modify.
You will then see a dialog box into which you should enter the password supplied by
the CIRRUS Helpdesk.

You will then be shown the list of existing materials ...
Click New Item to bring up the next dialog box:
Type Example Substance and Click OK; you will see the following empty table of
properties for the new material:

Type data into the cells, using the TAB key to move from cell to cell. Note that the
form will scroll up as the cursor reaches the bottom of the window.
The parameter Chemical defines whether the material will be shown as an option for
use in models which access the single component database, and would normally be set
to Y. The parameter Component defines whether the material will be shown as an
option for use in models which access the multi-component database, and may be set
to either Y or N.
When data entry is complete, Pull down Go-On and choose Continue. A new
dialog box is then displayed asking for which of the models the material is to be valid.
The names of these models refer to the model descriptors shown on the CIRRUS
model menu.
Select the models in CIRRUS in which you wish to use your new material followed by
Continue. The program will not allow you to continue if the data is not complete for
any of the models marked as valid.
After Continue the item selection dialog box will re-appear, with the new material
included in the list:
You would normally now just click Close to return to the model menu, but in this
case, we will delete the newly added material, by highlighting the material as shown
then clicking Delete, and then clicking OK in the resulting confirmation alert box:

Copying a material would require the highlighting of the required material to copy,
then clicking the Copy button. You would use this if you wished to create a new
material based on an existing material. A new dialog box would be displayed asking
for the new material name. On entry of a valid new name the material highlighted
would be copied to the new name. You can then edit the new material.
To print material properties to a printer, highlight the required material to print and
then click the Print button. When you have finished modifying the database you
should always exit by selecting Close.
2.2.5 The Multi-Component Database

The multi-component database is designed to enable the user to calculate the physical
properties of hydrocarbon mixtures. It is used by model S9 - Flash Fraction
Calculations, to estimate the average molecular weight, the liquid to vapour fraction
formed on isenthalpic expansion and the specific heat of the mixture. It also gives
guidance on the average molecular weight to use for a dispersion cloud formed by
release of a multi-component mixture allowing for aerosol entrainment. As the
physical properties of flashing liquids are used by model S7 - Release from a two-
phase pipeline, model S7 will also directly access the multi-component database. In
CIRRUS version 5.3 or previous versions no other model will make direct use of the
multi-component database.

2.2.5.1 Browsing an existing multi-component
From the model menu select Options, Multi-Component Database and View:
You will then be asked if you want to view the composition of existing material
definitions on a mole fraction basis or a weight fraction basis.
Click on OK and you will be presented with a dialog box from which you can browse
existing materials in the multi-component database (if any).

Select View Item and you will be able to browse the existing definition of the
material chosen:
To exit from this view mode select Continue and then Close from the material
selection dialog box.
2.2.5.2 Defining a new multi-component
To define a new multi-component material select Options, Multi-Component

Database and Modify from the model menu:
You will then be asked if you want to define the material composition in terms of
mole fraction or weight fraction:
Select Mole Fraction and then OK. You will be shown a list of the materials
which have already been defined within your copy of CIRRUS. You could select
the composition of one of these items for editing, copy an existing material to use
it as the basis for a new material, print the composition of an existing material, or
create a new material. We are going to create a new material click on New Item.

Enter the name of the new item. In this tutorial call it LPG - Test Example and
click OK.
You will be presented by a list of components from which to make up your

multi-component material composition. As you can see, some of these
components are defined in terms of boiling point ranges. If you have boiling
point ranges in your definition of the material which you wish to use, you should
split these into the ranges used by CIRRUS.
We are going to suppose that the composition of the material LPG - Test
Example is a 0.5:0.5 mole fraction mixture of Propane and n-Butane. From the
list of materials click on Propane and then click on Include Component.

At the next dialog box overwrite the default mole fraction (in this case 1) with
the mole fraction of Propane in the mixture; i.e. 0.5, as shown below:
Click on OK, and from the listing of materials select n-Butane and click on
Include Component. You will see that the component mole fraction is now
defaulted to 0.5, which is the difference between 1 and the fraction of materials
already defined. Click OK as you wish to input a mole fraction of 0.5 for
Butane.
Note: It is good practice to use the CIRRUS mole fraction default when entering the last component of
a composition. In this way you will ensure that the sum of the mole fractions of all components
adds up to 1.
From the component listing dialog box click Continue,
and from the Item Selection dialog box click Close. You will be returned to the
model menu. You have now created a multi-component material called LPG -
Test Example which will be saved for future use. As a trial you could now
select Source from the model menu and then model S9. From here you can
select the multi-component material LPG - Test Example and perform a flash
calculation on this material.

PART 2
SCENARIO
DEVELOPMENT
37
PART 2. SCENARIO DEVELOPMENT SECTION 3. INTRODUCTION TO SCENARIO DEVELOPMENT
3. HOW TO USE PART 2 - SCENARIO DEVELOPMENT
PART 2 of the manual looks at all the hydrocarbon release scenarios which might
occur on oil and petrochemical installations. You may have already noticed that
this part of the manual has been written in an unusual way; though hopefully no
less simple to follow than a more conventional manual. This has been done in
order to guide the less experienced user through the sequence of models required
to provide a reasonable representation of any particular scenario involving the
release of hydrocarbons into the atmosphere.
The history of modelling has meant that most models were designed to solve a
specific problem, with relatively concise boundaries in which they are valid.
Thus in order to model a full range of consequences we must resort to using
many models, and the purpose of this part of the manual is to allow the selection
of the most appropriate methodology by those unfamiliar with the detail of
models. For each containment type, one or more material release rate or source
models must be chosen, and then one or more consequence models which
calculate the effects of interest.
To begin modelling a scenario you must first select the type of containment
which is relevant to the problem, and thereafter select an appropriate model to
assess the effects of loss of material from this containment. In this part of the
manual this initial type of containment is represented by an Icon; a pictorial
representation of the hydrocarbon plant. The model type is also denoted by an
Icon and the appropriate CIRRUS model characterisation; e.g. model D2.
If you wish to know about the full range of containment types and models
relevant to the scenarios discussed in the manual and contained within the
CIRRUS package, then you should browse this part of the manual in its entirety.
However, if you are interested in only one specific scenario then select the
appropriate containment type from the contents list and turn to this section.
In PART 2 all the relevant scenarios are described in pictorial form. Further
details on the use of each model are contained in PART 3 and referenced
appropriately. Remember that some context specific information on all input
values is available in the code itself by striking function key F1. See the tutorial
in PART 1 for further details.
In this manual we have used the nomenclature of the professionals in this

subject. We hope that as you continue to work through this manual you will
become more comfortable with the terms used.

PART 2. SCENARIO DEVELOPMENT SECTION 4. MODEL TYPES
4. MODEL TYPES
The three main effects which need to be addressed are:
The dispersion of gases/vapours in the atmosphere,
Thermal damage resulting from fires, and
Damage from explosions caused by blast and flying missiles.
In upstream oil and gas exploration operations, a major interest is associated

with the latter two, although dispersion behaviour will define the area over
which flammable concentrations may extend. In oil refining and chemical
manufacture dispersion will play an even greater role in view of the toxic nature
of many of the materials handled.
Flammable releases may burn in a variety of ways which are determined by the
nature of the release and how, when and where they are ignited. In some cases
the damage will arise solely from thermal radiation from, or contact with, the hot
flame gases. In certain circumstances, however, combustion may be responsible
for generating overpressures sufficient to damage equipment - such events are
termed explosions.
At this stage it should be recognised that CIRRUS will not necessarily fulfil all
the requirements for consequence modelling. The models have by necessity
been selected on their ability to run in a PC environment with minimal
supervision or expert knowledge. In certain circumstances it may be necessary
to resort to more complex three dimensional modelling procedures when trying
to access effects which are dependent on flow through complex geometries (for
example in predicting the detailed overpressure environment inside an explosion
in an offshore modules, or the dispersion of smoke distributed by ventilation
systems). Expertise in the use of such models is available within the BP group.
Computer models generally produce effects boundaries i.e. they predict toxic
concentrations/doses or the level of thermal radiation or blast at a particular
location - not the target damage. For the latter some form of damage criterion is
needed. These criterion are discussed in detail later in this manual in section 4.6.

PART 2. SCENARIO DEVELOPMENT SECTION 4.1. SOURCE TERM MODELS
4.1 SOURCE TERM MODELS
Before we can use consequence models to estimate the effects of hydrocarbon

releases on adjacent plant, people or the environment, we must first estimate the
quantity of hydrocarbon or other material which will be involved in an incident
scenario, i.e. the source term for the consequence model. Often the rate of
release of hydrocarbon is as important as the total quantity released, and models
S1 to S8 in CIRRUS help determine the rate of release given a loss of material of
various different types associated with various types of containment at different
pressures and temperatures.
Hydrocarbons are in one of three different phases when stored and processed in
petrochemical installations. In modelling release rates associated with loss of
containment to ambient atmosphere we have to take into account any change of
phase. Phase changes are important when modelling consequences and are dealt
with in detail in section 6.1 of this manual.
Firstly some hydrocarbons are processed or stored as a liquid, and will remain a
liquid when released to the atmosphere. These 'non-volatile liquids' include for
example stabilised crude oil or LNG (liquefied natural gas) and are generally
stored at ambient pressure or slightly above ambient pressure. Models S1 and S2
in CIRRUS estimate the rate of release of non volatile liquids for various
different sizes of containment breach in tanks and pipelines.
Some hydrocarbons are processed or stored as a gas, and will remain a gas when
released to the atmosphere. These gases include for example methane at ambient
temperature and are generally stored at above ambient pressure. Models S3 and
S4 in CIRRUS estimate the rate of release of gases for various different sizes of
containment breach in tanks and pipelines.
Some hydrocarbons are processed or stored as a liquid, but will flash to a gas
when released to the atmosphere. These 'volatile liquids' include for example
Propane, and are generally stored at above ambient pressure. Models S5, S6 and
S7 in CIRRUS estimate the (two-phase) rate of release of volatile liquids for
containment breach in the vapour or liquid space of tanks and volatile liquid
pipelines.
Model S8 in CIRRUS will estimate the rate at which material will boil off or
evaporate from liquid which has been spilt onto the ground. As will be
demonstrated later, the rate of release from a tank containing materials such as
LNG, refrigerated LPG, gasoline etc. would be estimated using model S1, and
then model S8 would be used to calculate the diameter of the LNG pool formed
and the quantity of vapour released from the pool. This vapour release rate

would then form the input to the vapour dispersion consequence model, which in
this case would be the heavy gas model D2. The model to estimate flash
fraction, S9, is also included within this section. You can access the source term
models from the model menu by selecting Source from the model menu as
shown below:
The eight types of source term model which have been included in CIRRUS are
described by the pictorial icons shown overleaf.

ICON Description
Model S1 - Release From A Liquid Tank

This model calculates the release rate of liquid flowing from a hole in, for example, a
tank under a combination of hydraulic and applied static pressure. It is used to
describe the release of a material which is liquid at ambient temperature and pressure.
Model S2 - Release From A Liquid Pipeline

This model calculates the release rate of liquid flowing from a hole in, for example, a
pipeline under a combination of hydraulic and applied static pressure. It is used to
describe the release of a material which is liquid at ambient temperature and pressure.
Model S3 - Release From A Gas Vessel

This model calculates the release rate of gas emitted from a hole in, for example, a
pressurised vessel. It is used to describe the release of a material which is a gas at the
containment pressure and ambient temperature.
Model S4 - Release From Gas Pipeline

This model calculates the release rate of gas emitted from a hole in, for example, a
pressurised pipeline. It is used to describe the release of a material which is a gas at
the containment pressure and ambient temperature.
Model S5 - Release From Vapour Space of Two Phase Vessel

This model calculates the release rate of vapour emitted from a hole in, for example,
the vapour space of a pressurised sphere. It is used to describe the release of a
material which is a liquid at the containment pressure and temperature but a gas at
ambient temperature and pressure.
Model S6 - Release From Liquid Space of Two Phase Vessel

This model calculates the release rate of liquid which is emitted from a hole in, for
example, the liquid space of a pressurised sphere. It is used to describe the release of
a material which is a liquid at the containment pressure and temperature but a gas at
Model S7 - Release From Two Phase Pipeline

This model calculates the release rate of liquid and vapour which is emitted from a
hole in, for example, a pipeline. It is used to describe the release of a material which
is a liquid at the containment pressure and temperature but a gas at ambient
temperature and pressure.
Model S8 - Vapour Release From Boiling / Evaporating Pool

This model calculates the release rate of vapour emitted from an evaporating or
boiling liquid pool.

PART 2. SCENARIO DEVELOPMENT SECTION 4.2. DISPERSION MODELS
4.2 DISPERSION MODELS
When a gas or vapour is released into the atmosphere we say it disperses in the
atmosphere. This is a process whereby the original vapour cloud is diluted with
the surrounding air as time passes. As wind is defined as the movement of air
from one point to another, it naturally follows that any vapour released into the
air will travel with the wind; i.e. it is advected by the wind. The dispersion
process therefore consists of a dilution affect caused by turbulence in the air and
a dilution affect caused by advection by the wind. At the same time the vapour
cloud will move downwind. For a continuous release this process will reach an
equilibrium forming a conventional plume from the source. For an
instantaneous puff release the vapour cloud will continue to dilute and travel
downwind simultaneously until the cloud is fully mixed with (indistinguishable
from) the air.
Because atmospheric dispersion describes the way in which a release interacts

with, and is diluted in, the surrounding atmosphere, it is not surprising that both
the nature of the release and the prevailing atmospheric conditions are
instrumental in determining what concentrations of the released substance are
experienced at various locations and times.
The most important factors which determine how dispersion takes place are
related to:
The release type: e.g. source strength (release quantity and rate), initial
density and momentum.
The atmospheric conditions: e.g. wind speed, surface roughness, stability

category.
These terms are dealt with in more detail in section 7.1.2 of this manual.
For toxic releases, one of the most significant pathways to man is by way of a
vapour cloud dispersing in the atmosphere. Predicting dispersion behaviour is
therefore critical in defining the area affected and the extent of damage caused by
accidental releases of toxic materials. Dispersion behaviour will also play an
important role in determining when and where a flammable release ignites, how
it burns and what area is affected.

4.2.1 Types of dispersion models
There are essentially two different approaches to dispersion modelling - the

integral approach and the field theoretical (3D or finite element) approach.
The integral approach aims to model the overall properties of the flow without
too much concern for the details. In principle this is fine, since it provides the
kind of information which would be required in, for example, a quantified risk
assessment or a site emergency plan, and yet computing requirements are only
modest, in most cases well within the capabilities of a desk-top PC. Because of
the relative simplicity of the approach in dealing with a far from simple problem,
empirical knowledge gained from experiments is usually vital to the construction
of these models and they invariably contain a number of assumptions which may
impose restrictions. All the models within CIRRUS and similar packages are of
the integral type.
The field theoretical approach is more fundamentally based and relies on solving
the conservation equations for mass, momentum, energy and species over a
network of interconnected cells, with a suitable means of modelling turbulence
being a necessary requirement of the process. The field theoretical approach
need not idealise the problem to nearly the same extent as that required in
integral models, so that more complex situations may be addressed directly. The
drawbacks are that there is still uncertainty in defining many parameters, for
example in the ability to model atmospheric turbulence. Perhaps as important,
computing demand is high, limiting the application of these models at the
present time to specialist applications. However as technology develops and
computing power becomes less of an issue, it is feasible that these models may
become increasingly employed in more routine studies.
There are five dispersion models in CIRRUS, D1 to D5. You can access the
dispersion models from the model menu by selecting Dispersion as shown
below:
The five types of source term model which have been included in CIRRUS are
described by the pictorial icons shown on the following page.

ICON Description
Model D1 - Momentum Release Vapour Dispersion (Refer Section 7.3)

This model is applicable for modelling vapour jet releases where an important factor
is the momentum associated with the release, for example release from a pressure
safety valve. This momentum entrains air into the cloud. The model is used for
relatively accurate modelling close to the source of the release, say between 20m and
500m. The model calculates airborne concentration as a function of cloud width,
height and downwind distance for user defined weather conditions. To model
greater distances for buoyant or neutrally buoyant plumes the model D3 should be
used.
Model D2 - Heavy Vapour Dispersion (Refer Section 7.2)

This model is applicable for modelling dispersion of vapour which is heavier than
air. This situation may occur where the vapour is at a low temperature or where the
vapour has a high molecular weight. It can be used for modelling dispersion in the
range typically 100m to 10 km. The model calculates airborne concentration as a
function of cloud width, height and downwind distance for user defined weather
conditions.
Model D 3 - Passive Dispersion with Short Averaging Time (Refer Section 7.4)
This model is used to calculate the airborne concentration at ground level of buoyant
or neutrally buoyant gases or pollutants in the far field, say typically greater than
about 1km. It is less appropriate for modelling the near field accurately as it has a
simplified representation of the release mechanism. It is generally used in modelling
the effects on the environment from relatively low concentrations of pollutants over
relatively short periods of time, e.g. 1 hour or 1 day, and for specified weather
conditions.
Model D4 - Passive Dispersion with Long Averaging Time (Refer Section 7.4)
Similar to model D3 this model is also used to calculate the airborne concentration
at ground level of buoyant or neutrally buoyant gases or pollutants in the far field,
say typically greater than about 1km. Again it is less appropriate for modelling the
near field accurately. It is used to predict the ground level concentrations of
pollutants surrounding a particular source or number of sources averaged over all
wind and weather conditions. Typically the 'annual average' meteorological data for
a site is used. This data is contained in the CIRRUS code for a few specific sites,
but can be included for other sites and areas of the world on request. If you wish to
have data for your specific location incorporated please contact the CIRRUS
Helpdesk.
Model D5 - Particulate Dispersion

This model will calculate the airborne concentration of particulates at ground level.
The constraints are similar to D3.

PART 2. SCENARIO DEVELOPMENT SECTION 4.3. FIRES AND EXPLOSIONS
4.3 FIRES AND EXPLOSIONS
When a flammable material burns we say there is a fire. The process involved
is described as combustion which is really a highly exothermic oxidation
reaction - in the vast majority of cases the oxidant is air. A flame is the visible
sign of a fire and consists of the products of combustion radiating intense heat
and light due to their high temperature. Damage in a fire is caused by heat
emitted from the flame. A person or an object may be damaged either by direct
contact with the hot combustion products or, if outside the flame, by the radiated
heat. The models in CIRRUS estimate the quantity of radiated heat as a function
of distance from the fire.
For a material to burn in air, the fuel must be in the vapour phase, it must be
mixed with the correct amount of air and a source of ignition must be present to
initiate the process. For solids and liquids, fuel must be vaporised before it can
burn, although when the material is in the form of very fine particles (such as a
liquid mist or a solid dust) this vaporisation can take place almost
instantaneously in the presence of a flame. The range of fuel/air compositions
over which combustion can take place is termed the flammable range. For
each substance the range is bounded by an upper flammable limit, UFL, above
which there is insufficient oxygen and a lower flammable limit, LFL, below
which there is insufficient fuel to support combustion.
If air is not premixed with the fuel then the fuel will burn only on its surface
where a vapour layer exists. In this case air must diffuse to the fuel surface and
such combustion is generally (but not entirely accurately) described as a
diffusion controlled fire. Diffusion controlled combustion (fires) are by far the
most commonly occurring type of combustion, and models to estimate their
thermal radiation effects and are included in CIRRUS under fire models
We define an explosion as the rapid expansion of gases which generate a

pressure wave. This pressure wave (overpressure) propagates through the air.
Damage is primarily due to the interaction of this pressure wave with people or
with structures, although in some cases the explosion may also produce flying
vessel fragments or missiles which can cause damage to a people or objects on
impact. The models in CIRRUS estimate the overpressure as a function of
distance from the explosion centre.

A rapid expansion of gases leading to overpressure may be due one of the

following causes:
a) Fuel might be premixed with air before ignition occurs. Premixing of fuel and
air is only likely to occur in vapour clouds which have undergone dispersion in
the atmosphere prior to ignition. Ignition of this mixture can result in the very
rapid generation of gases caused by combustion of the fuel premixed in air. In
many circumstances the speed at which the flame can propagate into fresh
fuel/air mixture is insufficient to cause a significant development of overpressure
(this effect is defined as a flash fire). However under certain conditions much
higher flame speeds can occur which can result in a vapour cloud explosion (or
VCE).
b) The release of stored pressure energy when a vessel bursts. The vessel burst may
occur at normal operating pressures (e.g. due to fatigue, creep etc.) or it may
occur as a result of the vessel being subject to excessive internal pressures. The
latter might arise if there were an accidental combustion inside the vessel
(termed a confined or internal explosion) or if the vessel were accidentally
overpressured due to a process upset (e.g. high pressure gas breakthrough to a
vessel designed to withstand much lower pressures). In such explosions
missiles are always a potential source of further damage.
c) The very rapid generation of gases due to a chemical reaction not confined in a
vessel. In some cases the reaction may be self-accelerating (e.g. a chain reaction
or a highly exothermic decomposition reaction). Most of these types of reactions
occur in the liquid or solid phase and do not require the mixing together of two
reactants. Such explosions are known as condensed phase explosions.
(Conventional explosives such as TNT fall within this category).
When trying to estimate the effects of accidental release of hydrocarbons or other

materials we are concerned mainly with explosions from releases of flammable
materials (mechanism a), although pressure vessel failure (mechanism b) may
also be an important consideration. Models to predict the overpressure generated
by premixed combustion (explosions) are included in CIRRUS under explosion
models. A more detailed discussion of the modelling of fires and explosions is
given in sections 0 and 9 of this manual.

4.3.1 Types of fire and explosion model
There are three fire models in CIRRUS, F1 to F3. You can access the fire
models from the model menu by selecting Fire as shown below:
and there are two explosion models in CIRRUS, E1 and E2 and one missile
throw model M1. You can access these models from the model menu by
selecting Explosion as shown below:
The fire and explosion models which have been included in CIRRUS are
described by the pictorial icons shown on the following page:

ICON Description
Model F1 - Pool Fire Radiation (Refer to Section 8.1)

This model calculates the thermal radiation emitted from a pool of burning
liquid. The pool can be at ground level, e.g. a bund fire, or at an elevation
above ground, e.g. a tank roof fire.
Model F2 - Torch Fire Radiation (Refer to Section 8.2)

This model calculates the thermal radiation emitted from a burning torch of
liquid or vapour. This can be one of the types defined below:
Gas Jet - Normally associated with gas burning from a pipe flare or from a hole
in a gas vessel or pipeline.
Liquid Jet - Normally associated with a torch flame from a material which is a
liquid at the containment temperature and pressure but a gas at ambient
temperature and pressure (e.g. propane).
Indair Flare or Mardair Flare- Thermal radiation from a specific (Kaldair) type
of gas flaring tip.
Model F3 - Fireball Radiation (Refer to Section 8.3)

This model calculates the effects of thermal radiation resulting from a burning
fireball. A fireball is the description of the type of fire resulting from a
phenomena such as a BLEVE. (Boiling Liquid Evaporating Vapour Explosion.
Model F4 - Boilover
A boilover model is not currently available in CIRRUS, but it is intended to
include one when the science is available. It will calculate the effects of thermal
radiation resulting from a burning fireball created by a boilover occurring
during a mature tank fire.
Model E1 - Vapour Cloud Explosion (Refer to Section 9.1)

This model calculates the explosion overpressure which originates from the
rapid burning of a hydrocarbon or other flammable material in air within the
confines of a process plant (or other area containing turbulence inducing
obstacles).
Model E2 - Pressure Burst Explosion (Refer to Section 9.2)

This model calculates the overpressure which originates from the over
pressurisation of a vessel, for both static overpressurisation and that occurring
during BLEVE. It does not include the subsequent burning of flammable
material and the thermal radiation effects associated with BLEVE.
Model M1 - Missiles (Refer to Section 9.3)

This model estimates the distance to which a vessel fragment will be thrown
following violent fragmentation of a vessel. This phenomena is often associated
with BLEVE and with static overpressurisation of vessels.

PART 2. SCENARIO DEVELOPMENT SECTION 4.4. CONTAINMENT TYPES
4.4 CONTAINMENT TYPES
In this part of the manual the starting point to modelling an on-site release or
emission originating from an oil or petrochemical installation begins with the
type of plant and the material to be released, for example a pressurised sphere
containing liquefied propane. This is in contrast to the more traditional expert
user approach which would begin with a definition of the consequence to be
modelled, for example the dispersion of a dense gas. By looking firstly at
containment, it is hoped to relate the methodology used in the models with the
physical reality of accidental releases and environmental emissions. In addition,
you can capture all the various consequences which might be relevant to the
piece of plant if you wish, for example fires, explosions, dispersion of toxic
clouds, or the ground level concentration of environmentally damaging
particulates, with only one definition of the original source of release.
From the release origin we can usually describe several scenarios of potential
interest by combining a sequence of release rate and consequence models. As
you have read previously in this section, each model is also allocated a
descriptive icon with a letter and number. The letter defines the general type of
model which should be used, and the number the specific model. For example in
model S1, the S refers to a source term model and the 1 refers to the first model
in the list on the CIRRUS menu bar; in this case to a release of fluid under
hydrostatic pressure.
The detail of the modelling concepts involved in each of the model types is
explained in PART 3 of this manual.
The thirteen release origins which are used to describe scenario development and
defined using icons are shown on the following page:

ICON Description
A tank which contains/emits material which is liquid at ambient

temperature and pressure.
Tank
- Liquid
A pressure vessel which contains/emits material which is a

vapour at the containment pressure and ambient temperature.
Pressure Vessel
- Gas
A pressure vessel which contains/emits a (two phase) material

which is a liquid at the containment pressure and temperature
but a gas at ambient temperature and pressure.
Pressure Vessel
- Liquefied Gas
A pipeline which contains/emits material which is liquid at the

containment temperature and ambient pressure.
Pipeline
- Liquid
A pipeline which contains/emits material which is a vapour at

the containment pressure and ambient temperature.
Pipeline
- Gas
A pipeline which contains/emits a (two phase) material which is

a liquid at the containment pressure and temperature but a gas
at ambient temperature and pressure.
Pipeline
- Liquefied Gas

ICON Description
A relief valve emitting vapour which does not condense at

Relief Valve
- Vapour
A relief valve emitting condensing vapour or a (two-phase)

vapour and liquid aerosol.
Relief Valve
- Liquid / Vapour
A flare stack burning gas.

Flare Stack
- Gas
A flare stack burning liquid.

Flare Stack
- Liquid
A chimney emitting vapour.
Chimney
- Vapour
A chimney emitting particulates.
Chimney
- Particulates
A site with numerous fugitive vapour emissions.

Site
- Vapour

PART 2. SCENARIO DEVELOPMENT SECTION 4.5. SPECIFIC RELEASE SCENARIOS
4.5 MODELLING PARTICULAR SCENARIOS
To model the effects of a release of hazardous materials with accidental or

environmental consequences, we begin with the release origin. In order to use
this manual you must first select the section with a heading icon which describes
the release origin of interest, and then follow the modelling sequence which
calculates the effects in which you are interested. Details on how to use the
models are given in PART 3 of this manual; the section reference for each model
in PART 3 is given with the icon descriptions on pages 45 and 49.
The following sections describe scenarios based on the release origins.

54
4.5.1 A TANK CONTAINING MATERIAL WHICH IS LIQUID AT AMBIENT

TEMPERATURE AND PRESSURE
Tank
- Liquid

4.5.1.1 Scenario 1 - A hole in the tank, evaporation, dispersion, explosion

Effects of interest less than 10 km.
S1 S8 D2 E1
Tank
- Liquid
In this scenario liquid will flow out of the tank and form a pool of liquid which
can evaporate (or boil if the liquid is cryogenic). Vapour released from the pool
will disperse.
Firstly model the fluid release rate using model S1 to calculate the rate at
which fluid will flow out of the tank. This will give you an answer in terms of
the release rate as a function of time. Decide if you wish to use this transient
relationship, or if you want to approximate this to a continuous release. Now use
model S8 to calculate the liquid pool radius and the rate at which vapour is
released from the pool. Again decide if you wish to use this transient
model D2 to calculate the distance of vapour dispersion. If the cloud can drift
into process plant or other partially confined volumes then use model E1 to
calculate the explosion overpressure contours which would result on ignition.
4.5.1.2 Scenario 2 A hole in the tank, evaporation and dispersion

Effects of interest greater than 10 km.
S1 S8 D3
Tank
- Liquid
Follow the same procedure as above, but use model D3 in lieu of model D2 .

4.5.1.3 Scenario 3 Rupture of the tank, evaporation, dispersion, explosion

Effects of interest less than 10 km
S8 D2 E1
Tank
- Liquid
In this scenario the tank contents will be released almost instantaneously, and
then form a pool which will evaporate (or boil) to generate vapour which will
disperse.
Treat rupture of the tank as an instantaneous release of the total inventory.

Firstly calculate the size of the pool formed and the rate of vapour evolution by
using model S8 . Remember to use the bunded version of this model if
appropriate. Again decide if you wish to use this transient relationship, or if you
want to approximate this to a continuous release. Now use model D2 to
calculate the distance of vapour dispersion. If there is process plant or other
partially confined volumes then use model E1 to calculate the explosion
overpressure contours which would result on ignition.
4.5.1.4 Scenario 4 Rupture of the tank, evaporation and dispersion

Effects of interest greater than 10 km
S8 D3
Tank
- Liquid

4.5.1.5 Scenario 4 A hole in the tank and fire
S1
F1
Tank
- Liquid
In this scenario (flammable) liquid flows out of the tank and forms a pool. The
pool is ignited almost instantaneously to burn as a pool fire.
Firstly model the fluid release rate using model S1 . to calculate the rate at
model F1 to calculate the thermal radiation emitted by the fire on ignition.
Remember to use the bunded version of this model if appropriate.
4.5.1.6 Scenario 5 Rupture of the tank and bund fire
F1
Tank
- Liquid
In this scenario the (flammable) tank contents are released almost

instantaneously and form a pool. The pool is ignited almost instantaneously to
burn as a pool fire.
The thermal radiation emitted from a fire following immediate ignition of a

rupture of a tank containing flammable material can be calculated using model
F1 only. Remember to put the duration of the release to 0 seconds and to use
the total inventory of the tank as the release rate. Also remember to use the
bunded version of this model if appropriate.

4.5.1.7 Scenario 6 A tank roof fire
F1
Tank
- Liquid
In this scenario the tank catches fire to burn as a full surface roof fire.
The thermal radiation emitted should be calculated using model F1 . Use the
bunded version of this model with the bund diameter set to the diameter of the
tank and the height of flame base set to the height of the tank walls.
The effects of smoke should be calculated using model P4 (not yet available).
4.5.1.8 Scenario 7 A tank boil over
F4
Tank
- Liquid
In this scenario a mature tank fire suffers a boil over. This is a sudden eruption
of burning liquid from the tank. Thermal radiation levels are calculated using
model F4 (not yet available).

60
4.5.2 A VESSEL CONTAINING MATERIAL WHICH IS A VAPOUR AT THE

CONTAINMENT PRESSURE AND AMBIENT TEMPERATURE
Pressure Vessel
- Gas

4.5.2.1 Scenario 1 A hole in a gas vessel, dispersion

D1
S3
Pressure Vessel
- Gas
In this scenario a hole in the tank (or broken small bore fittings etc.) allows gas
to be released into the atmosphere.
Firstly model the rate of release of the gas emitted from the hole using model
S3 . Then you must convert the resulting transient relationship to an
approximate continuous release. Now use model D1 to calculate the dispersion
of the vapour by selecting the gas release option.
4.5.2.2 Scenario 2 A hole in a gas vessel, dispersion

D3
S3
Pressure Vessel
- Gas

4.5.2.3 Scenario 3 A hole in a gas vessel and jet fire
F2
S3
Pressure Vessel
- Gas
In this scenario a hole in the tank (or broken small bore fittings etc.) allows
(flammable) gas to be released into the atmosphere which is then ignited
immediately.
The thermal radiation following ignition of the release will result in a torch
flame. Firstly model the vapour release rate using model S3 then model the
thermal radiation by using the gas jet option of model F2 .

4.5.2.4 Scenario 4 Rupture of a gas vessel, pressure burst explosion and missiles
M1
E2
Pressure Vessel
- Gas
In this scenario overpressurisation of a gas vessel results in a pressure burst

explosion and missile generation.
The overpressure effects of the pressure burst should be calculated using model
E2 . The Missile throw distances should be calculated using model M1 .

