Handbook For Fire Calculations and Fire Risk Assessment in The Process Industry. 2003.

Scandpower
" SCANDPOWER
Risk Management AS
~ SINTEF
Norwegian Fire Research Laboratory
Handbook for Fire Calculations and Fire Risk

Assessment in the Process Industry
Date 17 September 2003

Handbook for Fire Calculations and Fire Risk Assessment in the Process Industry Page i
TABLE OF CONTENTS
Page
1. INTRODUCTION 1-1
1.1 General 1-1
2. PURPOSE OF THE HANDBOOK 2-1

2.1 General 2-1
3. HOW TO USE THE HANDBOOK 3-1

3.1 Introductory Note 3-1
3.2 Flow Charts 3-1
4. RELEASE OF COMBUSTIBLE FLUID 4-1

4.1 General 4-1
4.2 Release Mechanisms and Leak Sizes 4-2
4.2.1 Introduction 4-2
4.2.2 Rupture of Pipe, Valve, Flange or Instrument Connection 4-2
4.2.3 Leak from Valve, Flange or Instrument Connections 4-2
4.2.4 Leak from Piping 4-3
4.2.5 Rupture of Vessel or Tank 4-3
4.2.6 Leak from Vessel or Tank 4-3
4.3 Continuous Gas Release 4-4
4.3.1 Steady State Gas Release, No Friction 4-4
4.3.2 Time-dependent Gas Release 4-6
4.3.3 Gas Release, with Friction 4-6
4.3.4 Gas Release, Pipeline Rupture 4-7
4.3.5 Gas Release with Condensation 4-8
4.4 Continuous Liquid Release 4-8
4.4.1 Steady State Liquid Release, No Friction 4-8
4.4.2 Time-Dependent Liquid Release 4-9
4.4.3 Liquid Release, with Friction 4-9
4.4.4 Liquid Release, with Flash Evaporation 4-9
4.5 Continuous 2-Phase Release 4-10
4.5.1 Release through Hole in Vessel Wall 4-10
4.5.2 Release through Short Pipe 4-10
4.5.3 Release through long Pipe 4-11
4.5.4 Computer Codes 4-11
4.6 Instantaneous Release of Gas 4-13
4.7 Instantaneous Release of Liquid 4-14
5. BEHAVIOUR OF RELEASED FLUID 5-1

5.1 General 5-1
5.2 Meteorological and Topographical Conditions 5-1
5.2.1 Data required 5-1
5.2.2 Atmospheric Stability 5-2
5.2.3 Surface Roughness 5-3
5.3 Gas Release 5-3
5.3.1 Dispersion of Sonic Gas Jets 5-3
5.3.2 Atmospheric Dispersion of light and neutral Gas 5-5
5.3.3 Atmospheric Dispersion of heavy Gas 5-7
5.3.4 Condensation 5-11
Scandpower
iG»SINTEF Risk Management AS
Handbook for Fire Calculations and Fire Risk Assessment in the Process Industry Page ii
Page
5.4 Liquid Release 5-13

5.4.1 Flash Evaporation 5-13
5.4.2 Boiling Liquid onto Water 5-13
5.4.3 Boiling Liquid on Land, Unrestricted 5-14
5.4.4 Boiling Liquid on Land, Restricted 5-16
5.4.5 Non-Boiling Liquids 5-17
5.5 Underwater Release 5-18
5.5.1 Overview 5-18
5.5.2 Underwater Plume 5-18
5.5.3 Liquid Surface Layer 5-20
5.5.4 Dilution of Gas lighter than Air 5-21
5.5.5 Dilution of Gas heavier than Air 5-22
5.6 Ignition Mechanisms 5-22
5.6 .1 Flammability Characteristics 5-22
5.6.2 Sources of Ignition 5-25
6. GAS FIRE 6-1

6.1 General 6-1
6.2 Flash Fire 6-1
6.2.1 Scenario 6-1
6.2.2 Flame Shape 6-3
6.2.3 Flame Temperature 6-4
6.2.4 Heat Transfer Assessment 6-4
6.2.5 Smoke 6-5
6.2.6 Duration 6-6
6.2.7 Example 6-7
6.3 Jet Fire 6-8
6.3.1 Scenario 6-8
6.3.3 Flame Temperatures 6-17
6 .3.5 Smoke 6-26
6.3.6 Effects of Obstructing the Jet 6-27
6 .3.7 Experience from Jet Fire Experiments 6-28
6.3.8 Results from the "Blast and Fire Project" 6-31
6 .3.9 Example 6-35
6 .4 Diffusive Gas Fires 6-37
6.4.1 Scenario 6-37
6.4.3 Flame and Gas Temperatures 6-40
6.4.5 Smoke 6-40
6.4.6 Example 6-41
7. LIQUID FIRES 7-1

7.1 General 7-1
7.2 Pool Fire in the Open Air 7-1
7.2.1 Scenario 7-1
7.2.2 Pool Fire Diameter 7-2
7.2.4 Flame and Smoke Gas Temperatures 7-9
7.2.6 Smoke 7-11
Scandpow er
Risk M anagement AS (G)) SINTEF
Handbook for Fire Calculations and Fire Risk Assessment in the Process Industry Page iii
Page
7.2.7 Duration of Fire 7-11

7.2.8 Example 7-12
7.3 Pool Fire on the Sea Surface 7-15
7.3.1 Scenario 7-15
7.3.3 Flame Shape, Flame Temperature, Heat Transfer Assessments.
Smoke Production, Duration of Fire and the Effects of Wind 7-19
7.3.4 Hazard Assessment and Thermal Loads from Crude Oil Fires
on the Sea Surface 7-20
7.3.5 Example 7-26
7.4 Pool Fire in an enclosed Area 7-29
7.4.1 Scenario 7-29
7.4.3 Effects of Ventilation 7-30
7.4.5 Flame and Gas Temperatures 7-35
7.4.7 Smoke 7-40
7.4.8 Duration of Fire 7-43
7.4.9 OveNiew Findings from Fire Tests with Confined Pool Fires 7-43
7.4.10 Example 7-46
7.5 Fireball 7-49
7.5.1 Scenario 7-49
7.5.2 Diameter and Height of Fireball 7-50
7.5.3 Temperature of Fireball 7-51
7.5.5 Duration of Fireball 7-53
7.5.6 Example 7-53
8. RESPONSE TO FIRE 8-1

8.1 General 8-1
8.2 Exposure Criteria for Personnel 8-3
8.2.2 Thermal Effects 8-3
8.2.3 Obscuration of Vision 8-7
8.2.4 Toxic Effects 8-9
8.3 Fire Response of Steel Structures 8-17
8.3.2 Properties of Steel at elevated Temperatures 8-18
8.3.3 Methods to calculate Temperature Response of Fire exposed Steel
Structures 8-21
8.3.4 Simple Design of Steel Structures 8-30
8.4 Fire Response of Concrete Structures 8-31
8.4.2 Behaviour of Concrete at elevated Temperatures 8-31
8.4.3 Methods to calculate Temperature Response of Fire exposed
Concrete Structures 8-33
8.5 Glass Fibre Reinforced Plastics (GRP) 8-36
8.5.2 General Fire Properties of GRP Materials 8-40
8.5.3 Performance of selected GRP Products at elevated Temperatures 8-45
Scandpower
~SINTEF Risk Management AS
Handbook for Fire Calculations and Fire Risk Assessment in the Process Industry Page iv
9. FIRE PROTECTION 9-1

9.1 Passive Fire Protection 9-1
9.1.1 Introduction and Philosophy 9-1
9.1.2 Separation by Distance 9-2
9.1.3 Fire Partitions 9-3
9.1.4 Structural Fire Insulation 9-5
9.1.5 Materials for Passive Fire Protection 9-7
9.1.6 Environmental Effects/Explosion Effects on Passive Fire Protection 9-9
9.2 Active Fire Protection 9-11
9.2.2 Properties of Water and Common Fire Extinguishing Media 9-14
9.2.3 Deluge/Sprinkler Systems 9-16
9.2.4 Sprinkler Installation Rules 9-17
9.2.5 Deluge Systems - Water Sprays 9-18
9.2.6 Pressure - Water Flow Rate 9-28
9.2.7 Cooling by Water Droplets 9-29
9.2.8 Water Mist 9-3
9.2.9 Water Concentration and Droplet Sizes for Extinguishment 9-36
9.2.10 Foam Media 9-39
9.2.11 Carbon Dioxide, CO 2 (Inert Gas Systems) 9-40
9.2.12 Halon Systems 9-41
9.2.13 Dry Chemical Powders 9-42
9.2.14 Pyrotechnically generated Aerosols 9-42
9.2.15 Effect of Active Fire Protection Systems on Smoke Generation
and Escaping/ Evacuation Possibilities 9-43
10. SMOKE 10-1

10.1 Methods for Prediction of Smoke Behaviour 10-1
10.2 Smoke Generation 10-2
10.2.2 Mass Flow Rate 10-2
10.2.3 Temperature 10-3
10.2.4 Soot Production 10-3
10.2.5 Carbon Monoxide 10-4
10.2.6 Carbon Dioxide 10-5
10.2.7 Oxygen 10-6
10.3 Smoke Flow Dispersion 10-6
10.3.2 Basis for Prediction of Smoke Flow Patterns 10-7
10.3.3 Results of Flow Field Modelling 10-8
10.3.4 Scaling of Smoke Properties to other Fire Sizes 10-8
10.3.5 Smoke Infiltration into a Building 10-10
10.3.6 Aspects of Smoke Infiltration 10-13
10.4 Impact of Smoke 10-13
10.4.1 Reduced Visibility 10-13
10.5 Smoke Control Systems 10-17
10.5.2 Pressure Differences 10-17
10.5.3 Ventilation Systems 10-19
10.6 Comparison with the Piper Alpha Accident 10-20
Scandpower
Risk Management AS ~SINTEF
Handbook for Fire Calculations and Fire Risk Assessment in the Process Industry Pagev
11. FIRE RISK ASSESSMENT 11-1

11.1 General 11-1
11.2 Overall Approach used in a Fire Risk Analysis 11-1
Appendices:
A: Nomenclature
B: Properties of Steel, Concrete and insulating Materials at elevated Temperatures
C: Calculation of Section Factors
D: Supplementary Models on Release Calculations
D1. Effect of Pressure Loss on Leak Rates
D2. Gas Condensation
D3. Large non-propagating Crack in a Pipe
E: Hazard Checklist
F: Characteristic Mechanical and Physical Properties of GRP Materials
G: Temperatures and Velocities of Buoyancy-controlled Diffusion Flames
H: Heat Transfer from Buoyancy-controlled Diffusion Flames
I: Geometric View Factor
J: Smoke Production from Buoyancy-controlled Diffusion Flames
Scandpower
Handbook for Fire Calculations and Fire Risk Assessment in the Process Industry Page 1-1
1. Introduction
1.1 General
This "Handbook for Fire Calculations and Fire Risk Assessment in the Process
Industry" has been prepared by Scandpower A/S and SINTEF-NBL. The Centre for
Industrial Research (SI) has assisted in preparation of Chapter 8.5 relating to Glass
Fibre Reinforced Plastics (GRP).
The project has been sponsored by:
NTNF
Conoco Norway Inc.
Elf Aquitaine
Gaz de France
Ruhrgas
Gasunie
SINTEF-NBL
Scandpower.
The book has been prepared during the period 1990-1991 . The second edition of the
book was released in January 1994. This third edition was released in September
1997. It is the hope of the participants in this project that this book will contribute to
increased fire safety throughout the process industry, offshore as well as onshore.
We also wish to express our thanks to the sponsors of this project who did not only
contribute with financial support, but also by playing an active role in the preparation of
this book.
Kjeller, 16 September 2003
B.I. Bakken
(President)
Scandpower
~ SINTEF Risk Management AS
2. Purpose of the Handbook

2.1 General
The "Handbook for Fire Calculations and Fire Risk Assessment in the Process
Industry" is intended as a practical guide for engineers working in the field of Loss
Prevention and Risk Analysis.
Risk Analysis is a widely recognized and well established tool to identify, quantify and
reduce risk. In the petroleum, refining and chemical industry fires are a major risk
contributor. The physical phenomena leading to a fire and of the fire itself have to be
properly understood and quantified in order to carry out a valid, quantitative risk
analysis and to define adequate risk reducing measures. This handbook intends to
help in this process.
The Handbook does not address questions of probability, for example the probability
for occurrence of leaks based on historical experience. In other words, the models are
"deterministic", not "probabilistic". Given certain starting conditions - the probability for
which must be estimated by some other means - the consequences can be described
using the models of this Handbook.
The economic incentives for performing quantitative risk analysis and, in this way, to
reduce both the probability for and the consequences of accidents, are strong. A cost-
benefit analysis should evaluate the net present value (NPV) over the whole lifetime of
an installation which results from
reduction of repair costs

reduction of production downtime
reduction of insurance premiums
in order to arrive at the amount of money which can - and should - be invested in risk
reducing measures. These amounts can be very high. In addition there are such
benefits which cannot easily be expressed in money value, such as reductions of per-
sonnel injuries and environmental damage.
Scandpower
IGJ)SINTEF Risk Management AS
3. How to use the Handbook

3.1 Introductory Note
This "Handbook for Fire Calculations and Fire Risk Assessment in the Process
Industry" is intended as a practical guide for engineers who have to assess the con-
sequences of accidental release of flammable materials.
Flow charts are used to guide the analyst through the different steps of the assess-
ment, and to point to the sections of the Handbook where the calculation procedures
can be found. The flow charts are found in Chapter 3.2.
Whenever possible simple calculations which can be carried out on hand-held calcu-
lators are given. Where simple calculations are not available or where they produce
too inaccurate results, references are made to recognized computer codes. Again,
simple codes which run on PCs are given preference over more complicated codes
requiring main frame computers.
3.2 Flow Charts

Four different flow charts are shown on the following pages:
Chart 1: Starting Chart and Underwater Release

Chart 2: Liquid Release
Chart 3: Gas Release
Chart 4: Ignition and Fire
The charts are self-explanatory, but some comments should be made:
Chart 1: Starting Chart and Underwater Release
All assessments start at the top of this chart.
Question 1. 1: Flammable Material?
This handbook does not include fires in solids (e.g. wood), pastes or powder, and fires
in such materials are therefore sorted out at this stage.
Question 1.2: Release?
The phenomenon of a fireball or BLEVE (boiling liquid expanding vapour explosion) is

a special one since it involves non-released fluid, i.e. fluid still contained in a vessel. It
is therefore the only phenomenon under the "no"-exit from the release question. How-
ever, quite often a fireball is triggered as a secondary effect by an external fire caused
by released material.
Scandpower
(t;J) SINTEF Risk M anagement AS
Page 3-2 Handbook for Fire Calculations and Fire Risk Assessment in the Process Industry
Question 1.3: Under Water?
Releases under water, e.g. from blowouts or pipeline leaks, are discluded from this
question since special models are required to describe the behaviour of fluids that are
released under water and which rise to the surface.
Question 1.4 and 1.5: Fluid?
Type of fluid (gas or liquid) refers to fluid conditions before release. For vessels con-
taining both gas and liquid the location of the leak (above or below liquid level) deter-
mines the branching output. For pipelines operating in 2-phase mode, the liquid
branch should be chosen.
Question 1.6: Vapour Pressure?
If the vapour pressure is higher than the external hydrostatic pressure flashing will
occur already in the leak cross section and 2-phase release calculations have to be
performed.
Question 1.7: Vapour Pressure?
If the vapour pressure is low in question 1.6 (lower than hydrostatic pressure) but
higher than atmospheric pressure, the "high" exit is not to be included here. Flashing
will occur as the liquid rises to the surface and flash calculations have to be carried
out. The gas and the liquid which reaches the surface will be treated in separate
models.
Chart 2: Liquid Releases
Question 2. 1: Vapour Pressure?
This question decides if 2-phase release calculations have to be performed in order to

determine the release rate. In most cases of interest the release will be to atmos-
pheric pressure (0.1 MPa) and ambient temperature (appr. 15 °C). In principle, all
liquids boiling at these conditions will be released in 2-phase mode. In practice boiling
will be more or less vigorous. Using release calculations for 1-phase (liquid) flow will
produce somewhat high release rates. In most applications this will be a pessimistic
(conservative) approximation.
Question 2.2: Onto?
A distinction is made, at this stage, between release to land and to water because the
subsequent behaviour of the liquid will depend on this. In some cases liquid may be
released first to land, and then run off to water (e.g. spills on jetties). In most of these
cases it will be conservative to assume direct release to the water. In other cases
(overflow of bunding to sea) it will be necessary to make a split of the total released
liquid and to follow both routes.
Question 2.3: Shape?
The shape (bounded or non-bounded) will determine the area of contact between
liquid and ground.
Scandpower
Handbook for Fire Calculations and Fire Risk Assessment In the Process Industry Page 3-3
Question 2.4 to 2.6: Release Mode?
The size of liquids pools formed, either on land or on water, will vary with time and
depend on the release mode. Calculations can be carried out for two simplified situa-
tions:
Instantaneous release of a given amount (kg) of liquid

Steady-state release at constant release rate (kgls).
Most situations can be approximated by one of these two simplifying assumptions.

Models taking into account explicitly the time dependence of release might be required
in special cases. Such models are not included in this handbook, at present.
Question 2. 7: Evaporation?
If the liquid pool is not ignited and if the vapour pressure of the liquid is high (boiling is
not necessary), gas will evaporate from the surface and possibly form a gas clouq.
The decision whether to choose the "Yes" or "No" exit depends on the hazard poten-
tial of the resulting gas cloud. This in turn depends on the type of gas being formed,
the evaporation rate, the meteorological conditions, the topography etc. No simple
rule can be given. The safe approach is to choose "Yes" and to evaluate the gas
cloud hazard.
Chart 3: Gas Releases
Question 3. 1: Release Mode?
Instantaneous and continuous release modes are both simplifications of the real be-
haviour. Instantaneous release may be assumed if the release duration is short com-
pared to the time required to establish the dispersion phenomenon UeUcloud) following
the release.
The time required to establish a jet is the time it takes for a gas molecule to travel from
the release point to the point where the gas concentration equals the lower flamm-
ability limit (LFL). This is usually 1 second for a small yet (10 kg/s) and 10 seconds for
a large jet (1 ,000 kg/s).
For a gas cloud the formation time is related to the wind speed and the downwind dis-
tance to LFL. This time will vary between 10 seconds (10 mis, 100 m) and 500
seconds (2 m/s, 1,000 m).
The criterion for selecting an instantaneous or a continuous release model depends

therefore on the type of dispersion, jet or cloud , to be expected. For a jet release
duration < 5 seconds should be treated as instantaneous, for a cloud duration
< 50 seconds.
Question 3.2 and 3.3: Condensation ?
Condensation can occur if the temperature drop due to the Joule Thomson effect is
sufficiently large. The formed liquid may either
Form a spray (together with the gas)

Drop out as a liquid pool.
Scandpower
Noiwegian Fire Research Laboratory
If the formation of a liquid is believed to be substantial a condensation calculation has

to be carried out to determine the amount of liquid. The distribution of the liquid bet-
ween spray and pool depends on the gas release speed, the surroundings, the sur-
face tension of the liquid and other factors thus making it difficult to predict. (We will try
to establish a rule-of-thumb for this later.)
Question 3.4: P2 > Pent?
Different formulae are used depending on the release velocity. The release velocity is
sonic if the inside pressure P2 is
Eq. (3.1)
where
p1 = outside (ambient) pressure
lj/ = -2-)'~' "'0.55 for natural gas

( y+1
Question 3.5: As a Jet?
The geometry of the release determines if the gas should be treated as a jet or as a
cloud. Gas released from the end of a broken pipe or orifice, without interaction with
obstacles in the surroundings, can develop in a jet-like fashion. Whether a gas re-
lease should be considered as a jet or not will also depend on the speed of the gas at
the point of release. otherwise treatment as a cloud is more appropriate.
Question 3.6: Steady State?
Before selecting the cloud dispersion model a decision must be made if the situation
can be described by a steady-state model or not. If the source term varies strongly
during the duration of the hazard condition a time-dependent model is more appro-
priate. A typical example is evaporation of gas from a cryogenic liquid (e.g. LNG)
spilled on solid ground. The evaporation rate will decrease quickly as the ground gets
colder. Using the initial, high evaporation rate in a steady-state model would give very
pessimistic results.
Chart 4: Ignition and Fire
After completion of the calculations in Charts 1, 2 and 3, all evaluations start again at
the top of Chart 4.
Ignition models are used to calculate the ignition probability. If the ignition probability
is zero or very small (::: 10-3) the evaluations can be terminated at this point. The
question of a possible toxic exposure must of course be evaluated separately. This is
not included in this Handbook.
Question 4.1: Fluid?
See Question 1.4 and 1.5, Chart 1.
Scandpower
Risk Management AS IGi)SINTEF
Question 4.2: Onto?
See Question 2.2, Chart 2.
Question 4.3: Open or enclosed?
For pool fires it is important if the liquid is burning inside an enclosure or outside, in
the open air.
Question 4.4: Shape?
The base area of a pool fire will depend on the shape {bounded or non-bounded) of
the ground onto which the liquid is released. See also Question 2.3, Chart 2.
Question 4.5: Release Mode?
For an instantaneous release of gas a flash fire will result, while for a continuous
release the form of the fire will depend on the release speed, see the next question.
Question 4. 6: Release Speed?
For sonic release, i.e. p2 > P crit, the release speed is high. This is true for most cases
of direct release of pressurized gas. For gas being formed by evaporation of spilled
liquid the release speed is low. One may also compare the release speed to the pre-
vailing wind speed when deciding the answer to this question.
Froude number (Fr): The non-dimensional Froude number is a relation between iner-
tia and gravity forces. It is considered that a high Fr. no, above 100,000, indicates a
domination of inert forces, a jet fire. And a low Fr. no, below 20,000, indicates domi-
nation of gravity forces, a buoyancy controlled gas fire.
Scandpower
Legend:
<>
START
Question
D Intermediate
calculation Solid/ paste/ powder
D Model
calculation
Q Transfer lo
other chart
Release?
No
0 Termination
Yes
Fireball
Chaptcr7.5
1.3
No Under- Yes
water?
Gas Liq1..~d Gas

Fluid Fluid
Release
calculation
Chapter 4.3
No
liquid
Flash
calculation Gas
Liquid spread Bubble plume

on model
sea surface
Chapter 5.5 Chapter 5.5
CHART No.1
Starting Chart and
Underwater Release
Scandpower
Release
calculation
Chapter 4.4
Liquid
Sea
Non-bounded
Continuou s Continuous
Chapter 7.2.2 Chapter 7.2.2 Chapter 7.2.2 Chapter 7.2.3 Chapter 7 .32 Chapter 7.3.2
Liquid Pool Models
Evaporation
calculation
CHART No. 2
Liquid Release
Scandpower
~ SINTEF Risk M anagement AS
<Pa11
>P<:rit
Sonic
Release
calculation
release (2-phase)
Gas cloud with

instant source
Chapter 5.3.2
No
Gasdoud
Gas cloud with with time~ Subsonic Sonic
steady state dependent Gas jet Gas jet
source source
Chap:er 5.3.2 Chapter 5.3.2 Chapter 5.3.1 ChaptorS.3.f
CHART No. 3
Gas Release
Scandpower
Risk Management AS IGJ)SINTEF
No fire hazard.
Chemical/ toxic
hazards not included
Liquid Gas
Sea Land
Instantaneous
Contin-
eous
Pool fire Open
on sea
Flash
Chapter 7.3 fire
Chapter 6.1
Pool fire in
enclosed area
Chapter 7.4
Hioh
NM-bounded
Low
Open pool fire Open pool fire Buoyancy

within without controlled Jet fire
bounding bounding gas fire
Chapter 7.2 Chapter 7.2 Chapter 6.3 Chapter 6.2
CHART No. 4
Ignition and Fire
{GJ) SINTEF Scandpower

Risk M anagement AS
4. Release of Combustible Fluid

4.1 General
Most fires in the process industries result from the uncontrolled escape of a combus-
tible fluid, followed by ignition.
There is a large number of situations which might result in a fluid release from a pipe-
line or a process plant. In a fire risk assessment it will not be possible to include all
possible leak scenarios. The evaluation will normally be based on a limited number of
typical situations. It is, however, important to select the typical situations so as to
properly represent the variety of leak situations in real life. (For estimation of leak fre-
quencies it will of course be necessary to assess the probability of all identified leak
causes and to assign to it a typical leak size).
Leak causes may be divided in two categories:
1. Leak due to a component failure.