4.5.3 A VESSEL CONTAINING MATERIAL WHICH IS A LIQUID AT THE

CONTAINMENT TEMPERATURE AND PRESSURE BUT A GAS AT
AMBIENT TEMPERATURE AND PRESSURE
Pressure Vessel
- Liquefied Gas

4.5.3.1 Scenario 1 A hole in the vessel, dispersion

D1
S5
Pressure Vessel
- Liquefied Gas
In this scenario a hole in the tank (or broken small bore fittings etc.) allows
vapour to be released into the atmosphere which then disperses.
approximate continuous release. Now use model D1 to calculate the
dispersion of the vapour by selecting the gas release option of this model.
4.5.3.2 Scenario 2 A hole in the vessel, dispersion

D3
S5
Pressure Vessel
- Liquefied Gas

4.5.3.3 Scenario 3 A hole in the vessel, evaporation, dispersion, explosion

D1
S6
E1
S8 D2
Pressure Vessel
- Liquefied Gas
In this scenario a hole in the liquid part of the vessel will result in a two-phase
flashing release of liquid and vapour. Depending on the material and the
containment temperature, the release will consist of either a two-phase vapour /
liquid aerosol and/or rainout of liquid at the boiling point of the material.
To model the scenario you should firstly calculate release rate of the
vapour / liquid aerosol and the release rate of pure liquid using model S6 . You
must then convert the resulting transient relationship for the vapour / liquid
aerosol given by model S6 to an approximate continuous release.
To calculate the dispersion of the vapour / liquid aerosol use the two-phase
option of model D1 .
If model S6 predicts liquid rainout, firstly decide if you wish to use the
transient relationship given for the liquid release rate, or if you want to
approximate this to a continuous release. Then use model S8 to calculate the
liquid pool radius and the rate at which vapour is released from the pool in either
the continuous or transient form as appropriate.
Now use model D2 to calculate the distance of vapour dispersion from the
liquid pool. The concentrations predicted by this model should be added to those
of model D1 at ground level. If there is process plant or other partially
confined volumes within the flammable envelope of the cloud then you should
use model E1 to calculate the explosion overpressure contours which would
result on ignition.

4.5.3.4 Scenario 4 A hole in the vessel, evaporation, dispersion

D3
S6
S8
Pressure Vessel
- Liquefied Gas
Modelling of this scenario is similar to Scenario 1 overleaf, with the exception

that the dispersion models D1 and D2 are replaced by D3 . The vapour
release rate predicted by model S6 should be added to the vapour release rate
predicted by model S8 and used as a combined source for use with model
D3 . If model S6 predicts a vapour / liquid aerosol release then this should be
converted to a mass equivalent vapour release rate before following the above
procedure.

4.5.3.5 Scenario 5 Rupture of the vessel, evaporation, dispersion, explosion

E1
D2
S8
Pressure Vessel
- Liquefied Gas
In this scenario rupture of a pressurised sphere without immediate ignition will

result in an almost instantaneous two phase release of liquid and vapour. The
liquid may flash to form a pool of the material at its boiling point. The scenario
can be treated as a release of heavy gas in one of two ways.
If the vapour pressure of the material at ambient temperature is above about 5

bars (for example propane) a significant proportion of the release will form a
vapour/liquid aerosol mixture, and the remaining liquid pool will vapourise
rapidly. In this case use the instantaneous option of model D2 , with the mass
release rate equal to the total vessel inventory divided by 1 minute. [For
example a release of 6 te of material would be equivalent to a release rate of
6000/60 = 100 kg/s for 1 minute].
If the material vapour pressure is less than about 5 bars (for example butane)
then the material will flash to form predominantly a liquid pool. In this case
firstly run the instantaneous option of model S8 . If you wish you can now
convert the output of model S8 to approximate a continuous release, and use
the continuous option of model D2 . Alternatively use the transient output of
model S8 and use the transient option of model D2 to calculate the distance of
vapour dispersion.
If there is process plant or other partially confined volumes within the flammable
envelope of the cloud then you should use model E1 to calculate the explosion
overpressure contours which would result on ignition.

4.5.3.6 Scenario 6 Rupture of the vessel, evaporation, dispersion

D3
S8
Pressure Vessel
- Liquefied Gas
This scenario is modelled in a similar way to that above excepting that model
D3 is used in lieu of model D2 .

4.5.3.7 Scenario 7 - A hole in the vessel, vapour release and jet fire
F2
S5
Pressure Vessel
- Liquefied Gas
In this scenario the thermal radiation following ignition of a vapour release will
result in a torch flame on immediate ignition.
Firstly model the vapour release rate using model S5 . Then model the thermal
radiation by using the gas jet option of model F2 .
4.5.3.8 Scenario 8 - A hole in the vessel, liquid release and jet fire
S6
F2
Pressure Vessel
- Liquefied Gas
In this scenario the thermal radiation following ignition of a small liquid release
will result in a torch flame.
Firstly model the liquid release rate using model S6 then model the thermal
radiation by using the liquid jet option of model F2 . Note that if the torch
flame strikes the ground then use scenario 9 instead of scenario 8.

4.5.3.9 Scenario 9 - A hole in the vessel, liquid release and pool fire
F1
Pressure Vessel
- Liquefied Gas
In this scenario a hole in the vessel below the liquid level will result in flow of
liquid from the tank which, on immediate ignition, will result in a pool fire.
This scenario can be modelled directly using model F1 to calculate the thermal
radiation emitted from the pool fire. Remember that direct impingement of the
flame on the non-wetted part of the tank could result in a BLEVE as described in
scenario 10.
4.5.3.10 Scenario 10 - Rupture of the vessel, fireball and missiles
F3
M1
Pressure Vessel
- Liquefied Gas
In this scenario a BLEVE (boiling liquid expanding vapour explosion) of

pressurised liquefied flammable gas will result in a fireball, and probably the
generation of missiles due to the violent fragmentation of the vessel.
The thermal radiation generated by the fireball should be calculated using model
F3 , and the throw of associated missiles by model M1 .

4.5.4 A PIPELINE CONTAINING MATERIAL WHICH IS LIQUID AT

Pipeline
- Liquid

4.5.4.1 Scenario 1 A hole in the pipeline, liquid release, evaporation, dispersion

and explosion.
S2
E1
S8 D2
Pipeline
- Liquid
In this scenario liquid will flow out of the pipeline and form a pool of liquid
which can evaporate (or boil if the liquid is cryogenic). Vapour released from
the pool will disperse.
model S8 to calculate the liquid pool radius and the rate at which vapour is
released from the pool. Again decide if you wish to use this transient
model D2 to calculate the distance of vapour dispersion. If there is process
plant or other partially confined volumes then use model E1 to calculate the
explosion overpressure contours which would result on ignition.
4.5.4.2 Scenario 2 A hole in the pipeline, liquid release, evaporation, dispersion

and explosion.
S2 D3
S8
Pipeline
- Liquid

4.5.4.3 Scenario 3 Rupture of the pipeline, liquid spill, evaporation, dispersion

Effects of interest less than 10 km
S2
S8 D2 E1
Pipeline
- Liquid
In this scenario the pipeline will depack almost instantaneously and the
remaining contents will be released under hydrostatic pressure and pumping
pressure. A pool will form which will evaporate (or boil) to generate vapour
which will disperse.
Firstly calculate the release rate of the material using model S2 . Decide if you
wish to use this transient relationship, or if you want to approximate this to a
continuous release, then calculate the size of the pool formed and the rate of
vapour evolution by using model S8 using the continuous option or transient
option as appropriate. Again decide if you wish to use this transient relationship
calculated by model S8 , or if you want to approximate this to a continuous
release. Now use model D2 to calculate the distance of vapour dispersion. If
there is process plant or other partially confined volumes then use model E1 to
calculate the explosion overpressure contours which would result on ignition.

4.5.4.4 Scenario 4 Rupture of the pipeline, liquid spill, evaporation, dispersion

Effects of interest greater than 10 km
S2
S8 D3
Pipeline
- Liquid
4.5.4.5 Scenario 5 - A hole in the pipeline, liquid release, pool fire
S2
F1
Pipeline
- Liquid
remaining contents will be released under hydrostatic pressure to form a pool.
The pool is ignited almost instantaneously to burn as a pool fire.
which fluid will flow out of the pipeline. This will give you an answer in terms
of the release rate as a function of time. Decide if you wish to use this transient

4.5.4.6 Scenario 6 - Rupture of the pipeline, liquid release, pool fire
S2
F1
Pipeline
- Liquid
remaining contents will be released under hydrostatic pressure to form a pool.
The pool is ignited almost instantaneously to burn as a pool fire.
which fluid will flow out of the pipeline. This will give you an answer in terms
of the release rate as a function of time. Decide if you wish to use this transient

78
4.5.5 A PIPELINE CONTAINING MATERIAL WHICH IS A VAPOUR AT THE

CONTAINMENT PRESSURE AND AMBIENT TEMPERATURE
Pipeline
- Gas

4.5.5.1 Scenario 1 A hole in the pipeline, vapour release and dispersion

D1
S4
Pipeline
- Gas
In this scenario a hole in the pipeline (or broken small bore fittings etc.) allows
gas to be released into the atmosphere.
approximate continuous release. Now use model D1 to calculate the dispersion
of the vapour by selecting the gas release option.
If you wish to model a release into the ground then you will need to develop a
virtual source term. Instead of using the hole orifice diameter, calculate a
diameter through which the mass flow rate would result in a relatively low efflux
velocity, say 5 m/s. Now use this new diameter as the vent diameter in model
D1 at a height about 1m above the ground, with all other terms including the
actual mass release rate remaining the same.

4.5.5.2 Scenario 2 A hole in the pipeline, vapour release and dispersion

D3
S4
Pipeline
- Gas

4.5.5.3 Scenario 3 A hole in the pipeline and jet fire
F2
S4
Pipeline
- Gas
In this scenario a hole in the pipeline (or broken small bore fittings etc.) allows
(flammable) gas to be released into the atmosphere which is then ignited
immediately.
The thermal radiation following ignition of the release will result in a torch
flame. Firstly model the vapour release rate using model S4 . Then model the
thermal radiation by using the gas jet option of model F2 .
If the release is aimed directly into the ground then the scenario would result in
an effect more like a pool fire. In this case follow a similar procedure to that
described in Scenario 4.

4.5.5.4 Scenario 4 Rupture of the pipeline and pool fire
F1
S4
Pipeline
- Gas
In this scenario rupture of the pipeline allows (flammable) gas to be released

omni-directionally into the atmosphere followed by immediate ignition.
The thermal radiation following ignition of the release can be modelled as a fire
which resembles a pool fire. Alternatively if all the release is assumed to be
directed upwards it can be modelled as a jet fire.
Firstly model the vapour release rate using model S4 . Then you must convert
the resulting transient relationship to an approximate continuous release.
The continuous release can now be modelled as a pool fire using model F1
which will treat the release as an equivalent mass of burning liquid. If you wish
to model the release as burning within a crater formed by the pipeline burst
pressure then the bunded option of F1 can be used. The assumption that the
fire is similar to an equivalent quantity of burning liquid may underestimate the
length of the flame, which will probably lie somewhere between this and the jet
flame scenario.
To model the release as a jet flame use the rate predicted by model S4 , and
develop a pseudo vent diameter from which it is imagined the material is being
released. The magnitude of the vent diameter will depend on the pipeline size
and the mass flow rate through the rupture. For a large pipeline this will be of
the order of several metres. Use this pseudo diameter along with the mass
flowrate predicted by model S4 in the gas jet option of model F2 .

84
4.5.6 A PIPELINE CONTAINING MATERIAL WHICH IS A LIQUID AT THE

CONTAINMENT TEMPERATURE AND PRESSURE BUT A GAS AT
Pipeline
- Liquefied Gas

4.5.6.1 Scenario 1 A hole in the pipeline, evaporation, dispersion, explosion

D1
S7
E1
S8 D2
Pipeline
- Liquefied Gas
In this scenario a hole in pipeline will result in a two-phase flashing release of

liquid and vapour. Depending on the material and the containment temperature,
the release will consist of either a two-phase vapour / liquid aerosol and/or
rainout of liquid at the boiling point of the material.
To model the scenario you should firstly calculate release rate of the
vapour / liquid aerosol and the release rate of pure liquid using model S7 . You
must then convert the resulting transient relationship for the vapour / liquid
aerosol given by model S7 to an approximate continuous release.
To calculate the dispersion of the vapour / liquid aerosol use the two-phase
option of model D1 by allowing a proportion of the liquid to form an aerosol.
If model S7 predicts a liquid fraction, firstly decide if you wish to use the
transient relationship given for the liquid release rate, or if you want to
approximate this to a continuous release. Then use model S8 to calculate the
liquid pool radius and the rate at which vapour is released from the pool in either
the continuous or transient form as appropriate. Remember to subtract from this
release rate any liquid fraction which has already been assumed to be dispersed
as an aerosol in the vapour phase.
Now use model D2 to calculate the distance of vapour dispersion from the
liquid pool. The concentrations predicted by this model should be added to those
of model D1 at ground level. If there is process plant or other partially
confined volumes within the flammable envelope of the cloud then you should
use model E1 to calculate the explosion overpressure contours which would
result on ignition.

4.5.6.2 Scenario 2 A hole in the pipeline, evaporation and dispersion

D3
S7
S8
Pipeline
- Liquefied Gas
Modelling of this scenario is similar to Scenario 1 overleaf, with the exception

that the dispersion models D1 and D2 are replaced by D3 . The vapour
release rate predicted by model S7 should be added to the vapour release rate
predicted by model S8 and used as a combined source for use with model
D3 . If model S7 predicts a vapour / liquid aerosol release then this should be
converted to a mass equivalent vapour release rate before following the above
procedure.

4.5.6.3 Scenario 3 Rupture of the pipeline, evaporation, dispersion, explosion

D1
S7
E1
S8 D2
Pipeline
- Liquefied Gas
In this scenario rupture of a pressurised pipeline without immediate ignition of

the release will result in a two phase release of liquid and vapour. The liquid
will flash to form a pool of the material at its boiling point. This scenario should
be modelled in a similar way to Scenario 1.
4.5.6.4 Scenario 4 Rupture of the vessel, evaporation and dispersion

D3
S7
S8
Pipeline
- Liquefied Gas
Modelling of this scenario is similar to Scenario 3 , with the exception that the
dispersion models D1 and D2 are replaced by D3 . The vapour release rate
predicted by model S7 should be added to the vapour release rate predicted by
model S8 and used as a combined source for use with model D3 . If model
S7 predicts a vapour / liquid aerosol release then this should be converted to a
mass equivalent vapour release rate before following the above procedure.

4.5.6.5 Scenario 5 A hole in the pipeline and jet fire
F2
S7
Pipeline
- Liquefied Gas
In this scenario the thermal radiation following ignition of a (flammable) two-

phase flashing release will result in a torch flame.
Firstly model the liquid release rate using model S7 . Then model the thermal
radiation by using the liquid jet option of model F2 . Note that if the torch
flame strikes the ground then use scenario 6 instead of scenario 5.

4.5.6.6 Scenario 6 Rupture of the pipeline and pool fire
S7 F1
Pipeline
- Liquefied Gas
In this scenario rupture of the pipeline allows (flammable) two-phase material to

be released followed by immediate ignition.
The thermal radiation following ignition of the release can be modelled as a fire
which resembles a pool fire. Alternatively if all the release is assumed to be a
vapour/liquid aerosol it could be modelled (with care) as a turbulent fire similar
to a fireball.
Firstly model the vapour and liquid release rates using model S7 . Then you
must convert the resulting transient relationship to an approximate continuous
release.
Strictly the vapour will burn as an omni-directional turbulent fire, and the liquid
will form a burning pool on the ground.
If the release is predominantly liquid calculate the resulting thermal radiation

using model F1 . If the release is predominantly vapour then proceed in the
following way:
Calculate the mass of material released over a period of between 5 seconds and
20 seconds. Use this mass of material as input to model F3 using the thermal
radiation flux option. This flux level can be assumed to be that which will be
generated continuously by the fire. Check that the fireball duration given by
model F3 is similar to the assumption made for the period of material released
(the 5 to 20 second assumption). If not then redo the calculation to obtain a
more accurate fit.

4.5.7 A RELIEF VALVE EMITTING MATERIAL WHICH IS A VAPOUR AT

THE CONTAINMENT PRESSURE AND AMBIENT TEMPERATURE
Relief Valve
- Vapour

4.5.7.1 Scenario 1 RV lifts, dispersion

D1
Relief Valve
- Vapour
In this scenario the relief valve opens to release gas into the atmosphere.
Firstly the rate of release of the gas should be known from design calculations.
Use model D1 to calculate the dispersion of the vapour by selecting the gas
release option. The vent diameter used should be that of a circular hole with
equivalent area to that of the opened relief valve orifice.

D3
Relief Valve
- Vapour

4.5.7.3 Scenario 3 RV lifts, jet fire
F2
Relief Valve
- Vapour
In this scenario the relief valve emits (flammable) gas into the atmosphere which
is then ignited immediately.
The thermal radiation following ignition of the release to form a torch flame can
be calculated using the gas jet option of model F2 .

94
4.5.8 A RELIEF VALVE EMITTING MATERIAL WHICH IS A LIQUID AT

THE CONTAINMENT TEMPERATURE AND PRESSURE BUT A GAS
AT AMBIENT TEMPERATURE AND PRESSURE
Relief Valve
- Liquid / Vapour


D1
Relief Valve
- Liquid / Vapour
In this scenario the relief valve opens to release a condensing or two-phase

vapour into the atmosphere.
Firstly the rate of release of the material should be known from design
calculations. Use model D1 to calculate the dispersion of the vapour/liquid
aerosol by selecting the two-phase release option. The vent diameter used
should be that of a circular hole with equivalent area to that of the opened relief
valve orifice.

D3
Relief Valve
- Liquid / Vapour

4.5.8.3 Scenario 3 RV lifts, jet fire
F2
Relief Valve
- Liquid / Vapour
In this scenario the relief valve emits (flammable) two-phase vapour/liquid

aerosol or a pure (flammable) liquid into the atmosphere which is then ignited
immediately to form a torch flame.
If the release is predominantly vapour then the thermal radiation should be

calculated using the gas jet option of model F2 .
If the release is predominantly liquid then the thermal radiation should be

calculated using the liquid jet option of model F2 .

98
4.5.9 A FLARE STACK EMITTING MATERIAL WHICH IS A VAPOUR AT

THE STACK PRESSURE AND AMBIENT TEMPERATURE
Flare Stack
- Gas

4.5.9.1 Scenario 1 Flare out, dispersion

D1
Flare Stack
- Gas
In this flare-out scenario there is relief of gas into the atmosphere.
Firstly the rate of release of the gas should be known from design calculations.
Use model D1 to calculate the dispersion of the vapour by selecting the gas
release option. The vent diameter used should be that of an circular hole with
equivalent area to that of the flare orifice if related to a pipe flare or Mardair, or
the diameter of the tulip if an Indair.
4.5.9.2 Scenario 2 Flare out, dispersion

D3
Flare Stack
- Gas

4.5.9.3 Scenario 3 Flare thermal radiation
Flare Stack
- Gas
In this scenario the flare burns (flammable) gas.
The thermal radiation can be calculated using the gas jet option of model F2 if
the flare is a pipe flare, or alternatively the Mardair or Indair option of model
F2 if appropriate.
The tip diameter used should be that of a circular hole with equivalent area to
that of the flare orifice if related to a pipe flare or Mardair, or the diameter of the
tulip if an Indair.
The effects of a hot plume should be calculated using model P3 (not yet
available, please contact the CIRRUS Helpdesk for advice).

102
4.5.10 A FLARE STACK EMITTING MATERIAL WHICH IS A LIQUID AT

Flare Stack
- Liquid

4.5.10.1 Scenario 1 Flare liquid rain-out
F1
Flare Stack
- Liquid
In this scenario a gas flare suffers liquid break-through, or a liquid flare suffers
flare-out. The liquid is either ignited by the flare itself or is later ignited as a
pool on the ground. In either case model F1 can be used to calculate the
thermal radiation emitted from a burning pool on the ground, with the release
rate equal to that of the liquid emitted.
4.5.10.2 Scenario 2 Flare thermal radiation
Flare Stack
- Liquid
In this scenario the flare stack emits a (flammable) liquid into the atmosphere
which is then ignited to form a torch flame.
The release is predominently liquid thus the thermal radiation should be

calculated using the liquid jet option of model F2 . The effects of a hot plume
should be calculated using model P3 .
The tip diameter used should be that of an circular hole with equivalent area to
that of the flare orifice if related to a pipe flare or Mardair, or the diameter of the
tulip if an Indair.

4.5.11 A CHIMNEY STACK EMITTING MATERIAL WHICH IS A VAPOUR AT

Chimney
- Vapour

4.5.11.1 Scenario 1 Chimney flue gas

D3
Chimney
- Vapour
The concentration of pollutant in the atmosphere at ground level from a release

of vapour containing pollutants from a chimney or other similar source at
particular point downwind over a period of time, for example 1 hour or one day,
should be modelled using model D3 .

Effects of interest greater than 1 km. - Annual Average
D4
Chimney
- Vapour
The concentration of pollutant in the atmosphere at ground level from a release

of vapour containing pollutants from a chimney or other similar source at points
surrounding the source over a period of one year should be calculated using
model D4 . In this instance contact CIRRUS helpdesk to input the annual
average meteorological conditions for your site.

4.5.12 A CHIMNEY STACK EMITTING MATERIAL WHICH IS IN

PARTICULATE FORM
Chimney
- Particulates


D5
Chimney
- Particulates
The effects of release of vapour containing particulate pollutants from a chimney

or other similar source a particular point downwind over a period of time, for
example 1 hour to an annual average, should be modelled using model D5 .

4.5.13 A SITE EMITTING MATERIAL WHICH IS A VAPOUR AT AMBIENT

TEMPERATURE AND PRESSURE
Site
- Vapour

4.5.13.1 Scenario 1 Site fugitive emissions

Effects of interest greater than 1 km. - Short duration or Annual average
D3 or D4
Site
- Vapour
For environmental purposes it is sometimes necessary to calculate the effects of

chronic emissions from the site over a period of time. The effects of these
fugitive emissions can be calculated for say the annual average concentration of
pollutants at particular distance downwind or for all points surrounding the plant
using model D3 or D4 .

PART 2. SCENARIO DEVELOPMENT SECTION 4.6. DAMAGE CRITERIA
4.6 DAMAGE CRITERIA
4.6.1 PREDICTION OF TOXIC EFFECTS
Some chemicals cause harm to the function or structure of living organisms. The
degree and type of damage depends on factors such as concentration and
exposure time, the medium in which the chemical is transported, and the
susceptibility of the organism to the chemical.
To complicate the issue, from a predictive point of view, the response of an

organism is individual in nature. Different chemicals affect different organisms
in different ways, and the same dosage of chemical will effect different members
of a species in different ways. Combinations of chemicals will have different
effects to those of the individual substances.
All these factors make the prediction of toxicological response of human beings
to chemical substances both an inexact and a probabilistic science. Inexact
because human response is generally predicted from experiments on animals
which we expect to react in a different way to human beings, and probabilistic in
that different human beings will respond to similar chemical dosages in different
ways. For example a bee sting is a relatively short term irritant to most members
of the population, however to a few susceptible individuals it can be fatal.
A further complication arises from the fact that the 'damage' resulting from
exposure to toxic substances can take many different forms. Whereas thermal
radiation causes burns, and explosion overpressure causes physical damage,
toxic substances can have a multiplicity of both short term and long term
biological effects.
For example fatalities can arise in the short term from corrosive damage,
narcosis, or chemical asphyxiation, and in the long term from cancers or other
forms of systemic damage. In addition, whereas fatalities are a relevant specific
measure of damage from fires and explosions, this is not always the case for
toxic materials; consider for example materials which are mutagenic, but cause
no direct damage to the exposed person. In these cases the analyst must decide
what 'quantum' of damage is equivalent to a fatality. In addition we must
contend with the fact that information of the toxicity of a substance is often
incomplete or perhaps non-existent.
Quantitative Risk Assessments and emergency plans are normally restricted to

evaluating risks from sudden brief exposure to relatively high concentrations of
toxic materials. These 'major accident' scenarios are more likely to give rise to
offsite risks. However, the prolonged exposure to relatively low concentrations

of toxics over a working lifetime may give rise to considerable personal risk.
The former, single dose, type of exposure is described as acute exposure, while
the latter more continuous exposure is described as chronic exposure.
Toxic flux criteria are often used to define a specific level of 'safe' exposure,
normally in a regulatory context. They normally embrace the concept of
'threshold limit value' (TLV).
The TLV-TWA is the time weighted average concentration for a normal eight
hour workday or forty hour workweek to which nearly all workers may be
repeatedly exposed without adverse affect.
The TLV-STEL is the short term exposure limit representing the maximum
concentration to which workers can be exposed for a period of up to 15 min
continuously with at least a 60 min gap between exposures.
The TLV-C is the concentration which should not be exceeded even

instantaneously. Substances with a quoted TLV-C are normally fast acting.
For risk assessment purposes it is normal to work in terms of dosage, a pre-

defined function (often the simple product) of concentration C and time t, which
can be calculated directly from dispersion calculations for any specified point in
space. The concentration vs time dependence is related to the probability of
harm (e.g. fatal injury) using a probit relationship. A general form of the probit
equation describing a dosage-response relationship which is used in CIRRUS is:
Y = k 1 + k 2 ln [C n t]
where Y is the probability unit (probit) which is related to % lethality, the

constants k1, k2 and n are calculated from the data on the relationship between
the intensity of the causative factor and the degree of harmful response, C is the
concentration, and t is the exposure time. The CIRRUS database contains
published probit relationships which are given as default values when
performing dispersion calculations for toxic materials.
For further information on modelling vapour dispersion refer to section 0 of this

manual.

4.6.2 PREDICTION OF FIRE DAMAGE
Fire damage can be estimated if we can determine the rate at which a target
receives and absorbs heat (usually expressed as a thermal flux in kW/m2) and
the duration of exposure. If the flame characteristics are well defined, this
reduces to a relatively straight forward heat transfer calculation.
Flame temperatures are generally between 1000C and 1500C, and at these
temperatures thermal radiation is the dominant mechanism for heat transfer to
person or objects both inside and outside of the flame envelope (although, in the
former case, convective heat transfer can also make a significant contribution).
People and objects fully engulfed by fire will suffer harm. For people fully
engulfed by fire we assume the situation will be fatal. For objects fully engulfed
by fire the initial heat input can be approximated to the emissive power of the
flame and therefore the exposure time is critical as the material will heat up
depending upon its total thermal mass. Thin steel will fail after a shorter time
than thick steel when engulfed in fire. The object shape is also important as the
object will begin to re-radiate heat as it heats up. As a rule of thumb, during fire
engulfment thin steel objects will weaken within the first few minutes and thick
steel objects will weaken within the first few tens of minutes.
People or objects outside the flame envelope can also suffer harm. In this case,
heat is transmitted through the atmosphere by thermal radiation. Hydrocarbon
fires emit radiant heat at a wavelength which is almost totally absorbed by most
non-reflecting materials, and the damage inflicted on the material is a function of
both mass and chemical composition. Unfortunately, skin is considerably more
susceptible than steel.
Death from burn injury does not seem to depend on the source of the burn,
therefore the same thermal flux and duration will cause the same damage to
people irrespective of whether it comes from a pool fire or a fireball. Hence we
can use the same thermal flux to exposure time relationship fgor all fire
situations affecting people. Similar to toxic concentration to exposure time
relationship, we can use a probit equation to relate thermal radiation damage to
people, which takes the form:
Y = -14.9 + 2.56 ln [I(4/3) t]
where Y is the probability unit (probit) which can be related to a % lethality, I is

the radiation intensity in kW/m2, and t is the time in seconds. This relationship
is contained within the CIRRUS computer code.

Based on this probit equation, a radiant flux of 10 kW/m2 would result in a 1%

mortality level among an exposed population after a duration of 45 seconds, and
a radiant flux of 33 kW/m2 would result in a similar mortality level after only 10
seconds.
In addition to the probit relationship, we can associate thermal flux levels with
various degrees of damage as shown below:
Radiative Flux (kW/m2) Harmful Effect
Typically related to a 'safe' exposure assuming

6.3 shelter can be found within a minute.
Typically related to the flux required to weaken

15 light structural steel within tens of minutes.
Also spontaneous combustion of clothing.
Typically related to the flux required to weaken

24 heavy structural steel within tens of minutes.
For further information on modelling fires refer to section 0 of this manual.
4.6.3 PREDICTION OF EXPLOSION DAMAGE
Explosions generate a pressure wave in air which is a function of both

overpressure magnitude and duration with a typical pressure-time history. The
pressure wave is defined in terms of its peak positive overpressure, its duration
and its shape. The energy contained within the wave is known as the impulse.
Some objects are more susceptible to peak overpressure and some to impulse.
When the leading edge of a blast wave strikes an object, pressure builds up very
rapidly to a level known as the peak reflected pressure (normally two or more
times that of the free-field overpressure), and then decays as the pressure is
relieved by flowing around the object. The maximum differential loading occurs
before the pressure wave has travelled to the other side of the object. Therefore
the shape of the object will to a large extent determine the damage resulting from
the pressure wave.

In addition, the associated impulse will cause wind drag on the structure, which
itself will cause further damage depending upon the ductility of the object. At
longer range the reversal between the positive and negative phase of the pressure
pulse may be sufficiently rapid to cause damage (windows may be sucked-out
rather than blown-in).
The type of explosion will determine the overpressure-impulse relationship. For

example a 1 psi overpressure is typically related to a positive phase duration of
40ms in the case of a TNT explosion, and 4500ms in the case of a nuclear
explosion. If the person or object is a brittle window, the damage would be
equal in both cases as window damage is predominantly a function of
overpressure. On the other hand, if the person or object happens to be a ductile
storage tank the damage would vary markedly in relation to the different impulse
energies.
Hydrocarbon explosions have overpressure-impulse relationships between those

of TNT and nuclear explosions; it is often assumed that an explosion of a
reactive gas such as ethylene oxide will generate a larger impulse for a given
overpressure than a similar explosion for a less reactive gas such as propane.
Further, the overpressure-impulse relationship itself will vary with distance of
travel from the source.
Clearly when explosion damage is a function of not only object distance, but also
object properties and exploding source properties, only relatively tentative
damage estimates can be calculated from generic models. These estimates are
usually based on peak positive phase overpressure values.
People can be harmed either directly from the pressure effects of the wave, or
indirectly from falling masonry, missiles, etc. Overpressure levels resulting in
typical damage are given overleaf, with the direct damage to people who are out
in the open shown separately. Falling masonry could kill occupants of buildings,
so that in general people are more at risk from explosion overpressure when
indoors (unless the building has been specifically designed to resist blast
damage).
For further information on modelling explosions refer to section 9 of this

manual.

The relationship between damage and overpressure is shown below:
Damage Peak Side - On

Overpressure
(bars)
5% window breakage 0.01
50% window breakage 0.03
Injuries from flying glass
Cladding damage 0.07
Collapse of light buildings
Roofs lifted 0.15
Partial collapse of brickwork
Trees destroyed 0.2
Rail freight wagons derailed
Collapse of steel-framed buildings 0.3
Eardrum damage
Houses destroyed
Sphere bracing failure 0.5
People picked-up and thrown
Reinforced concrete structures severely damaged 0.7 to > 1
Rail tank cars ruptured and reactors overturned (within the
Fatal injuries exploding cloud)
4.6.4 Missile Hazards
Missiles is the term used to describe fragments of a vessel or other object

which may be thrown through the air by the force of an explosion. They can
escalate the damage caused by the explosion by impacting on other vessels
containing hazardous inventories. Missiles generated from the failure of
pressure vessels are of particular interest in the oil and chemical industry. Vessel
fragments weighing over a tonne have been known to travel in excess of 1000m.
Case histories show that damage to other plant on impact is likely. Pipework and
thin walled tanks are particularly vulnerable but pressure vessels can also suffer
severe damage. In a BLEVE, burning hydrocarbon may also be transported with
the fragment, possibly causing outbreaks of fire along the line of travel of the
missile.
Missile hazards are therefore often addressed by evaluating the risk of escalation
through impact on a vulnerable person or object.

The important factors in assessing this risk are:
the frequency of the event which produces the missiles,
the number of missiles produced, and,
the probability of a particular missile hitting a person or object.
The model in CIRRUS aims to estimate the last of these three parameters. For
further information on how to use this model refer to section 9.3 of this manual.

118
PART 2. SCENARIO DEVELOPMENT SECTION 4.7. UNCERTAINTIES IN CONSEQUENCE MODELLING
4.7 UNCERTAINTIES IN CONSEQUENCE MODELLING
The uncertainties in the results of consequence modelling are discussed on a

model by model basis in PART 3 of this manual. In this section we give a
general overview in relation to each of the model types.
4.7.1 UNCERTAINTIES IN MODELLING VAPOUR DISPERSION
Uncertainties arise from the way in which dispersion models are constructed.
Inevitably they will contain a number of simplifying assumptions and empirical
parameters which are based on limited experimental information. Accordingly,
hazard ranges predicted using these models must be treated with a degree of
caution and the assumption must not be made that immediately beyond a contour
predicted by a model there is no hazard; the contour is only a best estimate
based on available knowledge.
Although efforts have been made to evaluate the levels of uncertainty in

dispersion estimates, no definitive view can be given with any degree of
confidence. The approach we have taken in PART 3 of the manual is to indicate
where the greatest areas of uncertainty lie so that where the outcome of a
particular dispersion range prediction is critical, sensitivity studies, or even more
sophisticated approaches, may be adopted.
Two areas, in particular, are known to have an important effect on dispersion

behaviour and yet are not comprehensively modelled. The first of these is the
dispersion behaviour in the early development stages of a vapour cloud. The
nature of the release can significantly influence the dispersion behaviour at short
ranges. Understanding of these various processes is still incomplete, although
research on source term modelling has led to significant advances in the last few
years. The various effects which play a part are:
the initial momentum of the release and how it is dissipated either through air
entrainment or by impingement on obstacles,
for two-phase releases, the initial flash expansion process; in particular how this
effects initial air entrainment and what is the fate of the liquid fraction,
any chemical or physical reactions which occur at this stage (eg association,
reaction with atmospheric humidity).
For dispersion down to very low concentrations (i.e. those which are applicable
to toxic releases) a precise definition of the source term becomes less important.

Under these conditions dispersion is dominated by atmospheric turbulence which

is relatively well understood.
The second effect which can significantly influence hazard ranges is the
assumption that dispersion takes place over flat, unobstructed terrain.
The effects of obstacles can only be incorporated by the advanced numerical

field effect models. For all the other models, the neglect of obstacles is likely to
lead to an overestimate of the hazard in the far field. Obstacles generally detain
the progress of a cloud, whilst also inducing additional turbulence; both effects
tend to shorten the hazard range. However, in the near field the presence of
obstacles can also lead to underestimates of the hazard, particularly in the
gravitational slumping phase where the presence of obstacles of greater height
than the gas cloud can lead to channelling.
Topography will also affect dispersion behaviour. For sloping terrain, in the
gravity slumping phase, gas clouds will be stretched longer and thinner in a
downhill direction and conversely clouds will be shorter and fatter in an uphill
direction.
4.7.2 UNCERTAINTIES IN MODELLING FIRE AND EXPLOSION EFFECTS
All of the fire and explosion models described above contain various simplifying
assumptions, and rely upon empirical parameters to determine the physical
loadings (thermal or blast) at a particular point. Although important
determinants of the resulting effect they are often considered in a generic
manner.
These features all introduce an element of uncertainty into the results of a

particular fire or explosion consequence assessment. Because of the diverse and
complex nature of these factors, it is impossible to give numerical estimates on
the uncertainty levels. However, it is important to be aware of the influence of
some of the more important uncertainties, so that, where appropriate, sensitivity
testing can be carried out.
The greatest difficulties in predicting the effects of fires lie in:
Defining the precise flame/person or object geometry. This can be very

dependent on the characteristics of the release - particularly where momentum
effects are important. It will also depend on the local surroundings which
determine whether flame impingement occurs, whether a person or object is
shielded to any extent and, for liquid spillages, the shape of the pool feeding
the fire. Flame shape can also be greatly affected by the wind and, although

empirical correlations are available for predicting this effect, the assumption of a
steady flame of a simplified shape cannot fully reflect real conditions.
Defining the flame emissive power. This is taken from experimental information
and often displays a wide scatter. For a particular fire, emissive power can also
fluctuate in time and vary over different parts of the surface. In many models
these variations are not taken into account, but rather the flame is assumed to
radiate uniformly at a constant power. Again these variations will have a greater
impact on predictions in the near field.
For explosions, there are particular difficulties in predicting the effects of

hydrocarbon vapour in air explosions (VCE). The major uncertainties are:
Determining when an explosion will take place (as opposed to a flash fire) and
how big it will be. Both will be very dependent on the release and its
surroundings (the and without recourse to the highly complex numerical codes
these factors cannot be incorporated in anything but a very basic manner.
Determining blast loadings within or very near to the exploding cloud, for which
a detailed understanding of the fluid flows and combustion processes within a
complex geometry would be required. Again, some of the advanced numerical
models may offer this potential.
Determining the effects of blast loadings on structures. This is a complex

engineering problem which would be beyond the scope of a routine consequence
analysis. Nevertheless this approach would need to be adopted when designing
blast resistant structures.
The last two points are particularly relevant to the prediction of near field effects.
At large distances from an explosion, all blast waves tend towards a similar
shape and the details of the source of the overpressure become less important.
At these distances damage can be determined on a more generic basis.

4.7.3 CONSEQUENCE MODELLING ACCURACY
Modelling of the consequences of a potential accidental release of material from

a facility makes a significant contribution to the assessment of risk. This
assessment may provide the basis on which important policy decisions are made.
The need for quality assurance of these models and their application is clear.
Unfortunately various studies have shown a wide divergence in estimates of

hazardous effects when different models and modellers have been used to
analyse the same scenario. This is well illustrated by the results of the 1991
European benchmark exercise initiated and funded by the CEC (Commission of
European Communities) in which 11 teams carried out a series of analyses in
parallel.
The estimated concentration levels in the same weather condition predicted by

the different teams for a guillotine break on a pipe from a 60 tonne pressurised
ammonia vessel varied by three orders of magnitude.
The main origins of the variation were found to be:
i) A factor of about 9 in the source term depending on whether or not it was

predicted that a pool formed.
ii) A factor of 18 in evaporation rates.
iii) Rain-out estimates which varied from 0% to 83%.
iv) The nature of the dispersion of the cloud from the pool was variously assumed to
be dense or neutral.

Faced with this evidence we have put in a great deal of effort to validate and
correct where appropriate the codes which are included within BP CIRRUS;
some of which are from internationally renowned origins external to BP. If you
want a copy of our validation manuals please contact the BP CIRRUS helpline.
Note: This workbook also forms a part of the validation of consequence modelling by providing you,
the user, with guidance on how to apply the models to accident release scenarios.
The formal validation process consists of three aspects:
i) Is the physics embodied in the model appropriate, has all the necessary physics
been included, are the assumptions / approximations appropriate.
ii) Is the model coding correct.
iii) Does the model give answers which are in reasonable agreement with
experimental observations.
Of these processes the latter is often the most problematic. Firstly it is unusual
to conduct field tests which are of sufficient size to emulate real major accident
scenarios; secondly the variation between the results of various field tests as
performed by different groups can be extreme.
Note: When interpreting results you should always bear these factors in mind and warn against the
assumption of a spurious level of accuracy. As we go through the various models in PART 3 we
will give you some idea of the uncertainties associated with the model - and why they exist.

124
PART 3
DETAILED SCENARIO
MODELLING
125
PART 3. DETAILED SCENARIO MODELLING SECTION 5. INTRODUCTION TO MODELLING DETAILS
5. HOW TO USE PART 3 - DETAILED SCENARIO MODELLING
The purpose of PART 3 of this manual is to help you to model the effects of
accidental releases of hydrocarbons (or other chemical materials) from oil
refineries, oil exploration and production sites and chemical works. To simplify
this task we have used examples based on the models and user interface
contained within our software package BP CIRRUS. Application of this manual
is not however limited to use of BP CIRRUS, as many of the principles and
methods explained are equally appropriate if an alternative software product is
available to the user which contains similar functionality.
If you are relatively new to consequence modelling we recommend that, as a first

pass, you read the material in this manual from cover to cover in the order that it
is presented. In this way you will become familiar with all the concepts involved
in consequence modelling. If you have access to the BP CIRRUS program you
should be able to load the program and familiarise yourself with its use by
following the tutorial in PART 1 of the manual, browse PART 2 of the manual
which describes scenario development using source term and consequence
models and then work through this PART 3 of the manual by trying out the
examples. If you have any problems in interpreting the results of either these
examples, your own cases, or output from other codes, please contact the BP
CIRRUS Helpline. The contact numbers for the Helpline are given in the front
of this manual.
PART 3 is written in two sections:
Firstly we give explanations and examples of the use and function of each of the
types of consequence model, and shows you how to link models together.
Following these examples as a prototype you should be able to successfully
model the majority of accidental release scenarios and obtain meaningful
potential hazard distances.
Secondly we discuss how to model release scenarios from accidents involving

complex multi-component process plant streams.

5.1 INTRODUCTION TO PART 3
As described in PART 2 of the manual, mathematical models which are used to

estimate the effects of accidental releases of materials are called 'consequence
models'. In general they encapsulate a great deal of both experimental and
theoretical scientific work. They consist of mathematical functions which aim to
predict the physics and chemistry of the processes involved, and in some
instances require the input of experimentally determined parameters or other
types of coefficients. In BP CIRRUS many of these input parameters have
default values which are not presented to the user. We have done this to aid
consistent and appropriate application of the models within the package, and also
to simplify unnecessary complexity for the non-expert.
If you want to have further information about the models in BP CIRRUS which
are not covered in the basic description of each model given in PART 2 of the
manual, a validation document is available for each model which fully explains
the Physics contained in the model, the reasons for the default settings and the
application of the model in the program.
5.2 CONSEQUENCE MODELLING HISTORY
As you will have become aware from reading PART 2 of the manual, we have
available an assembly of computer programs, each of which was written at one
time or another to provide a solution to a specific problem. The basic set of
consequence models available to the industry are:
i) Source Term Models

ii) Vapour Dispersion Models
iii) Pool Fire and Torch Fire Models
iv) Explosion Models
Within each of these categories there is a sub-set of different models. Sometimes

there is not an absolutely appropriate model for use with a particular accidental
release scenario, as the models were not developed specifically for release
scenarios. Therefore we have to use approximations to be able to fit the
available models to release scenarios.

From reading PART 2 of the manual you will have become aware that within the
Vapour Dispersion category there are three types of general model available;
buoyant plume models, momentum jet models and heavy gas models. Direct
application of any one of these models to a release scenario can be erroneous - as
each was developed on the basis of specific need. For example, the buoyant
plume models were developed in order to predict the dispersion of pollutant
from a chimney stack, the momentum jet model was developed in order to model
the dispersion of a near ideal gas when released from a high pressure container,
and heavy gas models were developed in order to predict the dispersion of the
vapour released from a spill of Liquefied Natural Gas; which owing to its low
temperature results in a vapour cloud which is heavier than air and hugs the
ground.
If we wish to model the effects of a typical accident scenario, for example the
flashing release of a multi-component hydrocarbon, the material is not an ideal
gas, and may pass through both heavier than air and buoyant regimes as it
disperses to the chosen boundary condition. The purpose PART 3 is to describe
our 'best guess' on how to fit real scenarios into the set of specific models we
have available and shown in PART 2 of the manual. We give an explanation of
the function and examples of the use of each of the consequence models in
CIRRUS. We will also aim to demonstrate the pitfalls!

PART 3. DETAILED SCENARIO MODELLING SECTION 6. SOURCE TERM MODELS
6. SOURCE TERM MODELS
The quantity of material released from a vessel, pipework or other containment is

defined as the 'source term' on which further consequence modelling is based.
Consequence models require the source term to be defined in one of two generic
ways:
i) The total mass of material released from the containment.

or
ii) The rate at which mass is released from the containment.
If you highlight the 'SOURCE' menu in BP CIRRUS you will see that there are
eight categories of model to choose from, numbered S1 to S8. There is also an
additional model S9 called 'Flash Fraction Calculations'. This menu is shown
below.
Model S9 is not a source term calculation in itself. You will not be able to
calculate either the quantity of a material or the rate of release of a material from
a containment using this model. However model S9 is very useful in directing
your choice of the most appropriate source term model, it helps you determine
what phase your material is in when contained and what the phase change will be
when released to atmosphere. To explain its use we have included a few pages
of theory in this manual. The theory presented is actually quite simple and
important - so please at least skim through it.

PART 3. DETAILED SCENARIO MODELLING SECTION 6.1. FLASH FRACTION CALCULATIONS
6.1 FLASH FRACTION CALCULATIONS
6.1.1 Phase Equilibria - Ideal Gases
The materials which we deal with on a facility exist in one of three physical
phases: a solid, a liquid or a vapour. In relation to offsite major hazards it is
unusual to envisage scenarios involving solid substances, therefore we focus in
this manual on liquids and gases.
The definition of an ideal gas is one which obeys the relationship:
PV = nRT
Where P is the pressure of the gas, V the volume of the gas and T the
temperature of the gas. R is known as the gas constant and n is the number of
moles of the substance.
R, the gas constant, is defined as 8314.3 when P is given in Pascals (1 bar =

1x105 Pascals), V is defined in cubic metres and T is defined in degrees
Kelvin. From this definition at 273.15K (0C) and atmospheric pressure,
101,325 Pascals (1.013 bar), 1 kgmole of an ideal gas will occupy 22.414
cubic metres of volume.
Note: In order to convert a given mass of a substance into moles, take the mass in kilograms and
divide this by the molecular weight of the substance. The resulting quantity gives the number of
kgmoles of the material.
Therefore at 0C and atmospheric pressure, 1 kgmole of an ideal gas will occupy

22.414 cubic metres of volume. This is true irrespective of the material if the
assumption that the material behaves as an ideal gas in the temperature and
pressure range of interest is also true. For our purposes, at temperatures above a
material's 'critical temperature' this assumption is normally valid.
The ideal gas equation tells us that when a vapour can be treated as an ideal gas,
at constant temperature its pressure will be proportional to its volume. It also
tells us that as we increase its temperature, either its pressure will rise and/or its
volume will increase.
We have included this section on the definition and behaviour of ideal gases in
this manual because, as you will observe later, a large number of consequence
models are valid only when the assumption that a material behaves as an ideal
gas is valid.
Now take a look at the phase equilibrium diagram shown overleaf in Figure 3.1.
(This actually shows the isotherms of CO2 near its critical temperature).

Figure 3.1

6.1.2 Phase Equilibria - Non-Ideal Gases
You will see that at a temperature of 50C carbon dioxide shows a equilibrium
behaviour approaching that of an ideal gas, with the volume varying in
proportion to the pressure. However as the temperature of the gas is lowered
odd things begin to happen. Carbon dioxide no longer behaves anything like an
ideal gas. What is actually being shown by the isotherms below 31C is
liquefaction of the vapour.
Follow the 0C isotherm from the right hand side of the diagram to the left.
Initially the pressure increases as the volume decreases as expected by the ideal
gas equation. Then after a certain point the pressure no longer increases as the
volume decreases - but remains static. What is happening is that the vapour is
condensing to a liquid - which has a much reduced volume. This phase in the
isotherm is described as the 'saturated vapour' phase and the pressure is defined
as the 'Saturated Vapour Pressure' or SVP.
If we continue along the curve to a molar volume of about 0.4 litres, all the
vapour has condensed to a liquid. In this liquid only phase we see a small
reduction in volume for a very large increase in pressure.
In the saturated vapour phase the liquid is said to be in equilibrium with its
vapour.
The temperature of 31.04C represents the critical temperature of carbon

dioxide, above this temperature a vapour will not condense into a liquid on
application of pressure. The point at the apex of the dotted line is defined as the
'critical point', and is associated with unique values of pressure, volume and
temperature which are defined as the 'critical pressure', 'critical temperature' and
'critical volume' of the material.
At temperatures below the critical temperature a gas will condense into a liquid
under certain temperature and pressure conditions and freeze into a solid in other
temperature and pressure conditions. Clearly when these transitions take place
you can no longer assume the material behaves as an ideal gas; use of the ideal
gas equation given above to model liquids or the transition to liquids would be
totally erroneous.

6.1.3 The Affect of Temperature on Vapour Pressure
As you know (and shown overleaf for carbon dioxide), a change of temperature
or pressure can change the phase of a material between a solid, a liquid or a
vapour. At atmospheric pressure water changes from a solid to a liquid at its
normal melting point of 0C and from a liquid to a vapour at its normal boiling
point of 100C. Water boils at 100C because the partial pressure of water
vapour immediately above the water surface (its vapour pressure) becomes equal
to atmospheric pressure. At this temperature the water would prefer to be in the
vapour phase and undergoes the transition.
If water were put in a closed container and heated above its boiling point to say
120C, the pressure in the container would rise to the water vapour pressure at
that temperature. In point of fact at 120C the pressure in the container would
rise to about 2 bara (29 psia.). The water would be in equilibrium with its
vapour and form a 'saturated vapour'. Therefore the Saturated Vapour Pressure
(SVP) of water at 120C is about 2bara. In refining terms the saturated vapour
pressure of a material at the temperature of interest is called the bubble point
condition.
At its melting point the vapour pressure of water (or any material) will be
practically zero.
We can also see these effects occurring on the phase equilibrium diagram for
carbon dioxide shown in Figure 3.1. The saturated vapour pressure of carbon
dioxide at 21.5C is about 62 atm. and at 0C it is about 34.5 atm.
6.1.4 Using Cirrus to Determine Bubble Point (SVP) Conditions
In this section we are going to explain how you can use CIRRUS to determine
bubble point conditions.
The hydrocarbons which are dealt with in a facility undergo phase changes
similar to those of water and CO2. Normal Butane for example at atmospheric
pressure changes from a solid to a liquid at its melting point of -138C and from
a liquid to a vapour at its boiling point of -0.5C. Increasing the temperature of
n-Butane from -0.5C to 80C (176F) increases its SVP from 1 atmosphere to
about 10 atmospheres (10 bara or 147psia).

Therefore if you wished to process or store n-Butane in liquid form in a sealed

container at a temperature of 80C, the partial pressure of butane inside the
container would be 10 atmospheres.
In summary, we can say that the Saturated Vapour Pressure (SVP) of normal
Butane at its boiling point -0.5C is atmospheric pressure (1 bara). At 80C the
SVP of normal Butane is 10 bara.
Phase change boundaries are of great importance in separation and reaction

processes and in storage applications. Reaction vessels are often maintained at
temperatures close to the saturation vapour pressure of the hydrocarbon materials
being processed. Storage vessels are often maintained at the saturation vapour
pressure of their contents at ambient temperature.
In CIRRUS you can use model S9 to find out about the saturation vapour
pressure of single component and multi-component materials at reaction or
storage temperatures.
Select model S9, highlight the single component option and select Butane-n
from the material list. The CIRRUS User Guide will explain to you how to use
the models in CIRRUS if you unfamiliar with the package.

After input of a dataset name and description you should have a screen which
looks something like this:
Input an initial temperature of interest of 80C and an initial pressure of 10 bara,

and leave the final pressure at the default value of 1.013 bara.
Note: For most materials which you enter the initial pressure is not used in the calculation of the bubble
point. It is however used for multi-component materials which you define and which do not have
a true bubble point.
Run the model and you should end up with the following screen:
The top portion of this screen shows you the bubble point pressure (saturated
vapour pressure) for the temperature of interest; in this case
10.15 bara (147 psia).

Now try to reproduce the carbon dioxide saturated vapour pressures shown in
Figure 3.1. Select single component and then carbon dioxide as the material of
interest. Use a temperature of 0C, CIRRUS will return an SVP for this case of
34.47 bara. This is very similar to the 34.5 atm. (equal to 34.95 bara) shown for
this temperature in Figure 3.1.
Note: In CIRRUS the vapour pressure conditions are calculated using the Peng Robinson equation of state
and using a method by Harmens to estimate the specific heat of the material. For straight chain
hydrocarbons the results should be accurate to within 10%. For mixtures of complex materials
a process simulator should be used.

6.1.5 Using Cirrus to Determine the Proportion of Liquid to Vapour Immediately

Following a Release of a Material from Containment at SVP Conditions.
In the lower half of the screen of the previous run you will see the results of the
'flash' calculations. This indicates that if n-Butane was released from a reactor at
a temperature of 80C and if the pressure of the reactor is at least sufficient to
hold the material in the liquid phase (i.e. at or above its saturated vapour
pressure at that temperature), then some of the material will vaporise if suddenly
released to atmosphere. In point of fact, the expansion of the liquid to give
liquid and vapour will cause the temperature of the material to drop down to its
boiling point, in this case -0.5C. The fraction of vapour formed is 53.5% by
weight of the initial quantity released.
You will also see that the suggested proportion of the liquid fraction which
remains as an aerosol is 100%. Therefore we recommend that for a release of n-
Butane from a reactor vessel at 80C at a pressure of 10.15 bara, the total
quantity of released n-Butane should be used to form a vapour cloud in the
atmosphere. The temperature of this cloud would be -0.5C.
This information is useful to you when you want to model the dispersion of
vapour flashed from liquid stored under saturated vapour pressure conditions.
You would use release rate model S6 and the two-phase option of the dispersion
model D1 to accomplish this We discuss this further in the following sections of
this workbook.
Note: The proportion of released liquid which will break up into droplets which are sufficiently small to
remain within the vapour cloud is a very inexact estimate. The size of the liquid droplets formed
is not only a function of the degree of superheat (the temperature differential above the normal
boiling point), but also of manner in which the release occurs, the relative size and shape of the
release orifice etc. Thus our recommendation on the proportion of released single component
liquid which will break up into droplets and remain within a vapour dispersion cloud (to
subsequently evaporate) does not have a rigorous scientific basis, but is a reflection of typical
assumptions used by risk assessment analysts at the time of writing.

6.1.6 Flash Fraction Calculations For Multi-Component Materials under SVP

Conditions
We have just demonstrated how to use CIRRUS to perform a flash calculation

for a single component liquid. However very similar principles apply when
dealing with multi-component liquids which contain various different sorts of
hydrocarbon. Crude oil for example is a multi-component mixture of
hydrocarbons, as is gasoline, kerosene, etc.
Note: To estimate the vapour pressure and flash characteristics of your own multi-component materials,
it is necessary to add them to the CIRRUS database. If you are not familiar with this process you
will need to read the CIRRUS USER manual. When you add your own multi-component mixture
to CIRRUS, it is important to ensure that the proportions of the various component add up to
100%. To achieve this it is convenient to allow CIRRUS to default the last value input.
Add the following multi-component material to the CIRRUS multi-component

database giving it the name 'Gasoline Example'. If you have problems doing
this you should find that this mixture is already in the CIRRUS system and
called Gasoline 1. This is a European blend of gasoline which was marketed
at the time of writing this manual:
Component % by Weight
C1 (Methane) 0
C2 (Ethane) 0
C3 (Propane) 0.28
iC4 (iso-Butane) 2.08
nC4 (n-Butane) 5.62
iC5 (iso-Pentane) 10.7
nC5 (n-Pentane) 5.39
C6 (Hexane) 13.68
C7 (Heptane) 22.2
C8 (Octane) 24.98
B.Pt. 163C-224C (Nonanes to Dodecanes) 12.1
B.Pt. 225C-350C (Tridecanes to Eicosane) 0
B.Pt. 351C-454C (Uneicosane to Triacontane) 0
B.Pt. 455C+ (Untriacontane plus) 0
Benzene 2.97

It is common to define hydrocarbons by the number of carbon atoms which they

contain. Thus C1 represents a compound with one carbon atom, for example
Methane. When making up multi-component mixtures for process plant streams,
the proportion of low molecular weight material up to and including C8 are
defined as alkanes or alkenes. Thereafter the mixture is defined in terms of its
boiling point range of which there are five definitions when defining mixtures in
CIRRUS; 36C to 162C, 163C to 224C, 225C to 350C, 351C to 454C
and 455C or greater. The first of these boiling point ranges represents the
materials from Hexane to Octane. It is better to use the exact fractions of these
materials when available.
Note: The higher boiling point (higher molecular weight) components are lumped together in relatively
course categories for the purposes of flash calculations in CIRRUS, as there contribution to
vapour pressure when released into an atmosphere at ambient temperatures is small. CIRRUS
should not be used as a replacement for the more complex flash routines used in design work.
The incorporation of a multi-component flash capability in CIRRUS is to allow a consequence
modeller convenient access to a program which is fit for this purpose.
If you now run model S9 by firstly selecting the multi-component option and
then selecting the gasoline mixture called 'Gasoline Example' which you have
made up above. Input a temperature of interest of 15C. (The pressure is not
used in this example as the material has a true bubble point - so input a nominal
value, for example 10 bara). You will see that CIRRUS predicts that this
gasoline will have a vapour pressure of 0.4231bara. as shown below:

This value is less than atmospheric pressure, therefore if spilt onto the ground
this gasoline would evaporate rather than boil. If you increase the temperature to
42C you will see that the saturated vapour pressure increases to 1 atmosphere,
indicating that this is the boiling point of the mixture.
In the lower portion of the output screen CIRRUS also suggests that after a
release to atmosphere from 15C, a negligible proportion of vapour would be
emitted due to flashing. Thus in order to model such a spill we would assume
that the gasoline would form a pool on the ground at a final fluid temperature of
15C. In order to estimate the release rate for vapour cloud dispersion we would
then choose the boiling/evaporating pool source model - S8. More of this later.
Note: CIRRUS will assume that components with a molecular weight of greater than 100 will have less
of a contribution in the flashed vapour than those of lower molecular weight.
The only models in BP CIRRUS vs 5.1a which directly make use of a user
specified multi-component mixture are models S2, S7 and S9. In the following
sections of the report we discuss how to model multi-component mixtures as
pseudo single component mixtures for the other models in CIRRUS.
6.1.7 Flash Fraction Calculations For Multi-Phase Materials
The flash fraction and bubble point calculations which we have described above
refer to 'two-phase' multi-component materials where all components have
critical temperatures below the temperature of interest. As described previously
the 'critical temperature' is the highest temperature at which a gas can be
condensed to a liquid; i.e. above this temperature the material cannot be
liquefied by application of pressure and will therefore will always remain in a
gaseous state.
Note: Liquids which have a saturated vapour pressure above atmospheric pressure at their containment
temperature are sometimes known in consequence modelling as 'two-phase' materials and
sometimes known as flashing liquids. For example pure Propane in a storage vessel at a
ambient temperatures, say 15C is likely to be stored under saturated vapour conditions. The
vessel pressure will be about 7.3 bara (6.3 barg). In this condition liquid Propane is often
described as being 'two. If we refrigerate the Propane to a temperature below its boiling point, its
saturated vapour pressure will be below atmospheric pressure. When stored in this form liquid
Propane is often described as being a 'single phase' liquid.

We are now going to consider a situation where the temperature of interest is

above the critical temperature for some of the gaseous components in our
mixture and below that for others. In other words some components may behave
like ideal gases when flashed to atmosphere - i.e. they would change volume but
all their mass would remain in the gas phase, and other components would
behave like 'two-phase' materials - a proportion of their mass may be converted
into liquid and vapour when flashed to atmosphere.
As an example of a multi-phase mixture we could look at a mixture of

Hydrogen, Methane, Ethane and Propane. (This for example could be a simple
feed to a Hydrotreater).
Make up the following mixture and call it 'Hydrotreater Example. (Again if you
have problems doing this you should find a multi-component mixture called
Hydrotreater 1 in the CIRRUS system.
Component % by Weight Critical Temperature (C)

Hydrogen 6 -240.0
C1 (Methane) 17 -82.6
C2 (Ethane) 31 +32.3
C3 (Propane) 46 +96.7
In this mixture there is approximately a 50% mole fraction of Hydrogen and a

50% mole fraction split equally amongst the other hydrocarbon components. In
this table we have also included the critical temperature of the components for
comparison.
Note: You can read the critical temperature of a component by selecting the CIRRUS database,
followed by single component and then view. Using the arrow keys you can then move to the
right of the database spreadsheet until you reach the critical temperature column.
In consequence modelling this type of mixture is often referred to as a multi-

phase mixture. On its own Propane 60F (15C) in equilibrium with its
vapour is known as 'two-phase'. Adding gaseous hydrogen (with a critical
temperature well below that of the containment temperature) changes this
mixture definition to multi-phase, as would adding other inert materials to the
system such as water. In reality of course there are still only two phases,
liquid and gas, but the change in definition indicates that this is not a mixture
in equilibrium with it vapour.
At 15C hydrogen is well above its critical temperature and to a first

approximation will always remain in a gaseous phase - assuming that no
chemical reaction takes place. Propane on the other hand can be condensed in

equilibrium with its vapour at this temperature as can Ethane, - so what will
the mixture do?
Intuitively we could expect that if released to atmosphere, the hydrogen and

methane would be released as 'inert' gases and the Ethane and Propane may
flash to give vapour and liquids. We are now going to use model S9 with this
mixture to see what happens. Run model S9 with this mixture and input an
initial temperature of 15C and an initial pressure of 10 bara. You should get
a screen which looks something like this:
As this multi-phase fluid is not a liquid in equilibrium with its vapour, it does
not have a true bubble point. Therefore the model returns a bubble point
pressure of None. The flash calculation is performed using the enthalpy of the
mixture at the starting conditions, so in this case the initial pressure is used in the
calculations. Due to the Hydrogen the mixture the final flash temperature is
above that of the boiling point of Propane, so all the mixture remains in the
vapour phase. If you change the initial conditions to that of a real process vessel,
for example 230C and 50 bara, you will obtain a similar result excepting that
the final mixture specific heat capacity (Cp) and ratio of specific heats () will be
3.326 and 1.18 respectively. The information from the flash routine can be used
in subsequent release and dispersion calculations. It tells you the average
molecular weight, the Cp and the to use in these calculations as described in
section 3.4.2.

PART 3. DETAILED SCENARIO MODELLING SECTION 6.2. MODEL S1
6.2 SOURCE TERM MODEL - S1 RELEASE FROM A LIQUID TANK
After selecting the 'model' menu in BP CIRRUS you will be able to highlight the
'SOURCE' menu. You will notice that model S1 is called 'Release from Liquid
Tank'.
Note: In CIRRUS the source term models have an 'S' designation, the dispersion models a 'D'
designation, the fire models an 'F' designation, the explosion models an 'E' designation and the
missile model an 'M' designation.
Model S1 is designed to calculate the rate of release of a material which is a

liquid at the containment temperature and atmospheric pressure, i.e. the
containment temperature is such that the vapour pressure of the material is less
than atmospheric pressure. Model S1 is intended to model the release rate of
materials which are normally defined as liquids when stored at atmospheric
pressure or at pressures very close to atmospheric pressure.
For example this would include modelling the release from:
Floating roof oil storage tanks

Fixed roof product tanks
Oily water tanks
Refrigerated Liquefied Natural Gas storage tanks
Refrigerated Liquefied Petroleum Gas storage tanks
Note: Model S1 calculates the release rate of a liquid using the 'Bernoulli' equation, and taking into
account the hydrostatic head of the liquid and any externally applied pressure.
6.2.1 A Release of a Single Component Liquid
As an example of the use of this model we will look at a scenario based on the
spillage of refrigerated Butane from a flange at 1m from the bottom of a Butane
storage tank containing about 20,000te of Butane. The material temperature is -
5.5C and the tank is 40m in diameter, 35m high and filled to a height of 28m.
The tank has a maintained pressure head at atmospheric pressure. Let us assume
that the burst flange has a release area equivalent to a hole 250mm (10) in
diameter.

Therefore select the source Term model S1. Select a liquid release directly from
the tank itself (we are assuming that the burst flange is that which connects the
outlet pipework to the tank wall) and assume that the tank will have a maintained
pressure. You should see the following dialog box:
Note: There are three pressure related options for this model which have the following implications:
Tank open to atmosphere: This option would be used for tanks without a fixed roof. Therefore
conventional crude oil storage tanks with only a floating roof would come into this category. It
implies that the pressure above the liquid surface will always be atmospheric pressure, and that
liquid will drain from the tank under the force of gravity alone. This option should also be used
for fixed roof tank in scenarios which encompass fracture of the tank roof, or where vacuum
breakers are assumed to open and are of sufficient size to maintain the pressure above the liquid
at near atmospheric pressure.
Tank sealed from atmosphere: This option would be used for tanks or vessels which are isolated
from the atmosphere. It implies that liquid will drain from the tank under the force of gravity but
will draw a vacuum in the tank as the release continues. This option can also be used where a
purge gas is connected to the tank but is of insufficient capacity to prevent the tank from drawing
a vacuum.
Tank with maintained pressure head: This option should be used for modelling the release from
a fixed roof tank or vessel where the void above the liquid is purged with gas and maintained at a
pressure above the ambient atmospheric pressure. It implies that the flow of purge gas is
sufficient to compensate for the volume loss as liquid is released and would therefore maintain
the pressure differential in the void space.
Release from attached pipework: If you choose the option of a release from pipework attached to
the tank, the results will give a reduced outflow rate in comparison to a direct release due to the
frictional forces associated with flow through the pipework. If the release scenario occurs very
close to the tank wall (say a couple of inches) then selection of a release directly from the tank
would be the most appropriate option.

Now continue and select the material of interest to be Butane-n.
Note: You can either scroll down the dialog box or use the keyboard. To use the keyboard strike 'b'
two or three times until you get to Butane on the material list.
The model input screen will now appear, input the tank dimensions and
inventory level given above; your input screen should look like this:
Note: The temperature of the liquid has defaulted to 5C less than the material boiling point. You can
change this to an actual storage temperature (as long as it is less than the material boiling point) -
but it is unlikely that it will make a significant difference to the results.
The critical aspect of this input is the discharge coefficient (Cd). The release
rates calculated by code are a direct multiplier of this coefficient. A coefficient
of 1 will give the maximum discharge rate. For a sharp edged orifice and a
liquid of low volatility a Cd of 0.6 is appropriate and this is the coefficient which
we will use in this example. Access the CIRRUS context specific help for more
details (Click on the Cd box and then hit function key F1).