2. Leak due to maloperation.
Failure of a component may be rupture or partial fracture of pipes or vessels, leaks

from flanges or other types of joints, leaks from instrument connections and valves
etc.
The cause of the leak may be
external impact (including domino effect from other initiating accidents)

material defect
construction failure (weld defects etc.)
fatigue
corrosion
inadequate maintenance
combinations of the factors above.
Leaks due to system maloperation are also important in a fire risk assessment. A
leak may result from e.g.:
Maloperation of drain, vent or sample lines connected to pressurized compo-

nents
Valves left in wrong position after maintenance/testing etc.
Overfilling or overpressurization causing excessive venting or liquid overflow.
The failure mode and the type of component that is subject to failure will determine the
associated leak size. This will be further discussed in Chapter 4.2.
The fluid composition and its state (liquid or gas), the release rate and duration, as
well as the surroundings, will all affect the resulting fire scenario. Whether the release
is gaseous or liquid or significant flashing occurs will also affect the likelihood of igni-
tion and the time elapsed from the initial escape to ignition. This latter aspect will
influence the risk of vapour explosion (confined or non-confined) in case of ignition.
These considerations are further described in Chapters 4.3 and 4.4.
Scandpower
4.2 Release Mechanisms and Leak Sizes

4.2.1 Introduction
The range of leak sizes within a process system will vary from minor pinhole leaks to
rupture of large process vessels or storage tanks. To enable an efficient consequence
evaluation the number of leak sizes must be reduced to a reasonable number. In the
following the criteria for selecting these typical leak sizes will be discussed on a
general basis. For the evaluation of an actual plant the leak sizes will have to be
adjusted to reflect the specific features of that plant (e.g. pipe and vessel dimension,
pressure rating etc.).
4.2.2 Rupture of Pipe, Valve, Flange or Instrument Connection

For the failure mode "rupture" the leak size will simply be the cross-section of the
actual component. In some cases the ruptured component may be connected to two
reservoirs (e.g. a pipe combining two process vessels). In such cases the leak size
may be twice the cross-section of the ruptured component, depending on the use of
check valves. In practice the different components may be grouped in a few cate-
gories depending on size as shown in Table 4.1.
Tabe4.1 : Typical Leak Sizes due to Rupture in Process p·Ip1ng

. an d Joints
.
Diameter of Leak size Typical components
component (cm 2)
3/4" 3 Instrument connections etc.
2" 20 Drain, purge and vent lines. Minor process piping
320 Major process piping and large relief and vent con-
nections
The actual number of leak size groups as well as the leak sizes themselves must be
selected to reflect the design of the plant being examined. Normally 3-5 leak size
groups will suffice.
4.2.3 Leak from Valve, Flange or Instrument Connections

Normally, a valve is considered to consist of two pairs of flanges in addition to the
valve body (unless the valve is welded). External leaks from the valve body may
occur via the valve stem. However, such leaks are considered to be small (a few
mm2) and are normally not included in general fire risk assessments. From a leak
point of view valves are hence treated as two pairs of flanges.
A flange leak may be due to a gasket failure, or to a faulty mechanical connection of

the flange. The typical flange leak size is difficult to assess and will vary with flange
type and size as well as gasket type. For large flanges (> 8 - 10") it has been
common practice to assign a leak area of 2 cm2 corresponding to a situation where
::::: 1/4 of the gasket is blown out. For flange leaks from minor valves and instrument
Scandpower
Risk M anagement AS ~ SINTEF
connections it is normally considered that these will be very small. They are not
considered to represent a major hazard and are often neglected in fire risk analyses.
4.2.4 Leak from Piping

A leak may occur from a pipe due to several reasons (material defect, corrosion,
external impact etc.). The typical leak area of a pipe may conservatively be assessed
as the largest non-propagating crack of the pipe. The largest non-propagating crack
in a pipe has been evaluated by applying a fracture mechanical approach in Appendix
D3 and may roughly be calculated as:
A= 0.475 · R · t
where
A = leak area (m 2)
R = pipe radius (m)
t = pipe wall thickness (m)
4.2.5 Rupture of Vessel or Tank

In case of rupture of a vessel or tank the contents will be released almost immediately.
In addition the supply lines to the tank may feed the leak until successful isolation has
been accomplished. This latter contribution may be assessed based on the process
flow diagrammes (PFDs), the piping and instrument diagrammes (P&IDs) as well as
the operating procedures and shut down logic.
4.2.6 Leak from Vessel or Tank

Leaks from a pressurized storage tank or process vessel may vary in size depending
on the cause of the leak. Quite often the fracture of one of the nozzles will be the
cause of a leak, and in this case it is the nozzle size which determines the leak size.
Leaks from the vessel itself are rare. However, as for pressurized piping there will be
an upper limit of the leak (crack) before the vessel ruptures completely. This upper
leak size will be a function of the vessel type, size, pressure, wall thickness, material
properties etc. and must be determined by fracture mechanics of the actual vessels.
However, as an example, we refer to a study performed by SRO, Ref. /4-1/ which eva-
luated the critical crack size for a typical 100 t propane storage vessel. Three different
fracture mechanical models were used in this study. For the selected (typical) design
the critical crack size was found to be in the range 100 - 300 mm. By assuming a
length to width ratio of 10:1 the corresponding leak area will be in the range of
10 - 100 cm 2 •
Scandpower
<;J)SINTEF Risk Management AS
4.3 Continuous Gas Release

4.3.1 Steady State Gas Release, No Friction
The leak rate of gas from a pressurized storage vessel or a pipeline will depend on the
following factors:
Leak size
Gas composition and physical properties
Pressure of gas
Temperature of gas
Volume and geometry of the containment
Whether condensation will occur or not.
The question of leak size has been discussed in Chapter 4.2.
If the pressure drop at the release point is critical, i.e.
-P1 < ( -2-

P2 - Y + 1
)r~1
Eq.(4.1)
where
P1 = ambient pressure
P2 = pressure inside containment
y = cpf Cv ratio of spec. heat
the velocity of the escaping gas will be sonic at the leak point.
The sonic gas release rate may then be calculated as
2 ); -+~ P2
m,, = A · Co Y( y + 1 · JT ~ R Eq. (4.2)
where
mog = initial gas leak rate (kg/s)

A = release area (m 2 )
Co = coefficient of discharge
y = cp/Cv, gas spesific heat ratio
Cp = specific heat of vapour, constant pressure (J/kg °C)
Cv = specific heat of vapour, constant volume (J/kg °C)
P2 = pressure inside containment (N/m2 )
T = temperature in containment (K)
R = gas constant= 8,314 (kg m 2/K kmol s2 )
M = molecular weight (kg/kmol)
The discharge coefficient will be a function of the leak geometry and the Reynolds
number (Re) of the leaking fluid. The discharge coefficient for a sharp edged orifice is
Scandpower
Risk M anagement AS ~SINTEF
plotted in Figure 4.1 as a function of Reynolds number and restriction ratio of the
orifice.
For the case that
-P1> ( -2-
P2 y+1
J'\
one has a subsonic release. A correction fator lj/ has to be included in Eq. (4.2),
defined as
y+1
-2..(.L'.:.!)r-i'
y-1 2
1-00-----,-,-,-,-,-,---------~~~=-~~-1'"''",,...,.,,
SHA.,P-EOGED ORIFICE
01;..ME!ER RATIO
OR!f!CE. CIA. j i
iNS1DE ?!PE DIA.
~1"'< O·COr----,--~-,-,-',,'-r:-
:c
V
"'
G0-70
~
2 0·o0
w
V I.'•
: l ~\
' ..
. i:
: ; i :i
, i 'I
ii•, r •
'1, •,i
Ill
: ! ! t,i_· : ! I I!!1
.°t J!i
O·JO ~--+--'--'--'-"-+.l--,--'--,-'-'~-.!.',c!+'.,...-+--'--+-4'.!.'"-'',....-,----'----,--+-,
10 20 40 o0 . ,Cl id . 10.:. 10
5
R:YNOLDS ;--;c~srn THP.0liGH OR!F!C::
Figure 4.1: Discharge Coefficient for Orifices, Ref. /4-2/
Scandpower
For risk analysis purposes and accidental leak situations a discharge coefficient as for
nozzles may be applied. In accidental leak situations the leak rate will normally be
high and the Re-number will be > 10 . Based on this it may be concluded that for all
practical purposes a discharge coefficient of
Co = 0.62
may be considered adequate.
4.3.2 Time-dependent Gas Release

Eq. (4.2) is valid only for release from containments where the energy loss due to
friction is negligible (e.g. from storage tanks). The equation is also only valid for the
first minutes of the release, after the onset of the leak. As the leak continues the
pressure inside the vessel will decrease, as will the leak rate. The release rate decay
will basically be dependent on two factors:
Initial leak rate

Mass of gas inside the vessel.
The decay of the leak rate will be exponential and can roughly be described by the
following equation:
mog
mg = mog · exp ( - - t) Eq. (4.3)
w
where
mg = leak rate as function of time (kg/s)

mog = initial leak rate (kg/s), see Eq. (4.2)
w = total mass of gas in the vessel (kg)
t = time after onset of the leak (s)
Note: Any feed to the vessel after onset of the leak would delay the decay of the leak
rate.
4.3.3 Gas Release, with Friction

The Eq. (4.2 and 4.3) above are best suited for calculation of leak rates from pressure
vessels. However, the formulas may be applied for estimates of leak rates from pipes.
The formulas would in this case tend to overestimate the leak rates since the friction
loss is neglected. The error due to this simplification will be large for long small dia-
meter pipes, and small for short large diameter pipes. For further details on the effect
on friction loss reference is made to Appendix D.1 .
The gas leak rates may also be calculated by use of the pressure-enthalpy diagramme
of the gas in question, see Appendix D.2.
Scandpower
Norwegian Fire ResearchLaboratory
Handbook for Fire Calculations and Fire Risk Assessment in the Process Industry Page4-7
4.3.4 Gas Release, Pipeline Rupture

Full rupture of a pipeline is a special case since the pipeline can be considered as a
long, thin vessel, and friction effects are very important. Following the rupture the
pressure drop propagates with the speed of sound along the pipeline and the release
rate out of the broken end of the pipeline drops quickly.
An approximate formula has been given by Fannel121p (Ref. /4-3/)
q= A-M
R-T
~ (~+b•C·
.,ft
.,ft) Eq. (4.4)
where
q = release rate (kg/s)

A = cross sectional area of pipeline (m 2)
R = gas constant= 8,314 Nm/kmol K
M = molecular weight (kg/kmol)
C = ./r·R-T/M =speed of sound (mis)
y = epic,, ~ 1.3
T = ambient temperature ( K)
Rr = D/(4-f) (m)
D = int. diameter of pipe (m)
f = friction factor(~ 0.002)
a = pressure at fracture location, prior to fracture (N/m 2 )
b = pressure gradient, prior to fracture (N/m 2 m)
t = time since fracture (s)
This formula obviously overestimates the release rate in the very early phases be-
cause q ➔ oo as t ➔ 0.
The releases from both ends of the ruptured pipeline have to be calculated using Eq.
(4.4), and the two terms to be summed in order to get the total release rate.
For certain consequence evaluations it is not the instantaneous rate of release, but the
amount of released gas up to time t which is of interest. This can be obtained by inte-
gration of Eq. (4.4) yielding
Q2A·
=- M- - ~ ( 3a-t "-2 +b-c-t"'")
2 Eq. (4.5)
3R-T
Comparing results obtained with Eq. (4.4) with those obtained with more refined com•
puter codes it is seen that deviations of ± 20 % must be expected for t > 5 seconds.
Fort~ 5 seconds Eq. (4.4) should not be used.
For t > a/(b-c) seconds the correction term for the pressure gradient (i.e. the second
term in the paranthesis of Eq. (4.4) should be neglected.
If higher accuracies are required one of the more advanced computer codes such as
NORA-HC (Ref. /4-4/) should be used.
Scandpower
4.3.5 Gas Release with Condensation

High pressure gas released to atmosphere will experience a temperature drop due to
the Joule-Thomson effect. If the temperature drop is sufficiently large liquid may be
formed by condensation. In this case most of the liquid will take the form of small
droplets which are carried away by the high speed gas stream , warmed up by mixing
with air, and re-evaporated quickly. But in some cases local conditions may be such
that the liquid drops out, to form a pool on the ground.
For pure gases the amount of liquid formed can be determined from the pressure-
enthalpy diagramme of the gas. This is described in Appendix D.2, where methane is
used as an example. For multicomponent gases, e.g. associated gas from oil pro-
duction, a process simulation code such as PRO/II (Simulation Science Inc.) may be
used to calculate condensation.
4.4 Continuous Liquid Release

4.4.1 Steady State Liquid Release, No Friction
For a leak from a process vessel or storage tank the liquid velocity inside the tank and
friction can be neglected. The general Bernoulli equation
Eq. (4.6)
where
p = pressure (N/m 2 ) index 1: inside

p = density (kg/m 3 ) index 2: outside
v = flow velocity (m/s)
Z = static head (m)
g = gravity constant (9.82 m/s2 )
can then be simplified to
Eq. (4.7)
where also incompressibility of fluid has been assumed (p 1 = p2 = p). The liquid dis-
charge rate is
m, =Co· A · p2 • V2 (kg/s) Eq. (4.8)
where
A = leak area (m 2 )
C0 = discharge coefficient
Scandpower
Risk M anagement AS (G})SINTEF
Norweg ian Fire Research Laboratory
Inserting for v2 from the simplified Bernoulli equation one obtains
Eq. (4.9)
4.4.2 Time-Dependent Liquid Release

Eq. (4.9) shows that the release rate depends on p1 , the pressure above the liquid,
and LiZ, the vertical distance between the liquid surface and the discharge point. Both
will decrease as the leak continues.
The pressure inside the vessel is not a straightforward matter to determine. In a

storage tank containing liquid with dissolved gas (e.g. crude oil with a high vapour
pressure), gas will flash off as the pressure drops due to liquid outflow, and this may to
some extent compensate the pressure loss due to liquid outflow alone. In a storage
tank containing partly refrigerated liquid (e.g. propane at 5 bara and 2.5 °C) the liquid
is superheated. Upon loss of pressure due to a small leak the liquid inside the con-
tainment will evaporate rapidly to maintain the thermodynamic equilibrium of propane
at the actual temperature. In such a case the internal pressure of the containment as
applied in Eq. (4.9) can be considered constant.
For the purpose of risk analysis it will be a conservative approach to assume a con-
stant discharge rate until the vessel is completely empty.
4.4.3 Liquid Release, with Friction

If the liquid leak occurs through a pipe fitted to a vessel the energy loss due to friction
may become significant. Eq. (4.9) would then have to be extended to include the fric-
tion loss:
- · A ·P 2[cP,-P2)
m,-co •z ·
- p - +,, g-p
Lipf] Eq. (4.10)
where Lipr depends on the linear velocity and, hence, on the discharge rate. Some
iterations may be necessary in order to arrive at the final solution.
4.4.4 Liquid Release, with Flash Evaporation

Liquid stored or transported under pressure can, if released to atmosphere, flash to
the vapour phase. For a pure liquid the fraction being evaporated can be calculated,
assuming adiabatic conditions, i.e. no heat input from the surroundings.
The heat available in the liquid for evaporation is roughly
Cp1 ·(TR-Ts)
where TR= release temperature, T 8 = boiling temperature, c~ = heat capacity of liquid.

If this - and only this - heat is used for evaporation, the fraction of liquid evaporated is
Scandpower
f -1
- -exp [ • .....;....
Cpl _ (TR-Ts
__ ) ] Eq. (4.11)
hv
where hv = heat of vapourisation (kJ/kg).
Example: For TR =3.5 °C and propane (T8 =- 42.1 °C, cp1 = 2.34 kJ/kg °C,
hv = 427 kJ/kg) one finds
f = 0.34 (34 %)
This is not quite correct since cp1 is not temperature independent. It is better to use the
difference of internal energy at the release temperature and at the boiling temperature
in the nominator. For the example above this results in a slightly higher value,
f = 0.36 (36 %)
In reality f will be still higher due to heat influx from the surroundings. Depending on
the detailed geometry of the leak, the flow conditions etc. part of the liquid will be in
the form of small droplets or mist, and they will evaporate more quickly due to heat
supplied by the surrounding air.
For liquids composed of several components the flash calculations are more compli-
cated and one of the computer codes for process simulation, e.g. PVTOLGA, has to
be used.
If flashing does occur it will occur in and around the leak opening , leading to 2-phase
flow conditions. This complicates the calculation of leak rates. Some approximate for-
mulas and references to computer codes are given in the following chapter.
4.5 Continuous 2-Phase Release

4.5.1 Release through Hole in Vessel Wall
If the release is through a hole in the vessel wall itself the 2-phase effects will be small
and can be neglected. This means that Eq. (4.9) can be applied without correction.
4.5.2 Release through Short Pipe

If the release is through a pipe with a length/diameter ratio L/0 :s; 3, Eq. (4.9) can be
modified according to Fauske to:
_ · A ·pr ·
m-Co [2¥,P
-1-Pr
- Eq. (4.12)
Pr
where the symbols are as before, but the index r refers to the release point. (The
static head contribution is neglected).
Scandpower
Risk Management AS ~ SINTEF
Experience has shown that
Pr= 0.55 · P1· (1 - e-u30 )
gives good agreement between Eq. (4.12) and experiments (p 1 and Pr are gauge
pressures). The fluid density Pr at the release point is some average density
Eq. (4.13)
where
f = mass fraction of vapour formed

p9 = density of gas (kg/m )
pr = density of liquid (kg/m )
The mass fraction of vapour formed, assuming an adiabatic one-component flash, can
again be obtained by Eq. (4.11), but where the atmospheric boiling temperature T 8 is
replaced by the equilibrium temperature Tr corresponding to the release pressure,
using vapour pressure curves.
4.5.3 Release through long Pipe

If the release is through a pipe with a length/diameter ratio in excess of UD = 12, the
same Eq. (4.12) can be used, but the expression for Pr is now simply
Pr= 0.55 · P1
4.5.4 Computer Codes

This chapter will, in general terms, discuss some available computer codes for calcu-
lation of two-phase flow phenomena.
In two-phase flow, the fluid density will be a function of the local phase fractions as
well as pressure and temperature.
The critical velocity is also a function of the local phase fraction, the pressure, the tem-
peratures and the degree of interaction between the phases. With interaction is here
understood both momentum, mass and energy transfer.
In a leak situation, the velocities will probably be high with a rather chaotic flow situa-
tion which means a high degree of momentum transfer between the phases. The fluid
will expand rapidly in the direction of flow and the expansion rate is controlled by the
mass- and energy transfer. The expansion rate again controls the flow rate through
the acceleration pressure drop. In case of a broken pipe (guillotine break) maximum
velocities may be achieved at the location of the leak. Whether critical flow conditions
are achieved or not will depend on the local flow conditions at the leak. In two-phase
flow a critical flow velocity can be defined in different ways, according to the
assumptions made concerning momentum-, mass- and energy transfer.
Scandpower
IQ»SINTEF Risk Management AS
Norwegian Fire Research laboratory
In Figure 4.2 is illustrated how the pressure wave propagation velocity, c, varies with
the gas volume fraction , ex.
c; '
'U
Q)

a.
Cl)
(1)
C
0
;::;
;-
(ti
C)
ffl
a.
0
a:.
0 0.5 1
Void Fraction
Figure 4.2: Pressure Wave Propagation Speed as Function of Void Fraction a.
Curve 1 is the result of assuming no interaction between the two phases. A curve
similar to curve 1 would be the result in a separated flow situation for small amplitude
pressure waves.
In curve 2 complete momentum transfer between the phases (homogenous flow) is

assumed, but no mass or energy transfer. This is similar to the situation where the
two phases are well mixed, but the pressure changes are fast compared to the mass-
and energy transfer rates. This again would be close to the situation of choked or
critical two-phase flow.
Curve 3 is the result of assuming complete momentum, mass and energy transfer
between the phases. This curve shows how large the compressibility of a two-phase
fluid is compared to each of the single phases. Note that this curve is discontinuous
for void fraction equal to zero.
In a practical critical flow situation values of C would be observed somewhere bet-

ween curve 2 and 3 for lower and medium void fractions, probably closer to 2 than to
3. For higher void fractions where the gas is the dominant and continuous phase,
observations would be somewhere between curve 1 and 2.
Scandpower
The triangles, /',., are examples of typical experimental results.
The gas volume or void fraction influences strongly the critical velocity and so does
pressure, temperature and the degree of non-equilibrium between the phases. These
are parameters that must be known locally at the leak position. They can only be
determined through rather complicated flow calculations. In the case of tank ruptures
or in flow situations which are far from one-dimensional, simplified methods have been
developed (Ref. /4-5/) for calculating critical two-phase flow, but no verification of
these methods are known to us.
In situations where one-dimensional flow can be assumed, practical tools in the form
of computer codes have been developed for the calculation of critical two-phase flow.
Some of the most commonly used tools are discussed in Ref. /4-6/ with emphasis on
the computer codes used within the nuclear industry. The main conclusion is that the
calculation of critical two-phase flow is rather unreliable, especially the scaling up from
experimental results to a real situation. One therefore has to make rather conserva-
tive approximations in the calculations.
A rather extensive cooperation on nuclear safety took place within the Nordic coun-
tries from 1975 to 1983. Both experimental and theoretical studies were performed.
As a result of this work, a computer code, NORA, has been developed at IFE, Kjeller
(Ref. /4-7/, /4-8/). This code has been extensively compared and calibrated to both
small and large scale critical flow experiments (Ref. /4-7/, /4-9/). NORA has later been
converted to a general single-component two-phase computer code (Ref. /4-4/).
At present there exists three NORA options: For water, chlorine and methane. Exten-
tion to other hydrocarbons is possible.
OLGA is another computer code developed by IFE/SINTEF for simulating the flow of
hydrocarbon mixtures (Ref. /4-10/), either single or two-phase, in pipelines. It
assumes thermal equilibrium between the phases and is therefore not well suited for
calculating critical flow or blowdown rates, that is for cases where non-equilibrium
effects are important. Examples of such cases are pipe ruptures close to large reser-
voirs of gases and liquids.
For cases where the frictional pressure drop controls the flow, for example in case of
ruptures in long pipelines far away from large reservoirs, OLGA would be a useful tool
for calculating leak rates.
4.6 Instantaneous Release of Gas

Assuming instantaneous release of a mass of gas is a simplification which sometimes
is justified. In reality any release, even the complete rupture of a pressure vessel, will
require some short period of time.
The conditions under which the assumption of an instantaneous gas release is justi-
fied have been discussed in Chapter 3.2, Question 3.1. If these conditions are met
the calculation of release rates becomes unnecessary. In such a situation it is the
total mass (kg) rather than the release rate (kg/s) which determines the further deve-
lopment of the accident. In most cases this mass will be equal to the total contents of
gas in a vessel or system.
Scandpower
As in the case of a continuous release the Joule Thomson effect can lead to tempe-
rature drop and liquid formation by condensation. See Chapter 4.3.5 and Appendix
D.2 for more details. In a vessel rupture case there is no local choke effect.
The work done by the expanding gas is distributed over a large area, the heat influx is
larger, the process is less adiabatic and the temperature drop less pronounced than in
a choked gas outflow case.
Furthermore, a sudden and complete rupture of a pressure vessel will often be accom-
panied by immediate ignition due to metal-to-metal impact and spark formation. In this
case the possible condensation and formation of liquid is of less interest.
4. 7 Instantaneous Release of Liquid

As in the case of instantaneous gas release, the assumption of instantaneous liquid
release is a simplification which is sometimes justified. For most practical purposes a
release duration of 10 seconds or less can be considered instantaneous.
Release rate calculations are in this case not necessary. It is the total mass (kg) of re-
leased liquid rather than the release rate (kg/s) which determines the further develop-
ment of the accident. In most cases this mass will be equal to the total contents of
liquid in a vessel or system.
Liquids stored under pressure can , if released to atmosphere, flash to the vapour
phase. For a pure liquid the fraction being evaporated can be calculated , assuming
adiabatic conditions, i.e. no heat input from the surrounding. The mass fraction
vapourized can be calculated according to Eq. (4.11) in Chapter 4.4.4.
For liquids composed of several components the flash calculations are more compli-
cated and one of the computer codes for process simulations, e.g. PVTOLGA, has to
be used.
Scandpower
References Chapter 4
/4-1/ T.A. Smith: "An Analysis of a 100 te Propane Storage Vessel", SRD-R314,
1986.
/4-2/ Coulsen and Richardson: "Chemical Engineering", Vol. I and 11, 1957.
/4-3/ AmScan Associates: "Study of Hazards associated with sudden Breaks of

underwater Pipelines transporting high Pressure Gas", March 1978.
/4-4/ R. Moe, J. Rasmussen: "Development of the Computer Code NORA-HG for

Pressure and Flow Transients in Methane Systems", IFE, Kjeller, 1982.
/4-5/ J.A. Trepp, V.H. Ranson: "A chocked Flow Calculation Criterion for nonhomo-
genous Nonequilibrium two-Phase Flows", Int. Journal of Multiphase Flow, Vol.
8, 1982.
/4-6/ "Critical Flow Modelling in Nuclear Safety", a State of the Art Report by a
Group of Experts of the NEA Committee on the Safety of Nuclear Installations,
OECD, Paris, 1982.
/4-7/ D. Maines: "Flashing in Water Systems", IFE, Kjeller, 1977.
/4-8/ D. Maines, J. Rasmussen: "A short Description of the Slowdown Programme

NORA", IFE, Kjeller, 1977.
/4-9/ 0. Sandervag et al: "Comparisons between Experimental Data from the CFT-
Program and NORA-Calculations", IFE, Kjeller, 1980.
/4-10/ K.H. Bendiksen, D. Maines, R. Moe, S. Nuland: "The Dynamic Two-Fluid

Model OLGA: Theory and Application, J, of Petroleum Technology", SPE
19451, 1989.
Scandpower
(G))SINTEF Risk Management AS
5. Behaviour of released Fluid

5.1 General
This chapter covers the behaviour of fluids after the release from their containment,
and prior to ignition. This behaviour will, among other things, influence the ignition
probability because it will determine if flammable concentrations will interact with igni-
tion sources or not. If ignition is not immediate, it will also influence the type and
shape of the resulting fire because the distribution and mass of flammable material will
depend on the surroundings.
Effects that must be considered after the release of the fluid are amongst others:
Phase transitions
Dispersion
Gravitational flow
Mechanical interaction
Ignition mechanism.
They will be discussed in the following subchapters for gas and liquid releases.
It should also be noted that when considering flammable ranges of a gas release,
turbulence fluctuations may exist. In order to assess appropriate hazard ranges by
the models presented in the following one should calculate the distance to a gas con-
centration of ½ LFL (Lower Flammability Limit) to take this factor into account.
5.2 Meteorological and Topographical Conditions

5.2.1 Data required
After release from their containments the fluids will behave in a manner strongly de-
pendent on the surroundings: Gases will be mostly influenced by meteorological con-
ditions, liquids mostly by topography. However, one also has to consider e.g.: Gases
heavier than air will flow along depressions in the terrain, and evaporation of liquids
will depend on wind velocity.
The type of data needed will be maps and weather statistics. Maps will not always be
good enough or up to date, so visits to the site in question - taking photographs - are
an important part of any risk analysis.
Weather data important for most dispersion calculations are:
Wind speed
Wind direction
Temperature
Humidity
Atmospheric stability.
Scandpower
~SINTEF Risk M anagement AS
For underwater releases also
sea temperature
sea current
wave height/period
can be important.
The weather statistics should cover a time period of several years and be detailed
enough to allow distinction between seasons and day/night. The latter is important
because the number of exposed individuals, both inside and outside the facility being
analysed, will vary depending on the time of day.
5.2.2 Atmospheric Stability

The meteorological data most difficult to obtain are usually those on atmospheric
stability. Atmospheric stability is related to the vertical temperature gradient and it
rules the vertical mass transfer mechanisms in the air, close to the ground.
Stability is described by stability classes, also called Pasquille Class, according to the
scheme in Table 5.1.
Table 5.1: Atmospheric Stabilitv Classes

Stability Pasquille class Temperature gradient
(deg. C per 100 meter)
Very unstable A
Unstable B <- 1
Slightly unstable C
Neutral D -1 to 0
Stable E 0 to 1
Very stable F >1
The correlation to the vertical temperature gradient given in the last column is one
used by NILU (Ref. /5-1/).
If measurements of the temperature gradient are not available, a correlation to cloud

cover, wind speed and time of the day may be used. This is shown in Table 5.2.
Scandpower
Table 5.2: Determination of Stability Class (Ref. /5-2/)

Surface wind at Day Night
10 m height Incoming solar radiation Thin overcast < 3/8
(m/s) or > 4/8 cloud cloud
11
Strong Moderate Slight
<2 A A-B B
2-3 A-B B C E F
3-5 B B-C C D E
5-6 C C-D D D D
>6 C D D D D
The neutral class, D, should be assumed for overcast conditions during day or night
5.2.3 Surface Roughness

Topography also includes surface roughness, which is important for calculation of gas
dispersion. Roughness is measured in meters and it can be seen as effective ampli-
tude of the terrain over which the gas travels. Typical values are given in Table 5.3.
Table 5.3: Typical Surface roughness Values used in Dispersion Calculations

Ref. /5-3/
Type of area Example Zo (m)
Flat land Polderland with few trees Appr. 0.03 m
Farmland Airfield, arable land, polder with many trees Appr. 0.10 m
Cultivated land Glasshouse area, open area with thick over- Appr. 0.30 m
growth, scattered houses
Residential Area with dense but low buildings, wooded Appr. 1.0 m
area area, industrial site with not too large obstacles
Urban area City with high buildings, industrial area with Appr. 3.0 m
high obstacles
5.3 Gas Release

5.3.1 Dispersion of Sonic Gas Jets
Gas released at sonic speed from a nozzle into the atmosphere will be diluted by tur-
bulence, mixing with the surrounding air. Formulas giving
gas concentration
gas linear velocity
as a function of the distance along the jet axis from the release point have been
derived, using the basic physical principles of conservation of mass and conservation
of energy, and making some simplifying assumptions.
Scandpower
The formulas can also be used to calculate:
Distance to a given concentration, e.g. to LFL or 1/2 LFL

Transit time of gas molecules from release point to a given concentration
Mass of gas in the jet, which can participate in an explosion, if ignited.
The simplifying assumptions are:
The jet is cone shaped with top angle a

Across the cone, the gas concentration is constant and drops to zero at the cone
surface ("top hat" profile)
The cone radius at the top is an equivalent radius, which corresponds to the gas
jet radius after expansion to ambient pressure.
The last assumption yields an expression for the equivalent radius
Eq. (5.1)
where
ro = equivalent radius (m)

q = release rate (kg/s)
C = .Jy •R • T / M =speed of sound (mis)
Po = density of gas after expansion (kg/m 3
)
y = gas specific heat ratio c,Jc., =1.3 for methane

R = 8,314 (g m'/K mol s') =gas constant
T = temperature after expansion (K)
M = molecular weight (g/mol)
For calculating the distance x to a given volumetric concentration C (fraction) the
following equation can be used:
x =__!:L_[~
tan a C
·.J[C · p + (1-C) · p ]/ p
O 1 0
-1] Eq. (5.2)
where
r0 = equivalent radius (m), see Eq. (5.1)

a = top half angle of cone
Pr = density of air (kg/m 3)
From Eq. (5.2) C as a function of x can be obtained:
Scandpower ~
Risk Management AS u SINTEF
The linear gas velocity decreases from c (speed of sound) with increasing distance x
as
1
v(x) = c - - - - - - - , = = = = = = = = = Eq. (5.4)
~2 [1- EL) + ~2
2 2
Po
[1 - E!.) + 4 . EL (1 + 2:. tan ex)
Po Po ro
The mass of gas in the cone frustum between XuFL and XLFL can be calculated as
f C(x)•1t(r +x-tanex) dx
XLFL
2
MG=pG· 0 Eq . (5.5)
XUFL
where
PG = density of gas at ambient conditions (kg/m 3)

XLFL = distance to LFL (m), according to Eq. (5.2)
XuFL = distance to UFL (m), according to Eq . (5.2)
Inserting C(x) from Eq. (5.3) and performing the integration one obtains
Eq. (5.6)
The integral expressions l(XLFL) and l(xuFL) are calculated from
l=f✓ax 2 +bx +c dx
✓ax2 +bx+ c - b 4 ac • ~ • In (2ax + b + 2 -Ja ✓ax 2 +bx+ c ) 5
2
= 2 ax + b - Eq. ( .?)
4a 8a ~a
with the abbreviations
a=
4
. Car~ ex J·=~
b = 8 . tan ex . fr
ro Po
c=[1+ ::J
5.3.2 Atmospheric Dispersion of light and neutral Gas
Estimation of hazard distances after release of lighter-than-air gas can be obtained in
two steps. Initially the gas dispersion will be dominated by lifting, turbulence and wind
spreading until the gas density equals the air density. In the second step, when the
gas density has become neutral, i.e. equal to the air density, a Gaussian dispersion
can be assumed. For details on dispersion of light gases, see Ref. /5-1/, /5-2/, /5-3/
and /5-4/.
Gaussian spreading of released gas is based on the following assumption:
Scandpower
a) The gas present in the cloud is not lost. (Known losses may be taken into
account)
b) Through the action of turbulence in the atmosphere, the gas is distributed accor-
ding to a Gaussian distribution, both horizontally and vertically, with the para-
meters changing as a function of the distance
c) The parameters of the Gaussian distribution are derived from experiments and
are dependent on weather conditions.
In Figure 5.1 an illustration of the Gaussian distribution of the gas concentration is

presented as a function of distance from the release point.
x-direction
-- -- --- ------ ...
.' .''
I' I
,,''
I
Source
y-direction
Figure 5.1: Diagrammatic Gaussian Distribution of the Concentration as a Function

of Distance from the Release Point (Ref. /5-3/
The main limitations of the idealised Gaussian distribution model are:
It applies only for gases with the same density as air

It applies only for flat and open terrain
It is difficult to take into account the effect of obstacles
It is assumed that the meteorological and ground conditions are constant over
the entire distance over which the gas cloud is to be followed
The model has a fairly high inherent inaccuracy because of the large fluctuations
in the turbulence. (The model only gives time-averages for a selected period)
There must be some wind, e.g. Uw > 1 m/s.
Based on the Gaussian distribution model a so-called puff model has been developed,
Ref. /5-5/.
The puff model simulates the cloud with Gaussian-shaped puffs on a specified grid.
For every time interval i',.t a new puff is released, while those released before move
Scandpower
Risk Management AS G)SINTEF
and grow. Their growth is determined according to their distance from the point of re-
lease and the stability of the atmosphere (Pasquille class). The speed of the puffs is
equal to the wind speed measured at the release point. The shape, the explosive
mass and volume of the cloud are obtained by superimposing the puffs.
In Ref. /5-3/ (TNO) practical formulas and graphs are presented for hand calculation of
gas dispersion based on the Gaussian model. It is suggested to select a continuous
source model if
X< 1.8 · Uw • tern
and an instantaneous model if
X > 1.8 · Uw • tern
where
x = downwind distance from source (m)

Uw = wind velocity (m/s)
tern = duration of the emission (s)
For a continuous point source of height h (m) the following formula is suggested for
the gas concentration:
C(x Y z) = _ _ _q_ __ exp (- y2 ) . [exp(- (z - h )2) + exp(- _(z_+_ h~)~ 2

)]
Eq. (5.8)
' ' 21tuwcr y(x)cr z (x) 2cr~(x) 2cr~(x) 2cr~(x)
where:
C(x,y,z) = concentration as a function of x, y and z (kg/m 3)

q = source strength (kg/s)
h = source height (m)
Oy = standard deviation in y direction (m)
Oz = standard deviation in z direction (m)
The a-values depend on the meteorological conditions. Correlations can be found in

Ref. /5-3/ and other publications on atmospheric dispersion.
The term with (z+h) is the contribution of the so-called reflector source, which gives
the contribution of the gas, which would have disappeared under the earth's surface
with continuing diffusion, but which is reflected in reality.
Eq. (5.7) is valid for point sources. Corrections for actual dimensions of a source are
made with the help of a virtual point source.
5.3.3 Atmospheric Dispersion of Heavy Gas

5.3.3.1 General
The typical heavy gas dispersion process involves three more or less distinct regimes
of flow. After the initial release a cloud will be formed. The initial spreading motion of
such a cloud is governed by gravity effects and relatively independent of the charac-
(GJ)SINTEF Scandpower
Risk Mana gement AS
teristics of the wind field. The slumping and lateral spreading motion continues until
the kinetic energy of the gravity-driven flow is dissipated. The dispersion process
which follows can be described as a stable stratified plume (or cloud) embedded in the
mean wind fiow. As the dispersion proceeds, the stable stratification decreases until
the process can be represented as a neutrally buoyant plume (or cloud) in the mean
wind flow. The three regimes, gravity dominated, stable stratified and passive dis-
persion, which may overlap and be present in various degrees, must be taken into
account if a model is to be generally applicable.
Mixing of air into the cloud depends on the turbulence of the flow field. This turbu-
lence is generated by:
The formation of the initial cloud

The spreading effects which may create vortexes
The ground/cloud interface
The difference of velocity between the flow fields above and within the cloud
The atmospheric conditions
The heat transfer between the cloud (often cold) and the ground
The heat transfer between the atmosphere and the cloud.
The intensity of these physical processes, the vertical profile of density and the vertical
mixing within the cloud are dependent on the initial density. Several processes which
determine it, can be distinguished:
In gases such as propane, the molecular weight is higher than that of air and it
have an excess density at ambient temperature
In gases such as methane, the molecular weight is less than that of air, but the
temperature may be very low due to cryogenic storage. The excess density ob-
tained causes the formation of a cold, heavy gas cloud. A second phenomenon,
which might influence dispersion, is moisture, coupled with a low temperature of
the cloud. This may cause fog and influence the buoyancy of the air-vapour
mixture
In gases such as ammonia, a low temperature is not sufficient to produce an ex-

cess density. But simultaneous release of vapour and liquid can lead to forma-
tion of liquid droplets or mist, which in turn may lead to excess density.
Box models for heavy gas dispersion have been developed extensively over the last
few years. One of the first models was the Van Ulden Model (1974). To overcome
some of the limitations of the Gaussian models a gravity spreading phase is introdu-
ced with two physical effects:
A cloud front velocity to account for the gravity phase

An edge entrainment that is proportional to the front velocity.
Box models progress from the gravity phase to a non-buoyant dispersion phase when
specified criteria are met (e.g. when the gas density equals air density.)
Most box models consider humidity, heat transfer from the ground and difference of
speed between the cloud and the wind. They assume that the cloud starts as a cir-
cular cylinder in case of instantaneous releases. For continuous release the cloud
shape becomes elongated and has rectangular crosswind sections. The concentra-
tion variation in the field is assumed to vary with a Gaussian or a rectangular distribu-
Scandpower
tion in each direction. The computational time needed by box models is very short
due to the parameterisation of the behaviour of the cloud with simple functions.
The determination of the coefficients of these parameterisations is an important prob-

lem because of their sensibility to the surface roughness and to the density of the
heavy gas.
Other deficiencies of box models are their inadequacy in case of extreme meteoro-
logical conditions such as high stability combined with calm conditions , or when terrain
and obstacle effects are important.
5.3.3.2 DENZ
The dispersion model DENZ was developed by the Safety and Reliability Directorate
(SRO) in the U.K. as a computer code (Ref. /5-6/ and /5-7/). DENZ was developed for
the study of hazards from catastrophic failures of pressurised vessels. It is suited for
rapid (instantaneous) release of large quantities of dense toxic or flammable gases to
the atmosphere. In the following some characteristics of DENZ are summarised:
The initial gas cloud (a cylinder) is specified by the user taking into account
the quantity of air entrained during the initial formation of the cloud
the density of the gas cloud mixture
the initial velocity of the cloud (a function of wind speed)
the initial height and radius of the cloud .
The DENZ code calculates slumping of the initial gas cloud , and cloud heating from
the ground and the entrained air.
Entrainment of air during the slumping phase is related to the atmospheric stability
(which the user specifies). Both edge and top entrainment are considered and the
user may specify proportionality constants for this purpose. The default values in
DENZ for entrainment are:
a top = 0.5
a edge = 0
Literature published of late (Ref. /5-8/) suggests that
a top = 0.1 - 0.3

aedge = 0.6 - 0.7
agrees better with experiments.
The computer programme DENZ uses two alternative tests to determine whether the
plume may be considered passive. One is concerned with the rate of cloud radius in-
crease caused by gravity and/or atmospheric turbulence. The other is concerned with
the density difference between the cloud and the surrounding air.
The concentration within the cloud is distributed in a Gaussian manner across the puff.
Scandpower
5.3.3.3 CRUNCH
The computer code CRUNCH was also developed by SRO (Ref. /5-9/) and is based
on the same assumptions as DENZ. CRUNCH is, however, concerned with disper-
sion calculations for continuous releases of heavy gas. The characteristic release
time over which the release rate is constant should be compared with the time for dis-
persion, for the model to be valid.
The initial gas source in CRUNCH can be visualised as a window with given dimen-
sions and a constant flow rate (possibly with some initial air entrainment). The model
does not take into account the real source configuration or any initial release momen-
tum of the gas. Plume motion, entrainment of air, cloud heating transition to passive
phase and Gaussian passive diffusion are basically calculated as for the DENZ code.
5.3.3.4 HEGADAS
HEGADAS was developed by Shell Research ltd., Ref. /5-10/. Contrary to DENZ and
CRUNCH, this model has the possibility to describe a time-variable source strength.
This is particularly important in cases where a cryogenic liquid is released to the
ground. The evaporation rate and, hence, the vapour source strength, will be very
high in the beginning. As the ground cools down the source strength will decrease
strongly. This behaviour cannot be modelled properly by DENZ or CRUNCH. DENZ
can describe the early phase by assuming an instantaneous release of a quantity
which corresponds to the mass released during the first few seconds/minutes, while
CRUNCH can assume a constant release rate which can be chosen pessimistically as
the rate after the first, fast cool down has ceased. HEGADAS on the other hand has a
transient programme version, which can describe the actual decrease in source
strength.
Also HEGADAS has been compared with large-scale experiments performed on

Thorney Island (Ref. /5-11/).
5.3.3.5 WHAZAN
The WHAZAN programme (Ref. /5-12/) contains a number of mathematical models for
industrial hazard assessment and was developed by Technica International ltd. The
WHAZAN models run on personal computers.
The dense cloud dispersion model, which is part of the WHAZAN package is appli-
cable only as the spreading of the cloud, is due to gravitational forces. This model is
based on the so-called Cox and Carpenter model of 1980.
The model can calculate instantaneous and continuous release of dense gases.
When the time to reach a specified concentration is greater than 4 times the release
duration, WHAZAN suggests using the instantaneous model.
Lateral gravity-induced spreading is calculated with the same formula as in DENZ and
CRUNCH, but it is not possible for the user to specify the constant.
The top and edge entrainment of air is also calculated with the same formulas as in
DENZ and CRUNCH, but again the user may not specify the values of the entrainment
factors. Which values actually are used in the programme is not specified.
Scandpower
Risk Management AS Gl)SINTEF
The criterion for transition from gravitational spreading to a neutral diffusion model is
that lateral spreading due to turbulence exceeds that of gravitational spreading.
For forced or natural convection the model uses a heat flow equation. The gas con-
centration in the cloud is assumed to be uniform.
5.3.3.6 Three-dimensional Models
The most accurate numerical models to evaluate dispersion in the atmosphere of a

denser-than-air gas/vapour are the three-dimensional grid models. Geometric distur-
bance of importance for the dispersion requires use of these types of numerical
models. The models are based on finite element methods. This means that the re-
lease geometry and the surroundings are built up from small elements. The accuracy
of the models is strongly dependent on the size of these elements. For each element
the following system of equations must be solved:
Conservation of momentum
Conservation of mass
Conservation of energy
Equation for advection-diffusion of the concentration
Equation of state for the mixture air/gas
Equation for humidity.
Three-dimensional grid models require use of large computers and the computational
costs are quite high.
To limit the computing cost it will often be necessary to use elements of a certain size.
This might lead to difficulties in calculating the turbulence and the initial release phase
exactly.
Ref. /5-13/ gives a survey over major three-dimensional models.
5.3.4 Condensation
High pressure gas escaping from a vessel or piping experiences a pressure drop and,
due to the Joule-Thomson effect, a temperature drop. The effect is widely used in
cooling and gas liquidation applications, also in oil/gas separation processes. If the
temperature drop is sufficiently large, liquid can be formed and drop out to form a pool
on the ground.
For pure substances the pressure-enthalpy diagram can be used to determine if and
how much liquid is formed, see for example the diagram for methane in Figure 5.2.
If the methane is at 130 bar and -28 °C the state of the gas can be described as point
A in Figure 5.2. By isoentropic expansion to atmospheric pressure the state of the gas
can be described as point C. From the phase envelope curve it can be seen that con-
densation will occur at this point. The quantity of condensate and gas can in fact be
calculated based on the location of point C within the phase envelope.
Scandpower
~ SINTEF Risk M anagem ent AS
.. -,..
. "I .,.
l I 'l -·~
•
Figure 5.2: Pressure-Enthalpy Diagram for Methane
If, for example, methane with 130 bara and - 28 °C (point A) is expanded in a chock
without doing mechanical work the expansion is isoenthalpic, corresponding to the
vertical line to point B at 1 bara and - 125 °C. If, on the other hand, the expansion
occurs with mechanical work being done, as in an expansion turbine, the process is
nearly isoentropic. The final condition is then at point C, in the two-phase region at -
161.5 °C, the atmospheric boiling temperature of methane. The fraction of liquid for-
med can be determined by the position of point C within the two-phase region.
For gas mixtures flash-calculations have to be performed by means of PVT computer

programmes. Several such programmes are available, some of them can be run on
PCs.
For most applications within risk analysis formation of liquid pools from gas conden-
sation is not of great interest. If liquid is formed the high release velocity of the gas
will in most cases lead to the formation of mist or small droplets. They will either be
re-evaporated quite quickly due to heat provided by mixing with surrounding air or, if
ignition occurs, they will contribute as fuel to the resulting jet flame.
The Joule-Thomson is, however, important even if no liquid phase is formed. The
temperature drop of the gas can lead to an increase of gas density, making the gas
heavier than air. This will then result in a different dispersion mechanism compared to
the case where the gas is neutral (gas density equal to air density), or where it is
buoyant (gas density less than air density).
Scandpower
Risk Management AS (GJ)SINTEF
See also Appendix D.1 , Gas Condensation, for further details.
5.4 Liquid Release

5.4.1 Flash Evaporation
Liquid stored or transported under pressure can, if released, flash to the vapour
phase. See Chapter 4.4.4 for more details.
5.4.2 Boiling Liquid onto Water

Boiling liquid released onto a water surface will distribute itself under the influence of
gravity until a patch or pool of liquid has been formed. Heat supplied from the water
and the air will control the evaporation rate.
The size (diameter) of the pool is controlled by two conditions:
Rate of liquid supply

Rate of evaporation.
A third condition is based on the experimental observation that the layer will break up
when a certain minimum thickness is reached. The pool size will increase with time
until the evaporation rate equals the supply rate. The supply can be either from
a moderate leak (continuous source)

a sudden release of a large quantity (instantaneous source).
In the continuous case the maximum pool area is achieved when
total evaporation rate = leak rate

A - m =q Eq. (5.9)
where
A = area of pool (m 2)
m = evaporation rate (kg/m 2 s)
q = leak rate (kg/s)
The evaporation rate m is controlled by heat supply. Early theories assumed that ice
formation would occur underneath the boiling pool, but experimental observations
have not confirmed this under realistic conditions, not even for LNG. Convection in
the water ensures supply of warm water from the deeper layers. Under these condi-
tions also heat supply from the air may be neglected. The heat supplied from the
water can be calculated or experimentally determined. Values in the range 25 - 100
2
kW/m are reported in Ref. /5-14/ and /5-15/. For propane 40 kW/m2 is proposed in
Ref. /5-14/, leading tom= 0.094 kg/m 2 s. Values for LNG are higher. The average of
4 experimental values is 77 kW/m 2 (Ref. /5-14/), leading to m = 0.15 kg/m2 s.
Example:
Release of fully refrigerated propane from a ship at a rate of 1,000 kg/s. (At this
rate it takes 1 1/2 hour to empty a ship tank of 9,000 m3).
Scandpower
A= q/m = 1,000/0.094 = 10,600 m2
In case of circular pool shape the diameter would be 116 m.
The case of instantaneous release of a large quantity of liquid is more complicated.