Now run the model and you should get the following results:
As you can see the predicted outflow rate is about 410kg/s initially falling to
zero after 25 hours (9.0x104 seconds). As the liquid height in the tank decreases
the hydrostatic head of the fluid decreases and the flowrate through the hole
decreases accordingly. Now select Print to save a hard copy of the outflow rate
plot, we will make reference to this plot later.
Note: If you select Graph from the menu and select the cumulative outflow versus time option you
can check that all 20,000te of material has been released after 25 hours.
6.2.2 A Release of a Multi-Component Liquid
To model a multi-component liquid mixture using model S1 in CIRRUS you

should have to use a single component liquid as a surrogate for the multi-
component liquid of interest. In this case you should select a single component
material from the CIRRUS database which has a similar density to the multi-
component liquid of interest.
We have now modelled the outflow rate of refrigerated liquid Butane from a
storage tank. The vapour pressure of this material is less than atmospheric
pressure when released at a temperature which is less than its boiling point.
However on contact with a warmer surroundings the Butane will warm up and
begin to boil. We are now going to model the vapour release rate from the pool
formed using model S8.

6.3 SOURCE TERM MODEL - S8 VAPOUR RELEASE FROM A BOILING

OR EVAPORATING POOL
6.3.1 Modelling of a Multi-Component Material using Model S8
This model calculates the rate of evaporation and spreading of a pool of liquid on
the ground. We will continue the scenario of a spill of Butane from a storage
tank by assuming that the Butane spills into a concrete bund surrounding the
tank. For this example we will assume that the 40m diameter tank is located in a
bund of 60m diameter. The area of the bund is 2,830m2, the tank itself occupies
about 1260m2 of area leaving an area of 1570m2 for the Butane to spread into
and evaporate. This is equivalent to a circular bund area with a diameter of
44.7m.
Now select model S8. After you have input a dataset name and title you will see
a dialog box; select the bunded option and the transient release option as shown
below.
Note: There are three release options to choose from which have the following implications:
Instantaneous Release: The inventory is released instantaneously, with the associated spread
of the pool being very rapid.
Continuous Release: The inventory is released at a constant rate for a given time period.
Transient Release: The inventory is released at a variable rate for a given time period. For
our example above this is the most appropriate model, as the inventory is being released at a
decreasing rate.

Now input the data appropriate for this scenario. The discharge vent diameter is
250mm (0.25m), from our modelling of the release using S1 we know that the
spill duration is 1500min (9.0x104 seconds), the bund diameter with an
appropriate area for evaporation is 44.7m, use a roughness of 0.01m and a
ground type of 4, an air temperature of 20C and a wind velocity of 5m/s. From
the release rate versus time plot produced by model S1 in section 3.2.1 you can
read the following values:
Time (sec) Liquid Release Rate (kg/s)
0 410
15000 340
30000 280
45000 230
60000 160
75000 80
90000 0
Note: The Ground Roughness is related to the ground type upwind of the release and determines the
turbulence in the air which is blowing across the pool surface. This affects the evaporation
process. The ground type of 4 refers to concrete and this affects the heat conducted into the pool
from the ground. For explanation on the choice of these parameters, click on the appropriate
input box, strike F1 and the CIRRUS on-line help will appear.

You should have the following input screen:
Now run the code - this will take about 3 minutes on an average 486 machine.
The first output shown is the vapour release rate from the surface of the pool:
As you can see, after an initial almost instantaneous vapour release at a rate of
about 35kg/s, the evaporation rate settles down to quasi-continuous rate of about
20kg/s.

To identify these changes more easily use the Axes command and change the
plot to a log/log relationship as shown below:
Notice that the peak in the evaporation rate occurs in about 40 seconds.
Now go to Graphs on the Menu bar and select pool diameter. Notice this is
effectively constant at the input value of 44.7m. - the bund diameter. To see the
detail of pool development change the axes to log/log, you will see that the
diameter of the pool increases until it reaches the bund wall - again this occurs in
about 40 seconds.
So why does the maximum evaporation rate occur in under 40 seconds? It is

because the ambient ground is conducting the most heat into the pool as the pool
is still spreading over the unwetted surface of the bund floor. When the pool has
spread to the bund wall, in just under 40 seconds, all the area of the bund floor is
covered with cold Butane (at -5C) and the ground under the pool begins to cool
down to the temperature of the boiling Butane, heat conduction into the pool is
reduced resulting in a lower evaporation rate.

The next stage in the consequence modelling process is to take the vapour
evaporation and put this into a vapour cloud dispersion model. We can then see
how far the vapour cloud travels while remaining in a flammable state. To do
this we input the vapour release rate we have just modelled into model D2. An
explanation of this dispersion modelling is given in this manual in section 4.1.
You can either turn to this section now in order to continue modelling this
scenario, or preferably put it to one side as we are about to go on to another type
of Source Term model.
6.3.2 Modelling of a Multi-Component Material using Model S8
Similar to model S1, if you want to model a boiling/evaporating liquid which is a

multi-component mixture, you must use a pseudo single component. This is
where things get rather awkward. Model S8 uses several specific properties of
the liquid, for example its vapour pressure/temperature relationship and its
molecular diffusivity in air. As a first estimate choose a pseudo single
component with a similar boiling point and molecular weight.

152
6.4 SOURCE TERM MODEL - S3 RELEASE FROM GAS VESSEL
This model is used to estimate the rate of release of a gas from a container which
contains only gas. In our case we assume that it is a gas which behaves as an
ideal gas.
Note: For a definition of an ideal gas you should return to section 3.11
In the first instance we are going to model something relatively simple, a release
of methane from a sphere containing only methane at 50 bara pressure.
Following this we are going to take the multi-component vapour analysed in
section 3.17 of the report, which was an imaginary vapour phase mixture in a
Hydrotreater, and discuss how to use model S3 to estimate the release rate
following a breach in this unit.
6.4.1 Using CIRRUS to Model a Release of Single Component Gas
Go to the Model Menu of CIRRUS and select SOURCE followed by S3. After
input of a dataset name and description you should see the dialog box shown
below. Select Uninsulated Tank.
Note: Selection of Uninsulated Tank will imply that the contents of the vessel will undergo
isothermal depressurisation. This means that heat will flow from the surroundings to the gas
in the vessel and therefore the gas in the vessel will not cool down due to loss of material
through a hole in the vessel.
Selection of Insulated Tank will imply that the contents of the vessel will undergo adiabatic
depressurisation. This means that heat will not flow from the surroundings to the gas in the
vessel and therefore the gas in the vessel will cool down due to loss of material through a hole
in the vessel. If depressurisation occurs in a timescale shorter than that required to allow heat
to flow into the vessel contents, then this option should be chosen even when the vessel is
uninsulated.

The scenario we are going to model involves a pressurised bullet which is 4m

diameter and 20m in length, and contains methane at 50 bara pressure. The
volume of the vessel is therefore 250m3. Assume that the ambient temperature is
20C and that you lose a small bore fitting which crates a hole in the vessel of
100mm diameter. Your input screen should look like this:
Note: As with model S1, the discharge coefficient (Cd) is again very important in determining the
results of this model. In this case we are going to change the default Cd from 0.8 to 1. (Again,
select the Cd box and strike F1 for further information.
Run the model and your output should look like this:
As you can see the initial gas outflow rate is 65kg/s dropping to virtually zero
after 380 seconds. The outflow rate decays as the pressure in the vessel is
relieved. If you now go to the GRAPHS option you can look at some of the
other relationships.

Firstly note that the vessel pressure does decay as a function of time. Secondly
note that the vessel temperature remains constant at 20C, and thirdly note that
the temperature of the gas at the orifice has fallen to -19C. In later modelling of
the dispersion cloud, using model D1, we will use this latter temperature as the
material temperature.
Note: If you now re-run this case but first select the Insulated Tank option, you will note that the
outflow rate is lower and that the temperature of the vessel no longer remains constant, but
decreases as a function of time. Note also that the temperature of the gas at the orifice also falls
as a function of time, in this case reaching -150C.
6.4.2 Using CIRRUS to Model a Release of Multi Component Gas
We are now going to model a release of vapour from our theoretical Hydrotreater
which was introduced in section 3.1.7. You will remember that this fluid has the
following composition:
Component % by Weight Mole Fraction Molecular

Hydrogen 6 0.4866 2.02
C1 (Methane) 17 0.1736 16.04
C2 (Ethane) 31 0.1689 30.07
C3 (Propane) 46 0.1709 44.10
You will also remember that due to the results of the flash calculations
performed using model S9, we decided that this fluid was not a saturated vapour,
and the released vapour was 100% gas with an average molecular weight of
16.38. We are therefore going to treat it as an ideal gas and estimate the release
rate from our Hydrotreater using model S3.
Note: For any real process plant stream you could obtain a more accurate approximation to the real
molecular weight of a vapour mixture from your process engineering or technical department
We now have to choose a surrogate single component on which to model this

mixture. If you browse through the CIRRUS database you will see that Methane
has a molecular weight of 16.04. Therefore Methane will make a pretty good
surrogate for our above multi-component mixture.
Note: To browse the CIRRUS database for surrogate mixtures, go to the Database menu item, select
single component material and then the view option. This database is an Excel spreadsheet
which you can move around using the arrow keys.

If the Hydrotreater is operating at a temperature of 230C and a pressure of 50

bara when the release from a 100mm diameter hole in the reactor occurs, then we
can simply estimate the release rate using model S3. Assume the reactor vessel
is 4m in diameter and 20m tall, a volume of 250m3.
Select model S3, select uninsulated vessel and select Methane as the material of
interest. Now change the molecular weight to 16.38, a Cp of 3.326 and a of
1.18 as calculated in section 3.1.7, and input the appropriate volume,
temperature and pressure, change the name of the material to Hydrotreater
vapour. You should end up with an input screen which looks like this:
Now run the model. The results are shown below. Take a print of this curve
as we are going to use it later to model the vapour cloud dispersion for this
scenario. Also view the vessel temperature graph, as this is an isothermal
release it shows a constant 230C vessel temperature. Note also the orifice
temperature graph which gives an orifice temperature of 190C. As a source
to our dispersion model, D1, the vessel storage temperature at the release rate
of the gas should be used (in this case 230C) and not the orifice temperature.
The orifice is calculated from the vessel storage temperature by the input
routines of model D1.

Your output should look like this:
In model S3 we have estimated the release of vapour from a vessel when the
vapour can be treated as an ideal gas. When the vapour is actually a saturated
vapour we cannot use this ideal gas model. We must use instead model S5.

158
6.5 SOURCE TERM MODEL - S5 RELEASE FROM VAPOUR SPACE OF A

TWO PHASE VESSEL
6.5.1 A Single Component Material at its Saturated Vapour Pressure
As the first example in modelling the release of a saturated vapour we are going
to look at the example of a pure single component. Suppose for example you
have on your facility a Horton sphere of 20m diameter. The volume of this
sphere will be about 4,200m3, and we are going to assume that it is half full of
Propane liquid (about 1,000te) and about half full of vapour. The pressure of the
sphere will be that of the saturated vapour pressure at the ambient temperature.
If the ambient temperature is 15C, then using model S9, we can determine that
the saturated vapour pressure is about 7.3 bara.
You can check this saturated vapour pressure for yourself using the procedure
explained in section 3.1.4.
Now select model S5 from the SOURCE menu. Enter your dataset name and
description. You will then see the following dialog box;
The Uninsulated and Insulated options have similar implications to those in

model S9. Turn back to section 3.4.1 for a full description.
Note: If your Horton sphere had thermal insulation or a relatively thick fire protection coating, we
would advise you to select the Insulated option.
For the purposes of this example select an Uninsulated and Spherical vessel
configuration. Continue and you should see the input screen. Do not overwrite

any of the red values at this stage. Input a temperature of 15C, a vessel
diameter of 20m, a vessel wall thickness of 25mm, the height of liquid in the
vessel of 10m (i.e. half full), a release orifice diameter of 250mm, an orifice
centreline height above base of 20 m (i.e. at the top of the vessel), change the
default discharge coefficient to 1 and input an ambient temperature of 15C.
You would follow a very similar procedure to this if you wanted to model a
vapour release from a relief valve located on the top of a vessel.
Now change the liquid density from the default option of 582kg/m3 to 508kg/m3.
On changing the liquid density the mass of the vessel contents changes from
1,218,938kg to 1,063,953kg. These values are calculated for you by CIRRUS to
enable you to cross check your input data. When you overwrite the values in red
then the underlying relationships will disappear. (In reality you are overwriting a
formulae in an Excel cell with a constant value).
You will see from the input screen that the default saturated vapour pressure is
7.5bara. This has been obtained by using Antoine coefficients, and can be
compared with the estimate made from using model S9 of 7.3 bara. In reality the
S9 estimate is likely to be the more accurate, but a default has been included in
this model as an option for users who do not wish to bother with flash
calculations. If you are not using pure Propane or you are using Propane as a
surrogate material you may wish to overwrite this value. You may also observe a
difference in the fraction of aerosol which remains in the cloud between the two
models for the same reason.

Your input screen should look something like this:
and on running the code you should get the following results:
As you can see, the vapour outflow rate is constant at just over 100kg/s. We can
explain this on the basis that using the Uninsulated option, heat is freely
absorbed by the liquid from the surroundings and therefore the liquid in the
vessel remains at 15C. Propane from the liquid will boil to maintain the
saturated vapour pressure in the vessel, therefore the outflow of material from
the hole inthe top of the tank will remain constant. As a result all the contents of
the vessel will vaporise and be lost through the hole.
Note: Look at the vessel temperature graph and the cumulative mass graph to check that this is the case.

Now re-run the case using the Insulated vessel option. In this option heat
cannot be freely absorbed by the liquid from the surroundings and therefore the
liquid in the vessel will cool as the Propane boils off to replace the vapour lost
through the hole. As the liquid cools its vapour pressure will decrease, and the
pressure in the vessel will decrease until the Propane cools down to its boiling
point where significant vapour evolution will stop.
Note: This is because latent heat is lost during the evaporation process and not replaced by heat input
from the surroundings.
We would expect the flowrate to be the same as the previous case initially, but to
decrease to zero as time progresses. This is the result as shown below.
Note also that in this case not all of the Propane liquid is vaporised, only that
which boils off before the Propane reaches its boiling point of -44C. You can
view these effects by selecting the graphs which show cumulative mass lost and
vessel temperature.
Note: We have just seen that using the uninsulated vessel option will give the greater material lost per
unit time. As we will see later this is not necessarily the worst case, as increasing the velocity
of a vapour release through an orifice can result in lower dispersion distances due to the turbulent
mixing of air into the plume. In reality an uninsulated vessel will behave somewhere between the
two extremes presented here.

6.5.2 A Multi Component Material at its Saturated Vapour Pressure
A multi-component material at its vapour pressure would behave in concept in

exactly the same way as a single component material, with the exception that we
might expect the lighter ends of the mixture to boil off first, with the result that
the vapour released from a hole in the vessel would be made up of
predominantly the lighter fraction of the mixture. The vapour released would
then tend to get heavier and its rate fall as the SVP of the mixture reduced with
time.
Note: The crude oil distillation process itself is a very similar process, where the temperature
differential up the column allows different molecular weight fractions to condense out at different
stages.
Within the accuracy of consequence modelling we can ignore much of this detail.
All we are looking for is a surrogate single component material which is
expected to behave in a similar manner to the multi-component vapour released
to atmosphere from our vessel within the first few minutes of the release. Later,
for example in section 10, we will discuss in more detail how to model these
type of releases.

164
6.6 SOURCE TERM MODEL - S6 RELEASE FROM THE LIQUID SPACE

OF A TWO PHASE VESSEL
6.6.1 Release of Single Component Liquid at its Saturated Vapour Pressure
This model is a corollary to model S5, excepting that this time we are going to
model a hole in the bottom of the vessel which contains the liquid phase instead
of the top of the vessel which contains the vapour phase.
As a first example in modelling the release of a saturated liquid we are again

going to look at the example of a pure single component, Propane, from the same
Horton sphere used as an example in model S5. If you remember the Horton
sphere was 20m diameter with a volume of about 4,200m3. We have assumed
that it is about half full of Propane liquid (about 1,000te) and about half full of
vapour. The pressure of the sphere was that of the saturated vapour pressure at
the ambient temperature. If the ambient temperature is 15C, then using model
S9, we determined that the saturated vapour pressure was about 7.3 bara.
In this case we are going to assume that a breach equivalent to a hole with a
250mm diameter occurs in the flange connecting the vessel itself to the outlet
pipework at the base of the vessel.
Now select model S6 from the SOURCE menu. Enter your dataset name and
description. You will then see the following dialog box:

From this dialog box you can select the geometry of the vessel, the pressure
conditions of the vessel and whether the release occurs directly from vessel wall
or from pipework attached to the wall.
Note: The two pressure related options for this model have the following implications:
Vessel at Saturated Vapour Pressure: This option would be used where the material in the vessel
is at its saturated vapour pressure at the vessel temperature; i.e. the material has a vapour pressure
greater than atmospheric pressure at the vessel temperature and the pressure is not increased
above the SVP by the presence of an ideal gas in the vessel.
Vapour Pressure Maintained Above SVP: This option would be used where an ideal gas was
used to maintain the pressure of the tank above its saturated vapour pressure. In the example of a
Propane vessel at ambient temperatures, the pressure in the vessel would be increased if a gas
such as nitrogen or methane were injected into the vessel.
Release From Attached Pipework: Similar to model S1, if you choose the option of a release
from pipework attached to the tank, the results will give a reduced outflow rate in comparison to
a direct release due to the frictional forces associated with flow through the pipework. If the
release scenario occurs very close to the vessel wall (say a couple of inches) then selection of a
release directly from the vessel would be the most appropriate option.
To continue with our example you should select a spherical vessel, at its
saturated vapour pressure with the release occurring directly from the vessel
itself.
Continue, select Propane as the material of interest and you should then see the
input screen. Do not overwrite any of the red values at this stage. Input a
temperature of 15C, a vessel diameter of 20m, a vessel wall thickness of 25mm,
the height of liquid in the vessel of 10m (i.e. half full), a release orifice diameter
of 250mm, an orifice centreline height above base of 0 m (i.e. at the bottom of
the vessel), change the discharge coefficient to 1.

As you did for the example of model S5, change the liquid density from the
default option of 582kg/m3 to 508kg/m3 to reflect the density difference between
the default value of the density at the boiling point and the actual value of the
density at the temperature of interest. Your input screen should look like this:
Note: As with our other release rate models the release rate calculated is directly proportional to the
discharge coefficient (Cd). In this case a Cd of between 0.6 and 1 is appropriate. For more
information on discharge coefficients see the CIRRUS on-line help (press F1 when the discharge
coefficient input box is selected).
As with S5, you will also see from the input screen that the default saturated vapour pressure is
7.5bara. This has been obtained by using Antoine coefficients which do not give exactly the
same answer as the more rigorous approximation of 7.3 bara for these conditions used by the
flash calculation program, model S9.

Now run the model and you should get a result which looks like this:
The outflow rate as a function of time is approximately constant, which is a

reflection of the fact that the pressure in the vessel will be maintained at the
saturated vapour pressure (i.e. as the volume of liquid in the vessel decreases due
to outflow through the hole, Propane will vaporise (boil) to maintain the pressure
in the vessel. The slight drop in outflow rate is due to the decreasing pressure
exerted by the hydrostatic head component as the level of liquid in the vessel
falls with time.
The outflow rate calculated is approximately 1.3te/s for 800 seconds (13
minutes). During the release the Propane will flash and both the liquid and
vapour components will drop to the boiling point of Propane at the ambient
conditions (about -42C at atmospheric pressure).
From this point on modelling the dispersion of Propane vapour following this
release becomes a bit tricky!
6.6.2 Estimating a Vapour Release Rate from the Liquid Release Rate
In this scenario of a hole in the bottom a of a Horton sphere you have calculated
the liquid release rate through the hole in the vessel. What you will probably
want to know is how far the vapour from this release travels as a flammable
cloud.
The vapour release rate is made-up of two components: Firstly as the liquid
exits the orifice it will flash. This means that we will have to use model S9 to

calculate a flash fraction of the total release which will be in the vapour phase.
Secondly if any of the liquid reaches the ground before vaporising it will form a
liquid pool, therefore we will have to estimate the evaporation rate of the liquid
pool formed; using model S8.
Note: Use of model S9 for this type of scenario was explained in section 3.1.5, and use of model S8
was explained in section 3.3.
Now run the flash fraction calculation, model S9, for these conditions. Select
model S9, choose single Component and Propane as the material of interest and
input 15C as the temperature of interest. (Input a nominal pressure of 10 bara,
this is not used in this calculation). You should get a results screen which looks
something like this:
After flashing to atmosphere both the liquid and the vapour will be at the boiling
point of Propane, -42C. The flash fraction calculation shows us that 68% of the
liquid mass will remain as a liquid fraction and 32% will evaporate to form a
vapour fraction. Therefore of your original 1.3te/s release rate about 0.9te/s will
be liquid and 0.4te/s will be vapour.
Now if we assumed that all the liquid disassociated from the vapour cloud, i.e. if
all the liquid rained out of the cloud, then this would form a pool on the ground
which would then evaporate. Note however that CIRRUS suggests that 100% of
the liquid fraction remains in the cloud. Therefore for this case it is appropriate
to assume that no liquid will rain-out of the cloud.
In the Dispersion section of this report you will see that we are going to suggest
that the most appropriate way of modelling the vapour dispersion from this
release is to assume that all the liquid will be carried with the dispersing vapour
as an aerosol. If the release does not impinge on the ground the entire release

should be input to the two-phase option of model D1. For a release which does
impinge on the ground the entire release should be input to model D2 - but more
of this later when, hopefully, the underlying reasons for these apparently ad-hoc
suggestions will become clear.
Note: It is interesting to compare the difference if you assume that all the liquid rains out of the cloud.
This would form a pool of Propane at its boiling point. You can model the spreading and
evaporation of this pool using model S8; try it now. Select model S8, input a dataset name and
description and then select the continuous discharge option. Also for this example select the
unbunded option. Put in a discharge rate of 900kg/s for 13 minutes (780 seconds). Use a
ground roughness of 0.01, a temperature of 15C and a wind velocity of 5m/s. The results of this
option are shown below:
You will see that the pool spreads to a diameter of about 300m, but does not reach its maximum
diameter until after the discharge has finished. You will also see that the vapour discharge rate
increases to a maximum of 650kg/s after about 800seconds, plateau at this level and then starts to
diminish. This result is shown below. To model vapour dispersion assuming 100% rain-out you
would have to take these values, but also add the vapour flash fraction of 400kg/s to the release
rate for the first 800 seconds. In other words the total vapour release rate would start at about
400kg/s and increase to about 1250kg/s after 800 seconds.
Therefore this assumption would be less pessimistic than a continuous mass release of 1300kg/s
forming a vapour cloud from time zero as suggested previously.

6.6.3 Release of a Multi-Component Liquid at its Saturated Vapour Pressure
As with other models in order to model the release of a multi-component liquid

which is stored or processed at its saturated vapour pressure you must use as a
surrogate a single component material. In this case you would choose a single
component with a similar density and molecular weight to the density and
average molecular weight of the multi-component liquid. Following this you can
select the surrogate single component material as the material of interest in
model S6, overwriting the material name, vapour pressure, density and other
physical characteristics where known with those of the multi-component
material. You can obtain the vapour pressure of the multi-component liquid by
using model S9 as described in section 3.1.4 of the report.
We have now modelled the release from tanks or spheres of a normal liquid, a
normal gas, and the vapour and liquid fractions of two-phase fluids. We are now
going to model the release of these types of materials from pipeline systems.
The models in CIRRUS used to accomplish this are model S2 - liquids, S4 -
gases and S7 - two phase fluids.

172
6.7 SOURCE TERM MODEL - S2 RELEASE FROM A LIQUID PIPELINE
The purpose of this model is calculate the release of materials which are liquids
at ambient temperatures; i.e. materials which have a vapour pressure of less than
one atmosphere at ambient temperatures. We are going to look at the use of the
model when used with single component material and multi-component
materials. As a single component material we are going to model a pipeline
transporting Octane, and as a multi-component material we are going to choose
the gasoline mixture which you made up in section 3.1.6.
6.7.1 Modelling of a Single Component Material
Select model S2 from the SOURCE menu, input a dataset and description and
you should then see the dialog box show below:
Note: The options have the following implications:
Pipeline with risers: The model in CIRRUS can split a pipeline into three sections with
elevations A to B, B to C, C to D. Originally this was developed to enable the program to deal
with an offshore situation where a pipeline leaves an offshore platform at typically 20m above
sea level, travels vertically downwards to the sea floor, say at 100m below sea level, travels a
distance of say 20km along the sea floor to another offshore platform, where it travels vertically
upwards onto the new platform to an elevation of typically 20m above sea level. In total the inlet
and outlet elevation of this pipeline are the same but the pipeline is not horizontal. A hole half
way along the pipeline will have, in addition to any pumping pressure, a hydrostatic pressure due
to the 120m elevation at either end of the pipeline.

The portion of the pipeline from an offshore platform to the seabed is called a riser, hence the
question in the dialog box. This version of the code can be used for any other geometry of
pipeline which you may want to split into two or three sections; for example a pipeline which
runs along a flat plateau and then down the side of a mountain.
With valve: This option allows you to specify the location and time to closure of a valve in the
pipeline system. Following closure the model will not allow liquid to continue to flow from an
section of pipeline upstream of the valve to a breach in the pipeline.
Select the default option from the dialog box, pipeline without risers and
pipeline without valve, then select the material of interest to be Octane. Fill
out the input screen in a similar manner to that shown below.
Define a pipeline which runs from a relatively mountainous area to the coast.
The pipeline has a diameter of 500mm (20), with an inlet elevation of 500m and
an outlet elevation of 0m. Leave the default wall roughness of 1mm. Input a
flowrate through the pipeline of 100kg/s and a temperature of 15C. The
pipeline has an inlet pressure of 20bara. Select a fracture in the pipeline of area
equivalent to a hole diameter of 300mm located at a position 5,000m from the
inlet. Use 30 minutes as the time of interest.
Note: The elevations used are relative, therefore an inlet elevation of 1,500m and an outlet elevation of
1,000m would define the same pipeline as the above.

Using the suggested input conditions, you should end up with a screen which
looks like this:
Note: The wall roughness represents in mm the roughness on the inside of the pipe, and this governs the
pressure drop down the pipeline. The selected roughness should not be too large in relation to
the pipe diameter. If values which would make the code unstable are selected, then these are
automatically overwritten by the program. Ideally you should vary the wall roughness to best
estimate the pressure at the breach location during normal operations.
Run the model and you will see and intermediate screen with the following
message.
This indicates that the flow from the breach has reached steady state conditions -
probably the inlet flowrate which you used.

Take the output and change the axes to log:log; you should see a plot which is
similar to that shown below:
Firstly notice that the pressure at the breach is 36 bara - i.e. above that of the
inlet pressure. This is because the hydrostatic head of the fluid upstream of the
breach, which occurs halfway down the hill, adds to the inlet pressure.
Secondly notice the shape of the curve, from a very high initial instantaneous
flowrate of greater than 1te/s during the first second of the release, the flowrate
drops as material upstream of the hole drain through the hole due to a reduction
in the hydrostatic head. Note also that the flow from the hole reaches the steady
state inlet flow after about 600 seconds, i.e. all the stored liquid in the pipeline
has escaped. After this period until the time which you input as the time of
interest, 30 minutes, the flowrate remains constant at the inlet flowrate.

If you now look at the graph which shows the cumulative mass of liquid
released, you will see the following graph. Approximately 770te of Octane was
released in the first 10 minutes, with a total of 880te of Octane being released in
the 30 minute time of interest.
Note: In these types of calculation always use a relatively short time of interest in the first instance. In
the above example had you input a 24 hour time of interest, all the initial outflow detail would
have been lost on the plots.

6.7.2 Modelling of a Multi-Component Material
In model S2 you can input your own user defined multi-component material.
As an example, select the gasoline which you created in section 3.1.6. Use the
same pipeline conditions as those described above.
You should be able to reproduce the following log:log graph.
As you would expect, this is a very similar result to the Octane pipeline.

6.8 SOURCE TERM MODEL - S4 RELEASE FROM A GAS PIPELINE
This model estimates the release rate as a function of time from a pipeline
containing a gas or vapour which is not a saturated vapour; i.e. it lies within the
ideal gas region of the phase diagram discussed in section 3.1.
The most obvious example of this case is a Methane pipeline; either a low
pressure pipeline used to distribute fuel gas, or a high pressure cross-country
transmission pipeline. A less obvious example would be a low pressure Propane
pipeline which is operating at ambient temperatures at a pressure less than its
saturated vapour pressure.
Therefore as an example of using model S4, we are going to estimate the release
rate from a 150mm diameter pipeline operating at 20C and 5 bara pressure
which is attached to a Propane scrubber with a volume of 200m3. The pipeline is
5 km long and has a breach in it equivalent in area to a hole of 100mm diameter.
The time of interest is 10 minutes.
Note: The location of the breach in gas pipelines is less likely to affect the release rate conditions. In
model S4 the frictional drop down a pipeline system is modelled less rigorously than that in the
liquid or two-phase pipeline model.
Select model S4 from the SOURCE menu, enter a dataset name and description
and select Propane as the material of interest. You should then see the following
dialog box:

Note: The three choices have the following implications:
Constant Volume Reservoir Feeding Pipeline: This option should be used where the pipeline
system is attached to a vessel or other containment holding a large volume of gas. The model
will calculate the escape of gas from the pipeline itself and the large volume feeding into it.
Continuous Flow Feeding Pipeline: This option should be used where gas is being pumped into
and through the pipeline system. In this option an attached vessel or other containment holding a
large volume of gas would not feed into the pipeline system, the feed to the system would be that
of the pumped gas.
When the system to be analysed involves both pumped flow and attached vessels, run both of the
options separately to decide which is the more important influence on the result. In the majority
of cases over short time periods the feed from an attached vessel is likely to be of far greater
significance than that of pumped flow.
No Flow Feeding the Pipeline: This option should be used where the pipeline is effectively
isolated from an external vessels or pumps which could feed gas into the system.
Select constant volume feeding pipeline as this best represents the Propane
scrubber plus attached pipeline. Input the pipeline parameters described above
and you should end up with a screen which looks something like this:

The model output should look something like this:
From this result you can see that we expect an initial instantaneous release at
about 9kg/s falling to about 1.5kg/s after 3 minutes up to the 10 minute time of
interest. Print this graph as we will refer to it later when we analyse the
dispersion of this release using model D1 in the dispersion section of this
manual.
6.8.2 Modelling of a Multi-Component Material
Similar to the other models which consider an ideal gas type of release, we
cannot model a multi-component vapour directly but must choose a surrogate
single component material. It this case it is appropriate to choose a single
component whose molecular weight is similar to that of the averaged molecular
weight of the multi-component vapour.

182
6.9 SOURCE TERM MODEL - S7 RELEASE FROM A TWO PHASE

PIPELINE
This model has been designed to calculate the release of liquid and vapour from
a pipeline system transporting a material which is liquid at the containment
pressure and temperature but is all or part vapour at atmospheric pressure and
ambient temperatures.
Examples of single component materials which conform to this definition

include Propane or ethylene transported at moderate or high pressure. Examples
of multi-components which conform to this definition include crude oil fractions
which contain C2 or C3.
To make a change from examples using Propane, this time we are going to model
a scenario related to a breach in an Ethylene pipeline. We are going to base the
scenario on a horizontal Ethylene pipeline of 500mm diameter with a wall
thickness of 25mm and a length of 10km (10,000m). The pipeline flowrate is
100kg/s, the pipeline temperature is 8C and the inlet pressure is 50 bara. The
breach in the pipeline has an area equivalent to a hole of diameter 300mm which
is located at a distance of 5km (5,000) from the inlet. The time of interest is 30
minutes.
If you browse the CIRRUS database you will noticed that the normal boiling
point of ethylene is about -104C, therefore we expect the release to flash into
predominantly vapour.
Note: Ethylene is a particularly tricky material to model, as its critical temperature (9.3C) is around
ambient temperature, and its critical pressure (50 bara) is close to its saturated vapour pressure at
ambient temperatures.
Select model S7 from the SOURCE menu, input a dataset and description and
you should then see the dialog box similar to that described for model S2 in
section 3.7.1. The notes relating to this dialog box are repeated here for
convenience.

Note: The options have the following implications:
Pipeline with risers: The model in CIRRUS can split a pipeline into three sections with
elevations A to B, B to C, C to D. Originally this was developed to enable the program to deal
with an offshore situation where a pipeline leaves an offshore platform at typically 20m above
sea level, travels vertically downwards to the sea floor, say at 100m below sea level, travels a
distance of say 20km along the sea floor to another offshore platform, where it travels vertically
upwards onto the new platform to an elevation of typically 20m above sea level. In total the inlet
and outlet elevation of this pipeline are the same but the pipeline is not horizontal. A hole half
way along the pipeline will have, in addition to any pumping pressure, a hydrostatic pressure due
to the 120m elevation at either end of the pipeline.
The portion of the pipeline from an offshore platform to the seabed is called a riser, hence the
question in the dialog box. This version of the code can be used for any other geometry of
pipeline which you may want to split into two or three sections; for example a pipeline which
runs along a flat plateau and then down the side of a mountain.
With valve: This option allows you to specify the location and time to closure of a valve in the
pipeline system. Following closure the model will not allow liquid to continue to flow from an
section of pipeline upstream of the valve to a breach in the pipeline.