The liquid will spread under the influence of the gravitational force, being opposed by
a frictional force. The radius of the pool increases with time until it reaches a maxi-
mum. After this, it decreases as evaporation consumes the liquid. An expression for
the maximum radius has been derived in Ref. /5-14/:
118
R 2 3)
= 1 02 Ap. P, . gV
(m)
max ·
( Pw·m
2 Eq. (5.10)
where
Ap = Pw • p, (kg/m 3)
Pw = density of water (kg/m 3)
p, = density of liquid (kg/m 3)
g = gravitational acceleration (9.81 m/s2)
V = initial volume of liquid (m 3 )
Also the time until all liquid has evaporated is of interest:
2
le= 0.67 Pi · Pw · ~
)1/4 (s)
(
Ap•g-m
Eq.(5.11)
In deriving these expressions the existence of a minimum layer thickness was neglec-
ted, i.e. the maximum pool radius occurs at time le, just before the last (infinitesimally
thin) layer evaporates. In reality the layer will break up into patches when the thick-
ness approaches the minimum layer thickness which, for LNG on completely calm
water, is approximately 1.8 mm. For less calm water the break up will start earlier.
The effect of minimum thickness layer will lead to smaller pool size, lower evaporation
rate, longer duration. In most cases it will therefore be conservative (pessimistic) to
neglect this effect.
5.4.3 Boiling Liquid on Land, Unrestricted

This case is somewhat hypothetical since it requires a completely smooth, horizontal
surface. Boiling will be violent at first, as long as the ground is warm. As the ground
is cooled down into deeper and deeper layers, heat flux and boiling will be gradually
reduced.
The behaviour over time of the evaporation rate can be described as (Ref. /5-16/):
s
m = ✓t (kg/m s)
2
Eq. (5.12)
where the constant S contains all the necessary physical constants:
Scandpower
Risk Management AS IG))SINTEF
S = "1T · ✓P · c · 'A.Ire (kg/m 2 s112 ) Eq. (5.13)

hv
with
L1T = initial temperature difference between ground and boiling liquid ( C) 0
p = density of ground (kg/m 3

)
C = spec. heat of ground (J/kg °C)
'A, = heat conductivity of ground (W/m °C)
hv = heat of vaporisation (J/kg)
The equation for m is not valid for the very first few seconds. (For t ➔ 0 one gets
m ➔ ao).
For the case of a continuous liquid source with source strength q (kg/s) the pool
diameter will increase with time. Ref. /5-16/ gives the expression:
1/4
R = _jl_g__ . t3t4 Eq. (5.14)
(
p 1 • 2rc )
No equilibrium diameter is reached , because as the ground cools down with time, a
larger and larger area is needed to evaporate the released liquid. But the total evapo-
ration rate will approach the release rate after some time, and using the liquid release
rate as an upper estimate for the total evaporation rate is a conservative approach.
One has, however, to choose a reasonable value for R, for example for time t = 1,000
sec.
For instantaneous release of liquid the pool size will increase with time, go through a
maximum, and decrease again. According to Ref. /5-3/ the maximum diameter is
approximately:
1/3 v s112 9 1/12
_ 2 · P1 • •
Dmax - S113 (m) Eq. (5.15)
Example: V = 5.2 m propane, p1 = 580 kg/m 3, g = 9.81 m/s2 , concrete ground

3
with 15 °C, "1T =15 + 42.1 =57 °C, p =2,000 kg/m3 , c =1,000 J/kg °C,
')., =1 W/m °C, hv =427,000 J/kg . This gives S =0.11 and Dmax =84 m.
A pool of this diameter (84 m) and 5.2 m3 volume would have a layer thickness of only
1 mm. This is thinner than the observed minimum layer thickness. According to Ref.
/5-17I the following minimum layer thicknesses are appropriate:
Rough sandy soils: 25 mm
Farmland, pasture: 20 mm
Smooth sand, gravel: 10 mm
Concrete, stone: 5 mm
Calm water: 1.8 mm
Using 5 mm (concrete) the 5.2 m3 will produce a maximum diameter of 36.4 m only.
SRO (Safety and Reliability Directorate) has developed a computer code SPILL (Ref.
/5-18/) which describes spilling of liquid hydrocarbon on the ground. Ref. /5-19/ gives
a formula which fits the results from SPILL. For propane:
~ SINTEF R isk
Sca ndpower
Management AS
Dmax = 22.5 . M0443 (m) Eq. (5.16)
where M is the mass in tons. In our example M = 3.02 t and Dmax becomes 37 m,
which agrees well with the figure of 36.4 m from the minimum layer model.
The time to reach the maximum diameter is approximately
2
t- D (s) Eq. (5.17)
4Jg:°v
which in our example works out as 46 seconds. After this time the evaporation rate (in
the middle of the pool) will be down to:
s 0.11 1 2
m= -Jt = ,J4s = 0.0 6 kg/m s
while it will still be higher at the edge (newly covered ground). Ref. /5-19/ gives as a fit
to SPILL - results a total evaporation rate of
m. 2:. D 2 = 11.24- M 0·746 (kg/s) Eq. (5.18)

4
which in our example is 26 kg/s. Dividing by the maximum pool area the area-avera-
ged evaporation rate of
mav = 0.024 kg/m 2 S
is obtained, which corresponds to the evaporation rate after 21 seconds.
5.4.4 Boiling Liquid on Land, Restricted

Quite often storage vessels and other vessels with a large inventory of flammable
liquid are surrounded by a bunding. This will confine the escaping liquid to a well-
defined, limited area, or dike, thus reducing the evaporation rate.
The problem of calculating the evaporation rate is simplified since the surface area is
determined by the dike geometry. (If the escaping volume is so small that the bunding
is not reached, the methods of Chapter 5.4.3 for unrestricted escape are applicable).
In some cases the floor inside the dike will be sloped. This reduces the pool surface
area and the evaporation rate still further, and it makes conditions even more well-de-
fined.
As in the case of unrestricted outflow, the evaporation rate will decrease with time as
1/✓t. Because of the depth of the pool, evaporation may continue for a very long time.
According to Ref. /5-20/ the evaporation rate will approach the constant value:
LlT (kg/m 2 s)
m=k•- Eq. (5.19)
hv
The constant k, is varying with different types of substrations (for average soil and
concrete; k = 10.5).
Scandpower
Example: Propane, ~T = 15 + 42.1 = 57.1 °C, hv = 427,000 J/kg, gives

m = 0.0014 kg/m 2 s, which corresponds to the time-dependent evaporation rate
after 6,200 seconds (1. 7 hours).
The highest total evaporation rate will occur when the total bund area is just covered.
An estimate of the time necessary to cover the bund area A (m2 ) may be derived from
Eq. (5.14) for R (t) for continuous release q (kg/s).
tc = 2A . P1
2 )1/3 (s)
Eq. (5.20)
(
1t•g•q
Example: Propane into dike of area 20 x 20 = 400 m2 at a rate, q = 100 kg/s.

p1 = 580 kg/m 3 . This gives tc = 40 seconds. After this time the evaporation rate
is down tom= S/✓40 = 0.11/✓40 = 0.017 kg/m2 s for the parts which have been
covered from the start. The average m may be taken as that after 1/2 tc, i.e. 20
seconds, giving m = 0.025 kg/m2 s, which gives 10 kg/s for the whole dike.
For the special case of LNG being released into bunded dikes, both with flat and
sloped floors, nomographs have been worked out which simplify the necessary calcu-
lations (Ref. /5-21/ and /5-22/).
For instantaneous release of a volume V (m3 ) the time to cover the dike floor is:
A
tc= ~ v (s) Eq. (5.21 )
1t if g' V
Example: V = 7 m3 (corresponds to 40 sec fiow at q = 100 kg/s, example above)

into 20 x 20 m bunding: tc = 15 seconds. The average evaporation rate, taken
at 1/2 tc = 7.5 seconds ism= 0 . 11/✓7.5 = 0.04 kg/m 2 s, or 16 kg/s for the whole
dike.
5.4.5 Non-Boiling Liquids

For non-boiling liquids the spreading behaviour on the ground or water will be similar
to that of boiling of liquids, i.e. the size of pools as a function of time for continuous
and instantaneous releases may be calculated using the same formulas as above.
The main difference lies in the evaporation rates. Since these liquids will mostly be
stored at or near ambient temperatures they will, at least initially, have approximately
the same temperature as the ground on which they are spilled. Evaporation will not
occur by a boiling mechanism, but rather by vapour diffusion. Since the air above the
liquid surface is constantly renewed by the wind, no equilibrium is established and
evaporation will continue until all liquid is used. The heat necessary for the evapora-
tion is mainly taken from the air, but also the ground will contribute and cool down
somewhat.
An expression for evaporation rate is given in Ref. /5-3/.
m=2-10-J.vo.1s l -0.11 M·Pa 1n[1 + Pv - Pp] (kg/m 2 s) Eq. (5.22)

R-T Pa - Pv
Scandpower
~SINTEF Risk M anagement AS
where

V = wind speed (mis)
L = length of pool, parallel to wind (m)
M = molecular WElight (g/mole)
R = gas constant (J/kmole K) = 8,314 J/kmole K
T = tern peratu re (K)
Pa = air pressure (Pa)
Pv = vapour pressure of evaporation liquid (Pa)
PP = partial pressure of liquid in arriving air (Pa)
Example: Pentane at 15 °C in a 20 x 20 m dike, with 5 m/s wind parallel one

side; From vapour pressure curves for pentane Pv (288 °K) = 0.47-105 Pa,
Pa= 1-105 Pa, Pp= 0 Pa m = 0.01 kg/m 2s or, for the whole 400 m2 dike, 4 kg/s.
5.5 Underwater Release

5.5.1 Overview
Underwater release of gas, liquid, or multi-phase fluids is an important problem in the
offshore oil and gas industry. Underwater blowouts, pipeline ruptures and leaks from
subsea wells are accidents, which require models to describe the behaviour of the
released fluid
between the release point and the water surface

on and above the water surface.
Both oil and gas are lighter than water and will raise to the surface in the form of a
cone-shaped plume. Oil will spread out into a thin layer on the water surface, while
gas will be diluted by air. The dispersion of gas in the air will again depend on the
density of the gas, if it is light, neutral or heavy relative to air.
5.5.2 Underwater Plume

Models for underwater plumes have been developed by Fannel0p and coworkers
(Ref. /5-23/, /5-24/ and /5-25). They describe the geometry of the plume, the concen-
tration profiles, the rising speed etc. Figure 5.3 shows a sketch of the situation.
Scandpower
SUrfaee flow lbrteraction zone

I
Buoyancy - Plume
\\~=-
profile. r(r,z)
--'--'--+--4-'--,f---l--+1---/LL-L--'--
~
\ I/
Sea bed . /_ - 1,
~~~~:, ~_-;/f/·'.·5~~::~; ·I;/t:-'.·:J{'/
Virtual orf in ,.
---------'-----0-----------•r
Figure 5.3: Underwater Release with Plume Formation
Experiments have shown that the tip angle of the plume is approximately 18°, but most
of the gas will be contained in the central part with tip angle 0.7 x 18 = 12.6°. This
means that the diameter of the circular area where the plume breaks the water surface
is
D = 2 • tan (1 ~
6
} z (m)
Eq. (5.23)
=0.22 · z
where
D = diameter of plume breaks the water surface (m)

z = water depth (m)
The rate at which the fluid reaches the sea surface and which determines the source
strength for the further calculations depends on the underwater release rate and the
rise time to the surface. This rise time can be quite long (velocities are in the range
1-3 m/s) and it will tend to smooth rapid variations of the subsea release rate. For
sudden ruptures of pipelines, for example, it will not be appropriate to use the very
high initial release rate, but rather a release rate averaged over some minutes, depen-
ding on the water depth.
Scandpower
(r,» SINTEF Risk M anagement AS
Liquid containing volatile components which are in equilibrium at the hydrostatic

pressure at the release point (e.g. 30 bara at 300 meter water depth) will release gas
by flashing as it rises to the surface and the pressure decreases. Due to the long
rising time and ample heat supply from the sea water the mass fraction of gas
reaching the surface should be determined by assuming equilibrium at 1 bara
pressure and sea water temperature (- 5 °C in North Sea).
In some cases it is important to know the sideways displacement of the plume occurr-
ing due to water currents. In such cases it is necessary to know the plume rise velo-
city v and the water current j as a function of water depth z and to calculate the dis-
placement
6R =
0
Jv(z)
j(z) dz Eq. (5.24)
z
For v (z), the plume rise velocity, Ref. /5-25/ gives the expression
113 113 2
V= (25) [U 9o•g• (FJ+1)] _ 2 -113{ 1+ _!_! ( Z )+§.!! [ Z ] } E .(5 _25 )

24 zua:~ 39 Z 0 + 10 2 39 (Z 0 + 10) q
where
V = plume riser velocity (mis)

z = water depth
Ugo = source volumetric flow rate (m 3/s)
Zo = water depth (m)
g = gravitational constant (m/s2)
FG = ratio inner plume/total plume diameter - 0.143
The subscripts G on F and a indicate that they apply to a Gaussian distribution in the
radial direction. The values given were derived from experiments with rather small
flow rates (0.6 - 1.3 m3/s). In real accidents much higher flow rates occur, and indica-
tions are that F approaches a value of 1, and aG a value of 0.165.
Information on the sea current velocity profile must be obtained from measurements.
Typical values could be, for example:
0.6 mis at the surface

0.4 mis at 100 m water depth and below
with a linear decrease from the surface to 100 m depth.
Where only rough estimates are required it suffices to use average values for both j
and v and to calculate the displacement according to
6R=lV (m)
5.5.3 Liquid Surface Layer

Liquid not soluble in water such as crude oil, condensate, LPG etc. will reach the
water surface and behave in a similar way as liquid being spilled onto the water from
Scandpower
above. They will spread out under the influence of the density difference, the surface
tension and the viscosity, and the resulting layer will be exposed to
wind forces
water current
waves
evaporation.
See Chapters 5.4.2 and 5.4.5 or Ref. /5-23/ for more details.
5.5.4 Dilution of Gas lighter than Air

A very simple approach is to assume an immediate dilution of the gas to a concen-
tration , which corresponds to the ratio
C= area of gas bubbles

area of plume at surface
which can be written as
C = u/v Eq. (5.26)

7t
D2
4
w here
u = volumetric gas flow (m3/s)

v = bubble rise velocity (m/s)
~ 0.3 m/s
D = 2 • tan (6.3°)-z = 0.22-z (m)
A somewhat more refined approximation is given in Ref. /5-25/.
1/3
2. u2 . Pa
C= Eq. (5.27)
Scandpower
where
Pa = density of air (kg/m 3)

pg = density of gas (kg/m 3 )
(other symbols as before)
As pg ➔ Pa, C ➔ ao and the equation becomes invalid. Also, the model is only valid for
calm conditions, i.e. for zero wind velocity.
5.5.5 Dilution of Gas heavier than Air

A simple model for this case is to consider a cylindrical volume on top of the plume
circle and to assume homogeneous mixing of gas entering from below and air being
blown in from the side. In this cylinder the gas concentration w ill decrease with in-
creasing height above the sea surface as
C(h) = u Eq . (5.28)
D · h · Vw
where
h = height above sea surface (m)

D = 2 · tan (6.3°)-z = 0.22-z (m)
Vw = wind velocity (m/s)
u = volumetric gas flow (m 3/s)
z = water depth (m)
Eq. (5.28) can be used to obtain an estimate of the height above which the concentra-
tion can be expected to be lower than some critical value, for example 1/2 LFL (lower
flammability limit).
For more refined calculations one of the heavy gas dispersion models must be used,
using the gas flow reaching the water surface as the source term. (See Chapter
5 .3.3).
5.6 Ignition Mechanisms

5.6.1 Flammability Characteristics
Combustion of a flammable dust, mist or vapour/air mixture occurs if the composition
of the mixture lies in the flammable range and if the conditions exist for ignition. The
most important flammability characteristics are:
Upper/lower flammability limit

Flash point
Auto ignition temperature
Minimum ignition energy.
These characteristics are illustrated in Figure 5.4.
Scandpower
Boi l ing point
mmable limi t
upper fla
C
0
...., Autoignition
ltJ
I..
....,
C
Cl)
u
C F lammabi Ii t y range
0
u
1 Flash Auto ignition 1

:/point temperature~:
I
Tempe r ature
Figure 5.4: Typical Flammability Characteristics, Ref. /5-26/
Flammability Limits
The discussion of flammability limits below assumes that the flammable mixture con-
tains air.
A flammable mixture burns in air only over a limited range of composition. Below a
certain concentration, lower flammability limit (LFL), the mixture is too "lean", while
above a certain concentration, upper flammability limit (UFL), the mixture is too "rich".
Flammability limits are affected by, Ref. /5-26/:
Oxygen concentration
Gases other than oxygen
Inert gas concentration
Size of equipment
Direction of flame propagation
Pressure
Temperature
Scandpower
Turbulence
Mixtures.
The limits of a flammable gas may be determined experimentally, Ref. /5-27/. Simpli-
fied methods for estimating the limits do however exist, Ref. /5-26/.
The flammability limits of mixtures of gases may be calculated by Le Chatelier Equa-

tion, Ref. /5-26/, /5-27/ and /5-28/. For the lower limit:
LF L = ....!QQ_ Eq. (5.29)

±Ci
i = 1 Li
where
Ci = volume percent of fuel gas i in the fuel gas mixture

Li = volume percent of fuel gas i at its lower flammability limit in air alone.
For the upper limit:
UFL = ....!QQ_ Eq. (5.30)

±Ci
i =1 U i
where
Ui = volume percent of fuel gas at its upper flammability limit in air alone.
It should be noted that the Le Chatelier Equation is an empirical one and it is not uni-
versally applicable.
Flash Point
The flash point of a substance is the temperature at which the vapour pressure in air,
at equilibrium conditions, equals the lower flammability limit for that substance.
Auto Ignition Temperature
The auto-ignition temperature is the minimum temperature at which a substance will

selfignite and sustain a combustion process. It will depend on the following factors,
Ref. /5-26/:
Composition of vapour-air mixture

Oxygen concentration
Shape and size of combustion chamber
Rate and duration of heating
Time delay
Catalyst effects of materials.
The auto ignition temperature is liable particularly to variations caused by the nature of
the hot surfaces. The auto ignition temperature may be reduced by as much as
100-200°C for surfaces which are lagged or are contaminated by dust, Ref. /5-27/.
Scandpower
Ignition Energy
The ignition energy is the energy required igniting a flammable mixture. If a flam-
mable gas-air mixture is to be ignited by a local ignition source, it is not sufficient as
with auto ignition to raise a volume of mixture to a certain temperature for a certain
time. There is also a minimum volume of mixture so treated which is required in order
to give rise to a continuing flame throughout the rest of the mixture, Ref. /5-27/.
The minimum ignition energy for a selected substance usually occurs close to the
stoichiometric mixture.
5.6.2 Sources of Ignition

Sources of ignition include the following:
Flames, direct heat and hot surfaces

Rotating machinery
Combustion engines
Faulty electrical equipment
Mechanical sparks
Static electricity
Hot work, maintenance
Chemical energy
Auto ignition by the fracture mechanism.
Flames, direct Heat and hot Surfaces
Much process equipment operates at a high temperature. Although the equipment is

usually lagged, there may be some surfaces, which could be a source of ignition.
Rotating machinery, such as pumps/compressors etc. , with a faulty bearing, may run
hot and this hot spot is a potential source of ignition.
Combustion engines have been the source of ignition in a number of cases, Ref.
/5-26/. Ignition can result from the hot surface of the exhaust system raising the tem-
perature of the flammable gas-air mixture above the auto ignition temperature. Hyper-
speed/disintegration of the machinery due to gas exposure of the air intake may also
represent a possible ignition source.
Faulty electrical equipment. Electrical equipment in a plant handling flammable

materials is normally subject to area classification. This ensures that electrical equip-
ment is unlikely to provide a source of ignition from either electrical discharge or hot
surfaces.
Faulty electrical equipment may however represent a source of ignition. A break in a

live cable and arcing in the ionised air may initiate ignition. The over loading of a con-
ductor followed by overheating and arcing due to excessive voltage is a further cause
of ignition.
Mechanical Sparks
Sparks from metal tools, falling objects and cigarette lighter flints may be sources of
ignition.
Scandpower
Static Electricity
Static electricity is an important source of ignition in process plants. There have been
many apparently mysterious explosions where the cause eventually has been traced
to static electricity, Ref. /5-27/. Electrostatic charge generation occurs primarily when
two dissimilar materials are brought together and then separated. This occurs with
solid-solid, liquid-liquid or solid-liquid surfaces. It does not occur with gases except
where the gas carries a fine powder, in which case the mechanism is solid-solid.
Hot work, Maintenance
Hot work activities like welding, cutting , grinding, burning, etc. obviously represent
potential sources of ignition. In welding this applies not only to the welding flare or
arc, but to material ignited by the welding.
Chemical Energy
The decomposition of a chemical or the reaction of two chemicals may be sufficiently

exothermic to raise the temperature to the auto ignition temperature.
Scandpower
Risk M anagement AS G) SINTEF
/5-1/ Bjarne Sivertsen: "The Application of Gaussian Dispersion Models at NILU",
NILU Technical Note No. 11 /80 , August 1980.
/5-2/ US Department of Health, Education and Welfare: "Workbook of Atmospheric

Dispersion Estimates", PB-191 482, May 1970.
/5-3/ TNO Yellow Book: "Methods for the Calculation of Physical Effects of the
Escape of Dangerous Material", 1979.
/5-4/ Bjarne Sivertsen: "Estimation of Diffusive Hydrocarbon Leakages from Petro-

chemical Factories".
/5-5/ Y. Rion: "Methods for describing heavy Gas Dispersion in the Environment of
Industrial Sites", 5th International Symposium of Loss Prevention and Safety
Promotion in the Process Industries, July 1986.
/5-6/ L.S. Fryer: "DENZ - A Computer Program for the Calculation of the Dispersion
of Dense Toxic or Explosive Gases in the Atmosphere", SRO R 152, Safety
and Reliability Directorate, UKAEA, March 1985.
/5-7/ S.F. Jagger: "The Application of the Computer Code DENZ", SRO R 277,
Safety and Reliability Directorate, UKAEA, March 1985.
/5-8/ C.J. Wheatly et al.: "Comparison between Data from the Thorney Island heavy
Gas Trials and Predictions of simple Dispersion Models", SRO R 355, Safety
and Reliability Directorate, UKAEA, February 1986.
/5-9/ S.F. Jagger: "Development of CRUNCH - A Dispersion Model for continuous

Release of a Denser-than-Air Vapour into the Atmosphere", SRO R 229,
Safety and Reliability Directorate, UKAEA, January 1983.
/5-10/ H.W.M. Witlox: "User's Guide for the HEGADAS - Heavy Gas Dispersion Pro-
gramme", Shell Oil Company, August 1988.
/5-11/ J.S. Puttock: "Comparison of Thorney Island Data with Predictions of HEGA-
BOX/HEGADAS", J. Hazardous Materials, 16 (1987) 439-455.
/5-12/ "Manual of Industrial Hazard Assessment Techniques", World Bank, October

1985.
/5-13/ 5th International Symposium: "Loss Prevention and Safety Promotion in the
Process Industries", Societe de Chimic lndustrielle, July 1986.
/5-14/ G. Opschoor: "Investigations into the Evaporation of liquified Gases spilled on

Water", TNO Report 80-05624, April 1980.
/5-15/ P.J. Waite, R.J . Whitehouse, E.B. Winn, W.A. Wakeham: "The Spread and
Vaporisation of Crygenic Liquids on Water", J. Haz. Mat. 8 (1983) 165-184.
/5-16/ N.O. Jensen: "On Cryogenic Liquid Pool Evaporation", J. Hazardous Mat. , Vol.
8, p. 157 - 163, 1983.
Scandpower
/5-17/ Oyez Publishing Ltd.: "Hazard Identification and Control in the Process Indu-
stries", 1981.
/5-18/ AJ. Prince: "A Users Manual to SPILL", SRD R 210, July 1981.
/5-19/ M. Considine, G.C. Grint: "Rapid Assessment of the Consequences of LPG

Releases", Proc. of GASTECH Conference, Amsterdam, p. 187-195, Novem-
ber 1984.
/5-20/ F.P. Lees: "Loss Prevention in the Process Industries", Butterworth, London,
1980.
/5-21/ AD. Little: "Evaluation of LNG Vapour Control Methods", AGA Report No. 76
285, October 1974.
/5-22/ AD. Little: "Simplified Methods for estimating Vapour Concentration and Dis-
persion Distances for continuous LNG Spills into Dikes with flat or sloping
Floors", AGA Report No. X50978, April 1978.
/5-23/ T.K. Fannel0p, K. Sj0en: "Hydrodynamics of Underwater Blowouts", Report

R-97-80, NSFI (Norsk Skipsforskningsinstitutt), January 1980.
/5-24/ H. N0strud: "Strnmningsteknisk vurdering av gassutblasningen pa Halten-

banken", vedlegg nr. 3 i West Vanguard rapporten, NOV 1986:16, March 1986.
/5-25/ M. L0es, T.K. Fannel0p: "Concentration Measurements above an Underwater

Release of Natural Gas", SPE Drilling Engineerin, June 1989.
/5-26/ John Bond: "Sources of Ignition, Flammability Characteristics of Chemical and

Products", 1991.
/5-27/ Frank P. Lees: "Loss Prevention in the Process Industries".
/5-28/ "The SFPE Handbook of Fire Protection Engineering", September 1988.
Scandpower
6. Gas Fire
6.1 General
When a flammable gas is released into the atmosphere, different kinds of fires may
take place dependent on the release mode and the degree of delayed ignition. Thus,
it is convenient to divide gas fires into the following types:
Flash fire or cloud fire

Jet fire or flare fire
Diffusive gas fire.
"Flash Fires" or "Cloud Fires" result from a delayed ignition of a release of gas or
vapour forming a cloud, which disperses downwind. There is a low probability for
detonation of the cloud provided that no flame accelerating obstacles or confinement
are present in the location of the cloud. However, flash back or burn back may occur.
That is, the flame consumes the premixed portion of the cloud and propagates back
toward the fuel source without creating any pressure or blast effects. Thus, a flash fire
is a highly transient fire, which has a very short duration usually less than 1 minute.
The main hazard from a flash fire is thermal radiation to human beings.
A "Jet Fire" results from a high-pressure leakage of a flammable gas. Because of the
high exit velocities of the gas, jet fires are often termed as "momentum controlled"
fires due to the fact that the momentum force prevails over the buoyancy force in large
parts of the flame plume. However, at some distance from the exit, the gas velocity is
reduced to a level at which the buoyancy force is dominant. The jet fire is from that
point more like a buoyancy controlled fire. The term "Jet Fire" is also often used in
case of less momentum controlled fires. However, the term "Flare Fire" is more con-
venient for these types of fires. The terms "Jet Fire" and "Flare Fire" are often used
interchangeably.
A "Diffusive gas fire" is a fire, which results from a diffusive release of a flammable gas
through a comparatively large opening. A diffusive gas fire will emerge from the
opening if the gas is being ignited. In contrast with jet fires, a diffusive gas fire is a
"buoyancy controlled" fire because of the diffusive release of gas. The buoyant force
is the dominant force in the entire flame plume.
6.2 Flash Fire

6.2.1 Scenario
A flash fire may occur if there is a release of a flammable gas or vapour in the open
forming a gas cloud followed by a delayed ignition of the gas cloud. The gas or
vapour release can either be continuous or more or less instantaneous. The first case
may result from a continuous gas release or a liquid release from which there is a
considerable amount of gas or vapour flashing from a not ignited liquid pool. A failure
of a vessel containing LNG or LPG may cause this situation. There will be a consider-
ably flash evaporation from the liquid pool around the vessel. Thus, a continuous
vapour cloud is formed downwind from the fuel source (see Figure 6.1).
Scandpower
~SINTEF Risk M anagem ent AS
In recent years, there has been an increasing tendency to transport and store liquefied
gaseous fuels (e.g. LNG and LPG) in large vessels. The term "flash fire" has there-
fore become synonymous with the term "Vapour Cloud Fires" in spite of the fact that
the latter fire scenario may result from a pure gas release. In Figure 6.1, a schematic
view of vapour cloud burning is presented.
---- -- \
Figure 6.1: Model of Vapour Cloud burning (where v, is the Flame Velocity and W, is
the Width of the Flame Front, from Croce et al. Ref. /6-1 /)
If wind is present, the gas cloud disperses downwind and entrains air, which mixes
with the gas. Sooner or later the gas cloud may encounter an ignition source.
A presupposition for ignition of the cloud is that the concentration within the cloud is
within the flammability limit range of the released substance. If a flash fire takes lace
within a complex structure of obstacles or confinement causing an acceleration of the
flame front, it may result in an explosion with comparatively large pressure effects
compare to a flash fire.
A characteristic feature of flash fires is that they have generally low combustion rates
compares to gas or vapour cloud explosions. The combustion rate determines
whether blast effects will be present or not. Flash fires, apart from explosions, do not
cause any blast effects. The flame travels backwards to the fuel source through the
premixed regions of the cloud.
Mudan et al., Ref. /6-2/, report on the basis of experiments that with LNG on water
three stages of the flash fire development are identified. First, a transient turbulent
flame spreads through the cloud. The flame propagates in both upwind and downwind
Sc and power
direction. The flames are initially contained within the cloud, but subsequently extend
in the form of a flame plume above the gas cloud. Thus, the cloud burns back to the
source as a diffusion flame and a premixed flame simultaneously. The second stage
is the steady-state propagation toward the liquid pool. The third and final stage results
in a pool fire at the location of the spill. Mudan reports that the flame propagation
velocity relative to ground in case of LPG spills varies between 4 - 5 mis.
The flame propagation velocity depends mainly on the local gas concentration and the
wind speed. In cloud fires the flame moves through the cloud as a yellow wall-like
diffusion flame with a blue premixed flame propagating underneath. The height of the
diffusion flame depends on the height of the cloud. The thickness of the cloud varies
approximately inversely with the height of the diffusion flame. The flame wall propa-
gates in an unsteady way as it encounters varying gas concentrations.
In case of a more or less instantaneous release of a large amount of gas and provided
a comparatively early ignition of the formed gas cloud, a flash fire which has the shape
of a ball of fire 1> may result.
This type of flash fire tends to start to burn around their outer envelope and will "lift off'
to form a ball of fire. Such fires may be extremely dangerous for people in the vicinity.
Flash fires produce burning mushroom clouds, which form upward convection current
which can suck up and ignite debris, and scatter burning brands over a large area.
Further, flash fires radiate intense heat, which can be lethal. Another lethal effect is
the depletion of oxygen in the atmosphere caused by a flash fire. Such flash fires
may, according to Eisenberg et al. Ref. /6-5/, also cause considerable damage to pro-
cess plants. However, in this text, thermal radiation hazards will be considered only.
6.2.2 Flame Shape

Eisenberg et al. , Ref. /6-5/, have proposed a model which assumes the flash fire to be
a half ellipsoid. In this model the volume Vr and area of radiation Ar of a flash fire are
given by the following equations:
Eq.(6.1)
Ar = 21t (crx2 + cry2 + crz2) ( r1

2
+ ru2 ) Eq. (6.2)
3
with:
1n
2m
r1 = 21n
[ (c2,) 312
a, a, a, k,) ]
Eq . (6.3)
[ ((2•)
ru = 21n 3/2
2m
a , a , a, kJ l
1/2
Eq. (6.4)
1) This type of fire has much of the same fire characteristics as the final stage of the fire scenario termed "'Fireball"
which is dealt with under liquid fires in Chapter 7 (see Chapter 7.5). This type of flash fire is often termed as "Fireball"
~ SINTEF Scandpower
Risk Management AS
where:
k, = the concentration at lower explosion limit (kg/m 3)

ku = the concentration at upper explosion limit (kg/m 3)
m = total mass of gas or vapour released (kg)
r, = a parameter of gas cloud at lower explosion limit(-)
ru = a parameter of gas cloud at upper explosion limit (-)
Ox = dispersion coefficient in the downwind direction (m)
Oy = dispersion coefficient in the crosswind direction (m)
Oz = dispersion coefficient in the vertical direction (m)
In some cases the flammability limits and explosion limits are virtually the same, but in
other cases there are substantial differences (e.g. upper limits of methane and lower
limits of hydrogen). For most gases and vapours, only flammability data are available
(Zabetakis Ref. /6-6/). Generally, the flammability/explosion limits widen with increa-
sing temperature and pressure.
The dispersion coefficients ox, o,, and oz can be predicted by methods proposed in
Chapter 5.3.2 for light and neutral gases and in Chapter 5.3.3 for heavy gases.
6.2.3 Flame Temperature

The temperature in a flash fire varies very much during its duration. The initial flame
temperature may, according to Eisenberg, Ref. /6-5/, be set equal to the adiabatic
flame temperature. Since air is rapidly entrained into a flash fire, flame temperatures
are also accordingly decreasing. The flame temperatures are at the end of the fire
most likely approaching flame temperatures of ordinary diffusion flames, i.e. well
below 1,500 °C. In fact data (Ref. /6-4/) show that that the emissive powers (and
effective flame radiative temperatures) of flash fires are similar to those of for jet fires
and fireballs (see Chapters 6.3 and 7.5, respectively).
6.2.4 Heat Transfer Assessment

Since a flash fire has a very short effective fire duration (i.e. in the order of few
seconds), it is not necessary to consider the heat transfer to structures. Further, due
to the fact that flash fires usually are elevated from the ground, personnel will only re-
ceive incident radiative heat flux. It has been customary to assume that personnel
directly exposed to the flash fire will be killed immediately. Thus, only radiative heat
transfer to external targets will be considered.
Due to the very short duration and the highly transient characteristics of flash fires, the
average incident heat flux received by an observer is a complex function of many
factors. For example, it is hard to estimate an appropriate surface emissive power
and in what time interval this emissive power is valid.
The net heat loss from a flash fire, Q (kW), is mainly by radiation which is given,
according to Eisenberg et al. Ref. /6-5/, by the following equation:
Eq. (6.5)
Scandpower
Risk Management AS (Gi)SINTEF
where:
Eg = emissivity of the burning gas cloud (-)

Tg = effective radiation temperature of the flash fire 2) (K)
emissivity of the environment (-)
ambient temperature (K)
a = Stefan Bolzmann's constant= 56.7-10-12 kW/m2 K4
Since the emissivity of both the burning gas cloud and the environment can be set to
unity, Eq. (6.5) will be simplified to the following equation:
Q::::: Ar (1 (Tg 4- Ta 4) Eq. (6.6)
The net effective thermal radiation heat flux, q"r (kW/m 2), to a target at some distance
from the flash fire is given by:
Eq . (6.7)
where:
F = is the view factor between the flash fire and the target (for close target, the
view factor can be set equal to unity (-). Otherwise, the point source
model presented in Appendix H, Eq. (H.9) and Eq. (H.10) has to be used.
However, when considering the thermal response of human beings, the gross incident
heat flux has to be used. The gross radiative heat flux is given by:
"
qr.gross = crFTg4 Eq . (6.8)
6.2.5 Smoke
Due to the very short duration of flash fires, large amounts of smoke are produced
within a short period of time. Unfortunately, there is hardly any correlation available
which can predict the rate of smoke production in the duration of a flash fire. How-
ever, a simple estimate of average smoke production rate can be estimated by multi-
plying the total amount of gas or vapour released (in kg) with the stoichiometric air to
fuel ratio . Further, a production rate can be achieved by dividing the total smoke pro-
duction by the effective duration (in seconds) of the flash fire (see next chapter). The
resulting number is an estimate of the minimum smoke production rate of a flash fire.
It has to be pointed out that this is the minimum theoretical smoke production when
assuming stoichiometric combustion. In the real case a lot of excess air is entrained
into the cloud under the combustion process due to the strong upward convective
currents. Thus, the real smoke production can be many times the theoretical mini-
mum.
2) Can be set equal to the average flame temperature in the course of the effective duration of the flash fire
Scandpower
6.2.6 Duration
According to Eisenberg et al., Ref. /6-5/, the heat loss from the flash fire is given by:
Eq. (6.9)
where:
Cp = specific heat of the burning gas (kJ/kg)

p = density of the burning gas (kg/m 3)