Select the default option from the dialog box, pipeline without risers and
pipeline without valve, then select the material of interest to be Ethylene. Input
the pipeline parameters described above for this scenario. You should have an
input screen which looks like the one below:
Now run the model, and you will get the following message:
This indicates that all the mass in the pipeline has been depleted within the
30 minute time of interest. The outflow through the hole at the end of the time
period should be equal to that of the inflow; in this case 100kg/s.

Now carry on and look at the results to confirm that this is the case:
The default results screen above shows you the combined vapour and liquid
mass flowrate. If you interrogate the Graph menu item you will be able to
display the vapour release rate and the liquid release rate. As expected, the
vapour flowrate makes up the dominant proportion of the outflow. This graph is
shown below:
For consequence modelling purposes you can now follow a similar procedure to
that discussed for model S6. The vapour fraction can be entered into a vapour

dispersion code directly. The liquid fraction would form a pool on the ground at
the boiling point of ethylene (-104C) which would then boil/evaporate; use
model S8 to estimate this evaporation efflux rate. As the boiling point of
ethylene is low in comparison to most ambient temperatures this evaporation
would be rapid. For most practical applications you would put the entire mass of
the release directly into a dispersion code. We discuss this aspect further in the
dispersion section of this manual.
6.10 SUMMARY OF SOURCE TERM MODELLING
You have now come to the end of the section explaining the various types of
source term model, and you have calculated the release rate if a breach was to
occur in a variety of different types of containment for a variety of different
materials.
The next step on the process of consequence modelling is to determine where the
material would disperse given the release conditions, and if the material was
ignited, what type of fire would ensue.
In the next section of this manual we discuss vapour dispersion modelling.

188
PART 3. DETAILED SCENARIO MODELLING SECTION 7. INTRODUCTION TO VAPOUR DISPERSION
7. VAPOUR DISPERSION MODELLING
As discussed in the Introduction to this manual, there are specific types of model
available to us in order to model the dispersion of vapour through the
atmosphere. In this section of the manual we explain how to best use the
available models to predict hazard distances from facility release scenarios.
7.1 GENERAL CONSIDERATIONS
7.1.1 The Generic Types of Consequence Model
Overleaf we have listed the specific types of consequence model available, and
we have listed whether they accept a continuous steady state or transient
source term.
A continuous steady state source term is one where the release rate is assumed
to go on for ever at the same rate. This allows the model to be an equilibrium
model. You will soon demonstrate that the dispersion of a release into
atmosphere often reaches its maximum extent after about 10 minutes of the
release, and thereafter the dispersion cloud does not continue to grow as the
release continues.
A transient source term is one where the release rate changes as a function of
time. As demonstrated in the source term modelling section of this manual this
is often the case in realistic scenarios, especially those involving relatively major
releases from hydrocarbon processing plant.

Table 4.1 The Specific Types of Consequence Model Available
Description Type Continuous Transient

Steady State Source
Source
D1 Dispersion from a Yes No
Momentum Release
D2 Heavy Gas Dispersion Yes Yes
Buoyant or Neutrally
D3 Buoyant Gas Dispersion - Yes No
Short Term Average
Buoyant or Neutrally
D4 Buoyant Gas Dispersion - Yes No
Long Term Average
D5 Dispersion of Particulates Yes No
In some scenarios, especially those related to a major release of material from

high pressure plant, the initial outflow of hydrocarbon liquid or vapour will drop
markedly within the first few seconds of time. If these types of scenarios were to
be modelled as if they were continuous releases at the initial flowrate, the
amount of material lost in the first ten minutes of a release would be a gross
overestimate of that which would, in actuality, be lost. This type of assumption
is erroneous and can lead to a gross overestimate of the estimated dispersion
cloud.
Note: This error is especially prevalent when an analyst or consultant only has available a single point
estimate of release rate calculated from a simple formulae. In this formulae it is natural to use the
initial pressure of the vessel or pipeline of interest, which naturally leads to a calculation of the
highest release rate. Application of this release rate to a continuous dispersion model provides
us with, arguably, one of the most prevalent and largest sources of error found in major accident
risk assessment.
Therefore as the majority of models above only accept a continuous steady state
source term, we have to be very careful how we use these models to calculate
dispersion distances.

7.1.2 Parameters with Important Implications in Dispersion Modelling
There are certain generic parameters which are used by all forms of dispersion
model which can have major implications concerning the model predictions.
These include the parameters of atmospheric stability, surface roughness,
toxic and flammable limits and sampling time.
7.1.2.1 Atmospheric Stability and Windspeed
In any weather condition the speed of the wind is a function of height above the
ground. Virtually all dispersion models assume that the vertical wind velocity
profile conforms to a power law description, which depends on parameters such
as surface roughness, air stability and temperature.
Note: The air in contact with the ground will be still, therefore at the level of the ground the wind
velocity will be zero in all wind speed conditions. In dispersion codes ground level is actually
interpreted to be the roughness height.
For most dispersion modelling purposes the speed of the wind is that which is
measured at a height of 10m. Therefore when you input a wind speed of 5m/s
(about 11 mph) to any of the CIRRUS dispersion codes, this corresponds to the
wind speed of interest at a height of 10m. The windspeed at say 2m (person
height) may be lower than that at 10m by a factor of 50% or more depending on
the stability category chosen.
The change in the vertical temperature profile of the air (often referred to as the
lapse rate) determines the stability of the air. Air decreases in temperature until
the height of the inversion layer is reached. When the temperature profile is
changing the air below the inversion layer will either be stationary, rise, or fall
depending on the direction of change of the temperature profile.
Correspondingly any vapour cloud dispersing through this medium will rise and
fall with it.
Note: The inversion layer corresponds to a point in the atmosphere where there is a temperature
inversion. The air temperature cools as height increases up to the inversion layer. Thereafter the
air temperature can increase.

Another, more important, feature of stability class is the degree of turbulence in

the air. Air turbulence increases the dispersion process. The normal turbulence
in the air increases as the air is rising and decreases when the air is falling.
In neutral conditions, categorised using the Pasquill-Gifford definition as D

stability, the air is neither rising or falling. A dispersion cloud will rise or fall in
the atmosphere due to its own positive or negative buoyancy but not due to the
atmospheric conditions.
In unstable conditions, categorised as A stability (and to a lesser extent 'B and

C stability), the atmosphere is rising. Therefore a neutrally buoyant plume
would rise with the atmosphere. It will tend to disperse more rapidly due to the
increased air turbulence.
In stable conditions, categorised as 'F' stability (and to a lesser extent 'E'

stability), the atmosphere is falling. Therefore a neutrally buoyant plume would
fall with the atmosphere. It will tend to disperse more slowly due to the
decreased air turbulence.
Note: Choice of atmospheric stability will have most influence on the results of heavy gas models (e.g.
model D2), where 'F' stability tends to give a far wider and often longer ground footprint than
the equivalent run in 'D' stability. The choice of stability category will have a moderate effect on
both the height and width of a gaussian plume (e.g. model D3&D4), but will tend to have less
influence on the results of the momentum models (e.g. model D1).
Atmospheric conditions will always be neutral, D stability, in moderate to high

wind conditions, for example 5m/s (11 mph) or greater. This is because air is
mixed by the wind and a temperature gradient does not exist. Extensive cloud
cover will also result in neutral atmospheric conditions.
In the early afternoon under moderate or strong insolation (sunshine) the ground
heats up relative to the air above it. This sets up a positive temperature gradient
in the air which can result in unstable, A stability, conditions.
In early evening, after dusk, the ground cools relative to the air above it. This
sets up a negative temperature gradient in the atmosphere which can result in
stable, F stability, conditions.
Note: For full guidance on the choice of stability category look at the CIRRUS on-line help (key F1)
after entering a stability category on a dispersion model input screen.
At the time of writing the 'F' stability predictions of the heavy gas models are based on a physical
model, but as no field trials have been carried out in 'F' stability conditions there is little data
available to assess whether this sensitivity is real, or a spurious result arising from the particular
formulation of the model. As these predictions often dominate safety assessments, field trials
have been planned to test the validity of heavy gas modelling in 'F' stability conditions.

7.1.2.2 Surface Roughness
Surface roughness is a parameter which describes the roughness of the ground.

It affects dispersion modelling by influencing the turbulence and wind velocity
profile in the atmosphere. Surface roughness is described with a coefficient
with units of meters. The numerical value of this coefficient is approximately
one tenth of the height of obstacles on the ground. For example smooth ice
has a roughness parameter of 1 x 10-5 m and long grass has a roughness
parameter of 3 x 10-2. Cities or other conurbation of buildings (e.g. refineries)
are often allocated a roughness parameter of 1m or greater.
For ground level releases the roughness parameter chosen should reflect the
average roughness of the ground over which the cloud is dispersing. For
example if a cloud is predicted to disperse for 200m initially in the facility
complex and then for 2km over fields beyond the facility fence, then it would be
appropriate to use a roughness associated with the fields rather than the facility.
This case will include all scenarios using a heavy gas model.
For elevated releases the roughness parameter input should reflect the ground
roughness upwind of a release point. This is because the release is dispersing in
the air which has turbulence created by ground roughness upwind of the release
point. To be more exact at any point along an elevated plume, the roughness
which affects this air turbulence is not that of the ground underneath the point
but that of the ground upwind of the point at a distance about ten times the
height of the centreline of the plume at the point. Again to be pragmatic you
should choose the roughness which is relevant to the major portion of the plume.
As the roughness length chosen will affect the distance over which the plume
disperses, correct use of the above suppositions may require iterative use of the
models.
Note: The various types of model have different sensitivities to the ground roughness parameter.
Heavy gas models (e.g. model D2) are extremely sensitive to this parameter. Increasing the
roughness parameter from 0.1m to 1m can halve the dispersion distance predicted from heavy gas
models. Gaussian models (e.g. model D3&D4) and momentum models (e.g. model D1) are
less sensitive to ground roughness. As with the sensitivity of heavy gas models to stability
category, there is little data available to assess whether this sensitivity to roughness is an accurate
representation, especially in regard to heavy gas models. Perhaps because of this, at the time of
writing, the US EPA do not advise modelling using a roughness parameter of greater than 0.1m
for use with heavy gas models.

The dispersion models commonly used assume a free field process; i.e. they do
not account directly for obstacles on the terrain, but use roughness to influence
the air turbulence for dispersion above obstacles on the terrain. For a release in
the centre of a facility the cloud may initially disperse through and around
obstacles - this process is not modelled. Since heavy gas clouds have a
maximum height of only a few metres, in reality the gas cloud will be dispersing
through and around obstacles. Increasing the roughness parameter will not
model this process, but will influence only the dispersion above the obstacle
layer. For this reason it is recommended that a roughness parameter should be
used which reflects free field terrain.
7.1.2.3 Toxic and Flammable Limits
The Lower Flammable Limit (LFL) is defined as the minimum quantity of gas
which, when mixed with air, will ignite. Similarly the Upper Flammable Limit
(UFL) is defined as the maximum quantity of gas mixed which, when mixed
with air, will ignite. The LFL and UFL are normally presented in terms of the
percentage of flammable gas in a gas:air mixture on a volume basis. Thus the
LFL of Methane is presented as 5%vol:vol; i.e. a mixture of Methane and air
which contains 5% by volume Methane. Similarly the UFL of Methane is
15%vol:vol.
As the molecular weight of a hydrocarbon increases the trend is for both the LFL
and UFL to decrease. For example the LFL of Propane is 2.1%vol:vol. and the
UFL of Propane is 9.4%vol:vol.
Note: We model the dispersion of flammable mixtures in order to ascertain the maximum extent of a
cloud which, in all probability, could be ignited. From current knowledge the most appropriate
concentration to choose for this purpose is the LFL boundary. There is little scientific evidence
to suggest that gas clouds can be ignited at concentrations less than the predicted LFL boundary
if sensible sampling (averaging) times are chosen.
In CIRRUS default concentrations of interest are automatically read from the database for
flammable mixtures. These can be overwritten with alternative values when flammability is not
the feature of interest.

7.1.2.4 Sampling Time
Sampling time and averaging time are often used as a description of the time
over which a gas concentration is averaged. At a particular point in space the
concentration of a plume at equilibrium will vary for two reasons: Firstly as the
wind direction is not perfectly constant the plume will meander about a mean
value. Secondly there are in-cloud concentration fluctuations due to the
turbulence inherent in the atmosphere. As dispersion codes aim to show an
time-averaged concentration at a particular point, this average will depend on
the length of time over which the concentration was sampled.
The situation is complicated further by the fact that momentum (e.g. model D1)
and gaussian models (e.g. model D3&D4) include both a plume meander
component and an in-cloud concentration fluctuation component in their
formulation, whereas heavy gas models (e.g. model D2) often only include only
the latter component.
As a rule of thumb, for momentum; and gaussian models select a sampling

time equal to the period of interest up to a maximum of 1 hour. Use a similar
rule when using heavy gas models excepting for 'instantaneous releases, when
a much shorter time is appropriate.
Note: In CIRRUS the default setting for the former is 10 minutes and that for the latter is 20 seconds.
When using models which move into the passive phase, for example the use of model D1 to
predict low concentrations or when using models D3 or D4, the choice of sampling time can have
a large effect on the ground level concentration. This is because at short sampling times the
models predict that the plume will be relatively narrower but longer than those of longer
averaging times. These changes in plume shape will affect the distance at which a particular
concentration boundary will touch the ground. Therefore changing the averaging time can alter
the distance to a particular ground level concentration by up to a factor of five or greater.

y
196
PART 3. DETAILED SCENARIO MODELLING SECTION 7.2. MODEL D2
7.2 DISPERSION MODEL - D2 DISPERSION OF HEAVY THAN AIR

VAPOUR FROM A LOW MOMENTUM RELEASE AT GROUND LEVEL
To begin our discussion of dispersion models we are going to look at the use of
the class of model called Heavy Gas models. In CIRRUS this is model D2.
Note: In CIRRUS version 5 we have chosen to use a heavy gas model called HEGADAS which was
developed by Shell oil company. Similar models of the same type include SLAB, DEGADIS,
GASTAR and DRIFT.
This model is appropriate to predict the dispersion behaviour of a vapour cloud

which is heavier than air. A vapour cloud can be heavier (denser) than air for
several reasons:
Low Temperature: The density of a vapour will increase as the

temperature decreases. For example the density of Methane vapour at -
161.5C (the boiling point of Methane) is 1.73 kg/m3. The density of
Methane vapour at 15C is 0.68 kg/m3. The density of air at 15C is
1.23 kg/m3.
Therefore as vapour is released from a boiling pool of liquid methane

(LNG) at -161.5C it is heavier than air and will hug the ground. As it
disperses it will mix with air and warm up. At some stage the mixed
vapour will become lighter than air and will begin to lift off the ground.
Clearly these processes have dramatic implications for the concentration
of methane and similar materials at ground level.
Note: As we decrease the ambient temperature the density of air increases.

(The density of air at -50C is 1.58 kg/m3). Therefore the density differential between a
heavy vapour and air decreases as the air temperature decreases. In other words the
vapour becomes relatively more buoyant as the air temperature decreases. The opposite
is true as the air temperature increases.

High Molecular Weight: At the same temperature the density of the

vapour is greater than that of air. For example the density of Propane at
15C is 1.9 kg/m3. in comparison to that of air which is 1.23 kg/m3 at
15C.
Note: If hot Propane vapour (>180C) was released into air at 15C then its initial
density would be less than that of air; at least initially it would behave as a buoyant gas.
CIRRUS will not allow you to use a heavy gas model in this circumstance.
To calculate the density of a non-saturated vapour, you can use the Ideal Gas equations
given in section 3.1.1. At atmospheric pressure and 0C, 1kgmole (44.1kg) of Propane
will occupy 22.414m3 of volume; i.e. its ideal gas density at 0C is 1.97kg/m3 (= 44.1 /
22.414). As the temperature is increased from 273.15K (0C) to 288.15K (15C), the
volume occupied by 1kgmole of gas at atmospheric pressure increases to 23.645m3 ( =
[288.15 / 273.15] x 22.414), therefore the ideal gas density of Propane vapour at 15C is
1.87kg/m3 (= 44.1 / 23.645). This is less than the measured real value above but a
sufficiently good enough approximation for our purposes.
7.2.1 The Assumptions Inherent in Model D2
The mathematical model in D2 is based on the assumption that the initial source
is a heavier than air gas effusing from a rectangular area of ground with zero
momentum.
Clearly this has no exact corollary in the real world. The closest real scenario we
might envisage is that of a pool of liquid from which vapour is evaporating. The
vapour, of course, must be heavier than air.
Note: Following a release where the material flashes, the pool of liquid formed on the ground will
initially be at the boiling point of the material. This may well be significantly less than the
ambient temperature, and the vapour cloud formed may be heavier than air as a result of this low
temperature. For this reason that the default temperature of the vapour when using model D2 in
CIRRUS is set at just below the boiling point of the material.
Another basis of the mathematical model in D2 is that the cloud is initially

pancake shaped and spreads by gravity slumping and entrainment of air at its
sides, with very little entrainment at its top surface. This means that the cloud is
always of relatively low height, for example the lower flammable limit
concentration is typically less than 3m high, albeit the cloud may be a kilometre
in length and width. If the cloud reaches neutral buoyancy, due for example to
the entrainment of warm air into a cold low molecular weight vapour, the model
will change into a gaussian dispersion similar to model D3. This will allow the
cloud to rise.

7.2.2 Examples of the Use of Model D2 - Dispersion following Evaporation from a

Liquid Pool
The scenario you are going to model first is the evaporation of Butane from a
pool of refrigerated liquid Butane. In section 3.3 you modelled the evaporation
of a spill of refrigerated Butane into a tank bund. For convenience we have
repeated these results below. These two graphs show the same evaporation rate
plotted on Lin:Lin and Log:Log axes:

Note: If you want a closer look at the above graphs, but printed with both the major and minor axes
showing, this is relatively simple to do within CIRRUS. Rerun model S8 to obtain one of the
above graphs. Now select the Excel option on the menu bar. On selecting OK you will be
presented with the normal Excel menu bar. You can now unprotect the graph, select the
gridlines option and select both major and minor gridlines. Now print the graph. Select the
Cirrus option from the menu bar and this will return you to CIRRUS. Change the axes of the
graph and repeat the above.
The release shown is highly transient, with an evaporation rate of up to 34kg per
second in the first minute, falling to about 14kg/s after a week and 2.5kg/s after
two weeks as the temperature of the concrete bund cools down to that of boiling
Butane (-5C).
Note: As the material becomes more volatile the transient will become more extreme. For example a
similar Propane release would show a greater change in rate over the first few days, as in this
case the bund would cool from ambient conditions to the boiling point of Propane (-42C).
You might be interested in two aspects of this scenario. Firstly, what is the
maximum extent of a flammable vapour cloud, this is likely to occur within the
first ten minutes of the release. For example you could use this information to
determine who could be caught in a flash fire if the cloud were to be ignited.
Secondly what is the maximum extent of a flammable vapour cloud for the
remaining month. You might be interested in this as this would be your
sterilisation zone until all the spill had evaporated.
As a best guess you would model the first ten minutes of the release as a
transient release, and the equilibrium conditions after a few days as a
continuous release. We will do the latter first as it is the simplest.

Select Dispersion from the CIRRUS model menu, you should be presented
with a drop down box which looks something like this:
Choose the second selection, Model D2 - Heavy Vapour Dispersion. You will be
presented with the following dialog box:
Note: The various options have the following implications:
A Continuous Release: This will model equilibrium conditions. It effectively assumes that the
release will be constant and of sufficient duration to reach equilibrium; for practical purposes this
typically a 10 minute duration or greater.
A Transient Release: This will model a time varying source term. It does so by looking at the
result of an equilibrium calculation for a series of times of interest during or after the release.
An Instantaneous Release: To model an Instantaneous release we have assumed that the release
occurs at a constant rate over one minute. It is the transient model where the default averaging
time parameters have been adjusted to best fit this assumption. (We have opted for this
configuration in CIRRUS as in real scenarios there is no such thing as a truly instantaneous
release).
Concentration or Lethality: The concentration option will plot the dispersion boundary to a
%vol/vol of interest (i.e. the percentage of material in air on a volume basis). The lethality

option will plot the %lethality of an average population for a given exposure time according to
Probit equations published by the Dutch Government in their document CPR16E Methods for
the determination of possible damage.(often referred to as the TNO Green Book).
In short transient or instantaneous releases a toxic vapour cloud may pass over a given location
very rapidly. For example in a wind speed of 5m/s (measured at a height of 10m), a 500m
diameter vapour cloud will pass over a point 1 m above the ground in approximately 3 minutes.
The exposure of a person in the open to this cloud would be about three minutes. Care should be
taken when modelling lethality not to input totally unrealistic exposure times - this is often a
major source of error.
HF or an Ideal Gas: If the HF option is chosen this will use the special chemical
thermodynamic equations for HF incorporated in the HF HEGADAS model. Otherwise the
material is treated as if it behaved as an Ideal Gas.
The model in CIRRUS does not take into account the affects of barriers or
ground elevation. In reality the bund may be below the average ground level
over which the cloud is dispersing or the bund wall may affect the dispersion
process.
7.2.3 A Continuous Source Term
For our first example select the continuous steady state release and on land
option. Select Butane as the material of interest and Input the following
parameters: Roughness 0.01, wind velocity 5m/s, ambient temperature 15C,
stability category D, relative humidity 50%, vapour release rate 1.4 kg/s. Leave
the other parameters at their default values. For the time being leave the self-
calculated source diameter of 3.8m. We are going to investigate the influence of
this parameter in the next run.

Now run the code and you should see a plot showing the plume centreline
concentration as a function of distance.
The axes of the graph are Log:Log to allow you to read the values a little more
clearly. You should be able to see that a concentration of 8.4%vol:vol (the UFL
of Butane) occurs at a distance of about 14m and that the 1.8%vol:vol (the LFL
of Butane) occurs at a distance of just less than 40m.
Now go to the GRAPHS menu and select the Horizontal contour plot. This
shows you the shape of the cloud at ground level.

Notice that the UFL contour reaches 14m on the centreline, and the LFL contour
reaches just over 38m on the centreline.
You could conclude on the basis of this modelling that the sterilisation zone
around the spill two weeks after the incident should be 40m from the centre of
the pool; but remember we used a source diameter of 3.8m! If the original bund
diameter should have been about 45m, then the LFL would have been predicted
to occur within the bund.
Note: The default source diameter used in CIRRUS is an empirical guess which
provides reasonably sensible input to the underlying code given the release rate
of interest, and allows the model to start the dispersion process. If the code falls
over using this input you should try doubling or tripling this diameter. The
choice of source diameter will only affect the model results close to the source,
this facet is demonstrated below. Note however that the results of any dispersion
model are not reliable close to the source.
Using a small source diameter tends to concentrate the released vapour at the
source. Now rerun the model with the correct diameter of 44.7m which was
calculated by model S8. You will find that HEGADAS (the heavy gas dispersion
model in CIRRUS) cannot cope. This is because the initial release is spread over
the pool and is so small that the equations cannot be satisfied; i.e. the release rate
per unit area is too low. Now go back to the input screen and change the source
diameter to 19m, the largest source diameter with which the code will run
satisfactory for this release rate. Rerun the model and choose the horizontal
contour option.

You should get a graph which looks something like this.
In this case HEGADAS has represented this source by a square of dimensions

17m x 17m (the same area as a 19m diameter pool). The distance to the centre
of this square is 8.5m. In this case HEGADAS is predicting a distance to LFL
from the bund edge of about 30m. (The UFL is not plotted as it is within the
bund area).
In summary, because we are trying to use the results of a model close to its
source, all we can say is that the sterilisation zone is likely to be a few metres
from the bund edge, but might extend to 20m from the bund edge (as a distance
to LFL of 40m was predicted from the centre of a 3.8m diameter source, and the
distance from the centre of the real bund to its edge is 20m).
Now inspect the Log:Log plot from model the evaporation model which shows
the transient evaporation rate. If we had to average the release rate of the first
few days we might choose a value of about 20 kg/s. Go back to your input
screen and change the release rate from 1.4kg/s to 20kg/s. Use the correct bund
diameter at 44.7m.

Note this time the code will run with this diameter, and the output will look
something like this:
The UFL boundary is still within the bund area, but the LFL is predicted to occur
about 110m from the bund centre. On the basis of some approximations, using
the continuous model would lead you to the conclusion that the maximum
distance to the LFL boundary immediately following a the spill would be about
110m from the bund centre.
We are now going to repeat the above but this time using more exact transient
modelling. In CIRRUS go back to the model menu, select Dispersion and
select model D2 again. Enter your dataset name and description. This time
select Transient from the option dialog box, and select Butane and the material
of interest.

7.2.3.1 A Transient Source Term
Use the same Atmospheric/Ground conditions as the previous run. Put in a

maximum time of interest of 12 minutes and a duration of the release as 10
minutes. CIRRUS will default the time intervals for the transient release by
dividing the lesser of the time of interest or the duration of the release into six
intervals.
Note: CIRRUS will always split the release duration of interest into six linear
timesteps. It is possible to overwrite these timesteps to better match a transient
release. Do not however choose the initial couple of timesteps to be very close
together as this can create instabilities in the underlying code which can crash
CIRRUS or your PC. If you have a version of CIRRUS with the option to
choose a default log timestep you should be cautious in using this for the same
reason.
Change the first interval to 30 second, which will then catch the maximum
release rate from our evaporation. You can read the following rates from the
evaporation plot:
Time (seconds) Release Rate (kg/s)
30 34
200 24
300 23
400 22
500 21
600 20

Input these times, and change the source diameter to 44.7m. Your input screen
should then look like this:
Now run the model, you should see the following result.
Reading the results of a transient run will take some getting used to, but as you
become more familiar with the format you will find that they are self
explanatory.
Firstly note that the time of interest which you input, 12 minutes, has been split
into six linear timesteps; 120, 240, 360, 480, 600, and 720 seconds. In this case
there are no values for the 720 second timestep as the concentration is below the
lower value input. The plot shows the centreline concentration of the gas cloud
at each of the other timesteps.

Note: CIRRUS will always split the time of interest into six linear timesteps and
calculate the results of the transient release for each of these timesteps.
At the 2%vol:vol concentration level you will see that the plume reaches a
distance of about 175m at the 120s timestep, which reduces to about 140m at the
240s timestep and reduces slowly thereafter for the remainder of the 12 minutes.
If you now go to GRAPH and select the horizontal contour - lower concentration
option, you will see the footprint of these clouds at the LFL which confirms this
analysis.
This plot is shown below.
In this case, because the initial release rate was 34kg initially and then reduced,
the initial release dominates the maximum distance to the LFL concentration.
The dispersion cloud moves back towards the source as the release rate from the
pool diminishes.
Compare this result of 175m to LFL with our approximation using an average
source term as input to the continuous version of this model of 110m. - Our
approximation wasnt too good!
Note: In transient modelling it is possible to miss the maximum distance because this
may fall between two timesteps. It is prudent to rerun the model with a different
time of interest, which will result in different timesteps, to check that the
maximum has been observed.

Interestingly the above model used a release duration of 10 minutes and a time of
interest of 12 minutes. Therefore the duration in which the cloud dispersed in
the wind while not being fed by a source term was only two minutes. If a time
duration significantly longer than this was chosen, then at the first timestep the
plume could have already dispersed to a concentration lower than input as the
concentration of interest, and no plot will be produced. To demonstrate this
change the time of interest in the example to 30 minutes. Run the model and
notice that the LFL concentration (1.8%vol:vol) disappears between 600s (10
minutes) and 900s (15 minutes). This indicates that in less than five minutes
after the release stops all the cloud has dispersed to a concentration of less than
1.8% vol:vol.
7.2.3.2 An Instantaneous Source Term
Now let us look at different scenario where the release occurs instantaneously.
By using CIRRUS we are going to assume that this release occurs in 1 minute.
Go back to the model menu, select Dispersion followed by model D2. Input a
dataset name and description. At the dialog box select Instantaneous as the
release option and then select Propane as the material of interest.
For this example you should use similar Ground/Atmospheric conditions to that
of the last example, but use a release duration of 1 minute and a release rate of
70kg/s. Use the default source diameter. Assume that the time of interest is 3
minutes. You should have an input screen which looks like this:

Run the model and you should get a results screen which looks like this:
This plot show the centreline concentrations of the plume. From this graph you
could anticipate that after 2.5 minutes (150s), the LFL contour (1.8%vol:vol)
begins 275m away from the source centre and extends to about 430m from the
source centre. We can rationalise this result due to the fact that the release we
have input stops after 1 minute. Thereafter the plume will be blown downwind
to such an extent that at the source, the concentration of propane in the air is less
than LFL. Away from but downwind of the source, the plume concentration is
above LFL.
Now go to the GRAPH menu and select the plot which shows horizontal
contours - lower concentration to confirm this. From this plot you can track the
LFL contour as it moves downwind.

This plot is shown below.
Out of interest now go back to your input screen and change the time of interest
to 30 minutes. Notice that the plume disperses to below the LFL contour before
the first calculated timestep of 5 minutes.
We have now dealt with Heavy Gas models. The next type of dispersion model
which we are going to discuss comes under the category of Momentum Release
models.

7.3 DISPERSION MODEL - D1 DISPERSION FROM A MOMENTUM

RELEASE
You will remember that one of the criteria for using a Heavy Gas model was
that the release should have zero momentum (for practical purposes this means
low momentum). We are now going to look at the category of model which is
used to calculate dispersion of a release which has some momentum associated
with it. In CIRRUS we have designated this model as D1. It is used, for
example, to model the dispersion of gas from a pressurised vessel or pipeline
where the gas is emitted at high velocity. A significant feature of this type of
model is the ability to allow for the additional entrainment of air due to the
velocity of the release.
Note: In CIRRUS version 4 we have chosen to use the model called PLUME which
was developed by Shell oil company.
7.3.1 Dispersion of a vapour which behaves as an Ideal Gas
As our first example in using model D1 we are going to predict the dispersion of
a high velocity methane release. We are going to use the release scenario
analysed in section 3.4.1; a 100mm hole in the side of a gas vessel which is 4m
in diameter and 20m in length, containing Methane gas at 50bara. We are going
to put the hole in the centre of the 4m diameter end; i.e. 2m from the bottom of
the vessel. We are also going to assume that the vessel is standing on a plinth
1m from the ground, therefore the total height of the hole from the ground is 3m.
The gas will exit horizontally. The rate of the release which was calculated using
model S3 dropped from an initial rate of 65kg/s to virtually zero after 380
seconds. For convenience we have reproduced this relationship overleaf:

You will remember from the general introduction to dispersion modelling that
momentum dispersion models only take a steady state continuous release rate as
an input. If you look at the above relationship you will conclude that the above
release is anything but steady state!
You will also remember that, as a rule of thumb, we said that a continuous
release was one of 10 minute duration or longer. The above release is all over in
about 10 minutes! Therefore we now have the problem of fitting a release which
in reality is a transient non-steady state release into a model which is designed to
accept only a continuous steady state release as input.
Note: This is a very common situation in modelling dispersion from refinery release
scenarios.
The release duration is about 10 minutes and in this time the rate varies from
65kg/s to virtually zero. A pragmatic approach to yield a best estimate of the
dispersion of the gas from this relatively smooth transient release is to take the
analytical average of the release rate, 33kg/s, and input this as a continuous
release rate in model D1.
In CIRRUS go to the model menu and select Dispersion followed by model D1.

You should get an input screen which looks like this:
Note: The various options have the following implications:
Released Vapour: This option should be used to model the dispersion of a

material when this material is the only component of the released vapour.
Contaminant in Released Vapour: This option should be used to model the

dispersion of a material when this material is the one component of the released
vapour.
Concentration or Lethality: Similar to the Heavy Gas Model, the concentration

option will plot the dispersion boundary to a %vol/vol of interest (i.e. the
percentage of material in air on a volume basis). In CIRRUS the lethality option
will plot the %lethality of an average population for a given exposure time in
accord with Probit equations published by the Dutch Government in their
document CPR16E Methods for the determination of possible damage (often
referred to as the TNO Green Book).
For this example select the Concentration option and the released vapour
option. Continue and select Methane as the material of interest. You should now
see the input screen for the model. If you remember the gas temperature at the
orifice was calculated in model S3 as -19C, therefore we will use this as the
upstream temperature for our release. Therefore input a release rate of 33kg/s,
a gas temperature of -19C, and orifice diameter of 100mm (0.1m). You will
now see that the vessel pressure has been given a default. This is less than the
initial vessel pressure as it is associated with our artificial continuous release
rate, which is less than the initial release rate. Leave the default pressure as it is.
Note: You can overwrite the default vessel pressure, but having done this you cannot
then go back to the default calculation without beginning again. What you are
actually doing is overwriting an Excel formulae in this cell. If you select your
own pressure then CIRRUS checks to see if the orifice diameter, vessel
temperature and pressure are consistent with the input flowrate by calculating the
implied discharge coefficient. Outside certain limits CIRRUS will not let you
run with unrealistic input data.