By setting Eq. (6.7) equal to Eq. (6.9), results in the following equation:
dTg = - k (T4 - T4) Eq. (6.10)

dt 9 a
where:
k = A,o/pV,.
Eq. (6.10) has the following solution:
. t-
- -1- {[ tan -1(Tg)
- 11 (Tg·Ta)]
-n
2kT~ Ta 2 T 9 +Ta
Eq. (6.11)
9
• [tan•1(T ;) ...!.1n( Tg; ·Ta)]}
Ta 2 Tg,+Ta
where the subscript i means the initial value. The initial temperature of the hot gases,
T9;, is by Eisenberg et al., Ref. /6-5/, given to be the adiabatic flame temperature.
Eq. (6.11) may be rewritten in terms of the half-life time, 1112 (sec.), of the flash fire, or
the time at which:
Eq. (6.12)
Then the following expression for 11 12 may be deduced:
1 [ tan-1(J3+1)
1112=-- - - -tan ·1 P--ln
1 (J3+1)]
-- Eq. (6.13)
2kT~ 2 2 J3+3
where:
Scandpower
Risk Management AS (G»SINTEF
The effective duration of the flash fire, tett (in sec.), is by Eisenberg, Ref. /6,5/, given to
be:
teff = 3x t112 Eq. (6.13b)
6.2.7 Example
A total mass of 10,000 kg of LNG is released into the atmosphere: The vapour cloud
of LNG is formed. The cloud encounters an ignition source approximately 100 m from
the release point. The volume and area of radiation of the resulting flash fire shall be
predicted as well as its effective duration and average radiation intensity on the
ground just below the flash fire.
Solution
1. Define necessary input data:
Concentration at lower explosion limit: 3.5 kg/m 3

Concentration at upper explosion limit: 12.4 kg/m 3
A diabatic flame temperature (Methane): 1,950 °C = 2,223 K
Ambient temperature: 10 °C = 283 K
Density of hot gas layer (air at T 9 = 1,500 K): 0.25 kg/m 3
Dow nwind dispersion coefficient: 5.6 m
Crossw ind dispersion coefficient: 4.0m
Dispersion coefficient in the vertical direction: 3.8 m
2. Prediction of volume and area of radiation:
Eq. (6.3) predicts r1= 1. 70 and Eq. (6.4) predicts ru = 0.61 . Thus, from Eq. (6.2)
the area of radiation of the flash fire is predicted to: 423.5 m 2
and Eq. (6.1) predicts a volume of the flash fire of: 833.4 m 3
3. Effective duration of the flash fire:
From Eq. (6.13) combined with Eq. (6.13b), the effective duration of the flash fire
can be predicted to: 24.5 sec.
4. Radiation intensity to a target just below the flash fire:
In this case the flash fire is rather close to the target and the view factor can be
taken equal to unity. For an average flash fire temperature of (1 ,950 + 10)/2 =
980 °C, the incident radiation intensity will be according to Eq. (H .11) when
assuming an emissivity of the burning vapour cloud of unity: 140 kW/m 2
Human beings with no protective clothing will achieve full blister within a second .
A heat flux of that level will be lethal within a very short time.
(t,J) SINTEF Scan dpower

Risk M anagem ent AS
6.3 Jet Fire

6.3.1 Scenario
A jet fire may result from a high-pressure leakage of gas from process plants or
storage tanks. An accidental leakage of gas may occur from flanges or ruptures in
pipes. Storage tanks or process vessels containing for example liquefied petroleum
gas (LPG) which is exposed to an enveloping fire, will after a very short period of time
vent their contents through a relief valve. If the released gas is ignited, a jet fire of
length 5 - 50 m can be the result. Even worse is the case of a gas blowout above drill
floor on an offshore production platform creating a free jet. Gas discharge rates even
above 100 kg/s may be burned almost completely in the jet fire.
Jet fires are characterised by a high momentum flow, which is highly turbulent. The
flame is lifted above the exit opening from which the gas is discharged generally at
very high pressures. This distance, which is often termed as "lift-off", appears be-
cause the combustion process can only take place when the flow velocity is reduced
sufficiently to allow stable combustion.
If the exit opening is sufficiently small, the flame plume will not stabilise before the gas
and the air mixture is below the lower flammability limit of the gas. Hence, the result
will be the phenomenon termed "blow-off''.
In practical situations this phenomenon can be observed by igniting a jet fire at low
exit pressure and then increasing the pressure gradually. The lift-off distance will
increase, until the fire turns off or "blows off''. There is a maximum exit diameter at
which blow-off may take place. For methane gas the maximum exit diameter is
approximately 25 mm. Gas leaks through smaller diameters may burn even though
the combination of the exit velocity and the diameter should result in blow-off for an
unrestricted discharge. Restrictions in the flow field may act as flame stabilisers and
lead to burning of the jet anyhow. In practise an ignited gas discharge from process
plants will always create a jet fire.
Another feature of such fires is the high entrainment of air into the flame plume due to
highly turbulent flame. An air entrainment from 4 - 5 times of that required for stoichio-
metric combustion is reported. Due to the efficient mixing of fuel and air, such flames
may exhibit higher flame temperatures than ordinary buoyancy controlled diffusion
flames. Thus, such fires may create very high heat radiation zones in the vicinity of
the fire.
Due to the extremely high flow velocities, the convective heat transfer to objects en-
gulfed by flames may be significant compare to radiative heat transfer.
The jet fires scenarios previously described are a result of an accidental release of
gas. Similar fires may also occur in case of intentional disposal of unwanted gas in
flares. Flaring is the combustion process that has been the traditional method for dis-
posal of large quantities of unwanted gases and vapour in both the offshore industry
and the petroleum industry on land.
"Production flaring" is a method for safe disposal of unwanted gas in the oil industry,
while "Process flaring" is a method for disposal of gas, which has passed through
safety valves protecting various process equipment or tanks. The case initially de-
scribed in which large amounts of gas have to be disposed of safely in case of an
Scandpower
emergency situation such as fire, power failure, or overpressure in a process vessel, is

termed as "Emergency flaring".
Fires, which are a result of flaring , are often termed as a "Flare fires" rather than "Jet
fires". However, there is no sharp distinction between these two terms and they are
often used interchangeably. In flaring the gas is released through nozzles which may
cause supersonic exit velocities of the gas. It is impossible to achieve supersonic exit
velocities in cases of accidental gas releases in which the gas is discharged from rup-
tured pipes, flanges or through holes in a process vessel. The maximum exit velocity
in this case will be the sonic velocity (the velocity of sound of the gas at the gas exit
conditions). The exit velocity is sonic if the upstream pressure (i.e. pressure of fuel
source) is higher than 1.7 - 1.9 bar for most gases3l. Consequently, a sonic exit velo-
city is achieved in most actual leak situations from process plants and storage tanks.
y = isentropic exponent of the fuel (-) =cp/Cv
6.3.2 Flame Shape

In practice a fire resulting from a high-pressure gas release will have a shape of a
frustum of a cone with an opening angle of 20-25°. In case of large jet fires there may
be smoke production from regions of the flame plume, which to certain extent may
cover or shield large parts of the flame plume.
Unfortunately, almost all analytic methods available for predicting jet fire characteri-
stics are developed for the design of flare systems. S0nju and Hustad, Ref. /6-7/,
have developed empirical correlations for turbulent jet diffusion flames. However,
these methods are primarily applicable to subsonic jets released into quiescent air.
The methods of Brzustowski, Ref. /6-8/, Kalghatgi, Ref. /6-9/, Chamberlain, Ref. /6-
10/, and McMurray, Ref. /6-11 /, all include wind effects. These latter methods are all
primarily developed for designing of flare systems in order that radiation intensities or
safe distances with respect to incident radiation to human beings can be predicted.
Thus, it is necessary to calculate the wind effects on the size and tilt of the jet or flare
fire.
In case of wind, two well-known and recognised methods will be presented here,
namely the method of Brzustowski, Ref. /6-8/, and Kalghatgi, Ref. /6-9/. The former
method is very simple in use because it uses the "point source" model (see Appendix
3) Sonic exit flows are achieved when the following relation is satisfied:
-P. < [- 2
Pv Y + 1
r:,
where: Pv = pressure offuel in the exit (Pa)
Po = ambient pressure (Pa)
y = isentropic exponent of th e fuel (-) = c,Jcv
Gp = specific heat at constant pressure (kJ/kgK)
Cv = specific heat at constan t volume (kJ/kgK)
Minimum upstream pressures in order to ensure sonic releases for different gases provided ambient pressure of
1.o.1a5 Pa:
propane: 1. 74 bar
m eth ane: 1.85 bar"
b utane: 1. 72 bar
Scandpower
H) for predicting the incident radiation intensities to external targets. The method of
Kalghatgi, Ref. /6-9/, is intended for the "solid flame model" (see Appendix H) which is
more complex in use (mainly due to the view factor calculation), but yields more accu-
rate results in the near field of the fire. It is important to predict accurate results near
the fire when evaluating safe distances for human beings or separation distances for
storage of other hydrocarbon fuels.
1. A calm wind situation
In a calm wind situation the well known model of Hawthorne et al., Ref. /6-12/, may be
used. It predicts the effective visible flame length, L - s (i.e. the distance from the
orifice to the visible flame tip minus the lift-off distance}, of a turbulent jet released into
still air. This model is given by the following expression:
1/2
L-s 5.3 Tad Mo
- = -[
De C, C<Tv
-[c,+(1-c,>-]
Mr ]
Eq. (6.14)
where:
De = the orifice or exit diameter (m}

s = the lift off distance (m}
c, = mole ratio of fuel to reactants (-)
Cl = mole ratio of reactants to products (-)
Tad = adiabatic flame temperature (K)
Tv = temperature of the fuel before it is released (K)
Mo = molecular weight of air =29 (kmol/kg}
Mr = molecular weight of fuel (kmol/kg)
Since c, << 1, a "' 1 and Tad/Tr varies between 7 and 9 for most actual fuels 4l, Eq.
(7.14) may be simplified to the following approximated equation:
1/2
L-s = ~ M0
De C 1 Mr ( J Eq. (6.15)
McCaffrey et al., Ref. /6-13/, state that for choked releases (i.e. release with an effec-
tive Mach number, Mer, greater than unity) the exit diameter, De, has to be replaced by
a hypothetical nozzle (i.e. a convergent-divergent nozzle) which has a diameter, Der,
given by the following equation:
y+1
4
Der = _1~[2 + (y -1) M~r] (y-1)
Eq. (6.16)
De fr,jJ:; y+1
4) Brzustowski, Ref. /6-81, gives the following values for a, C1 and T,ct!Tt for three common fuels:
Fuel a C
Methane: 1.00 0.091 7.2
Ethylene: 1.04 0.074 9.0
Propane: 0.96 0.038 7.6
Scandpower
where the effective Mach number, Met, is given by the following equation:
Eq. (6.17)
where:
y = =
isentropic exponent of the actual gas cp/Cv (-)
Pv = upstream pressure at the event orifice (bar)
Po = atmospheric pressure (bar)
The lift-off distance, s, may be determined from the following equation by Kent, Ref.
/6-14/:
Eq. (6.18)
w here:
Ue = exit velocity (m/s)

Ua = the average jet velocity (m/s) ~ 0.4•Ue
The effective exit velocity which is the effective exit Mach number, Met, times the local
sound speed , is given by the following relation :
Ue = Met y -1 2)
y RTv / ( 1 + - - Met Eq. (6.19)
2
where Tv is the upstream gas temperature and R is the individual gas constant of the
actual gas which can be found by dividing the universal gas constant(= 8.314 kJ/kgK)
by the molecular weight of the gas.
The following equation for the diameter of the jet fire flame plume is given by Baron,
Ref. /6-15/:
112
Di = 0.29x [1n (~)] Eq. (6.20)
and
D jmax =0.12L at X =0.16L Eq. (6.21)
where:
L = the distance from the orifice to the visible flame tip (m)
X = the axial distance from the orifice (m)
Scandpower
2. The wind situation
A. Method of Brzustowski
Brzustowski, Ref. /6-8/, has developed the following stepwise procedure in order to
determine the co-ordinates of the jet fire flame tip (see Figure 6.2):
1. Required information
Molecular weight of fuel: Mr (kmol/kg)
Molecular weight of the ambient air M0 (kmol/kg) = 29 kmol/kg
Exit diameter De (m)
Density of fuel: Pr (kg/m 3 )
Exit gas velocity ui (m/s) (often set to 220 mis or predicted by Eq. (6.19))
Wind speed Uw (m/s)
Ambient density p0 (kg/m3 } = 352.17/T0
Ambient temperature TO (K)
Lower flammability limit of the actual gas CL(-)
z
Flame
t .,,
h . ✓ I
Stack ✓ I
* , , X
,,
,' Target
. / (x,y,zt)
I ,
__________________________ J'
Figure 6.2: Model of a tilted Jet Fire (from Ref. /6-10/)
Scandpower
2. Calculate dimensionless lean limit concentrationSl:
Eq. (6.22)
3. Calculate the dimensionless co-ordinate SL of the concentration CL on the

flame axis:
If CL ~ 0.5: SL = 2.04( CL r1 °3
If CL > 0.5: SL = 2.51(CL)-0625
4. Calculate the downwind co-ordinate XL:
If SL > 2.35: XL = L - 1.65

If SL ~ 2.35:
Eq . (6.23)
In this case Figure 6.3 shows XL as a function of SL.
5. Calculate the dimensionless rise or vertical co-ordinate, ZL, of the tip of the jet
fire relative to exit opening:
28
Z L = 2.04 Xt Eq. (6.24)
5) The lowest concentration of fuel at the flame axis at which combustion is sustained
Scandpower
1-0~-----------~
01
t_,
I><
001
Figure 6.3: The Relation between Ek and XL for SL < 2.35 (from Ref. /6-8/)
6. Calculate the co-ordinates (X,Y,Z) of the flame tip in relation to the exit accor-
ding to the following equations:
Eq. (6.25)
Eq. (6.26)
The length of the flame, L, measured from the exit to the flame tip is then given by
(usually Y = 0):
L= ✓x2 + y2 + z2 Eq. (6.27)
Flame Tilt
By combining the Eqs. (6.24), (6.25) and (6.26), the flame tilt (i.e. the angle with res-
pect to the vertical), ex, is given by the following equation:
Scandpower
0 72
tga =( ; ~) =0.49 XL · Eq. (6.28)
B. Method of Kalghatgi
Kalghatgi, Ref. /6-9/ has on the basis of a series of large-scale experiments developed
empirical correlations for the jet fire characteristics by the influence of wind. This
method includes a collection of correlations describing the variation of flame shapes
and flame tilt under a wide range of ambient and gas flow conditions. This model
idealises the flame plume as a frustum of a cone (see Figure 6.4).
The following jet fire characteristics are determined by the method of Kalghatgi, Ref.
/6-9/ (see Figure 6.4):
The angle (in degrees), as, subtended by the burner tip and the tip of the flame
with respect to the vertical
The angle (in degrees), ex, subtended by the flame axis with respect to the verti-
cal
The vertical length of the flame, Lsv, measured from the exit hole to the flame tip
The width of frustum base, W 1
The width of frustum tip, W 2 .
VERTICAL
_,.._HOLE
., AXIS
Figure 6.4: A simplified Sketch of the Flame Geometry of a Jet Fire idealised as a
Frustum of a Cone which shows the Parameters defining the Size and
Location of the Flame (from Ref. /6-10/)
Scandpower
(r,J) SINTEF R isk Management AS
Flame Tilt
16
eta =94- · -35R Eq. (6.29)
R
1.1
a= 94 ---30R Eq. (6.30)
R
where:
R = UwfUe
Uw = wind velocity (m/s)
u. = exit or discharge velocity (given by Eq. (6.19)) (m/s)
Flame Dimensions
The following equation applies to the flame height, frustum tip and frustum base in
case of wind all relative to exit:
Vertical flame length Lav:
Lav _ 2.35 R Eq. (6.31)

- - 6 +--+ 20
Ds R
Width of frustum base w,:

0 22
W, = 49 - · - 380R + 950R 2 Eq. (6.32)
Ds R
Width of frustum tip W2 :
57
W 2 =BO- 0. -570R + 1470 Eq. (6.33)
Ds R
where the combustion source diameter is given by:
Ds =D e fi Po
Eq. (6.34)
where Pi is the density of fuel in the jet. This parameter can be predicted by the
following expression:
y-1
Pi = PoM, (Pv)' Eq. (6.35)

Ro Tv Po
where:
p0 = ambient pressure (bar)

M, = molecular weight of the gas (kmole/kg)
Scandpower
Ro = universal gas constant= 8.314 kJ/kgK

Tv = upstream temperature of gas (K)
The net or visible flame length, L, of the jet fire flame plume can be predicted by the
following expression using the dimensions in Figure 6.4:
L= Lav sin a.a

Eq. (6.36)
sin(90 - a.a) sin (180 - a.)
The lift-off distance, s, is given by:
2
s = L2 + L~ - 2LLa cos{a.-a.) Eq. (6.37)
where:
L - Lav Eq. (6.38)

a - sin (90 - a.a)
6.3.3 Flame Temperatures

After ignition of high-pressure gas release a conical flame plume may result in a
region in which the gas velocity is high. The combustion process occurs in a turbulent
mixing zone where the gas fuel and the air are efficiently mixed. This will lead to
effective combustion and to high temperature gradients within the flame plume; i.e.
from the cold air entrained at the flame envelope, via high temperature in the most
effective combustion zone, to a core zone where the fuel gas is in excess and with
reduced combustion. In the centre of the flame plume, close to the exit opening there
may even be a cold zone of not burned gas, i.e. in the "lift-off zone".
An unknown factor when it comes to enclosed jet fires offshore, is the temperature
distribution of the flame plume. The combination of a minimum heat loss to the en-
vironments and the local air to fuel ratio will determine how close to the adiabatic tem-
perature (a situation with no heat loss to the environment) the flame temperature will
be. In experiment carried out by the Norwegian Fire Research Laboratory in 1997 the
maximum temperature was as high as 1,380 °C, Ref. /6-16/ (see Chapter 6.3.8).
Enclosed fires may result in high temperature due to the reduced heat loss and more
ideal rates of mixture.
Table 6.1 shows the effective flame radiation temperature for several common gase-
ous fuels. This temperature can best be compared with the average flame tempera-
ture. However, the flame temperature will vary to a large extent both in the axial and
in the radial direction. Odgaard, Ref. /6-17/, reports some experimental results ob-
tained by Lenz et al. Ref. /6-18/. In these experiments the fuel was natural gas which
left the nozzle with a velocity of 70 mis. The diameter of the nozzle was 8 mm. Tem-
perature profiles in burning gas jet are shown in Table 6.5. According to Odgaard,
Ref. /6-17/, it appears from that the zone of maximum mean axial temperature is loca-
ted at about 60 % of the total flame length. Further, it can be seen that the maximum
flame temperature is located off-axis and that the maximum of the temperature fluc-
tuations is located further downstream than the maximum of the mean temperature.
This latter feature demonstrates, according to Odgaard, Ref. /6-17/, clearly the dyna-
mic character of the flame. Finally, Odgaard, Ref. /6-17/, concludes that a maximum
jet fire flame temperature of 1,500 °C is most realistic for most common gaseous
Scandpower
hydrocarbon fuels. However, as already mentioned, in experiments with enclosed jet

fires carried out by the Norwegian Fire Research Laboratory in 1997 the maximum
temperature was measured to 1,380 'C, Ref. /6-16/ (Ref. Chapter 6.3.8), which is a
somewhat lower temperature.
Table 6.1: Effective Flame Radiation Temperatures for several common gaseous
Fuels
Hydrocarbon fuel Effective flame
radiation temperature
(OC) .
Methane 1,016
Ethane 1,317
Ethylene 1,449
Propane 1,288
lsobutane 1,281
Normal butane 1,339
Propylene 1,217
lsobutylene 1,136
'f, rms (Kl 'i', rms (Kl

i600
·r
• rms
,200
600
L.00
0 1.0 60 120 150 200 0 2.S 5 7.5 10

x /d 0 r/d 0
( G) (bl
Figure 6.5: Axial (a) and Radial Distribution (b) of the Mean Temperature and RMS
(Root Mean Square) of the Temperature Fluctuations of Natural Gas Jet
Flems (from Odgaard, Ref. /6-17/)
An illustration of the temperature distribution in a vertical jet in a crosswind is shown in

Figure 6.6. It is shown that it may be difficult to select or predict a representative or
effective flame temperature for calculation of the heat radiation from the fire plume. If
a representative flame temperature shall be selected, it should be the integral of all
(E-T4) over the entire volume of the flame. An averaged effective heat radiation tempe-
Scandpower
Risk Management AS
~SINTEF
4 114
rature would then be (E•T ) . This temperature may be used in Eq. (H.1) instead of
the term d T + Ta.
Temperature
contours °C
600
800
1000
1200
1400
1600
Figure 6.6: The Temperature Distribution in a vertical Jet in a Crosswind. The Tem-
peratures are given in a vertical Cross Section through the Centreline of
the Exit Pipe, Ref. /6-23/

1. Heat Transfer to external Targets
The two methods proposed in Chapter 6.3.2 for the prediction of the flame dimensions
in case of wind namely the method of Brzustowski, Ref. /6-8/, and Kalghatki, Ref.
/6-9/, are well suited for the so-called 'point source model' and the "solid flame model",
respectively. The latter model is also most suitable for the calculation procedure
suggested for the calm wind situation.
"The Point Source Model"
'The point source model' that is described in Appendix H, assumes that the flame
plume is represented by a point source of thermal energy that normally is concen-
trated to the centre of the flame plume. This simplification is a better approach for jet
fires than for pool fires due to the usually lesser extent of a jet fire which is a result of
more intense burning. The Eq. (H.9) and Eq. (H.10) must be applied when predicting
radiative heat flux intensities. The fraction of the heat release resulting in thermal
Scandpower
Norwe gian Fire Research Laboratory
radiation, fR, in Eq. (H.10), will according to Brzustowski, Ref. /6-8/, attain the values
depending on the fuel given in Table 6.2.
Table 6.2: The Fraction of the total Combustion Energy resulting in Thermal Radia-
tion !from Ref. /6-8/\
Type of gas Chemical formula Radiative fraction
fR (·)
Hydrogen H2 0.2
Methane (C1) CH4 0.2
Ethylene (C2) C2H2 0.25
Propane (C3) C3Ha 0.3
Butane (C4) C4H10 0.3
C5 and higher 0.4
By means of the method of Brzustowski, Ref. /6-8/ which was described in the pre-
vious section, the co-ordinates of the flame tip relative to the orifice, (X,Z), can be
predicted. The co-ordinates of the flame centre, (Xe, Ze), are approximately given by:
Xe= 1/2 · X Eq. (6.39)
Ze=0.82-Z + h Eq. (6.40)
Where the factor 0.82 is½ raised to the power 0.28 in accordance to Eq. (6.24) and h
is the stack height according to Figure 6.2. The distance, S, from the flame centre to
the target with co-ordinates (X1, 0, Z1) is then given by:
Eq. (6.41)
When the distance S is predicted from Eq. (6.41) and the fraction radiated, fR, is taken
from Table 6.2, the incident radiation intensity, q to an external target is given by Eq.
(H.9) where x = S. The combustion efficiency, 1leomb., can in case of turbulent sonic jet
fires attain a value between 0.95 and unity. 1lcomb. = 1 will give a conservative result,
which is recommendable in this case.
"The solid Flame Model"
Even though the point source model yields satisfactory results even pretty close to a
jet fire, it is unsuitable in the near field. "The solid flame model" which is described in
Appendix H, idealises the flame plume as a cylinder. However, the method of
Kalghatgi, Ref. /6-9/ idealises the jet fire flame plume as a frustum of a cone. In order
to apply the calculated results to the solid flame model, the plume is transformed to a
cylinder with an equivalent diameter, Deq, which gives the same surface area as the
frustum of cone of same length L. Deq is given by:
Scandpower
Risk Management AS IG))SINTEF
Deq = ../L2 + 2A -L
r
where: Eq. (6.42)
2 1
A= :(wf +W~)+ ;(W1+W2) L +(Wz~W
where W1 is the width of the frustum base and W 2 is the width of frustum tip. The error
made by doing this approximation is by no means large. The uncertainties of the
models applied are assumed to be much larger.
When calculating the incident heat flux, q, to a target at a given distance, x, from the
exit orifice of the gas, the Eq. (H.11) in Appendix H has to be used. The atmospheric
transmissivity, -c, in Eq. (H.11) , is given by Eq. (H.13) when the distance, x, and the
relative humidity, r, of the atmosphere are given. The surface emissive power, Ep, is
given by Eq. (H.12) if the effective flame radiation temperature, Tr (from Table 6.2) , is
known. Table 6.3 shows T, and the corresponding emissive power, Ep, by assuming
black body radiation for different gaseous fuels.
Table 6.3: Effective Flame Radiation Temperatures and Emissive Powers by assu-
ming black Bod Radiation (from Mud an et al. , Ref. /6-4/)
Type of hydrocarbon fuel Temperature Emissive power
(K) (kW/m2)
Methane 1,289 157
Ethane 1,590 362
Ethylene 1,722 498
Propane 1,561 336
lsobutane 1,554 330
Normal Butane 1,612 383
Propylene 1,490 279
lsobutylene 1,409 223
When the gas release rate is small resulting in a flame plume which is not optically
thick, the emissive powers given in Table 6.3 must according to Eq. (H.12) be multi-
plied with the flame emissivity, e , given by Eq. (H.4) where Lm is equal to Deq• Accor-
ding to Croce et al. , Ref. /6-1/, the absorption-emission coefficient, k, which is included
1
in Eq. (H.4), varies between 6 and 10 m- for most hydrocarbons. Therefore, it can be
established from Eq. (H.4) that even jet flames of diameter 0.5 - 1 m may be assumed
to be optically thick.
Alternatively, the average surface emissive power, Ep (kW/m2), may also be calcula-
ted according to the follow ing equation:
t: = farht~Hc Eq. (6.43)

P At
~ SINTEF Scandpower
Risk Management AS
where:
fR = Fraction radiated of total energy released(-)

m = Gas release (kg/s)
£\He = Heat of combustion (kJ/kg)
A1 = Flame surface area (m 2)
The flame surface area for a frustum of a cone is given by:
Eq. (6.44)
Chamberlain, Ref. /6-10/, has found that fR correlates with the jet velocity, ui (m/s),
according to the following expression:
fa= 0.21e•o.ooa2au, + 0.11 Eq. (6.45)
As it appears from Eq. (6.45), the radiative fraction decreases when the jet velocity
increases. The advantage of this latter calculation procedure for prediction of Ep given
by Eq. (6.43) is that it takes into account the dependence of the exit velocity on the
radiative properties of the jet fire flame plume expressed by Eq. (6.45).
View Factors:
The view factor F is given by the Eqs. (1.2) - (1.6) or Figure 1.2 and Figure 1.3 in
Appendix I. The formulas given in Appendix I apply strictly speaking to circular pool
fires which may be idealised as tilted cylinders and to targets which are located at the
same level as the pool. This is the main reason for transforming the geometry frustum
of a cone predicted by the model of Kalghatgi, Ref. /6-9/, to a cylinder of equal radia-
ting surface area. Thus, the energy emitted per unit area will be the same.
Figure Figure 6.6 shows a simplified sketch of the approximated flame geometry and
the fire exposed target. One has to divide between two principally different vertical
locations. That is, above and below the horizontal line (termed A - A in Figure 6.6)
which goes through the point of intersection between the axis through the exit orifice and
the cone axis (termed P). This line will be the ground level in the procedure for calcu-