Now input the remaining parameters; use a angle to horizontal of 0, a height

from grade of 3m, a roughness parameter of 0.01, a humidity of 50%, an ambient
air temperature of 15C, a wind velocity of 5m/s and a stability category of D.
Your input screen should look like this:

Now run the model and you will get the following result:
This is the centreline concentration of the plume. The 5%vol:vol (LFL of

Methane) is predicted to occur at about 58m. For a more graphic representation
of the plume shape select GRAPH from the menu and then vertical cross-
section contours.
You should now have plume profile with respect to the ground as shown below.
This example shows how to model a release of a material which will behave as
an ideal gas. Our next example will show you how to model the release of a
material which is processed or stored at conditions close to saturated vapour
pressure conditions.

7.3.2 Dispersion of a vapour which is close to its saturated vapour conditions
You are able to model the release of vapour close to its saturation conditions
using the two-phase option of model D1. We can therefore continue with the
release scenario modelled in section 3.5.1, which considered the release rate
from the vapour space of a Horton sphere through an orifice with an equivalent
diameter of 250mm.
Note: You should also use model D1 to calculate the dispersion of a release of vapour
from a relief valve. In the above case of a Horton sphere containing Propane
liquefied under pressure, you would choose the two-phase option. For relief
valves which are connected to a stand-pipe, the diameter of the stand-pipe end
should be used as the vent diameter (as should the pressure at the stand-pipe
end). For relief valves which vent directly to atmosphere from the seal faces you
will have to estimate the equivalent area of the orifice during relief conditions.
Remember however that the model will treat this as a unidirectional release from
a vent of this diameter.
We will use the isothermal (uninsulated vessel) release rate. Note however that
dispersion can be treated as a continuous release if it occurrence at the same rate
for about ten minutes. The first ten minutes of an adiabatic (insulated vessel)
release is not very different form the isothermal case. Therefore this assumption
is not too pessimistic.
For convenience we have repeated the results of this modelling below:
In this case the assumption of a continuous release is a good approximation to

the actual release rate. Therefore to model the dispersion of this release we can

use a scenario based on a 250mm hole in the top of a Horton sphere (or relief
from a relief valve with a 250mm stand-pipe diameter). If the diameter of the
Sphere is 20m and sits on a plinth 5m high, then the release will occur at a height
of 25m. Assume that the release I directed upwards (90 to the horizontal).
Allow the ambient temperature to be 15C with a 5m/s windspeed and D
stability. Input a ground roughness of 0.01, which represents the typical
roughness for 250m upwind of the sphere and 20m downwind of the sphere.
Note: In section 4.1.2.2 we gave an explanation of why the ground roughness upwind
of an elevated release should be considered.
Go to the model menu of CIRRUS and select Dispersion, choose model D1

and then input your data, the screen should look like this
Now run the model.
CIRRUS has detected that this is actually a two phase release because the input
pressure is greater than the saturated vapour pressure of the material. Hit OK to
the intermediate dialog box.

You should see a screen which looks like this:
The upstream pressure has now been reduced to the vapour pressure of the
material.
Note: If you are interested in viewing the actual parameters which CIRRUS is using as
input to the PLUME code you can do so by selecting the Expert user level
when you first start CIRRUS. In this case you will notice that the temperature of
the release has been reduced to -42C. Most of the release would be liquid
aerosol, and consequently the appropriate vapour molecular weight has been
calculated as 119. Due to this aerosol the plume will tend to slump. The exit
velocity has been calculated as 307m/s and the exit diameter expanded slightly to
260mm. These modifications are necessary to use PLUME, which treats the
dispersion as if it were an ideal gas. Refer to the CIRRUS validation manual for
the mathematical formulae and logic behind these conversions.

The vertical contour graph for this case should look like this:
The plume remains elevated but you can see evidence of slumping due to the
higher density associated with the aerosol.
7.3.3 Dispersion of a contaminant
As the final example of using model D1 we will pretend that we have

contaminated the Propane in our Horton sphere so that the vapour now contains
10%vol:vol if hydrogen sulphide. We are using this as an example of not only
the use of the contaminant option in D1, but also as an example of the
automatic transition of the PLUME momentum model into the HEGADAS
heavy gas model when the vapour plume comes down to ground due to the
negative buoyancy of the released vapour. (The Propane-Hydrogen Sulphide
mixture has a molecular weight greater than that of air).
Go to the CIRRUS model menu and then select Dispersion and D1. At the
option dialog box select the options contaminant in release vapour and
concentration output. Select Propane as the carrier gas material and Hydrogen
Sulphide as the material.
Note: If in other scenarios the carrier gas is not a single component material, use a
surrogate single component with the same molecular weight as the vapour which
will disperse away from the release source.
Input the same release conditions as for the previous run with Propane as a pure
vapour, but this time include 10% of contaminant. Also, you will have to input

concentrations of interest. For our example we are interested in H2S

concentrations of 100ppm (0.01%vol:vol) and 10 ppm (0.001%vol:vol). You
should have an input screen which looks like this:
At the intermediate dialog box select OK to continue as a two phase model.
You should see a screen like this:
Now run the model.

You should get the following centreline concentration versus distance plot:
Note a break in this plot occurs at about 600m. This is where the PLUME
momentum code transfers to the HEGADAS heavy gas code. Now choose
horizontal cross-section contour from the GRAPH menu. You should see plot
like this:
You can see from this plot that from this release an H2S concentration at ground
level is predicted to disperse downwind to a distance of 7km with a plume width
of about 1km.

For an alternative view go to the AXES command and choose floating. For
more information about the cloud geometry take a look at the other plots under
the GRAPH menu.
Note: When the PLUME model transfers to the HEGADAS heavy gas model or the
PGPLUME gaussian model (the latter occurs when the plume loses all
momentum but is still buoyant), the transition can fall over. CIRRUS normally
plots the results that have been calculated up to this point. You can tell that the
run has not been completed satisfactorily by looking at the centreline
concentration graph and noting that the minimum concentration is still greater
than the minimum concentration of interest which you asked for. In addition a
plume contour will seem to stop in mid air and not close to a point. If this
happens to you when modelling scenarios unfortunately there is no rule of thumb
to correct the situation. All we can suggest is that you change some input
parameters of interest; for example change the stability category or windspeed.
(The fault lies with the underlying code rather than with the CIRRUS front end).
This brings us to the end of our examples of the use of the momentum model
D1. We are now going to discuss model D3 and D4, which are known as
gaussian or passive models. They are used to specifically model dispersion of
vapour which when released into air is buoyant or neutrally buoyant (i.e. less
dense or the same density as air), and where the release has little or no
momentum. A typical application of this type of model would be the dispersion
of pollutants from a chimney. In use they overlap model D1, but are preferable
when looking at dispersion to low concentrations (100 ppm or less) and have a
different form of output.
7.3.4 Using model D1 with Multi-Component Materials
To use model D1 with multi-component materials, firstly add the multi-

component material to the database as described in section 2.2.5. When using
model D1 you can select then this multi-component material from the material
dialog box. The material will appear at the bottom of the list under the list of
single component materials.

PART 3. DETAILED SCENARIO MODELLING SECTION 7.4. MODEL D3 & D4
7.4 DISPERSION MODELS - D3 & D4 DISPERSION OF A VAPOUR

WHICH IS BUOYANT OR NEUTRALLY BUOYANT WHEN RELEASED
INTO AIR.
You will remember that the criteria for using a Heavy Gas model was that the
material released should be heavier than air, have low momentum and occur at
ground level. Model D3, and a variant of this model D4, are a corollary to the
heavy gas model. They take as input elevated or ground level releases of
materials which are buoyant or neutrally buoyant and where the release has little
or no momentum (i.e. low velocity releases). In addition it is also applicable to
use model D3 for an elevated release of material which is heavier than air where
the release has little or no momentum.
Note: You should use the momentum model D1 for all releases which occur at
moderate or high velocities, irrespective of whether the release is elevated or
occurs at ground level or whether the material is more dense or less dense than
air. In this context a moderate velocity would be greater than about 5 m/s.
These types are often called gaussian or passive models. They are particularly
applicable to the emission of hot vapours where plume rise occurs, but are
limited in that the model is accurate only in the far field, i.e. greater than about
100 release orifice diameters downwind of the source point.
In CIRRUS we have described two types of source category. One category we

have described as a stack source. This is intended to be used for elevated
releases and is appropriate for the modelling pollutant concentrations at ground
level some way away from for example a chimney or turbine exhaust. The other
category we have described as an area source. This is intended to be used for
releases randomly dispersed over an area (normally at ground level but which
can be elevated) and is appropriate for the modelling of fugitive emissions from
a refinery complex or other industrial site.
Note: In CIRRUS version 5 we have chosen to use the model called ISC (Industrial
Source Complex) which is released by the US Environmental Protection Agency
(EPA). There are two versions of the code; the short term average (ISCST) is
model D3 within CIRRUS and the long term average (ISCLT) is model D4
within CIRRUS

Model D3 accepts input which is very similar to that of model D1 and D2. In
CIRRUS you define release rate, wind speed and atmospheric stability
conditions and the code calculates the centreline concentration downwind of the
release point. In this model only the centreline concentration is calculated at
ground level and/or another height of interest.
Model D4 is somewhat different in that it calculates an annual average pollutant

concentration for the area surrounding the release point. To do this it accesses a
database which contains the annual average wind speed and atmospheric
stability conditions at the site of interest. To use D4 in CIRRUS this site specific
data has to be input by the CIRRUS helpdesk for the site location.
7.4.1 Emissions From Stack Sources
As a first example of using model D3 we are going to calculate the downwind

concentration of H2S in air at ground level 1km away from turbine exhaust. The
turbine exhaust which emit gas with the same molecular weight as air, 29, but on
occasion can contain 10% by volume of H2S which has a molecular weight of
34. The exhaust emission total mass is 1kg/s, the diameter of the stack is 1m and
its height is 10m. For this scenario we will choose an ambient temperature of
10C, and a windspeed of 5m/s in stability D atmospheric stability. In this EPA
model you have only two choices of roughness parameter; R (rural) and U
(urban). For our scenario we will assume that the cloud disperses over a rural
environment.

From the CIRRUS menu select Dispersion. Then select model D3 and the
Stack Release option as shown below:
Enter a dataset and description.
Note: Due to the formulation of the model the averaging time used is one hour.
Input the data given previously leaving the height of interest blank, you should
end up with the input screen shown below:
Note: In CIRRUS the height of interest refers to calculation of results for a specific
height in additional to that at ground level. This should be left blank if you are
interested in only ground level concentration to avoid over complication of the
output graphs.

The results should look like this.
From this plot you can see that the pollutant concentration at ground level at a
distance of 1km from the source is about 1ppm. This plot is equivalent to a plot
of centreline concentration similar to those given by models D1 and D2. In
model D3 we do not have the capability of graphing the plume shape.
7.4.2 Emissions From Area Sources
In this example we are going to calculate the concentration of fugitive

emissions from an imaginary refinery plant. In this example we will assume that
the plant has a total area of 100mx100m (1x104 m2) from which 1 tonne of
Propane is released as a fugitive emission per day. This equates to a release rate
of 1.16x10-6 kg/m2/s. Use an ambient temperature of 10C, a windspeed of 5m/s
and D atmospheric stability conditions. Leave the default roughness condition as
R (rural).
Select Dispersion from the CIRRUS menu and then select model D3 and the
area source strip-off. Input a dataset name and description. Input the above
conditions.

You should have an input screen like the one shown below:
Run the model to view the maximum ground level concentration as a function of
distance downwind of the site. Notice that the Propane concentration 1 km from
the site is about 0.2ppm. This graph is shown below.
Note: Although in our example we have used the example of a large area source to
calculate the concentrations of fugitive emissions, this model has many other
applications. For example it could be applied to calculate the dispersion of
pollutant emitted from the top of a tank containing hydrocarbon, or of gas
bubbling up to the sea surface from a rupture of an underwater pipeline, etc.

7.4.3 Using Model D4 to Estimate the Annual Average Pollutant Concentration

Surrounding a Site
Model D4 is a very similar model to D3 with the exception that the annual
average wind direction, windspeed and stability categories are input from a
CIRRUS helpline defined dataset. The output is a density plot showing the
average concentration of pollutants around the site area. We can also look at this
same annual average along a single bearing from the site.
Note: To input the annual average meteorological data for your site please contact the
CIRRUS helpline.
As an example we are going to use the meteorological conditions appropriate for

the Sullom Voe site; a dataset which is already contained within the CIRRUS
package.
Let us imagine that we are interested in calculating the annual average

concentration of NO2 from three turbine exhausts located 300m apart. The
average molecular weight of the exhaust gas is that of air, MW=29, but all three
turbines emit NO2 (MW=46) which makes up 10% by volume of the exhaust gas
stream. The total mass release rate from each stack is 1kg/s at a temperature of
70C. We wish to know if any of the surrounding area will suffer annual
concentrations of NO2 of greater than 1ppm at ground level.
The exhausts are lined up at 90 to the wind in a row 30m apart in an East west
direction. The model takes the north direction as the x axis with co-ordinates in
metres. Therefore for this example they will have the same x co-ordinate of 0m,
y co-ordinates of -30m, 0m, +30m and z (height) co-ordinates of 10m.

Select Dispersion from the model menu and then select model D4 and the
Stack strip-off. Select the Annual weather window and the planar plot
option. Change the number of release sources to 3.
Note: As the meteorological information is split on a month by month basis, we can

look at the annual average in each month if need be. This is useful where the
predominant wind direction or stability class changes markedly for one or more
seasons.
Input the above data and your screen should look like this:
Note: The output is displayed as a density plot using the dimensions shown in the
bottom part of the screen. If you want to view a surrounding area which is not
20km by 20km (-10,000m to +10,000m in the x and y directions) then change
these values on the input form. If you want a greater resolution then you should

change the number of grid sections analysed to greater than the default 41. Note
however that this will increase the computing time requirements.
Run the model and view the output. Note that the maximum annual average
concentration calculated is 0.3 ppm - well below our hurdle value of 1ppm.
(This value is shown in the top left hand side of the screen). You can change the
plot from zero to maximum value to zero to median value in order to obtain a
better resolution of the far field contours; this has been done below. In addition
you can select a 3D view. Although this is presentational appealing, in practice
this form of plot is more difficult to read.
Maximum Average
Concentration is 3.04E-01 Theoretical NO2 concentrations Concentration of Contaminan
ppm Annual Average Air at Ground Level
(ppm)
Location : Sullom Voe
10000
8000 > 0.01 ppm
6000 9.00E-03-1.00E-02
4000 8.00E-03-9.00E-03
2000 7.00E-03-8.00E-03
0 6.00E-03-7.00E-03
-2000 5.00E-03-6.00E-03
-4000 4.00E-03-5.00E-03
C -6000 3.00E-03-4.00E-03
o -8000 2.00E-03-3.00E-03
n -10000 1.00E-03-2.00E-03
-10000
-7500
-5000
-2500
10000
2500
5000
7500
0
c
e 0.00E+00-1.00E-03
Distances in metres N
Note: In CIRRUS there are two plot types which range from zero to maximum value
and from zero to median value. This is because in Excel we cannot control the
band selection - which is a linear selection. For our purposes we would like to
use a Log representation as there is likely to be an exponential fall off of
concentration with distance. To overcome this problem we plot firstly the entire
range calculated. This tends to show the high near-field concentrations with the
great majority of the far-field area being encompassed by a single low value
band. Secondly we plot the zero to median values. This tends to show the far
field concentrations with the near field encompassed by a single high value band.
For your plots you might want to experiment with the grid size to best fit the
concentrations of interest.

Although we have demonstrated an example using closely spaced turbine

exhausts, model D4 is again quite versatile, and can be used to model pollutants
from chimney stacks over an industrial conurbation, simultaneous fugitive
emissions from a number of different area sources on a site etc.
7.4.4 Using Model D3 and D4 with Multi-Component Materials
Models D3 and D4 can be used with multi-component materials if you know the
average molecular weight of the escaping vapour. This is entered directly in the
input screen for the model.
To calculate the molecular weight of a multi-component material, firstly add the multi-
component material to the database as described in section 2.2.5. Now use
model S9 to flash the material as described in section 6.1.6. The average
molecular weight of the multi-component material and the suggested molecular
weight of a vapour/aerosol mixture for use in dispersion calculations (if
appropriate) are shown on the results page.


PART 3. DETAILED SCENARIO MODELLING SECTION 7.5. SUMMARY OF VAPOUR DISPERSION MODELS
7.5 A SUMMARY OF VAPOUR DISPERSION MODELS
From the previous discussion you are now probably well aware that dispersion
models have originated as a set of solutions to particular problems, rather than as
a comprehensive set of tools for modelling accidental releases from refineries.
However we have attempted here to give guidance on how to use the available
types of model in various scenarios. Below you will see a table which can act as
a first reference to enable you to select the most appropriate model for any
particular scenario.
Elevated Source Ground Level Source

Buoyant or Buoyant Gas Dense Gas
Dense Gas
Low Velocity D1 - Near Field D1-Near Field D2

Release D3 - Far Field D3 - Far Field
Moderate or High D1 D1 D1
Velocity Release
We have used here examples based on the particular models in the CIRRUS
package. Many other packages have a similar selection of the generic types of
model. The same basic principles discussed here are equally applicable to these
packages.

236
7.6 DISPERSION MODEL D5 DISPERSION OF A PARTICULATE

MATERIAL
Model D5 is designed to model the airborne dispersion and ground deposition of

particulate matter. It will take as input either ground level or elevated sources.
The model assumes that the particulate matter will be advected downwind in a
gaussian manner similar to the D3 and D4 models. As they are carried
downwind the particulates will settle and ultimately may deposit on the ground.
The particle distribution in the plume is dependent on the settling velocity and
the quantity deposited is dependent on the deposition velocity.
Note: The rate of settling is dependent on the type and density of the material and its particle size.
The quantity deposited on the ground In CIRRUS we have used a settling velocity calculated
for each of the particle size categories used in the model calculated from standard equations, as
settling velocities for different materials are often unknown. If the settling velocity calculated
is grater than 1m/s, then ballistic equations have been used to calculate the settling velocity.
Deposition velocity describes the rate at which material will deposit onto the ground, which is a
function of windspeed and particle size, and again we have used values calculated from
standard equations in CIRRUS.
As input to the model, you have to choose the maximum and minimum limits of
the particle sizes you wish to model in the release. If there is a single particle
size then the maximum limit should equal the minimum limit. It is assumed that
the particle distribution within the maximum and minimum limits you have input
will be gaussian.
Note: During execution the model will divide your size range into various size categories and calculate
the proportion of the release in each category. There are 11 size categories in the model defined
in microns; 0.01-0.03, 0.03-0.10, 0.10-0.30, 0.30-1.0, 1.0-7.0, 7.0-10, 10-30, 30-100, 100-300,
300-1000, 1000-3000. For settling velocity and deposition velocity purposes the mean particle
size in each category is chosen. Your maximum and minimum sizes input must span at least one
of these ranges excepting when defining a single particle size (when this single value is used to
define settling and deposition velocities).
7.6.1 Example of usng model D5 A Particulate Release from a Stack Source

In this scenario we are going to imagine that we are trying to model an
instantaneous release of Kaolin from a 30m stack, the Kaolin is released as a
particulate in the size range between 500 microns and 1 micron. In this case all
the particulate release is Kaolin.
From the CIRRUS menu select Dispersion and then model D5. Input a dataset
name and description.

You will then see the dialog box shown below:
Input the data as shown and run the model. You will note that for this model
CIRRUS only allows you to select either stability D windspeed 5m/s or
stability F windspeed 2m/s weather conditions. You should have an output that
looks like this:
This first graph shows the concentration of all particles that are still airborne at
ground level. You will see that the maximum concentration of 300kg/m3 occurs

at 300m downwind from the source. The next graph shows the concentration of
respirable particles; i.e. those particles which can be drawn into the lungs.
If we now look at the mass deposition graph, as shown below, we can see that
the highest concentration of particulate matter which is deposited on the ground
occurs at a distance of 400m from the source and has a value of 4.3 x 10-6 kg/m2.
The remaining graph shows the cumulative mass deposited as a function of

distance.
The next category of model we are going to discuss are those which calculate the
thermal radiation from fires.

240
PART 3. DETAILED SCENARIO MODELLING SECTION 8. FIRE THERMAL RADIATION MODELS
8. POOL FIRE, TORCH FIRE AND FIREBALL MODELS
Heat is transferred away from a fire in three ways; by conduction, convection and
radiation. In the context of fires from accidental releases of hydrocarbon on
refineries:
Conduction is normally associated with direct impingement of flame onto

structures, where the structure heats up due to conduction of heat into the
material.
Convection occurs from the heating of air close to the fire and the mixing in of
hot combustion products. This hot gas then rises resulting in the convection of
heat away from the fire.
Thermal radiation is the transfer of energy away from a fire in the form of
electromagnetic radiation. This is caused by the excitation of the molecules in
the combusting material. In this way thermal radiation acts in a very similar
manner to light.
The models which have been developed to estimate the effects of fires on
structures and human beings deal almost exclusively with the estimation of the
thermal radiation component of the fire. In consequence modelling we generally
assume that a human being who suffers direct impingement by flame, or who is
located in an area affected by very hot convected air or combustion gases will be
either fatally or seriously injured.
Note: When interpreting the results of fire radiation models it is important to note that only the thermal
radiation component of the heat release is being modelled. For example, at a distance above a
pool fire the model might predict that the thermal radiation would be insufficient to injure people.
This may be true, however you cannot conclude that this would be a safe area as in this case
convection of hot gases from the fire would add to the heat burden above the fire.
In consequence modelling there are typically three types of generic model:
The first of these is called a pool fire model. This model is used to assess
fires from pools of hydrocarbons lying on the ground or in a bund, berm or
dike. They are also used to model crude oil storage tank fires or other
hydrocarbon tank fires. These types of model estimate the flame height and
surface emissive flux of the fire (the quantity of heat radiated from the
surface of the fire) to characterise the fire, and then use a view factor
calculation and an atmospheric attenuation algorithm in order to estimate the
thermal radiation burden at a specific point some distance from the fire.
The second of these generic models is a torch fire model. These are used to
model the radiation from flares or the ignited release of a jet of flammable
vapour or liquid ejected from a hole in a pressurised vessel or pipeline. They
work in a similar way to the pool fire models by first estimating the length of
the flame and the proportion of the combustion heat which is radiated, and
then using a view factor calculation and an atmospheric attenuation

PART 3. DETAILED SCENARIO MODELLING SECTION 8. FIRE THERMAL RADIATION MODELS
algorithm to estimate the thermal radiation burden at a specific point some

distance from the torch flame. The same considerations concerning
convective heat flux apply to these models.
The third of the generic models is a fireball model. These types of model
were designed specifically to estimate the thermal radiation effects of the
fireball which occurs on BLEVE of a pressurised Propane or Butane vessel.
As you will read later, this model can be adapted to estimate the radiation
from a turbulent fire which might result if an ignited jet strikes the ground or
other object close to the release point.
Note: BLEVE stands for Boiling Liquid Expanding Vapour Explosion. It is a misnomer as this
phenomenon is not an explosion in the scientific sense of the word. A BLEVE can occur when a
fire impinges on a vessel containing a material which is a gas at atmospheric pressure and
temperature, but is a liquid at the process or storage conditions. For example Propane is often
stored as a liquefied gas at ambient temperature and elevated pressure. When the flame impinges
on the unwetted surface above the level of the liquefied gas, the metal of the vessel wall heats up.
Above about 350C the yield strength of steel diminishes rapidly and at some point the vessel
will fail. It can fail at pressures below the relief valve set pressure. Therefore a BLEVE can
occur even if the relief valve is operating correctly. Note also that an ignited torch flame from
the relief valve itself may add to the thermal burden on the top of the tank.
When the vessel fails the contents expand and burn rapidly. A fireball is formed which lifts off
the ground. The radiation from the fireball is of very high intensity and very damaging to human
beings, although the time taken to burn the vessel contents is quite short. The models calculate
the fireball duration and its intensity, and estimate the % likelihood of a person at a specific
distance from the fireball being killed by the resulting thermal radiation.
We are now going to look at example scenarios using the pool fire, torch fire and
fireball models. In CIRRUS these are models F1, F2 and F3 respectively.

PART 3. DETAILED SCENARIO MODELLING SECTION 8.1. MODEL F1
8.1 FIRE MODEL F1 - POOL FIRE RADIATION
To begin our discussion of fire radiation modelling we are going to look at the
capabilities and application of the pool fire model. In CIRRUS this is model F1.
8.1.1 The Assumptions Inherent in Model F1
The model can be used to predict the thermal radiation at any point surrounding
the fire. It works by assuming that the fire can be modelled as a cylinder which
emits radiation from the sides of the cylinder. The cylinder can be tilted by the
wind and the flame dragged beyond the original pool diameter by the wind.
The radiation flux assumed for this type of model is material dependent and
based on experimental observations. Unfortunately the agreement in the
measured data between different sets of experiments is poor. In addition only a
few materials have been the subject of extensive experimentation.
Note: CIRRUS makes use the empirical observations taken in the China Lake set of experiments
commissioned by the US coast guard and cross-checked against other sets of experiments. For
validation documentation contact the CIRRUS Helpline.
For materials which have not been subject to extensive testing CIRRUS contains default
empirical constants. You cannot use a user defined multi-component mixture with this model for
this reason. To model multi-component mixtures using CIRRUS you must choose a like
surrogate from the database; e.g. stabilised crude oil, gasoline, Propane etc.
As an example of using the pool fire radiation model we are going to look at two
scenarios. Firstly an ignited spill of refrigerated Propane onto water, and
secondly a crude oil storage tank full surface fire.

8.1.2 Example of using Model F1 - A Spill of Refrigerated Propane onto Water.
In this scenario we are going to imagine a hose burst incident while loading a
refrigerated Propane tanker at a refinery docks.
Note: If the scenario involved modelling a fire which occurred after a hose burst incident while loading
pressurised Propane, you should first use model S9 to determine the proportion of the release
which would flash to form a vapour, and only spill the remaining liquid, which would now be just
below its boiling point of -42C, onto the sea to form a pool. (You can refer back to section 3.1
for a reminder on how to use model S9).
We are going to assume that the loading rate is 200kg/s, all of which is spilt onto
the sea. As the maximum consequence event we are also going to assume that
shutdown does not occur for some considerable time thereafter.
From the CIRRUS menu select Fire and then model F1. Input a dataset name
and description. You will then see the dialog box shown overleaf:
In this case select the options on water and unconfined. At the input screen
input the source of the release as 200kg/s, the release durations as 1800s, tab
down through the default values and then input a relative humidity of 80%, an
ambient temperature of 15C and a wind velocity of 5m/s. Input the thermal
fluxes of interest of 2kW/m2, 6kW/m2 and 12kW/m2. Input plane height of
interest of 0m, 35m and 70m.

Note: In CIRRUS, if you want to model a pool fire due to immediate ignition of an instantaneous
release of material, put the total quantity of material released in the rate of liquid release input
box and 0s in the duration of release input box. For further details highlight either of these two
boxes and select on-line help (F1).
There are a number of important parameters on this screen which are explained
in more detail overleaf.
8.1.2.1 Sooty Flame:
The model treats the flame as a cylinder which can tilt in the wind. The height
of the flame is calculated on the basis of experimental correlations, as is the
average surface emissive flux. This latter parameter describes the amount of
thermal radiation per unit area emitted from the flame surface, which when
combined with the target view factor and attenuation by absorption in the
atmosphere gives the thermal radiation at the target position. The thermal
radiation at the target is therefore directly proportional to the thermal radiation
emitted by the flame.
The thermal radiation emitted by the flame depends on parameters such as the
flame diameter and how much soot is formed in flame. Soot absorbs thermal
radiation and therefore reduces the surface emissive flux of the flame. In
general, burning of materials with a molecular weight less than about 100 will
burn with a non-sooty flame. (Although these flames often appear to be sooty,
the soot density is often not sufficient to significantly affect the emissive flux).

CIRRUS will default the non-sooty flame option where it is believed to be

appropriate. Where the sooty flame option may be appropriate no default is
given.
As a rule of thumb, materials with a molecular weight above 100 will burn close
to their base with a non-sooty flame, with increasing soot density above this area.
For these materials the following rules are appropriate:
For a target in-line with the base of the flame, the view from the target is
predominantly of the lower part of the flame and the non-sooty flame option
should be chosen to model this case.
For a target at a higher level or at a distance (say 5 diameters) away from the
flame, the view from the target will be predominantly of the sooty portion of the
flame, and the sooty flame option should be chosen.

8.1.2.2 Thermal Flux of Interest:
This parameter defines the flux contour of interest. Damage to either people or
plant depends upon the duration of the exposure. For action/escape levels API
521 gives the following guidance:
Radiation Level (excluding solar radiation) API 521 Design Basis

Description
Btu /ft2h kW /m2
Emergency actions lasting up
2000 6.31 to 1 minute by personnel
without shielding but with
appropriate clothing
Exposure must be limited to a

3000 9.46 few seconds, sufficient for
escape only
Refer to section 4.6 for a more detailed explanation of damage criteria.
8.1.2.3 Plane Height of Interest:
In CIRRUS the flux contours for a horizontal plane in space are plotted for the
input plane heights of interest. In the example we are currently running the
contour plot at ground level, at a level of 35m, and at a level of 70m have been
chosen. You will see later that the height of the flame predicted in this example
is 70m, therefore the plane height of interest in this case correspond to those at
the bottom of the flame, mid way up the flame and at the top of the flame.
Now continue with the modelling and run the case. The first set of results shows
the flux of interest at the various plane heights. As this is a little difficult to
interpret until you are familiar with the output you should ignore this graph, and
instead select the flame shape graph from the GRAPH menu.

Having done this you should see the following picture:
You can see that the flame has a predicted height of 70m and has been tilted in
the downwind direction. The calculated pool diameter is about 32m, and the
flame itself has been dragged downwind by a further 15m.
Now select the graph showing the thermal radiation flux contours at the 0m
plane height.
The graph should look like this:
Notice that this ground level contour shows the flame from the pool in the centre
of the plot with the flame being dragged asymmetrically downwind. The
radiation contours reach distances downwind which are further than those
upwind. This last factor is because the view factor of a point on the ground from
a cylinder of flame tilted towards it is greater than that when the cylinder of
flame is tilted at the same angle but away from it.

At the plane which is level with the mid-point of the flame the view factor will
be the same for points located upwind and downwind of the flame. For points
on the plane which is level with the top of the flame the view factor is greater
upwind than downwind. To demonstrate this effect the contour at the 70m level
is very close to that which cuts the of the top of the flame. If you select this
contour from the GRAPH menu you will see the following plot:
Notice that the centre of the tilted cylinder of flame is no longer in the centre of
the plot but is 50m downwind of the centre. Refer back to the vertical cross-
section to corroborate the translation of the flame at the 70m level. Notice also
that the radiation contours extend further upwind at this level than downwind.
This is the view-factor effect described above.
We are now going to model a tank fire scenario.

8.1.3 Example of using Model F1 - A Crude Oil Tank Full Surface Fire
In this scenario we are going to model a full surface fire on a crude oil tank
which is 60m in diameter and filled to a level of 15m. We are interested in
radiation at ground level and whether the fire will impinge on adjacent tankage.
As our targets of interest are close to the base of the flame we are going to
choose the non-sooty option for the model.
From the CIRRUS menu select Fire and then model F1. Input a dataset name
and description, and select on land and confined in dialog box which follows:
Continue, and then on the input screen type in a fire base diameter of 60m, a
height of flame base of 15m, a relative humidity of 80%, an ambient temperature
of 15C and a wind velocity of 5m/s. Input the thermal fluxes of interest of
2kW/m2, 6kW/m2 and 12kW/m2. Input plane height of interest of 0m, 15m and
50m. Your input screen should look like this:

Now run the model and from the GRAPH menu select the flame shape. Your
results should look like those below.
Notice the flame has a height of about 50m above ground level and is tilted by
the wind. The flame base is dragged over the edge of the tank by the wind for a
distance of about 20m.
From GRAPH select the contour plot at ground level (0 metres plane height).
Notice that the downwind approach distance for a person without special
clothing (the 6kW/m2 contour) is predicted to be 140m from the edge of the tank
(170m from the centre of the tank), while upwind the radiation levels are quite
low. Again this effect is due to the difference in view factor described
previously.

At the level of the flame base the predicted radiation contours are more
symmetrical as shown below:
Note: The above calculation assume a steady burning rate which is an average measured in
experiments. The calculations also assume a steady wind speed. In reality tank fires rarely have
a steady burning velocity, nor do winds stay constant. These factors should be borne in mind
when interpreting the results of tank fire models.
Further, the experimental basis of most tank fire models is taken from experiments at ground
level. Relationships for flame drag and wind vortex effects on elevated tank fires are in general
unknown, and are approximated from less appropriate ground level empirical correlations.
The next category of Fire models are those relating to Torch fires and Flares, and
are discussed in the following section.

8.2 FIRE MODEL F2 - TORCH FIRES AND FLARES
Torch fires are the generic name given to the long pencil shaped flame which
results following ignition of an accidental release of a flammable gas or liquid
from a pressurised vessel or pipeline. They have similar characteristics to
refinery flares, and the thermal radiation from both torch fires and flares is
estimated using the same model. In CIRRUS this is model F2.
Torch fires and flares can ignite because the jet of hydrocarbon can entrain air
and burn at its edge, they remain ignited because the burning velocity of the
flame is greater than the velocity of the hydrocarbon jet, in other words the flame
is able to burn back towards the source of the jet. These factors have two
consequences:
Above a certain velocity a jet of flammable hydrocarbon will not sustain

ignition. Quite often the torch fires of high velocity jets do not burn back
completely to the source of the jet but stand-off from the source. This
is because the jet can only sustain ignition when the jet velocity has
diminished.
Liquids of low volatility will not sustain ignition as a torch fire because
the burning velocity of the flame at the edge of the jet is very low; less
than the jet velocity. However a pool of liquid will form which may
sustain ignition.
Note: For practical purposes materials with a vapour pressure less than 1 barg at ambient temperatures
can be assumed not to sustain ignition as a torch fire. To calculate the vapour pressure of a
material at ambient temperatures refer to section 3.1 of this manual.
An accidental high velocity release of gas or a gas containing hydrocarbon liquid

aerosol are classed as gas torch fires for modelling purposes. In general all
refinery flares (unless suffering gross liquid carry-over) should be modelled as a
gas torch fire.
Accidental high velocity releases of flashing liquids, for example liquid

propane from a hole below the liquid level of a pressurised propane storage
vessel at ambient temperature, are classed as a liquid torch fires for modelling
purposes.

The model is used to predict the thermal radiation at any point surrounding the
torch flame. The flame can be elevated, as for example with a refinery flare, and
pointed in any direction in relation to the wind with the exception that the flame
cannot be pointed downwards. High windspeeds bend the flame.
In the gas torch fire model the flame is assumed to be the frustrum of a cone,
and to radiate heat from the entire surface with uniform surface emissive power.
In the liquid torch fire model the flame is assumed to be a long thin cylinder of
constant radius which radiates heat from a uniform distribution of heat sources.
This radiation decays as it passes through the flame due to the flame opacity.
Note: The gas torch fire model in CIRRUS is the Chamberlain model which was developed by the
Shell Oil company. The liquid torch fire model was developed by BP and calibrated using
liquid propane tests. For light hydrocarbons there is a great deal of torch fire and flare radiation
experimental data against which the models have been developed. But, similar to pool fires,
experimental data for more complex materials is limited, and the parameters used as a default in
CIRRUS are based on similarity judgements.
To model multi-component mixtures using CIRRUS you must choose a like surrogate from the
database; e.g. Ethane, Propane etc. The molecular weight, heat of combustion and other known
parameters of the surrogate should be changed to match those of the material being studied.
8.2.1.1 F Factor
In both models the heat radiated as a proportion of the total heat released is
estimated from an empirically derived fraction called the F-factor. In the gas
torch fire model the F-factor is calculated internally and depends on parameters
such as the jet velocity. In CIRRUS the user can over-ride the internal
calculation and input an F-factor directly. In the liquid torch fire model the
user supplies the F-factor to be assumed, a default is suggested by CIRRUS.
The F-factor is very important in flare radiation models as it is directly related to

the thermal radiation which is assumed to be emitted from the flame. Smoky
flames will emit more radiation. For clean burning high pressure methane flares
the F-factor should be about 0.7. For normal steam assisted refinery flares the F-
factor should be about 0.2. For smoky flares burning heavy unsaturated or
aromatic materials, or where the steam assistance is insufficient, the F-factor
should tend towards 0.4.

As an example of using the pool fire radiation model we are going to look at two
scenarios. Firstly a refinery flare of predominantly ethane, and secondly the
torch fire which would result for an ignited release from a hole in the lower
liquid portion of a propane Horton sphere. This second scenario is based on the
calculations, using model S6, previously performed in section 3.6.1 of this
manual.
8.2.2 Example of using model F2 - A Refinery Flare
In this scenario we are going to model the thermal radiation at ground level from
a refinery flare. From the CIRRUS menu select Fire and then model F2. Input
a dataset name and description, you will then see the dialog box below.
In CIRRUS you can model the combined radiation from more than one flare.
however in this example we are going to model a single flare, therefore leave the
default as 1 and click on OK.

You will then come to the torch fire model input screen. In CIRRUS the input
for the torch fire model is different to that of the other models as there are two
input screens, the first of these describes the flare details which is shown below.
Leave the default torch fire model type as Vapour Torch Fire or Pipe Flare. Use
the arrow on the right of the Material Name box to select ethane from the
database. The material properties of ethane will be displayed on the screen and
the pointer moved to the Z Height box. Input a height of 50m, and angle to the
horizontal of 90 (i.e. vertical), a flare tip diameter of 0.5m and a gas flowrate of
20kg/s.
Note: If Ethane was a surrogate material you would overtype the physical properties of ethane with
those of your actual material. You should also change the name in the name box.
The X-coordinate and Y-coordinate would be required for each flare if you were modelling the
combined thermal radiation from multiple flares.
Notice the F-factor for this gas torch fire model is defaulted to internal calculation. If your
material was not ethane and for example you suspected a smoky flame, then you could overtype
this input with your judged F-factor.
The horizontal angle off downwind direction is to model the radiation when the flare is not in
the plane of the wind. For further details you should read the context specific help (function key
F1). Note that help for this input screen is related to all parameters on the screen so you will
have to use the hypertext finger to select the horizontal angle off downwind direction
parameter.

Hit OK. You should now see the input screen showing atmospheric conditions
and your output requirements. Put in a wind speed of 5m/s, a humidity of 0%
and change the attenuation box to No. Input a temperature of 15C, plane
heights of interest of 0m (ground level), 50 m and 100m. Notice you are
required to guess the distance to which your input radiation levels will reach. In
CIRRUS this defines the boundary over which the matrix of radiation points is
calculated. The Try value is only an empirical estimate, in this case input a
value of 300m.
Note: Similar to pool fire radiation, allowing atmospheric attenuation due to the absorption of thermal
radiation by water vapour and carbon dioxide in the atmosphere, will reduce the predicted
thermal radiation burden at distances from the flame.
Your second screen should look like that shown below. If you wanted to go back
and amend your flare details you would click on the Flare Details button. Now
click on OK to run the model.

The first results screen shows you the flame length, its tilt in the wind, the stand-
off between the 50m high flare tip and beginning of the flame, and the radiation
contours. This is the vertical projection in the plane of the wind.
Notice that the radiation at ground level exceeds 2kW/m2 between about 0m to
almost 100m downwind. If you now select GRAPH you can examine the
horizontal contour at ground level (the horizontal plane at 0m) to see this contour
plot.
Note: Sometime you may get strange shapes within the contours. This is not a real effect but due to the
instability of the contouring program. To remove the shapes select a different Maximum
distance of interest on the second input sheet.
8.2.3 Example of using model F2 - A Liquid Torch Fire
In this example we are going to continue the scenario of a release from a 250mm
(10) hole in the lower liquid part of a Horton sphere containing propane at its
saturated vapour pressure at 15C (section 3.6.1). The calculated outflow rate
was 1.3te/s. Clearly this is a massive flowrate therefore we are expecting a very
large predicted torch flame.

Go to the Model menu and select Fire and model F2. Select a new case and
input a dataset name and description. Select to model a single flare or torch
flame and on the following input screen select the Liquid Torch Fire or Pipe
Flare option. Select Propane from the material box, input a height of 10m and
a Vertical Angle from grade of 0 (i.e. a horizontal release). You should have
the screen shown below.
Input the same atmospheric conditions as the previous case, model the 2kW/m2,
6kW/m2 and 12kW/m2 thermal radiation boundaries at plane heights of 0m, 10m
and 30m. Input a distance of interest of 1500m. Your screen should look like
the one below.

Now run the model and you should get the following output.
Notice the predicted flame length is about 500m. Clearly in the real case you
would take into account any obstructions in the path of the flame. On striking an
object the torch flame will become a turbulent fire where the jet momentum is
lost and the remainder of the unburnt hydrocarbon burns as a ball of flame. In
some instances it is possible to estimate the radiation resulting from a turbulent
fire by using the fireball model.
Note: The liquid torch fire model has been calibrated against relatively low liquid release rates - up to
50 kg/s. Even these releases can give very long flame lengths. We could conclude that the
model is well outside its calibration experiments - but we could not conclude that the results are
definitely incorrect.
8.2.4 Using Model F2 with Multi-Component Materials
To use model F2 with multi-component materials, firstly add the multi-

component material to the database as described in section 2.2.5. When using
model F2 you can select then this multi-component material from the material
dialog box. The material will appear at the bottom of the list under the list of
single component materials.
The next fire model of interest to us is the Fireball model. In CIRRUS this is
model F3

8.3 FIRE MODEL F3 - FIREBALL
Fireballs are a phenomenon normally associated with BLEVE, the mechanism

for which is discussed in the introduction to section 5 of this manual.
When a material which is a gas at ambient temperature and pressure but is stored
or processed as a liquid under pressure suddenly loses containment due to the
action of an impinging fire, the material boils, flashes and expands rapidly and
entrains air at its edge. It burns as a fireball with a high surface emissive power.
The fireball increases in radius and lifts off the ground. It will burn for a
relatively short duration until all the hydrocarbon is consumed by the fire.
Fireball models are calibrated against real incidents, the majority of which have
occurred with propane and butane bullets or spheres. In general it is thought that
if a quantity of material released is greater than about 300 tons then the models
are likely to overpredict the emitted radiation, as very large fireballs are likely to
burn less efficiently than relatively small fireballs.
Note: It is often repeated that a BLEVE event gives rise to blast overpressure - in general this is not the
case. A certain small overpressure will be developed due to the pressure burst of the
containment. You can model this using model E2 as discussed in section 6.2 of this manual.
Alternatively if the gas is released into an obstructed area a vapour cloud explosion might occur
in lieu of or in addition to a fireball. Vapour cloud explosions are separate phenomenon which
are modelled using model E1 and discussed in section 6.1 of this manual.

8.3.2 Example of using Model F3
In this scenario we are going to model the radiation effects from a fireball
generated by BLEVE of a 20 ton propane bullet. You are interested in the
distance at which 5% of persons would be fatally injured and the distance at
which 1% of persons would be fatally injured.
From the CIRRUS menu select Fire and model F3. Input a dataset name and
description. On the Input screen type in the above parameters. Your input
should look like this:
Now run the model and examine the results. The first graph shows you the
percentage lethality as a function of distance from the centre of the fireball at
ground level (which can be assumed to be the location of the propane bullet).

From this graph you can see that the 5% lethality boundary occurs at about 170m
from the source and the 1% lethality boundary occurs at about 190m from the
source. Select GRAPH and Lethality Contours to see an alternative view of
this information. Notice also that the maximum fireball radius is 80m and its
duration is predicted to be about 13 seconds.
If you now select Max Energy versus Distance you will see the energy in kJ/m2
displayed as a function of distance. You can change the axes to Log:Log as
shown below by choosing AXES from the menu.
This relationship is useful when you want to use criteria other than lethality
levels.
8.4 FLASH FIRES
Flash fire is the name given to the burning of a flammable vapour cloud at
normal flame speeds. For example the speed of a flame burning through a
flammable mixture of methane in air will be typically ten metres per second. No
overpressure is developed. Under some conditions the flame can accelerate to
speeds of several hundred metres per second, and in this case we would say that
the burning cloud explodes and overpressure would be generated. Explosions
and the modelling of explosions are discussed in the next section of this manual.

264
PART 3. DETAILED SCENARIO MODELLING SECTION 9. EXPLOSION MODELS
9. EXPLOSION MODELS
An explosion is characterised by the development of overpressure which travels

through the air as a pressure wave. In modelling hazards on refineries we can
differentiate two entirely separate types of phenomenon which result in pressure
waves; A vapour cloud explosion and a pressure burst explosion.
Vapour cloud explosions can arise when hydrocarbon vapour or aerosol forms a
flammable mixture with an oxidant. The models in CIRRUS assume that
the oxidant is air. As described previously in section 4.1.2.3, a
hydrocarbon/air mixture has a minimum proportion of hydrocarbon
which will allow the mixture to sustain ignition. This limit is called the
Lower Flammable Limit (LFL). Similarly a hydrocarbon/air mixture has
a maximum proportion of hydrocarbon which will allow the mixture to
sustain ignition which is called the Upper Flammable Limit (UFL).
Between these limits lies the stoichiometric concentration, which is the
concentration of hydrocarbon which just uses all the oxygen in the
hydrocarbon/air mixture. For most hydrocarbons the combustion process
has maximum efficiency at the stoichiometric concentration, but a
mixture is defined as being flammable if within the LFL and UFL limits.
A pressure burst explosion is caused by failure of a pressure vessel or other

pressurised containment. A pressure vessel can either spontaneously fail
due, for example, to a material defect, or alternatively a pressure vessel
can fail if overpressured. In this latter case the vessel will fail at its yield
pressure which will be some way above its design pressure.
Note: The terms Lower Flammable Limit (LFL) and Lower Explosive Limit (LEL) are interchangeable
and refer to the same concentration. Likewise the terms Upper Flammable Limit (LFL) and
Upper Explosive Limit (LEL) are interchangeable.
During the process of a pressure burst explosion the pressure vessel will break
into fragments which can be thrown through the air. In modelling terminology
these fragments are often called Missiles.

PART 3. DETAILED SCENARIO MODELLING SECTION 9. EXPLOSION MODELS
The BLEVE phenomenon, described in the introduction to section 5 of this

manual, is also a particular form of pressure burst in that the pressurised
containment fails. In this case the vessel will probably fail at a pressure below
its design pressure (due to the impingement of flame on unwetted metal).
Because of this the pressure wave developed during a BLEVE is usually
insignificant. In addition during BLEVE the pressure vessel will also break into
fragments which can be thrown through the air. Therefore BLEVE is also
associated with missile generation.
Note: During BLEVE of a propane or butane bullet the end of the bullet often breaks away from the
body of the bullet. In this case the body, which still contains rapidly expanding material can
rocket. The missile model in CIRRUS does not take into account the rocket phenomenon.
Therefore CIRRUS contains a vapour Cloud Explosion model - E1, a Pressure

Burst model - E2, and a Missile model - M1.

PART 3. DETAILED SCENARIO MODELLING SECTION 9.1. VAPOUR CLOUD EXPLOSION
9.1 EXPLOSION MODEL E1 - VAPOUR CLOUD EXPLOSION
When released into an open field and ignited, virtually all flammable mixtures of
hydrocarbon in air will burn with a slow flame speed as a flash fire. (Flash fires
are described in section 5.4 of this manual). Under the right circumstances
however most hydrocarbon vapours and/or aerosols when mixed with air can
explode and generate overpressure. Overpressure is developed because either
the combustion products of the explosion are confined, and/or the flame speed in
the mixture increases by several orders of magnitude due to turbulence induced
by obstacles in the path of the flame. Therefore one of two circumstances are
required to facilitate overpressure development.
The mixture must be confined in a relatively air-tight volume. When a

stoichiometric mixture of hydrocarbon in air burns, the combustion
products occupy a volume about eight times that of the original mixture.
If the combustion products are confined to the original volume the
pressure in the volume will increase to about 8 bar (approximately eight
times atmospheric pressure). For combustion of a flammable mixture in
buildings the expansion of the combustion products will be sufficient to
damage the building if the additional volume cannot escape in sufficient
time. This type of combustion product volume expansion overpressure
development is not modelled in CIRRUS.
In addition to volume expansion, reflection of the pressure from the sides of the
confined volume and the increase in the pressure itself can increase the
flame speed in the mixture. In this case the transient overpressure
recorded during the explosion can exceed 8 bar.
The flame speed in increased due to turbulence in the hydrocarbon/air mixture.

One reason for turbulence generation is the presence of obstructions such
as pipework, vessels and other equipment in the path of a flame front.
Due to the turbulence generated the flame front accelerates over the first
obstacle, this accelerated flame then accelerates even more over the
second obstacle etc. The result is an exponential rise in the flame speed
and the consequent development of a pressure wave.
Another reason for the development of turbulence is when the combustion of a

hydrocarbon/air mixture in a confined area exhausts into a larger volume
of the flammable mixture.
The degree of overpressure which is developed due to turbulence depends on the

degree of turbulence. This is related to the density of the obstructions
and the number of obstructions. Therefore the input to these models

require the obstructed volume and a parameter which describes the

density of the obstructions. The model in CIRRUS treats vapour cloud
explosions in this manner.
In summary, excepting ignition by detonation (a special case which is discussed

in section 6.1.1 below), a vapour cloud explosion will only occur if there are
obstructions in the path of a flame front. Therefore if you imagine a very large
flammable gas cloud with a small process plant in the centre of it containing
obstructions, and ignition of the vapour cloud at one edge away from the process
plant, the cloud will burn slowly towards the process plant with any overpressure
development, it will accelerate (explode) through the process plant and then burn
slowly to the other edge of the cloud. Therefore the explosion overpressure is
related to the obstructed volume and not the volume of hydrocarbon in the cloud
(i.e. the quantity released).
Another type of model called the TNT equivalence model assumes that explosion overpressure
is related to the quantity of hydrocarbon in a cloud. Experiments in the last decade have proved
this to be incorrect. In general only hydrocarbon clouds in an obstructed area will explode.
The explosion model used in CIRRUS is the TNO Multi-Energy model.

9.1.1 The Assumptions Inherent in Model E1
The vapour cloud model in CIRRUS assumes that a vapour cloud explosion
occurs due to the obstructions in the path of a flame front. The overpressure
generated by the vapour cloud explosion is related to the obstructed volume The
actual volume of hydrocarbon released is unimportant as long as there is
sufficient to fill the volume with a flammable mixture.
Note: As a minimum the flammable cloud should fill one eighth of the chosen obstructed volume, as
combustion products will push unburnt hydrocarbon into areas of the obstructed volume not
filled by the original cloud.
The overpressure is also related to the obstacle density and type of material. In
CIRRUS this is encapsulated within an energy coefficient which ranges
between 1 and 10. Use of a coefficient of 10 represents detonation, use of
coefficients of less than 10 represent a deflagration of corresponding relative
energy.
Note: A detonation is defined as an explosion where the flame speed is equal to that of the speed of
sound in the medium, (i.e. it is travelling as fast as it can do, the upstream material is not affected
by the flame front until it is combusted in the flame front). The pressure wave consists of a shock
front followed by a decay. Once a detonation commences it should (in theory) continue until all
the material is consumed. Therefore if a hydrocarbon cloud is ignited by a detonation then it will
continue to combust as a detonation.
A deflagration is defined as an explosion where the flame speed is less than that of the speed of
sound in the medium, (i.e. it could speed up to detonation if the right conditions prevailed, or
alternatively slow right down to the normal burning velocity found in a flash fire with no
overpressure generation. The pressures and temperatures of the upstream material will increase
prior to combustion in the flame front). The pressure wave consists of a ramp up to a maximum
pressure followed by a decay.
9.1.2 Example of using model E1 -
In this scenario we are going to imagine that a stoichiometric mixture of propane

in air rolls along the ground as a dense gas cloud and into a process plant of
dimensions 20m x 20m x 5m. As this is a dense vapour cloud its height is
assumed to be about 2m, but since combustion products may push unburnt
hydrocarbon into the upper part of the plant we will assume the entire confined
volume can take part in the explosion. As there is a moderate obstacle density in
the plant we will use an energy coefficient of 7. We are interested in the

overpressure at the location of an administration building 100m from the plant

centre.
Note: For a normal process plant the default energy coefficient in CIRRUS is 7. Work is on-going to
provide a better description of the energy coefficient to obstacle density relationship. As a worst
case an energy coefficient of 10 would assume that detonation occurs - historically a very
unlikely event.
The model redefines your input confined volume in terms of a hemi-spherical confined volume.
Therefore overpressures close to the source (one confined volume length) will not be calculated
correctly
From the CIRRUS menu select Explosion and then model E1 - Vapour Cloud
Explosion. Input a dataset name and description and select Propane as the
material of interest. Now input your data. The volume of the obstructed plant is
2,000m3 (20m x 20m x 5m), Leave the default percentage of fuel in fuel air
mixture as 3.82% (the stoichiometric proportion) and leave the default heat of
combustion, energy coefficient and overpressures of interest. Input 100m as the
point of interest.
Your input sheet should look like this:
Now run the model and you should see the overpressure versus distance
relationship.

Change this to log:log using the AXES command as shown below.
Note that the overpressure at 100m is just greater than 1 bar.
For confirmation, from GRAPH select static overpressure and duration at the
point of interest (impulse) as shown below:
An overpressure of 0.11 bar would correspond to masonry damage at the

administration building. To examine the contours of the three overpressures
which you left as default values on the input screen you should go to GRAPH an
select static overpressure contours.
Note: The pressure wave shape shown in CIRRUS is an isosceles triangle and is used as a
representation of the overpressure and duration. In reality the shape will vary from a smoothed
triangle to vertical edged smoothed triangle (a shock wave).

These are shown below.
Note: Static overpressure is the maximum peak overpressure of the pressure wave at the point of
interest. There are other types of overpressure. Reflected overpressure would be experienced by
a building wall and is exactly twice the static overpressure. Dynamic overpressure is a parameter
which is related to the momentum of the air following the blast wave. In hydrocarbon explosions
this is lower than the static overpressure.

PART 3. DETAILED SCENARIO MODELLING SECTION 9.2. PRESSURE BURST EXPLOSION
9.2 EXPLOSION MODEL E2 - PRESSURE BURST EXPLOSION
This model calculates the yield strength of a pressure vessel from its diameter,
wall thickness and the Ultimate Tensile Strength (UTS) of the steel used to
fabricate the vessel. From these parameters the model calculates the burst
pressure and the energy released during the burst. This energy is used to
calculate the overpressure experienced in the surrounding atmosphere and its
decay as a function of distance.
9.2.1 The Assumptions Inherent in Model E2
The model assumes that the pressure vessel has not been weakened by either
material defect or by flame impingement (which would increase the temperature
of the vessel and reduce its strength).
9.2.2 Example of using model E2 - Overpressurisation of a Pressure Vessel
In this scenario we are going to feed high pressure methane gas into a pressure
vessel with a yield strength of about 50 bar (design pressure of about 30 bar).
We are interested in the overpressure at an administration building 100m from
the vessel. The vessel is a horizontal bullet which is 3m in diameter (height) and
8m long with a wall thickness of 15mm.
From the model menu select Explosion and model E2 - Pressure Burst
Explosion. Input a dataset name and description. Select Methane as the
material of interest. In the dialog box select a cylindrical vessel and click on
OK.
Input the vessel dimensions and leave the UTS and pressure contours of interest
at the default value.

PART 3. DETAILED SCENARIO MODELLING SECTION 9.2. PRESSURE BURST EXPLOSION
You should have an input screen which looks like this.
Note: The default UTS is quite high, look at the context specific help (click on the UTS box and hit
function key F1) for alternative values.
Now run the model and you should see the predicted relationship between
overpressure and distance. If you change this to Lin:Log using the AXES menu
you will be able to better interpret the information as shown below.
Note that the maximum burst pressure calculated is 50 bar, but this decays
rapidly over the first few tens of metres. If you change the graph to log:log using
the AXES menu you will be able to read the overpressure at the administration
building 100m from the vessel as 40 millibars, which would result in window
breakage.
The final model in the explosion section predicts the distance to which fragments
of the vessel will be throw.

PART 3. DETAILED SCENARIO MODELLING SECTION 9.3. MISSILES
9.3 EXPLOSION MODEL M1 - MISSILES
This model predicts the likely maximum range to which fragments of a pressure
vessel will be thrown if a pressure vessel bursts. It also predicts the probability
that fragments will land at other distances.
In the case where a pressure vessel is not weakened by fire, the model first
predicts the burst pressure in a similar way to model E2, and then uses a
proportion of the energy released to calculate a probabilistic distribution of
fragments from the vessel.
A similar calculation is performed for a vessel which undergoes BLEVE,

however in this case the vessel is weakened by an impinging fire, and therefore
the user has to estimate the burst pressure of the vessel which could be well
below its yield or design pressure.
9.3.1 The Assumptions Inherent in Model M1
The missile model has been developed on the basis of empirical data. It assumes
that fragments will be thrown from the vessel and will frisbee at there optimal
angle, and therefore the results are likely to be an overprediction.. A proportion
of the energy released from the vessel is used to throw the fragments.
9.3.2 Example of using model M1 - Fragment Throw Distance following

Overpressurisation of a Pressure Vessel
We are going to continue with the pressure burst scenario shown modelled in
section 6.2.2. This was the overpressurisation of a Methane bullet. From the
model menu select Explosion and then model M1 - Missiles. From the first
dialog box change the default vessel type to cylindrical and select gas as the
state of interest.

Your dialog box should look like this:
Click on OK and input the vessel parameters as described in section 6.2.2.

Assume that the vessel is 3m in diameter and 8.8m in length with a thickness of
15mm. Select Methane as the material of interest and use a gas temperature of
15C and assume that the vessel splits into 10 fragments. Your input screen
should look like this:
Now run the model.

Firstly you will see the probability distribution of fragments as a function of

distance. This relationship is shown below.
You will see that the majority of fragments land within a distance of 1.3km, with
the most popular landing distance being about 100m.
If you now go to graphs and review probability plots you can view an
alternative representation of this information as shown below.
The other graph, Cumulative Probability versus Distance shows you the
probability that a certain number of fragments will have landed by a certain
distance.

9.3.3 Example of using model M1 - Fragment Throw Distance following a

BLEVE
In this scenario we are going to use the BLEVE which we envisaged when
calculating Fireball radiation damage in section 5.3.2 of this manual. This was
BLEVE of a 20 ton propane bullet.
From the model menu Select Explosion and then model M1 - Missiles.
From the first dialog box change the default vessel type to cylindrical and select
two-phase as the state of interest.
On the following input screen you should input vessel dimensions of 2.4m
diameter and 7.3m length (approximately equal to a 20 ton propane bullet).
Input a wall thickness of 15mm. As the vessel can fail below its design pressure
input the pressure at failure to be 10 bar. In this BLEVE scenario assume that
the vessel breaks into 5 fragments.
Note: Using a vessel burst pressure of 10 bar in this instance assumes that the vessel will probably
BLEVE even before its relief valve has lifted. An alternative assumption would be to use the
relief valve set pressure on the basis that if the relief valve lifts, it will maintain the vessel at this
pressure.
If you now run the code you will see output which is similar in kind to that
described for example 6.3.2.

Below is the probability distribution of fragments.
You can see that the majority of fragments are likely to have landed within 600m
of the vessel, with 50% of the fragments having landed within 80m of the vessel.
We have now come to the end of our examples of using CIRRUS to model
typical scenarios.

280
PART 3. DETAILED SCENARIO MODELLING SECTION 10. FURTHER EXAMPLES
10. FURTHER EXAMPLES TAKEN FROM RFINERY AND E&P

FACILITIES
10.1 INTRODUCTION
In this section we are going to augment the examples discussed previously in

Part 3 of this manual to model release scenarios from example refinery units and
E&P facilities. There is an almost unlimited set of cases from which to choose
from in relation to the types of mixtures of materials contained within storage
and process vessels and pipelines, the temperatures and pressures of this
containment and the size of hole from which you envisage the release can occur.
In any scenario development you should try to look at a range of release sizes in
order to encompass the worst case scenario and cases which are more likely to
occur in reality. For refinery units or other types of process plant you should
choose various sections of the process unit in order to encompass all the types of
fluid mixtures and containment conditions within the unit. Also, you should run
the calculations for a range of weather conditions. Finally you should examine
the results in relation to both the consequence distances predicted and the
timescales over which the incident is predicted to occur in order to gain a better
sense of the reality of the scenario development in the real environment.
The first example we are going to use is taken from a refinery Hydrofluoric Acid
(HF) alkylation unit. We are going to assume that the worst case scenario is that
defined by the US RMP rule; a release of the largest vessel inventory over a
period of ten minutes. This would be associated in reality with for example
failure of a large fitting. An alternative worst case scenario which would be
more appropriate for modelling under the EC Seveso directive would be
catastrophic failure of the vessel itself; a release of all the vessel contents in 1
minute. Although not demonstrated here, the same principles would apply to
modelling this scenario.
As a second example we are going to take the release from a crude oil cross-
country transmission pipeline and model the gas dispersion of the heavy vapours.
As a final example we are going to model a blowout of an oil well.
Note that many examples of modelling have already been given in the previous
sections of the manual and which cover aspects such as; vapour dispersion from
an evaporating pool (7.2.2) or a relief valve (7.3.1 and 7.3.2), thermal radiation
from a tank fire (8.1.3) or flare (8.2) and an explosion in a process plant (9.1.2).

PART 3. DETAILED SCENARIO MODELLING SECTION 10.2. HF ALKYLATION UNIT
10.2 A WORST CASE RELEASE FROM AN HF ALKYLATION UNIT

STORAGE VESSEL
Overleaf you can see a schematic diagram of an HF Alkylation unit reactor

section. Part of this is a two section column containing at the bottom acid
storage and at the top an acid settler.
The scenario we are going to develop is a release from the acid storage section of
the column. The unit and the location have the following characteristics:
Parameter US Units SI Units
Diameter 23 ft 7m
Height of liquid level 31.5 ft 9.6m
Max. inventory 800,000 lb 3.63 x105 kg
Storage pressure 30 psig 15 barg
Storage temperature 70F 21C
Ambient temperature 68F 20C

Ambient humidity 50% 50%
Windspeed 1 (at 10m height) 4.5 mph 2 m/s

Windspeed 2 (at 10m height) 10.5 mph 5 m/s
Terrain roughness coefficient 0.01 0.01
Concentration of interest (ERPG-2) 20 ppm 20 ppm
Note: CIRRUS has a unit conversion facility to paste SI units into the input data screen. Select the
required box and from the Menu select Units, or alternatively use the shortcut key Ctrl U.
In the RMP rule the worst case scenario is defined as that which would release
the total contents of the vessel over ten minutes. This would in this case result in
a release of HF at an average rate of 605 kg/s (4.8 x 106 lb/hr). Firstly, using
model S6 in Cirrus, we can perform a few iterations to find out what hole size
this release would correspond to. Select model S6 and input the above
information.


Note that only the vessel diameter and liquid height are required as the program
will calculate the inventory mass. We are going to guess a hole diameter 400mm
will give a release of about 10 minutes.
The hole has to be at the bottom of the vessel to release all the contents. Now
run the model to obtain the following result:
The release rate decreases as the pressure falls to the saturated vapour pressure
and then as the liquid height falls. In this case the average rate of 605 kg/s
occurs about half way through the 10 minute duration.

We now want to disperse this material. Evaporation from an HF liquid spill

occurs initially in the HF hexomer form which changes to the monomer form.
As the spill is onto the ground and the vapour is initially a heavy vapour, we
should use model D2, the heavy gas model, as the appropriate dispersion model.
Refer to section 0 for details.
Note: In CIRRUS the model D2 is the Shell model HEGADAS. It therefore takes into account the
molecular behaviour of HF vapour.
We now have several different choices in the way we model the release:
We can assume that all the liquid spilt will vapourise instantaneously. Our first
option is to put a vapour release rate of 605 kg/s directly into the continuous
option of model D2. The concentration of interest for this example is 20ppm
therefore the continuous model is likely to overestimate the dispersion
distance. This is because the time taken for a vapour cloud to reach
equilibrium conditions at this low concentration will be far in excess of ten
minutes. Therefore a continuous release assumption is not a good
approximation for this case.
2. As above we can assume that all the liquid spilt will vapourise
instantaneously, but our second option is to put a vapour release rate of 605
kg/s directly into the transient option of model D2. If we use this option we
still have a further two choices; we can approximate the release by
assuming a constant release rate of 605 kg/s over ten minutes, or we can
input the actual transient developed in model S6 and shown previously.
3. We can assume that all the liquid spilt will not vapourise instantaneously,
but form a pool on the ground which then evaporates to give us the vapour
release rate feeding the dispersion cloud.
Firstly we are going to use the third option to model the release, and then
compare the results with the other two options.
Select model S8, assume that the unit is not bunded and select a transient release.

You S8 option box should look like this:
Input the release data, and you should have a screen which looks like this:
In this case notice we are using a 2 m/s windspeed.