lating the view factor to a target not located at the ground level accounted for in Appendix
I.
The tilted cylindrical flame.plume, which has a diameter, which is given by Eq. (6.42),
is an approximation of the frustum of a cone. This cylinder is further approximated to
a cylinder with a horizontal base and horizontal top shown by the horizontal dotted lines.
The base of this tilted cylinder is coincident with the line a - b as shown in Figure 6.6.
The view factor between this cylinder a-b-c-d and the target is a reasonable approxima-
tion for the view factor of the jet fire flame plume. This view factor is equal to the view
factor of the cylinder e-f-c-d minus the view factor of the cylinder e-f-b-a. The target is a
ground level target for these two latter cylinders and can therefore be calculated by the
expressions given in Appendix I.
Scandpower
u
L
'·Jet Fire
Figure 6. 7: A Sketch of the approximated Flame Geometry where 0 = a. (from Ref.

/6-10/
2. Heat Transfer to Objects engulfed or hit by a Jet Fire
The convective heat transfer can, as already mentioned, be very high onto objects
which the jet fire is impinging onto due to the high flame temperatures and high flow
velocities. The rate of convective heat transfer is given by Eq. (H.3) in Appendix H.
The convective heat transfer is very dependent on the geometry end extent of the
exposed surface.
Eq. (H .7) which is an expression for the convective heat transfer coefficient for forced
turbulent flows, may be used in this case. However, due to the extremely high flow
velocities of sonic jets, this expression is not really intended for this case. Special
expressions have been developed for the convective heat transfer coefficient depen-
dent on the geometry and extent of the exposed surface, namely for pipes and large
surfaces at which the jet is impinging on more or less perpendicularly.
A. Heat Transfer to Pipes
According to Zukauskas et al. , Ref. /6-19/, the convective heat transfer coefficient on
the outside of the pipe is given by the following expression:
Eq. (6.46)
where:
Re = Reynolds number= puaD/µ (-)

Nu = Nusselt number= hD/k (-)
Pr = Prandtl number= µcp/k (-)
Tu = turbulence intensity(-)
Scandpower
p = density of gas 352.17/Tg (kg/m 3)

Tg = temperature of reacting gases (i.e. jet fire flame temperatures shown in
Table 6.2)
h = convective heat transfer coefficient (kW/m 2K)
D = diameter of the pipe (m)
k = thermal conductivity of the pipe (kW/mK)
µ = dynamic viscosity of air at the jet fire temperature (kg/ms)
cp = specific heat capacity of air at the jet fire temperature (kJ/kgK)
A turbulence intensity, Tu = 0.02 (2 %), has been found to agree quite well with expe-
rimental results.
B. Heat Transfer to infinite large Surfaces
One has to distinguish between circular shaped exits and exits which have the shape
of a slot.
Circular Exits:
Martin, Ref. /6-20/, has developed the following expression for average values of the
convective heat transfer coefficient, h, from stagnation point up to a radial distance r:
h _ De 1-1.1 De Ir F(Re) Pro.,2..'S_ Eq. (6.47)

r 1+0.1(x/De- 6)Delr D
where:
De = diameter of the exit hole (m)

x = the distance between the exit hole and the exposed surface (m)
r = is the radial distance along the surface hit by the jet from the centre of the
impingement centre (m)
k = thermal conductivity of the exposed surface (kW/mK)
Re = Reynolds number= pu.D/µ (-)
Pr = Prandtl number= µcp/k (-)
The function F(Re) may be represented by the following smooth curve approximation:
F(Re) = 2Re½(1 + Re0·55/200t Eq. (6.48)
The range of validity of Eq. (6.47) is given to be'l:
2,000 < Re < 400,000

2.5 s; r/De s; 7.5
2 s; x/De s; 12
Slot Exit:
For single slot exit holes the integral mean value of the convective heat transfer coef-
ficient may be correlated by the following empirical expression:
sJ The two dimensionless numbers Reynolds number, Re, and the Prandtl number, Pr, have the same definition as in
Eq. (6.47)
Scandpower
h= 1.53 Rem<xts. rts> pro.42~ Eq. (6.49)

xis + r/s + 1.39 0
where:
m(x/s,r/s) = 0.695 - [r/s + (xis }1- 33 + 3.06 r

1
Eq. (6.50)
where:
s = hydraulic diameter of the exit (m)

= 21U(I + t) where I and t are the length and width of the slot exit hole, res-
pectively (m)
x = the distance between the exit hole and the exposed surface (m)
r = is the radial distance along the surface hit by the jet from the centre of the
impingement centre (m)
The range of validity of these expressions is given to be:
3,000 ~ Re ~ 90,000
2 ~ r/s ~ 25
2 ~ xis ~ 10
The total heat transfer to the target, q, is given by Eq. (H.1) and Eq. (H.2) in Appendix
H.
Validation of Computer-based Prediction Models with Experimental Results
A major objective of the project termed "Blast and Fire Engineering for Topside Struc-
tures", Ref. /6-16/, was to provide data for validation of existing computer based pre-
diction models (both CFO and zone models) for predicting the thermal loads from fires
on offshore platforms (see Chapter 6.3.8). Three following three fire tests were selec-
ted for the model evaluation exercise:
An unconfined horizontal oil/gas jet fire

A vertical liquid condensate fire in a 135 m3 enclosure
A vertical liquid condensate fire in a 415 m3 enclosure.
All model predictions were submitted before the tests were conducted.
The level of detail in the model predictions varied significantly between different
models. The greatest detail was provided by the computational fluid dynamic (CFO)
codes. The predictions from each model were compared with the experimental results
for the following parameters:
1. Unconfined fire:
Thermal radiation
Flame geometry.
2. Confined fire:
Gas velocity
Gas and steel temperatures
Radiative and total heat fluxes.
Scandpower
The predictions of most parameters were in the range of 0.5 ,,; {Predicted over Obser-
ved} ,,; 2.0, with some significant exceptions in certain cases with certain models.
Predictions of heat flux were less accurate with most predictions within the range 0.2 ,,;
{Predicted over Observed} ,,; 5.0, with some significant exceptions in certain cases
with certain models. This model evaluation exercise did not attempt to rank the
models, but it did, however, provide information on the performance of models to
assist users in selecting a model that is fit-for-purpose.
The following limitations of the fire model evaluation exercise must be recognised:
Only three experiments were used in the comparisons and none of these experi-
ments were repeated
Small differences exist between the specified test conditions used by the model-
lers and the actual test conditions
A degree of interpretation was required for both experimental measurements
and some predicted values used in the evaluation exercise
The exercise was not repeated after the release of test data to modellers; nor
was it repeated using other data sets from the experiments.
Interactions between the Heat Loads, Object and Pressure Relief
If the distance from the exit opening of the jet and the exposed object is large, the jet
fire may be observed as a free jet. If a high velocity jet fire impinges on an object,
there may be a "cold" spot in the centre of the impingement point of the object due to
unburned gas or liquid spray.
In some cases a pool fire may be created due to rainout of liquid fuel. High exit pres-
sures and small exit openings may, however, lead to smaller droplets of the dis-
charged liquid spray. Consequently, the chances of rainout will be smaller.
A leakage of propane gas of 0.3 kg/s into a structure with a shape of a square box,
2
created maximum total heat fluxes of 300 kW/m 2 and means values of 200 kW/m on
the rear wall with the dimensions of 1.5 m x 1.5 m.
For enclosed jet the corresponding maximum heat flux onto the object will be equal or
greater than the heat fluxes mentioned above. This is mainly due to the fact that an
enclosed jet of same size will have a larger flame volume that exposes the object with
heat radiation. Maximum heat fluxes in the range 350-400 kW/m 2 has been measured
in enclosed jet fires in enclosed spaces with a volume of 415 m2 and a leakage rate of
1.0 kg/s of condensate fuel, Ref. /6-16/ (for more details see Chapter 6.3.8).
6.3.5 Smoke
There exist hardly any methods for predicting the smoke production from momentum
controlled fires such as gas jet fires which is the case of buoyancy controlled diffusion
flames. Since it is already established that the air entrainment is 4 - 5 times of that
required for a stoichiometric mixture of air and fuel, the flow of air and combustion
products, which form what we usually define as smoke, the mass smoke production at
the flame tip will be approximately 60 to 75 times the gas leakage rate m1 with a
temperature of 500 - 600 °C.
Due to further entrainment of air from that point, the mass production of smoke will be
increased, and the temperature of the smoke will decrease correspondingly. The re-
Scandpower
Risk Management AS !Gi)SINTEF
duced smoke temperature will decrease the volumetric flow rate of gas correspon-
dingly. However, the entrainment of more air into the smoke plume will cause an in-
crease in the volumetric smoke production of smoke.
As already mentioned in case of large jet fires there may be a smoke production from
regions of the flame plume, which to certain extent may cover or shield large parts of
the flame plume. This screening effect of the smoke may reduce the heat radiation
onto an external target considerably compared to the situation with no smoke produc-
tion from the flame plume. In large or full-scale experiments the measured surface
emissive power (SEP) at a certain distance from the fire may be significantly lower
compared to the calculated SEP by using Eq. (H.1) and the measured temperature as
the effective temperature of flame plume (~T + T0 ) .
6.3.6 Effects of Obstructing the Jet

Cowley, Ref. /6-21/, has carried out an extensive experimental programme in order to
study the impingement of large scale natural gas and propane jet fires onto cylindrical
targets. The gas flow rates varied from 2.6 - 20 kg/s, the orifice diameter varied from
10 - 150 mm and the nominal drive pressure varied from 1.2 - 58.3 bar. Both sonic
and sub-sonic gas releases were obtained. The flame length of the sub-sonic release
was typically 20 m, whilst the flame length of the sonic releases was in the range
30 - 35 m.
A 900 mm pipe target was located at distances of 9, 15 and 21 m from the orifice. Jet
flame impingement was not localised, while substantial engulfment of the pipe occur-
red . This behaviour was similar for both the propane and natural gas flames. The
smallest releases still covered a significant area with the flame wrapping around the
pipe, so that heating occurred on all surfaces of the target and not only on the front
surface. Peak heat fluxes were achieved within a few seconds.
At a source distance of 9 m from the target the maximum heat flux of 250 kW/m2 was
on the rear side of the target. This maximum heat flux was a combination of radiation
from the substantial downstream portion of the flame and a convective heat flux of
2
160 kW/m in the highly turbulent wake of the cylindrical obstruction. The heat flux to
the stagnation point on the front of the target was relatively low since partially burned
gases gave rise to a lower convective flux and a smaller radiative contribution.
At source distances of 15 and 21 m the maximum heat flux of 300 kW/m 2 was on the
front of the target at the stagnation point. At a distance of 15 m from the orifice the
target is within a fully established flame and the heat fluxes are more uniform over the
target circumference. At 21 m the rear of the target is subjected to lower heat fluxes
due to lower back radiation from the smaller portion of the flame extending down-
stream of the target.
Results from a number of sonic natural gas experiments are summarised in Figure 6.8
terms of target area exposed to heat fluxes ranging from 50 - 300 kW/m 2 for different
distances from the release point (all data are for 8.5 kg/s release through a 75 mm
orifice). It appears from the figure that the engulfed area is generally independent of
source to target distance. However, when the target was in its most distant remote
position, the source to target distance was equal to the flame length and the flame did
not wrap completely around the target. It was reported that from analysis of the video
records the area engulfed by the flame was found to correspond to heat flux levels
greater than approximately 50 kW/m2.
~ SINTEF Scandpower
Risk M anagement AS
N
20 >50 ............_
E I
~
>100 C
-,.,...
<
w >1S0
cc
<
C 10
w
LL
-1 >200
::,
~
z
w >300
0 10 20 tn
Figure 6.8: The Pipe Area subjected to Heat Fluxes greater than the given Value for
a natural Gas Flame as a Function of Distance from the Source (from
Cowley et al., Ref. /6-21 /)
The propane flames showed similar behaviour with a cold core giving low heat fluxes
near the stagnation point for the smallest source distance (i.e. 9 m). The maximum
heat flux of 200 kW/m 2 (of which 80 kW/m 2 was convective heat flux) was measured
on the rear of the target. Total incident heat fluxes ranged from 50 - 250 kW/m 2 •
These heat fluxes were, according to Cowley, Ref. /6-21/, largely radiative because
the strong soot emission increased the radiative component, whilst the lower gas tem-
peratures and velocities of propane jet fires resulted in lower convective heat fluxes.
6.3.7 Experience from Jet Fire Experiments

SINTEF has done tests with medium-scale propane jet fires impinging various objects.
One of the basic observations is that the maximum flame temperature does not coin-
cide with the maximum velocity within
the flame. The jet leaving a nozzle
with internal pressure above the criti-
cal for sonic flow expands as it en-
trains air, and the combustion takes
□---
---
place at a lift-off distance from the
nozzle. When the flames impinge on
an object from a relatively short dist-
ance, a doughnut-shaped ring of re-
-
circulating flames is created. In the
centre of this ring there is no visible
flame, and the temperature is low.
Further out from the centre, the flame
temperature increases, and the velo- Figure 6.9: View through the Centre Plane
city along the wall is at its maximum. of Jet Fire impinging a flat Plate
The corresponding flame temperature
Scandpower
NolVv'egian Fire Research Laboratory
varies along the surface of the wall, depending on the distance between the wall and
the nozzle, and on the gas flow rate. There is no simple model to describe this wall jet
with combustion. However, by experiments of medium and large scale some of the
worst case situations have been traced.
An assumption is that the maximum heat flux density occurs where the convective
heat transfer is at its maximum. The maximum radiative heat transfer is additive to
this, assuming that the flames in front of the wall are optically thick and at a high flame
temperature.
To obtain this in a real case, the flow rate of gas has to be large enough to provide a
burning gas layer at a thickness of several meters in front of the wall. For the situation
shown in Figure 6.9 this means that the flame thickness where the flame is at its most
shallow part must be in the order of meters. This is obtained only in large fires, and
possibly in confined jet fires only.
To represent a larger jet fire with the

same leakage rate, a box-like con-
- Visible flame
struction is used to produce a thicker
flame. The same velocity field from D Unignited gas
the centre and outwards at the wall is
Orffloo
obtained, but the flame is much
thicker due to recirculation set up by
the side walls. In this case the maxi-
mum convective heat transfer occurs
at some position along the wall , as
the radiative heat transfer origins
from a flame with a thickness of up to
two meters. An experiment set up to Figure 6 _10: A Jet Fire impinging into a
produce this type of flame configura- Semi-confined Enclosure
tion is shown in Figure 6.10. This is
a way to reproduce the worst case conditions in a laboratory scale jet fire.
Experiments where velocity, temperature and heat flux densities were measured in
various positions in this semi-enclosed box were done by SINTEF in 1990-91 . Typical
values for these situations are compared with results from large-scale jets impinging
on a tubular, done by Shell Research Ltd and British Gas pie, at a test site called
Spadeadam. Temperatures, velocities and heat flux densities are in the same order of
magnitude, even if the British tests comprise leakage rates up to 20 kg/s of propane
(Wighus Ref. /6-22/).
Velocity Assessment:
The decay of velocity along the jet axis from the exit opening is controlled by the con-
servation of momentum. The initial momentum of the gas is transferred to the sur-
rounding air by entrainment. In the immediate vicinity of the opening an expansion
takes place. The gas expands from the initial pressure in the containment to atmos-
pheric pressure. If any entrainment of air takes place in this region is not verified, a
possible way to treat this is to assume no air entrainment. Mccaffrey, Ref. /6-13/ has
advised a method to replace the original exit diameter with a hypothetical nozzle with
an effective diameter, from which a free burning jet can be predicted, see Eq. (6.16).
Scandpower
~ SINTEF
Risk M anagement AS
Jet Fire Characteristics:
In the experiment propane jet was fired into a box-like target with dimensions height
H = 1.5 m, width W = 1.5 m, depth D = 0.45 m. The jet originated from a nozzle posi-
tioned at various distances from the wall of the target, at two heights from the lower part
of the box, and also at two angles of attack, Ref. 16-221.
Measured velocities in the SI NTEF experiments with a distance of 1.5 m between nozzle
and wall of the target are:
Unignited jet: "' 250 mis at centreline at a distance 0.15 m from exit
Flame velocity: 70-80 mis just before the jet hits the wall of the target
20-60 mis along the surface of the target
Corresponding temperatures:
Unignited propane:"' 30 °c .
Flame temperature: From 600 °c where the velocity is 70-80 mis,
to "' 1,150 °C where the velocity is in the order of 20-60 mis
Temperatures in flames, vertical distr.

IMPINGING JET, 1.5 M, 1/2 H, 1990-11-26
1 .s
1.4 .
□
1.J -
E 1.2 - □
.; -
•I.
0
1.1
□
•
"
1.0 -
►
0 0.9 - □
E
0 o.a .
••"0 □
0 0.7 -
E .
0 O.S □
I.
►
-
u
C
• 0.5
□
•
"'j•
0.4 -
0.3 - □
0.2 -
□
0.1 -
0.0
0 200 400 600 BOO ' ' 1000 1200
Te,rperat.ure c:C
Figure 6.11: Flame Temperatures of an impinging Jet Fire
A typical temperature profile representing the vertical distribution just in front of the
target is shown in Figure 6.11
Scandpower
Risk Management AS !Gi)SINTEF
Heat flux density measured at - Visible flame

various positions of the target has
been measured as total and
D
Unignited gas
radiative heat flux density. Total

heat flux density has been mea- Orifice
sured from 200 to 300 kW/m 2 .
The average total heat flux den-
sity over the wall surface was
about 240 kW/m 2 . The radiative Figure 6.12: Jet Fire impinging a Tubular
part of the total flux was in the
2
order of 150 kW/m close to where the maximum flux was measured.
These heat fluxes are indications of what can be expected in large-scale jet fires
impinging structural members and vessels, as shown in Figure 6.12. Higher heat
fluxes may be expected where the flames become thicker or get a higher temperature.
This may occur for enclosed or confined jet fires. This has not yet been experimen-
tally verified.
6.3.8 Results from the "Blast and Fire Project"

The Jet Fire Test Programme of the Blast and Fire project, Ref. /6-16/, comprised of
eleven vertical and four horizontal releases. However, all of the horizontal releases
and ten of the vertical releases were of condensate liquid with a molecular weight
averaging10 kg/kmol and a liquid density of 740 kg/m3 . Hence, only one vertical re-
lease with gaseous propane fuel was carried out for scaling and dependence compari-
sons.
In practice a high pressure condensate liquid fire, which may be termed as a "spray
fire", will broadly speaking have the same burning behaviour and have the same fire
characteristics as a gaseous jet fire at sonic releases. A sketch of the experimental
set-up is shown in Figure 6.13. The steady state discharge pressures at the jet exit
nozzle were typically in the order of 10 bar or greater.
hermocouple 20 T he rmocou ple 26 Therm ocouple 44
•
Release point Heat flux mete r Vertical target
Figure 6.13: A Sketch of the experimental Set-up of vertical Jet and Spray Fires in
the 135 and 415 m3 Compartments
~ SINTEF Scandpower
Risk M anagement AS
Early Stages of the Fire
During the early stages of the spay fires, a large diameter droplet formation and drop-
let rainout were created. The rainout formed as shown in small distributed pools on
the compartment floor, which burned as short duration pool fires.
During this period of the fire there was sufficient air for the jet to burn as it would in the
open. The heat release rates were similar to those in the corresponding open jet fire
experiments. The incident radiative and convective heat fluxes to surfaces were also
similar to those expected for surfaces placed close to or impinged by open jet fires, i.e.
typically in the range 200-300 kW/m 2 •
At a certain smoke layer temperature, which appeared around 900 °C, copious
amounts of soot particles were produced under ventilation controlled conditions. This
resulted in a time dependent reduction in the heat released by combustion, and, thus,
also the temperature rise of the smoke layer gases and heat transfer to surfaces.
Build up to "Steady" Fire Behaviour
As the mean temperature of the smoke and walls continued to rise, the dominant heat
loss mechanism from the compartment became mass heat transfer through the vents.
This feature was particularly important for insulated compartments where the heat loss
through the ceiling and walls are negligible.
During this period, which was typically of the order of 2 to 10 minutes, the rate of tem-
perature rise in the compartment reduced and the fire behaviour changed much more
slowly compared to the early stages of the fire.
Steady Fire Behaviour
After a period 10 to 12 minutes from jet fire ignition, the heat fluxes, wall and ceiling
temperatures, and smoke layer chemical decomposition tended toward "steady"
values. The net heat transfer to the ceiling and walls inside the compartment tended
toward zero and the depth of the smoke layer reached steady values. At this stage
the total net energy from combustion closely matched the total heat losses.
Values for all relevant parameters of the "steady" phase of the enclosed jet fire tests
will be presented in the subsequent sections.
Roof Temperatures
For all vertical releases the temperature rise of the region immediately above the
release nozzle was substantially lower than elsewhere in the upper regions of the
compartment, typically in the range 100-400 °C. This was true for all vertical conden-
sate releases independent of release height, and was due to unburned cool conden-
sate liquid fuel. However, in the propane gas jet fire test the ceiling immediately
above the jet attained temperatures in the region of 800 °C.
In the vertical release tests the maximum roof temperatures outside the region above
the release nozzle were between 900-1, 100 °C, while roof temperatures in the hori-
zontal jet fire tests were comparable in magnitude with those measured in the vertical
jet fire tests, typically ranging from 850-1,050 °C.
Scandpower
Wall Temperatures
In the vertical jet fire tests steady state wall temperatures ranged between 300 to
1,100 °C. The 300 °C temperature corresponded to the lower wall region in the full
vent test. The top region of the wall attained temperatures between 600 and 1,000 °c.
In horizontal jet fire tests, wall temperatures were typically between 800 and 950 °c
generally, but did reach 1,200 °C at the bottom of the wall.
Target Tern peratures
Two steel targets were located in the inner part of the compartment, near the west
wall. The hottest zones of the compartment were in the southwest corner, reflected by
the highest target temperatures. This was also supported by the damage of the rig
after this test. The maximum temperature reached in this test was about 1,320 °C,
which led to sagging of the tubular.
The temperature distribution of the vertical targets shows that for a ventilated fire, the
highest temperatures were measured at the upper part, shifting to the maximum tem-
peratures at an intermediate level for under ventilated fires. The temperatures of the
horizontal target show a trend of decreasing temperature with decreasing equivalence7
ratio .
Gas Temperatures
The gas temperatures ranged between 750 and 1,100 °C in the vertical jet fire tests. It
could be seen that for >1 higher gas temperatures were measured close to the fuel
release point than for >1, where maximum combustion intensity took place close to
the vent opening. It is reasonable to assume that as decreases the combustion
zone shifts from the region dominated by the jet towards the vent.
The highest gas temperatures measured was, however, 1,380 °C. Very high wall tem-
peratures and heat fluxes were also measured in this test resulting in many of the
thermocouples melting and partial destruction of the test rig.
Vent Temperatures
In the vertical jet fire tests steady temperatures measured at the bottom of the vent
ranged from 150 to 500 °C. The temperatures at the top of the vent ranged between
1,050 to 1,300 °C. Since no flames were present in the lower region of the vent, the
thermocouple readings in this region were mostly due to the radiation from the flames
emerging from the upper part of the vent opening.
7) The "Equivalence Ratio''. </J, referred to in the text is an average value based on arithmetic mean of independent
values of stoichiometry calculated separately from air flow into and gas flow out of vent area. The "Equivalence Ratio"
</J is defined as:
where:
ma = mass of air (kg)

mt= mass of fuel (kg)
r = mass ratio of air required for stoichiometric burning ( - )
~ SINTEF Scandpower
Risk Management AS
The main trend shows that temperatures throughout the vent rise with decreasing .
For >1 external flaming did only occur if the unconfined jet fire length was greater
than the distance between the nozzle and the vent plane.
Wall and Roof Heat Fluxes
The results of the vertical jet show that the position of maximum heat flux shifts from
near ceiling to near cold layer interface, as the fire becomes more ventilation con-
trolled. The heat flux is predominantly radiative in nature. The propane gas jet fire
tests indicate a larger convective component as expected because of the higher gas
velocities.
Similarly, the horizontal jet tests suggest that radiation dominates the heat transfer
and similar values of the maximum heat flux were observed. However, the roof values
of the heat fluxes appear to continue to rise with under ventilation unlike those derived
from the vertical jet fires.