Now run the model and obtain the following result:
Look at the results. We are predicting that the spill will take in excess of 50
minutes to evaporate when the wind speed is 2 m/s. The maximum evaporation
rate from this unbunded spill is about 130 kg/s, not the average 605 kg/s
predicted if the spill formed a vapour cloud instantly.
If you examine the pool diameter graph, you will see that the maximum diameter
of this unbunded spill is predicted to be 280m. If the unit had a bund which was
a smaller diameter than this, then the evaporation rate would be reduced further.
The results in these F2 conditions are shown overleaf, a distance to the 20 ppm
concentration of about 140 km (87 miles).

Note that we have used a time of interest of 1200 min (20 hours) in order to
catch the maximum distance of this dispersion, indicating that in F2 conditions
this ten minute release will not have dispersed to below the ERPG-2 level until
almost a day later.
If we now change the weather conditions to D stability and a 5 m/s windspeed,

we can obtain the following result. We have to change the maximum time of
interest to 120 minutes.
Firstly note that the maximum distance to the ERPG-2 level is 19 km (12 miles)
and occurs about 1 hour fifty minutes (6,000s) after the release occurs.
Thereafter the cloud diminishes in size and we can see it disappearing after 2
hours (7200s). Contrast this result with the F2 results.

Note: In this example we have used the input data from the previous S8 evaporation run. More exactly
we should have repeated this evaporation run with a windspeed during the evaporation of 5 m/s.
If we were to model this release as an instantaneous cloud formation at a rate of

605 kg/s over 10 minutes the results would not be very different. This is because
in this case the time taken to reach the chosen endpoint is significantly greater
than the release duration if modelled as an evaporating pool. However if we
modelled the release as a continuous spill of 605 kg/s then we would get a
markedly different and grossly pessimistic result.
The following table overleaf shows the results of this modelling:

Option Maximum distance to 20 ppm Time taken to disperse to less than 20 ppm
F2 D5 F2 D5
Continuous release 430,000 km 28 km continuous continuous

at 605 kg/s
Transient release if spill 125 km 21 km 17 hours 1 hour

evaporates instantly
Transient release if spill 140 km 19 km 20 hours 2 hours

evaporates over time

PART 3. DETAILED SCENARIO MODELLING SECTION 10.3. A CROSS COUNTRY PIPELINE
10.3 A RELEASE FROM A CRUDE OIL CROSS COUNTRY TRANSMISSION

PIPELINE.
In this example we are interested in an accident involving a release from a large

hole in a buried cross-country pipeline containing live (spiked) crude oil . As the
oil is a flashing liquid, the first thing we must do is to input the composition of
the crude oil into CIRRUS. To do this we create a new multi-component
material as described in section 2.2.5. The composition of the oil is shown
below. We will use the mole fraction method of describing the composition.
Component Mole Fraction
Ethane 0.14
Propane 0.08
Butane - n 0.04
Butane - iso 0.03
Pentane - n 0.12
Pentane - iso .0097
Hexane 0.11
Heptane 0.15
Octane 0.05
B. pt. 163C - 224C (Nonanes - Dodecanes) 0.11
B. pt. 225C - 350C (Tridecanes - Eicosane) 0.1556
Nitrogen 0.0045
Carbon Dioxide 0.0002
To examine the properties of the material you should use model S9 as described
in section 6.1.7.

Assume the material in the pipeline is at a pressure of 65 bara and 30C. Below
is the output from the model S9.
As you can see, the material vapour pressure is about 8.3 bara at this temperature
with a vapour flash fraction of 9.2%, a vapour a molecular weight of 41.2 and a
predicted temperature of 13.9C. (As the release is at ground level and the
vapour is heavier than air you will be using a heavy gas model to predict the
vapour dispersion distances - refer to section 0). To allow for entrained aerosol
CIRRUS suggests that 50.8% of the total release should be assumed to form a
vapour cloud and that the average molecular weight of the cloud should be 89.46
(i.e. an even heavier gas).
The actual vapour dispersion cloud will lie between these two ranges depending
on the method of release. For example if the hole is in the bottom of the pipe,
the former may be more appropriate as impact with the surrounding earth will
aggragate the aerosol and leave only flashed vapour to form the cloud. A clear
jet of flashing liquid from a hole in the top of the pipeline will clear the earth
above it leaving a jet involving aerosol formation.
Note: CIRRUS uses an algorithm appropriate for light hydrocarbons when making assumptions
regarding aerosol entrainment. In future versions this may be changed to allow for a lower
entrainment of higher molecular weight fractions.
We are now going to use model S7 to estimate the release of material from the
pipeline as described in section 0. Select S7, input a dataset name and
description and select the material from the multi-component database which
you have created. In this case we are going to assume that the pipeline is 100 km
long and is flat. However when modelling pipelines from offshore platforms
there is an option to allow you to specify risers at each end of the pipeline system
and a valve which closes to isolate inventory after a specified time.

Select a pipeline without risers and valves:
and input the following data for a 100,000m pipeline assuming that the hole
occurs at 50,000m and has a diameter of 300mm. Input a time of interest of 100
minutes.
Note that all the inventory of the pipeline is not exhausted within 100 minutes as
indicted by the subsequent message:

Now review the output
Note: The pressure drop between the rupture pressure and initial pressure is related to the input
pressure by only the wall roughness. In real systems bends and other flow obstructions add to
this differential. If you know the pressure of the pipeline at the position of the hole, change the
input pressure until you achieve the correct pressure at the hole position.
In order to obtain a better representation of the behaviour you can select different
scales from the Axes command. Also, the graph you are looking at is the total
mass released, from the Graphs option you can select the vapour release and
you should see that the flash fraction of vapour is about 9% of the total release.

Below is the Log:Log view of the vapour release.
In a wind speed of 5 m/s vapour will travel 500m in 100 seconds. As this is the
order of magnitude distance you are interested in then the first 100 seconds of
the release will determine the downwind hazard range. From the graph you
could approximate this to a continuous release of about 200 kg/s or put in the
real transient release rate as shown in section 7.2.3.1 if the release is into ground
(i.e. momentum is lost due to impact from the ground and model D2 would be
appropriate).
10.3.1 A hole in the top of the pipeline
To model a hole in the top of the pipeline we want to use model D1 (which
includes momentum affects - section 7.3) and find a surrogate for the material
which has a molecular weight of 42, ( a molecular weight of 89 inclusive of
aerosol entrainment) and a vapour pressure of about 8 bara at 30C.
Note: Currently CIRRUS will only accept single components. A future version may become available
which accepts multi-component materials and which uses the same flash parameters as model S9.
Propane is a reasonable surrogate for this material. The name, MW, Cp and
Cp/Cv can be changed on the input screen to match those of the material which
is being analysed (though the saturated vapour pressure versus temperature
relationship will remain as Propane).
Select model D1 and select Propane as the material of interest. On the input
screen change the material name to that of your material. Input a vapour release

rate of 200 kg/s and a temperature of 18C (this is to better match the vapour
pressure of our material at 30C). Change the other properties to those of the
crude oil as given by model S9. Use a roughness coefficient of 0.001 and
whether conditions of D5 (refer section 7.1.2 for details). Leave the pressure as
that which is calculated for you; the pipeline pressure of 62 bara refers only to
the pressure at the first moment of the release and not the pressure which is
equivalent to a release of 200 kg/s of vapour. You should have the following
input screen:
Now run the model.
Click on OK at the warning that this is a two-phase release, and you will get the
following input screen which gives appropriate upstream input for the case you
are trying to model.

Run this input and select the vertical cross-section from the Graphs menu for
a pictorial representation of the flammable plume.
10.3.2 A hole in the bottom of the pipeline
You should model this scenario in a similar way to the above except that this
time momentum will be lost by impact with the ground underneath the pipeline;
hence you should select model D2 - again using Propane as a surrogate material.
Choosing the continuous option you should end up with an input screen similar
to the one shown below.

On running this case, you should obtain a rather different flammable envelope
than that for the case when the hole was in the top of the pipeline.
This marked difference in behaviour is due to the differences in the physics

between the dispersion of heavier than air vapours dispersing on the ground and
those of vapours dispersing from an elevated jet. In the former case the vapour
entrains air only from the edges of the low cloud, whereas in the latter case the
plume entrains air from all edges and is diluted initially due to momentum
entrainment from the turbulence created when the jet is forced into the air.
For further details on these phenomena refer to section 7.1.

PART 3. DETAILED SCENARIO MODELLING SECTION 10.4. BLOW-OUT OF AN OIL WELL
10.4 BLOWOUT OF AN OIL WELL.
The consequences of blowouts are modelled in two ways:
As an gas dispersion cloud from an unignited release.
As a torch fire from an ignited release.
How we model these effects depends very much on the material being
considered. The first step is to input the composition of the wellhead fluid as a
multi-component material as described in section 2.2.5, and then use model S9
to flash this material to determine its average molecular weight (section 6.1.7).
10.4.1 Gas Wells
If the blowout is from a gas well, then the molecular weight will low (less than
30), and the dispersion can be modelled using model D1 (refer section 7.3).
Methane will often be an appropriate surrogate for this material and an
adjustment should be made to the molecular weight and specific heat of the fluid
as given by the output of model S9.
The torch fire from an ignited gas blowout will be similar to an open ended pipe
flare. Therefore the gas jet option of the flare torch fire model should be
selected using an appropriate surrogate (usually methane) which most accurately
matches the molecular weight of the real fluid. Section 8.2 gives details of this
process. Fine tuning can be applied by an adjustment to the heat of combustion.

PART 3. DETAILED SCENARIO MODELLING SECTION 10.4. BLOW-OUT OF AN OIL WELL
10.4.2 Oil Wells
The wellhead fluid of an oil well is a multi-component mixture of gas an oil and
can cover a diverse range of compositions. The fluid can be a mixture of gas,
flashing liquids and liquids.
To model the vapour dispersion from an oil blowout the same procedure as
outlined in section 10.3.1 relating to the dispersion from a release from the top of
a live crude oil pipeline should be followed.
The torch fire from an ignited oil blowout will be similar to an open ended pipe
flare, but in this case the liquid jet option of the flare torch fire model should be
selected. A surrogate material should be selected based on the average molecular
weight of the material given by the model S9 - the flash calculation. Again fine
tuning can be performed by altering the heat of combustion of the fluid, which is
often available from well test information.

CONTENTS
301
SECTION 11. CONTENTS
11. LIST OF CONTENTS
1. INTRODUCTION TO CIRRUS ................................................................................ 1
1.1 PURPOSE AND DESCRIPTION..................................................................................................... 1
1.2 ABOUT THIS MANUAL.................................................................................................................. 2
1.3 INSTALLATION............................................................................................................................... 3
2. PART 1 - TUTORIAL FOR NEW USERS................................................................ 6
2.1 A WORKED EXAMPLE.................................................................................................................. 8

2.1.1 Step 1 - Creating a folder in which to store your data..................................................................... 8
2.1.2 Step 2 - Choosing the model to use................................................................................................. 8
2.1.3 Step 3 - Creating a new Dataset for input to the model................................................................... 9
2.1.4 Step 4 - Checking data and trying to run the model ..................................................................... 12
2.1.5 Step 5 - Running the model........................................................................................................... 15
2.1.6 Step 6 - Reviewing the results....................................................................................................... 16
2.1.7 Step 8 - After modelling................................................................................................................ 22
2.2 OTHER FEATURES....................................................................................................................... 26

2.2.1 Definition and usage ..................................................................................................................... 26
2.2.2 User level ...................................................................................................................................... 26
2.2.3 Unit conversion facility................................................................................................................. 27
2.2.4 The Single-Component Database.................................................................................................. 28
2.2.5 The Multi-Component Database ................................................................................................... 32
2.2.5.1 Browsing an existing multi-component .................................................................................... 33
2.2.5.2 Defining a new multi-component ............................................................................................. 34
BP CIRRUS User Manual version 7.0 i

3. HOW TO USE PART 2 - SCENARIO DEVELOPMENT..................................... 38
4. MODEL TYPES......................................................................................................... 39
4.1 SOURCE TERM MODELS ........................................................................................................... 40
4.2 DISPERSION MODELS................................................................................................................. 43

4.2.1 Types of dispersion models........................................................................................................... 44
4.3 FIRES AND EXPLOSIONS ........................................................................................................... 46

4.3.1 Types of fire and explosion model................................................................................................ 48
4.4 CONTAINMENT TYPES............................................................................................................... 50
4.5 MODELLING PARTICULAR SCENARIOS............................................................................... 53

4.5.1 A TANK CONTAINING MATERIAL WHICH LIQUID AT AMBIENT TEMPERATURE
AND PRESSURE ...................................................................................................................... 55
4.5.1.1 Scenario 1 - A hole in the tank, evaporation, dispersion, explosion effects < 10 km. .............. 56
4.5.1.2 Scenario 2 A hole in the tank, evaporation and dispersion effects > 10 km. ............................ 56
4.5.1.3 Scenario 3 Rupture of the tank, evaporation, dispersion, explosion effects < 10 km ............... 57
4.5.1.4 Scenario 4 Rupture of the tank, evaporation and dispersion effects > 10 km........................... 57
4.5.1.5 Scenario 4 A hole in the tank and fire ...................................................................................... 58
4.5.1.6 Scenario 5 Rupture of the tank and bund fire ........................................................................... 58
4.5.1.7 Scenario 6 A tank roof fire ....................................................................................................... 59
4.5.1.8 Scenario 7 A tank boil over...................................................................................................... 59
4.5.2 A VESSEL CONTAINING MATERIAL WHICH IS A VAPOUR AT THE CONTAINMENT
PRESSURE AND AMBIENT TEMPERATURE ........................................................................ 61
4.5.2.1 Scenario 1 A hole in a gas vessel, dispersion effects < 10 km. ................................................ 62
4.5.2.2 Scenario 2 A hole in a gas vessel, dispersion effects > 10 km. ................................................ 62
4.5.2.3 Scenario 3 A hole in a gas vessel and jet fire ........................................................................... 63
4.5.2.4 Scenario 4 Rupture of a gas vessel, pressure burst explosion and missiles .............................. 64
4.5.3 A VESSEL CONTAINING MATERIAL WHICH IS LIQUID AT THE CONTAINMENT
TEMPERATURE AND PRESSURE BUT GAS AT AMBIENT TEMP. AND PRESSURE ..... 65
4.5.3.1 Scenario 1 A hole in the vessel, dispersion effects < 10 km..................................................... 66
4.5.3.2 Scenario 2 A hole in the vessel, dispersion effects > 10 km..................................................... 66
4.5.3.3 Scenario 3 A hole in the vessel, evaporation, dispersion, explosion effects < 10 km. ............. 67
4.5.3.4 Scenario 4 A hole in the vessel, evaporation, dispersion effects > 10 km................................ 68
4.5.3.5 Scenario 5 Rupture of the vessel, evaporation, dispersion, explosion effects < 10 km. ........... 69
4.5.3.6 Scenario 6 Rupture of the vessel, evaportaion, dispersion effects > 10 km.............................. 70
4.5.3.7 Scenario 7 - A hole in the vessel, vapour release and jet fire ................................................... 71
4.5.3.8 Scenario 8 - A hole in the vessel, liquid release and jet fire..................................................... 71
BP CIRRUS User Manual version 7.0 ii

4.5.3.9 Scenario 9 - A hole in the vessel, liquid release and pool fire.................................................. 72
4.5.3.10 Scenario 10 - Rupture of the vessel, fireball and missiles ........................................................ 72
4.5.4 A PIPELINE CONTAINING MATERIAL WHICH IS LIQUID AT AMBIENT
TEMPERATURE AND PRESSURE.......................................................................................... 73
4.5.4.1 Scenario 1 A hole in a liquid pipeline, evaporation, dispersion explosion effects < 10 km ..... 74
4.5.4.2 Scenario 2 A hole in a liquid pipeline, evaporation, dispersion explosion effects > 10 km. .... 74
4.5.4.3 Scenario 3 Rupture of the pipeline, liquid spill, evaporation, dispersion effects < 10 km ....... 75
4.5.4.4 Scenario 4 Rupture of the pipeline, liquid spill, evaporation, dispersion effects > 10 km ....... 76
4.5.4.5 Scenario 5 - A hole in the pipeline, liquid release, pool fire .................................................... 76
4.5.4.6 Scenario 6 - Rupture of the pipeline, liquid release, pool fire .................................................. 77
4.5.5 A PIPELINE CONTAINING MATERIAL WHICH IS A VAPOUR AT THE CONTAINMENT
4.5.5.1 Scenario 1 A hole in the pipeline, vapour release and dispersion effects < 10 km................... 80
4.5.5.2 Scenario 2 A hole in the pipeline, vapour release and dispersion effects > 10 km................... 81
4.5.5.3 Scenario 3 A hole in the pipeline and jet fire ........................................................................... 82
4.5.5.4 Scenario 4 Rupture of the pipeline and pool fire...................................................................... 83
4.5.6 A PIPELINE CONTAINING MATERIAL WHICH IS A LIQUID AT THE CONTAINMENT
TEMPERATURE AND PRESSURE BUT A GAS AT AMBIENT TEMP. AND PRESSURE . 85
4.5.6.1 Scenario 1 A hole in the pipeline, evaporation, dispersion, explosion effects < 10 km. .......... 86
4.5.6.2 Scenario 2 A hole in the pipeline, evaporation and dispersion effects > 10 km. ...................... 87
4.5.6.3 Scenario 3 Rupture of the pipeline, evaporation, dispersion, explosion effects < 10 km. ........ 88
4.5.6.4 Scenario 4 Rupture of the vessel, evaporation and dispersion effects > 10 km........................ 88
4.5.6.5 Scenario 5 A hole in the pipeline and jet fire ........................................................................... 89
4.5.6.6 Scenario 6 Rupture of the pipeline and pool fire...................................................................... 90
4.5.7 A RELIEF VALVE EMITTING MATERIAL WHICH IS A VAPOUR AT THE CONTAINMENT
4.5.7.1 Scenario 1 RV lifts, dispersion effects < 10 km ....................................................................... 92
4.5.7.2 Scenario 2 RV lifts, dispersion effects > 10 km. ...................................................................... 92
4.5.7.3 Scenario 3 RV lifts, jet fire....................................................................................................... 93
4.5.8 A RELIEF VALVE EMITTING MATERIAL WHICH IS A LIQUID AT THE CONTAINMENT
TEMPERATURE AND PRESSURE BUT A GAS AT AMBIENT TEMP. AND PRESSURE . 95
4.5.8.1 Scenario 1 RV lifts, dispersion effects < 10 km ....................................................................... 96
4.5.8.2 Scenario 2 RV lifts, dispersion effects > 10 km. ...................................................................... 96
4.5.8.3 Scenario 3 RV lifts, jet fire....................................................................................................... 97
4.5.9 A FLARE STACK EMITTING MATERIAL WHICH IS A VAPOUR AT THE
STACK PRESSURE AND AMBIENT TEMPERATURE ........................................................ 99
4.5.9.1 Scenario 1 Flare out, dispersion effects < 10 km. .................................................................. 100
4.5.9.2 Scenario 2 Flare out, dispersion effects > 10 km. .................................................................. 100
BP CIRRUS User Manual version 7.0 iii

4.5.9.3 Scenario 3 Flare thermal radiation ......................................................................................... 101

4.5.10 A FLARE STACK EMITTING MATERIAL WHICH IS A LIQUID AT
AMBIENT TEMPERATURE AND PRESSURE ....................................................................... 103
4.5.10.1 Scenario 1 Flare liquid rain-out.............................................................................................. 104
4.5.10.2 Scenario 2 Flare thermal radiation ......................................................................................... 104
4.5.11 A CHIMNEY STACK EMITTING MATERIAL WHICH IS A VAPOUR
AT AMBIENT TEMPERATURE AND PRESSURE................................................................. 105
4.5.11.1 Scenario 1 Chimney flue gas effects > 1 km. ......................................................................... 106
4.5.11.2 Scenario 2 Chimney flue gas effects > 1 km. - Annual Average ............................................ 106
4.5.12 A CHIMNEY STACK EMITTING MATERIAL WHICH IS IN PARTICULATE FORM...... 107
4.5.12.1 Scenario 1 Chimney flue gas effects > 1 km .......................................................................... 108
4.5.13 A SITE EMITTING MATERIAL WHICH IS A VAPOUR AT
AMBIENT TEMPERATURE AND PRESSURE .................................................................... 109
4.5.13.1 Scenario 1 Site fugitive emissions effects > 1 km. - Short duration or Annual average ......... 110
4.6 DAMAGE CRITERIA .................................................................................................................. 111

4.6.1 Prediction of Toxic Effects ......................................................................................................... 111
4.6.2 Prediction of Fire Damage .......................................................................................................... 113
4.6.3 Prediction of Explosion Damage ................................................................................................ 114
4.6.4 Missile Hazards........................................................................................................................... 116
4.7 UNCERTAINTIES IN CONSEQUENCE MODELLING ......................................................... 119

4.7.1 Uncertainties in Modelling Vapour Dispersion........................................................................... 119
4.7.2 Uncertainties in Modelling Fire And Explosion Effects ............................................................. 120
4.7.3 Consequence Modelling Accuracy.............................................................................................. 122
5. HOW TO USE PART 3 - DETAILED SCENARIO MODELLING................... 126
5.1 INTRODUCTION TO PART 3 .................................................................................................... 127
5.2 CONSEQUENCE MODELLING HISTORY ............................................................................. 127
6. SOURCE TERM MODELS ................................................................................... 129
6.1 FLASH FRACTION CALCULATIONS ..................................................................................... 130

6.1.1 Phase Equilibria - Ideal Gases.................................................................................................. 130
6.1.2 Phase Equilibria - Non-Ideal Gases.......................................................................................... 132
6.1.3 The Affect of Temperature on Vapour Pressure ......................................................................... 133
6.1.4 Using Cirrus to Determine Bubble Point (SVP) Conditions..................................................... 133
6.1.5 Using Cirrus to Determine the Proportion of Liquid to Vapour Immediately Following
a Release of a Material from Containment at SVP Conditions.................................................. 137
6.1.6 Flash Fraction Calculations For Multi-Component Materials under SVP Conditions ................ 138
BP CIRRUS User Manual version 7.0 iv

6.1.7 Flash Fraction Calculations For Multi-Phase Materials .............................................................. 140
6.2 SOURCE TERM MODEL - S1 RELEASE FROM A LIQUID TANK................................... 143

6.2.1 A Release of a Single Component Liquid ................................................................................... 143
6.2.2 A Release of a Multi-Component Liquid .................................................................................... 146
6.3 SOURCE TERM MODEL - S8 VAPOUR RELEASE FROM A BOILING OR

EVAPORATING POOL .............................................................................................................. 147
6.3.1 Modelling of a Multi-Component Material using Model S8....................................................... 147
6.3.2 Modelling of a Multi-Component Material using Model S8....................................................... 151
6.4 SOURCE TERM MODEL - S3 RELEASE FROM GAS VESSEL ......................................... 153
6.4.1 Using CIRRUS to Model a Release of Single Component Gas .................................................. 153
6.4.2 Using CIRRUS to Model a Release of Multi Component Gas ................................................... 155
6.5 SOURCE TERM MODEL - S5 RELEASE FROM VAPOUR SPACE OF TWO PHASE
VESSEL ........................................................................................................................................ 159
6.5.1 A Single Component Material at its Saturated Vapour Pressure................................................. 159
6.5.2 A Multi Component Material at its Saturated Vapour Pressure .................................................. 163
6.6 SOURCE TERM MODEL - S6 RELEASE FROM THE LIQUID SPACE OF A

TWO PHASE VESSEL ................................................................................................................ 165
6.6.1 Release of Single Component Liquid at its Saturated Vapour Pressure...................................... 165
6.6.2 Estimating a Vapour Release Rate from the Liquid Release Rate............................................... 168
6.6.3 Release of a Multi-Component Liquid at its Saturated Vapour Pressure .................................... 171
6.7 SOURCE TERM MODEL - S2 RELEASE FROM A LIQUID PIPELINE............................ 173

6.7.1 Modelling of a Single Component Material................................................................................ 173
6.7.2 Modelling of a Multi-Component Material................................................................................. 178
6.8 SOURCE TERM MODEL - S4 RELEASE FROM A GAS PIPELINE .................................. 179
6.8.2 Modelling of a Multi-Component Material................................................................................. 181
6.9 SOURCE TERM MODEL - S7 RELEASE FROM A TWO PHASE PIPELINE .................. 183
6.10 SUMMARY OF SOURCE TERM MODELLING..................................................................... 187
BP CIRRUS User Manual version 7.0 v

7. VAPOUR DISPERSION MODELLING............................................................... 189
7.1 GENERAL CONSIDERATIONS................................................................................................. 189

7.1.1 The Generic Types of Consequence Model ................................................................................ 189
7.1.2 Parameters with Important Implications in Dispersion Modelling.............................................. 191
7.1.2.1 Atmospheric Stability and Windspeed ................................................................................... 191
7.1.2.2 Surface Roughness ................................................................................................................. 193
7.1.2.3 Toxic and Flammable Limits.................................................................................................. 194
7.1.2.4 Sampling Time ....................................................................................................................... 195
7.2 DISPERSION MODEL - D2 DISPERSION OF HEAVY THAN AIR VAPOUR FROM

A LOW MOMENTUM RELEASE AT GROUND LEVEL ..................................................... 197
7.2.1 The Assumptions Inherent in Model D2 ..................................................................................... 198
7.2.2 Examples of the Use of Model D2 - Dispersion following Evaporation from a Liquid Pool...... 199
7.2.3 A Continuous Source Term......................................................................................................... 202
7.2.3.1 A Transient Source Term ....................................................................................................... 207
7.2.3.2 An Instantaneous Source Term .............................................................................................. 210
7.3 DISPERSION MODEL - D1 DISPERSION FROM A MOMENTUM RELEASE................ 213

7.3.1 Dispersion of a vapour which behaves as an Ideal Gas ............................................................ 213
7.3.2 Dispersion of a vapour which is close to its saturated vapour conditions ................................... 218
7.3.3 Dispersion of a contaminant........................................................................................................ 221
7.3.4 Using Model D1 with Multi-Component Materials .................................................................... 221
7.4 DISPERSION MODELS - D3 & D4 DISPERSION OF A VAPOUR WHICH IS

BUOYANT OR NEUTRALLY BUOYANT WHEN RELEASED INTO AIR........................ 225
7.4.1 Emissions From Stack Sources ................................................................................................... 226
7.4.2 Emissions From Area Sources .................................................................................................... 228
7.4.3 Using Model D4 to Estimate the Annual Average Pollutant Concentration Surrounding a Site. 230
7.4.4 Using Model D3 and D4 with Multi-Component Materials........................................................ 233
7.5 A SUMMARY OF VAPOUR DISPERSION MODELS ............................................................ 235
7.6 DISPERSION MODEL D5 DISPERSION OF A PARTICULATE MATERIAL ................ 237

7.6.1 Example of usng model D5 A Particulate Release from a Stack Source.................................. 237
BP CIRRUS User Manual version 7.0 vi

8. POOL FIRE, TORCH FIRE AND FIREBALL MODELS.................................. 241
8.1 FIRE MODEL F1 - POOL FIRE RADIATION ....................................................................... 243

8.1.1 The Assumptions Inherent in Model F1...................................................................................... 243
8.1.2 Example of using Model F1 - A Spill of Refrigerated Propane onto Water. .............................. 244
8.1.2.1 Sooty Flame: .......................................................................................................................... 245
8.1.2.2 Thermal Flux of Interest:........................................................................................................ 247
8.1.2.3 Plane Height of Interest:......................................................................................................... 247
8.1.3 Example of using Model F1 - A Crude Oil Tank Full Surface Fire ............................................ 250
8.2 FIRE MODEL F2 - TORCH FIRES AND FLARES................................................................ 253

8.2.1.1 F Factor ................................................................................................................................. 254
8.2.2 Example of using model F2 - A Refinery Flare .......................................................................... 255
8.2.3 Example of using model F2 - A Liquid Torch Fire..................................................................... 258
8.2.4 Using Model F2 with Multi-Component Materials ..................................................................... 258
8.3 FIRE MODEL F3 - FIREBALL................................................................................................. 261

8.3.2 Example of using Model F3........................................................................................................ 262
8.4 FLASH FIRES ............................................................................................................................... 263
9. EXPLOSION MODELS ......................................................................................... 265
9.1 EXPLOSION MODEL E1 - VAPOUR CLOUD EXPLOSION .............................................. 267

9.1.1 The Assumptions Inherent in Model E1...................................................................................... 269
9.1.2 Example of using model E1 -...................................................................................................... 269
9.2 EXPLOSION MODEL E2 - PRESSURE BURST EXPLOSION............................................ 273

9.2.1 The Assumptions Inherent in Model E2...................................................................................... 273
9.2.2 Example of using model E2 - Overpressurisation of a Pressure Vessel ...................................... 273
9.3 EXPLOSION MODEL M1 - MISSILES ................................................................................... 275

9.3.1 The Assumptions Inherent in Model M1 .................................................................................... 275
9.3.2 Example of using model M1 - Fragment Throw Distance following Overpressurisation of
a Pressure Vessel........................................................................................................................ 275
9.3.3 Example of using model M1 - Fragment Throw Distance following a BLEVE.......................... 278
BP CIRRUS User Manual version 7.0 vii

10. FURTHER EXAMPLES TAKEN FROM RFINERY AND E&P FACILITIES 281
10.1 INTRODUCTION ......................................................................................................................... 281
10.2 A WORST CASE RELEASE FROM AN HF ALKYLATION UNIT STORAGE VESSEL.. 282
10.3 A RELEASE FROM A CRUDE OIL CROSS COUNTRY TRANSMISSION PIPELINE. ... 291
10.3.1 A hole in the top of the pipeline.................................................................................................. 295
10.3.2 A hole in the bottom of the pipeline............................................................................................ 297
10.4 BLOWOUT OF AN OIL WELL.................................................................................................. 299

10.4.1 Gas Wells.................................................................................................................................... 299
10.4.2 Oil Wells..................................................................................................................................... 300
11. LIST OF CONTENTS............................................................................................. 301
BP CIRRUS User Manual version 7.0 viii

Cirrus 7.0 Manual

Uploaded by

Copyright:

Available Formats

Cirrus 7.0 Manual

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Cirrus 7.0 Manual

Uploaded by

Copyright:

Available Formats

CIRRUS version 7.

to download a copy of the latest version

access the CIRRUS helpline

access latest news

Alternatively: for CIRRUS HELPLINE enquiries contact:

Dr Steve Hall Tel +44 (0)1932 76 3574

or send correspondence to:-

Group HSE Resource

1.1 PURPOSE AND DESCRIPTION

CIRRUS is a software package which was developed by BP International

CIRRUS aims to provide a consistent interactive shell around these individual

Microsoft is a registered trademark and Windows is a trademark of Microsoft Corporation.

BP CIRRUS User Manual version 7.0 1

1.2 ABOUT THIS MANUAL

CIRRUS is a Microsoft Windows application, and in using CIRRUS you engage

An INDEX is given at the end of the manual.

BP CIRRUS User Manual version 7.0 2

When installation is complete, CIRRUS is started by double clicking on the

BP CIRRUS User Manual version 7.0 3

2. PART 1 - TUTORIAL FOR NEW USERS

BP CIRRUS User Manual version 7.0 6

BP CIRRUS User Manual version 7.0 7

2.1 A WORKED EXAMPLE

In this example you will model a Non-Momentum Release (Spillage) of Propane

2.1.1 Step 1 - Creating a folder in which to store your data

2.1.2 Step 2 - Choosing the model to use

BP CIRRUS User Manual version 7.0 8

2.1.3 Step 3 - Creating a new Dataset for input to the model

Now create a new Dataset. Select New from the menu:

BP CIRRUS User Manual version 7.0 9

BP CIRRUS User Manual version 7.0 10

Select 'Propane' as shown, then click the continue button.

BP CIRRUS User Manual version 7.0 11

For the example, enter the data as shown below.

2.1.4 Step 4 - Checking data and trying to run the model

BP CIRRUS User Manual version 7.0 12

BP CIRRUS User Manual version 7.0 13

BP CIRRUS User Manual version 7.0 14

2.1.5 Step 5 - Running the model

BP CIRRUS User Manual version 7.0 15

Soon the display will be as follows:

2.1.6 Step 6 - Reviewing the results

BP CIRRUS User Manual version 7.0 16

In this particular case select the log/lin option:

A dialog box will appear:

BP CIRRUS User Manual version 7.0 17

BP CIRRUS User Manual version 7.0 18

The result is a representation of the cloud in equal proportions:

BP CIRRUS User Manual version 7.0 19

BP CIRRUS User Manual version 7.0 20

Provided a suitable printer (e.g. an HP LaserJet) is connected and properly installed in

BP CIRRUS User Manual version 7.0 21

2.1.7 Step 8 - After modelling

BP CIRRUS User Manual version 7.0 22

BP CIRRUS User Manual version 7.0 23

BP CIRRUS User Manual version 7.0 24

BP CIRRUS User Manual version 7.0 25

2.2 OTHER FEATURES

2.2.1 Definition and usage

2.2.2 User level