The high heat fluxes of 330 kW/m 2 (time averaged) was amongst the highest mea-
sured in jet fires. This heat flux corresponds to a radiative temperature of 1,277 °C,
when disregarding the convective heat flux.
Target Heat Fluxes
The general trend was that the heat fluxes on targets were less than 200 kW/m 2 . For
certain conditions considerably higher heat fluxes were seen, i.e. in the order of
350-400 kW/m 2 . However, in all the other tests with similar conditions the heat fluxes
never exceeded 180 kW/m 2 .
Table 6.4 and Table 6.5 show a summary of the test results obtained from the jet fire
test programme of the blast and fire project.
Table 6.4: A Summary of the Test Results obtained from the Jet Fire Test Pro-
qramme of the Blast and Fire Proiect
Typical temperature values
Effect of
Effect Gas Wall Ceiling
temperature of
("C) ("C) ('C)
500-600
No significant dependency except 'cold
Scale 1120->1370 (bottom) 600-900
spot' temperature higher in small scale
750-950 (top)
550-700
No significant dependency. No cold spot
Fuel 1120->1370 (bottom) 600-950
observed for propane fuel
700-950 (top)
Weak dependency on <l> for gas and wall 300-600
Stoichiometry temperatures. No dependency for ceiling 1120->1370 (bottom) 1000
and vent temperatures 600-950 (top)
Vent configuration No significant dependency 900-1200 600-900 900-1000
550-650 800-1000
Release height No significant dependency 1200->1370 (bottom) (100-150
700-1000 (top) cold spot)
High effect on 'cold spot' temperature, no
Release pressure significant dependency for other tempera- 850-1350 750-850 750-1000
lures
Release No significant dependency. No cold spot in
1100-1300 750-1050 850-1050
orientation horizontal test
Scandpower ~
Table 6.5: A Summary of the Test Results obtained from the Jet Fire Test Pro-
gramme of the Bl ast and Fire Project
Effect of Typical heat flux values
Effect
temperature of Ceiling (kW/rn~) Wall (kW/rn~)
Scale No significant dependency 40 (cold spot) - 160 50-160
Fuel No significant dependency 90 40-170
Weak dependency on <l> for gas 50 for <1>=2.28 100 for
Stoichiometry ceiling heat flux. Strong dependency 130 <l>0.82 160 for =0.6
on for wall fluxes 170 for <t>=0.58
Vent configuration Significant dependency at 430 sec <150-250 10-100
Release height No significant dependency 110 40 (top of wall) - 170
No significant dependency (large
Release pressure 120 160 at top of wall
scatter in data)
Release No significant dependency (large 30-200 typically, but 30-
50-200
orientation scatter in data) 280 in one test
6.3.9 Example
Due to a fire exposure of a pressurised tank containing butane at 5 bar, the contents is
vented into the atmosphere from the top of the tank through a pipe of diameter
0.025 m. The gas is immediately ignited forming a turbulent gas jet. The exit orifice is
located 2 m above the ground. The heat flux onto the ground 10 m from the tank shall
be predicted by using the method of Brzustowski for wind speeds of 2 and 1O m/s.
The effects of increasing the height of the exit to 10 m on the incident heat flux on the
ground can be predicted. For comparison, the incident radiation will also be predicted
by using the method of Kalghatgi.
Solution:
Define necessary input data:
Molecular weight: 54
Heat of combustion: 45,000 kJ/kg
Lean limit concentration: 0.02
Gas vapour density: 2.25 kg/m 3
Discharge coefficient: 0.61 -
2. Calculate flame tip co-ordinates, flame length and flame tilt:
Eq. (6.22) predicts a dimensionless lean limit concentration of: 4.1 at 2 m/s wind
and 0.82 at 10 m/s wind
Since CL< 0.5, the dimensionless co-ordinate SL can be predicted to 1.04 and 2.84 for
wind speeds of 2 and 10 m/s, respectively .
Foe Uw = 2 mis, SL < 2.35 and the downwind co-ordinate XL has to be taken from
Figure 6.3: 0.1
For Uw = 10 m/s, SL > 2.35 and XL = SL - 1.65: 1.19
Scandpower
The dimensionless co-ordinate z, can be predicted from Eq. (6.24) to:

1.08 at 2 m/s wind
and: 2.14 at 10 m/s wind
The Eq. (6.25) and (6.26) predict the following co-ordinates (x,z) of the flame tip with
respect to exit orifice:
2 m/swind: X = 0.4 m
z = 4.1 m
10 m/s wind: X = 0.9 m

z = 1.6 m
Eq. (6.27) predicts a flame length of: 4.1 m at 2 m/s wind

and: 2.0 mat 10 m/s wind
Flame tilt:
5.3° at 2 m/s wind
and: 30° at 1O m/s wind
Prediction of incident heat flux on ground 10 m from the tank:
The Eq. (6.39) and (6.40) predict the following co-ordinates of the centre of the flame
plume (xc,Zc) of the flame tip:
2 m/swind: Xe= 0.2 m

Zc = 5.3 m
10 m/s wind: Xe= 0.5 m

Zc = 3.3 m
The distance S from flame centre to target is given by Eq. (6.41) and yields:
12.2 m at 2 m/s wind

and: 10.9 mat 10 m/s wind
The discharge rate of gas is predicted by Eq. (4.3) to: 0.56 kg/s
When assuming a burning efficiency of unity, Eq. (G.3) predicts the rate of heat
release to: 25,210kW
The incident heat flux is given by Eq. (H.9) and yields for a fR = 0.3 according to Table
6.2. The following results are obtained: 4.1 kW/m 2 at 2 m/s wind
5.1 kW/m 2 at 10 m/s wind
Effects of increasing the stack height:
When the stack height is increased by 10 m, the distance from flame centre to target
is: 16.5 mat 2 m/s wind
and: 18.0 m at 10 m/s wind
The incident heat flux in this case is according to Eq. (H.9):

Scandpower
Risk Management AS (Gl)SINTEF
Method of Kalghatgi for a wind speed of 10 m/s and a height of the exit orifice above
ground of 2 m:
Eq. (6.30) yields tilt angle with respect to vertical of: ex= 68.4°
Eq. (6.34) and Eq. (6.31) yield a vertical gross flame length: Lsv = 2.0 m
Eq. (6.32) yields a frustum base of: W1 = 1.0 m
Eq. (6.33) yields a frustum tip of: W2 = 1.5 m
Eq. (6.36) yields a visible flame length of: L = 3.3 m
Eq. (6.37) and Eq. (6.38) yield a lift-off distance of: s = 0.6 m
Eq. (6.42) yields an equivalent diameter of a cylinder with same
surface as the cone of: Deq = 1.26 m
Since the lift-off distance is predicted to 0.6 m, the target is located 2.6 m below the fuel
source. Hence, this is a target of type T2 as accounted for in Appendix I.
Here x = 10 m, z 2 = 2.6 m, 0= 68.4°, L = 3.3 m. According to the procedure presented in

Appendix I for targets of type T2 along with the data above, they yield a view factor of
0.012. According to Table 6.3, the surface emissive power of a butane jet fire flame
plume is 383 kW/m2 . The incident radiation heat flux is then according to Eq. (H. 11 )
4.6 kW/m2 .
The slightly higher incident radiation heat flux obtained by the method of Kalghatgi (the
Solid Flame Model}, i.e. 4.6 kW/m2 vs. 4.1 kW/m2 is solely due to the larger predicted
flame length (3.3 m vs. 2.0 m). As concluded in Appendix H, the Point Source Model
which has to be used when using the method of Brzustowski , will overestimate the inci-
dent heat flux in particular for close targets.
According to experience from jet fire experiments reported in Chapter 6.3.7 where the
gas release rate was 0.3 kg/s (vs. 0.56 kg/s in this example), a flame length of 3.3 m
(exclusive lift-off) seems to be more likely than a flame length of only 2 m (inclusive lift-
off) predicted by the method of Brzustowski.
6.4 Diffusive Gas Fires

6.4.1 Scenario
A diffusive gas fire may arise from a massive release of a flammable gas or vapour in
an enclosed space, which escapes diffusively out from one or several openings of the
enclosed space. The concentration of fuel gas is too high in the enclosure to cause a
fire within the enclosed space. The gas is burning outside the enclosure from its
openings where the fuel comes into contact with sufficient air. Buoyancy controlled
diffusion flames will emerge from the openings of the enclosure.
The preceding leakage scenario may in addition to a massive gas release also be a
continuous or more or less massive instantaneous release of an unstabilised oil that
flashes off considerable amounts of gas as well as light oil fractions due to a high tem-
perature of the oil.
Scandpower
6.4.2 Flame Shape

The flame plume correlation of Heskestad, Ref. /6-23/, is probably best suited for pre-
diction of the mean flame length from both a horizontally and vertically orientated
opening through which there is a massive flow of gas or vapour. This correlation was
originally developed for flows above a point source (e.g. a gas burner). However, it
has been shown that it also can predict the flame length of large pool fires as well. A
diffusive gas fire has usually a far more concentrated gas flow than a pool fire. Con-
sequently, the flame plume correlation should be well suited for this fire scenario.
The correlation for the average flame length to diameter ratio UD of buoyancy con-
trolled flame is given by the following expression:
_I:_= -1.02 + 15.6XN115 Eq. (6.51)

D
with:
Eq. (6.52)
where:
Cp = specific heat of air (kJ/kg)

To = ambient temperature (K)
Po = density of the ambient air (kg/m 3 )
g = acceleration of gravity = 9.81 (m/s 2)
iiHc = heat of combustion of the fuel (kJ/kg)
r = stoichiometric air to fuel ratio (-)
Q = rate of heat release by the fire (kW) (see Eq. (G.2))
D = equivalent diameter of the fuel source (i.e. the actual opening through
which the fuel will flow diffusely) (m)
This correlation is experimentally correlated with a non-dimensional parameter, N,

which may attain values in the range from 10-5 to 105 .
The stoichiometric air to fuel ratio, r, of a fuel with the general fuel molecule CmHnOp
(i.e. with m carbon atoms, n hydrogen atoms and p oxygen atoms) can be predicted
by the following equation:
r =34.33 4 m +n- 2 p Eq. (6.53)

12m + n + 16p
Assuming the value iiHJr = 3, 10081 and TO = 293 K, the following simplified equation
will be the result:
L Q21s
-=-1.02+0.23- Eq. (6.54)
D D
8) This value applies strictly speaking to methanol. However, this ratio does not vary appreciable among a number of
fuels. According to Heskestad, Ref. lo-221, it varies within 2,900 - 3,200 kJ/kg for a number of combusffbles
Scandpower
The equivalent diameter, D, of the actual opening from which a diffusive gas fire will
emerge is given by the following equation:
D= 2AB Eq. (6.55)

A +B
where:
A = the width of the opening for vertical openings or the largest dimension for
horizontal openings (m)
B = two-thirds of the height of the opening for vertical openings or smallest
dimension for horizontal openings (m)
If there is a wall above the vertical opening of length A, the following equation for the
equivalent diameter applies:
D= 4AB Eq. (6.56)

A+2B
The width, w , and depth, d , of the diffusive gas fire are defined as:
w=A
Eq. (6.57)
d = B
Flame Tilt:
If A> 2B the follow ing expression apply ing to trench pool fires may be used:
0.56 for F'r ~ 0.22

cos 0 = 0.36(~'r r 0 32
· for 0.042 ::; F'r ::; 0.22 Eq. (6.58)
{
for F'r ::; 0.042
where the modified Froude number, Fr', is given by:
F'r=~ Eq. (6.59)

2/go
If A~ B, the following correlation should be used:
1 for u· ::; 1
cas e = 1 f • Eq. (6.60)
{ JJ oru ~ 1
where the non-dimensional wind velocity, u·, is given by:
•_ Uw
Eq. (6.61)
113
u -( 4gmt ]
(A+ B)p9
Scandpower
~ SINTEF Risk Managem ent AS
where:
Uw = the wind speed (mis)

pg = gas density (kg/sm 3)
g = acceleration of gravity (m/s2)
When there is more than one opening in the enclosed space in which there is a
massive release of gas or vapour, the gas can be assumed to escape through all
openings which are located above the location of the exit orifice of the gas release for
light gases (i.e. lighter than the ambient air). The opposite is true for heavy gases.
The gas release rate through one specific opening will be approximately equal to the
total gas release rate times the fraction this opening represents of the total area of all
the openings.
6.4.3 Flame and Gas Temperatures

The correlations and calculation procedures presented in Appendix G can be used
when predicting the centreline temperature of the flame plume as well as the buoyant
gas plume. Eq. (G.3) in Appendix G has to be applied when predicting the total rate of
heat release by the fire. This is a key parameter in these correlations. The leakage
rate of gas, mr, which is included in Eq. (G.3), is taken as the total amount of gas re-
leased diffusively through the actual opening from which the diffusive gas fire is to be
predicted. Hence, a burning efficiency of 100 % may be assumed in case of diffusive
gas fires because the fuel is already in the gaseous state and there will be a good
mixing of oxygen and fuel molecules. The error by assuming this will by no means be
, large. The predicted flame and gas temperatures are supposed to be only slightly
/ conservative.

Since a diffusive gas fire is a buoyancy controlled fire, the calculation methods propo-
sed in Appendix H are to be used for predicting the heat transfer both to external
objects and objects which are engulfed by the fire.
When calculating radiative heat transfer intensities incident on targets some distance
away from the fire, it is most convenient to use the solid flame model and Eq. (H.11).
The surface emissive power, Ep, included in Eq. (H.11), can be set equal to a value in
the range 130 - 180 kW/m 2 . A value of 150 kW/m 2 is most likely.
In spite of the fact that the point source model is usually more suitable for diffusive gas
fires than for pool fires, it is not recommended to use this method when calculating
radiative heat fluxes from the fire.
6.4.5 Smoke
The method proposed in Appendix J may be used for the prediction of the smoke
production from diffusive gas fires.
Scandpower
6.4.6 Example
Within an enclosed space there is a release rate of 30 kg/s of a hydrocarbon fuel of
average chemical formula C 9.5 H21- The gas is ignited and a diffusive gas fire appears
from the only opening of the enclosure, which has a width of 15 m and a height of 3 m.
There is a high wall above the opening. The maximum incident heat flux onto a target
located 10 m from the opening in the horizontal direction and 4 m vertically below the
lower edge of the opening shall be predicted due to this diffusive gas fire.
Solution:
Define necessary input data:

Burning efficiency: 1 -
Ambient temperature: 20 °C
Density of the ambient air: 1.2 kg/m 3
Specific heat of air: 1.0 kJ/kgK
Relative humidity of the ambient air: 50 %
Prediction of flame geometry:
Eq. (6.53) predicts an air to fuel ratio of: 15.1 -
Since there is a high wall above the opening, Eq. (6.56) has to be
used in order to predict the following equivalent diameter of the fire: 6.3 m
Eq. (G.3) predicts a total rate of heat release of: 1,317,600 kW
Eq. (6.51) and Eq. (6.52) predict a flame length of: 60.2 m
Alternatively, the simplified Eq. (6.54) predicts a flame length of: 58.1 m
The error is by no means large by using the simplified correlation of Eq. (6.54) rather
than the more complicated expressions of the equations (6.51 ) and (6.52) .
Width of fire: 15 m
Depth of fire (two-thirds of the opening height): 2m
Since the width of the opening is much larger than its height (15 m vs. 3 m), Eq. (6.58)
and Eq. (6.59) have to be used to calculate the flame tilt: 0 = 56°
Prediction of incident heat flux:
Since the flames will fill the upper 2 m of the 3 m height opening, the vertical co-
ordinate of the target (i.e. of type T2 shown in Figure 1 in Appendix I) with respect to
the fuel source will be: =
z 2 -(4 m + 1 m) =-
5m
Horizontal co-ordinate with respect to the target: x = 10 m
By transforming the "flame cylinder" of rectangular base dimensions 15 m x 2 m to a

cylinder with diameter equal to the equivalent diameter (i.e. 6.3 m), the calculation
procedure applying to the calculation procedure for type T2 targets accounted for in
Scandpower
Appendix I have to be used. The equation (1.4) and (1.5) or alternatively the figures 1.2
and 1.3 predict a horizontal view factor:
Fh = 0.23
and a vertical view factor: Fv = 0.05

The maximum view factor will be according to Eq. (1.6): Fmax = 0.24
An atmospheric transmissivity of 0.89 is predicted by Eq. (H.13).
By using Eq. (H.11) with a surface emissive power of the diffusive gas fire of
2
150 kW/m and atmospheric transmissivity of unity: 32 kW/m 2•
'
I
I
Scandpower
/6-1/ P.A. Croce, K.S. Mudan: "Calculating Impacts for large Hydrocarbon Fires",
Fire Safety Journal 11 , pages 99 - 112, 1986.
/6-2/ K.S. Mudan, P.A. Croce: "Fire Hazard Calculations for large open Hydrocarbon
Fires", The SFPE Handbook of Fire Protection Engineering, ISBN 0-87765-
353-4, September 1988.
/6-5/ N.A. Eisenberg et al: "Vulnerability Model. A Simulation System for assessing
Damage resulting from Marine Spills", Nat. Tech. Inf. Service Report AD-A015-
245, Springfield, 1975.
/6-6/ M.G. Zabetakis: "Flammability Characteristics of Combustible Gases and

Vapours", U.S Bureau of Mines, Bulletine 627, Washington, 1965.
/6-7/ O.K. S0nju and J. Hustad: "An Experimental Study of Turbulent Jet Diffusion
Flames, Dynamics of Flames and React Systems", AIAA Progress in Astro-
nautics and Aerionautics, Vol. 95, pp. 320-339, 1984.
/6-8/ T.A. Brzustowski: "Flaring in the Energy Industry, Progress in Energy and
Comb.", Sci., 2, pp. 129-141 , 1976.
/6-9/ G.T. Kalghatki: "The Visible Shape and Size of a Turbulent Hydrocarbon Jet
Diffusion Flame in Crosswind", Comb. and Flame 52, pp. 91-106, 1983.
I
/6-10/ G.A. Chamberlain: "The Shell/Thornton Flare Radiation Model", 4th Int. Flare {
System Seminar, Trondheim, 1986. l
/6-11/ R. MucMurray: "Flare Radiation estimated", Hydrocarbon Processing, Kaldair
ltd., Feltham, England, November 1982.
/6-12/ W.R. Hawthorne et al.: "Mixing and Turbulent on Turbulent Gas Jets", 3rd Int.
Symp. on Comb., Comb. Inst., Pittsburgh, PA, pp. 266-288, 1949.
/6-13/ B.J. Mccaffrey and D.D. Evans: "Very large Methane Jet Diffusion Flames",
Centre for Fire Research, National Bureau of Standards, Gaithersburg, MD
20899, USA.
/6-14/ G.R. Kent: "Find Radiation Effects on Flares, Hydrocarbon Processing", Vol.
47, No. 6, 1968.
/6-15/ T. Baron: "Reactions in Turbulent Free Jets - The Turbulent Diffusion Flame",
Chem. Eng. Prag. 50, 1973.
/6-16/ Selby, C.A. and Burgan, B.A.: "Blast and Fire Engineering for Topside Struc-
tures - Phase 2, Final Summary Report", Steel Construction Institute, Silwood
Park, Ascot, Berkshire UK, 1998.
/6-17/ E. Odgaard: "Characteristics of Hydrocarbon Fires in the Open-Air", Veritas

Report No. 82-0704, Det norske Veritas Research Division, Norway, January
1983.
/6-18/ W. Lenz, J. Gunther: "Measurements of Fluctuating Temperatures in a Free-

Jet Diffusion Flame", Combustion and Flame 37, 63, 1980.
Scandpower
{r,J) SINTEF Risk Management AS
/6-19 / A. Zukauskas and J. Ziugzda: "Heat Transfer of Crossflows", p. 88, Hemi-

sphere, Washington D.C., 1985.
/6-20/ H. Martin: "Heat and Mass Transfer between impinging Gas Jets and Solid
Surfaces, Advances in Heat Transfer, 13, pp. 1-60, 1977.
/6-21/ L.T. Cowley et al.: "Thermal Impact on Structures from large Scale Jet Fires",
Institute of Mechanical Engineers.
/6-22/ R. Wighus: Paper to be presented, 7th International Symp. on Loss Prevention

and Safety Promotion in the Process Industries, SINTEF Norwegian Fire
Research Laboratory, Trondheim, 1992.
/6-23/ Heskestad: "Engineering Relations in Fire Plumes", Fire Safety Journal 7, pp.
25-32, 1984.
Scandpower
7 Liquid Fires
7.1 General
Various types of liquid fires may occur when a liquid fuel is released depending on the
environment(s) of the release and the release mode. It is convenient to classify liquid
fires into the following types:
Pool fire in the open air

Pool fire on the sea surface
Pool fire in an enclosed area
Fireball
Spray fires (not treated in the handbook)
Running liquid fires (not treated in the handbook)
A "Pool fire in the open air" may take place when there is an ignition of a liquid spill
which is released on a horizontal solid surface in the open air (e.g. on the ground). If
this spill takes place on the sea surface, a "Pool fire on the sea surface" may result.
Likewise, if the liquid fuel is released within an enclosed space, a "Pool fire in an en-
closed area", which may suffer from more or less air deficiency, will result.
A "Fireball" is a fire event which results from a "BLEVE" (Boiling Liquid Expanding
Vapour Explosion) in which an immediate ignition of the pressurized and liquified fuel
occurs. A "Running liquid fire" is achieved when the liquid fuel is released on a sur-
face which is not horizontal (e.g. the mantel walls of a tank container). The fuel burns
as it flows down the surface. If the liquid fuel is released under high pressure so that it
is dispersed into droplets, a so-called "Spray fire" may occur.
Owing to the fact that the types of fires mentioned above will exhibit widely different
burning characteristics, it is necessary that they are treated separately. All these fires
except for the latter will be described in this chapter with respect to prediction methods
for important fire characteristics. "Spray fires" and "Running liquid fires" are not dealt
with here due to the fact that there are very few prediction methods for these fires.
7.2 Pool Fire in the Open Air

7.2.1 Scenario
The "Pool fire in the open air" scenario may occur if there is an ignition of a liquid spill
of a combustible fuel on a solid substratum in the open air. Hydrocarbon liquid spill
scenarios may be classified on the basis of the spill rate and the duration of the spill.
Thus, the classification will fall into one of the following three categories:
1. Instantaneous spills - where the spill occurs in a very short time. This is the
case for a vessel or tank rupture in which the liquid contents is released almost
immediately
2. Continuous spills - where the spill continues at a specified rate (either constant
or time dependent) for a long time which may be the case when there is a leak
from a pipe connected to a reservoir
Scandpower
3. Quasi-continuous spills - where a given volume of liquid is spilled over a given

duration of time.
Although most spills will fall into the third category, simple analytic expressions descri-
bing such a scenario are not available. A criterion for classification of a given spill as
either instantaneous or continuous is to calculate the following expression for a non-
dimensional critical time given by Mudan, Ref. 17-1/:
_ tsVf
Tcr- vr3 Eq. (7.1)
where:
Tcr = non-dimensional critical time(-)

ts = the duration of the spill (s)
Vt = fuel regression rate (m/s)
VL = total volume of spilled liquid (m 3)
The spills are, according to Mudan, Ref. /7-1/, classified as instantaneous for spills
resulting in Tcr < 2-10-3• Otherwise, they are classified as continuous.
In case of ignition of the spill, the resulting fire may be very dependent on whether the
spill is instantaneous or continuous provided liquid spill is not contained in a dike.
A burning instantaneous liquid spill will spread until a barrier is reached or all liquid
fuel is consumed. A burning continuous spill will spread and increase the burning
area until the burning rate is equal to the spill rate. Thus, an equilibrium pool diameter
is achieved. It is maintained as long as the release or spill rate is constant. On the
other hand, if the liquid spread is restricted by physical restrictions capable of forming
a dike in which the fuel is contained, the fire may not be notably dependent on
whether the release is instantaneous or continuous.
7.2.2 Pool Fire Diameter

As already accounted for the pool fire diameter is dependent on whether there are
physical restrictions or barriers present on the substratum where the liquid fuel is
released, and whether the spill can be categorized as instantaneous or continuous.
A. Instantaneous Release/No Barriers Are Present
The growth of the spreading diameter with the time t of an instantaneous release on a
smooth horizontal surface is given by the following expression (Mudan, Ref. /7-1/):
Eq. (7.2)
The maximum diameter:
3 ]""
Dm=2 [ Vv~g Eq. (7.3)
Scandpower
The time to reach the maximum diameter:
1/4
tm=0.6743
[
v\
gvt ]
Eq. (7.4)
where:
De = the pool diameter at time t (m)

Dm = maximum diameter (m)
tm = time to reach the maximum diameter (s)
VL = total volume of spilled liquid (m3)
v, = fuel regression rate (m/s)
2
g = acceleration due to gravity (m/s )
t = time from the instantaneous release (s)
The maximum diameter, Dm, occurs when almost all the fuel is consumed by the fire.
Thus, it will exist only for a very short time. Use of Dm will result in an overestimation
of the fire hazard. A time averaged pool diameter can be obtained by integrating Eq.
(7.2) up to time tm. Use of average pool size for an instantaneous release provides
more realistic results for the thermal radiation hazards. By integrating the expression
for the maximum pool area, 1tD/4 (where De is given by Eq. (7.2)), the following ex-
pression will result for the average pool diameter, Da, of an instantaneous liquid re-
lease:
[ l
112
Da=; ~+1 Dmi:::::0.683Dm Eq. (7.5)
B. Continuous Release/No Barriers Present
The equilibrium diameter of a continuous release on a smooth surface and the time to
reach this diameter are given by the following expressions (Mudan, Ref. /7-1/):
112
Deq =2
[
...!'.L] Eq. (7.6)
7tVt
teq = 0.564 Deq 113 Eq. (7.7)

[g Vt Deq]
where:
= equilibrium diameter (the spill rate equals the burning rate) (m)
= leak rate of the liquid fuel (m 3/s)
= time to reach the equilibrium diameter (s)
C. Physical Barriers Are Present
If the extent of the rectangular dike (of lengths A and B) formed by the physical
barriers are smaller than the resulting liquid spread in case of no such barriers, the
spread of the liquid fuel will depend on the dimensions of the dike. Such barriers may
be curbs surrounding the process area, walls, etc. Pool areas of different shape may
be formed. However, only rectangular shaped pool areas will be considered. The
Scandpower
equivalent diameter is given by the following expressions dependent on whether one

or two walls is adjacent to the rectangular pool area or not:
2AB
- - for no walls
A+B
4AB
- - - for one wall of length A
A+2B
4AB
---for one wall of length B
2A+B
4
o.= AB for two perpendicular walls A, B
A+B
Eq. (7.8)
2B for two parallel walls of length A

2A for two parallel walls of length B
D for circular pool with diameter D
The equivalent diameter is an "effective diameter" which is used in the models for pre-
dicting the flame height (see next chapter). The equivalent diameter will in all sub-
sequent text simply be termed as "D".
From the expression above it appears that the equivalent diameter is larger if there
are one or two walls present adjacent to the pool in order to reflect that the flame
height will be higher. A wall restricts the air supply to the flame plume which causes
the flame to rise higher in order to get sufficient air supply for the combustion.
It should be noted that if the extent of the dike is larger than the liquid spread due to
either an instantaneous or continuous release, Eq. (7.2) and Eq. (7.6) should be
applied for the equivalent diameter.
7.2.3 Flame Shape

The flame shape or geometry of pool fires are generally determined by assuming the
flame plume as a cylinder which may be tilted in the wind blown situation. The shape
of the flame is characterized by the equivalent diameter, visible flame length and,
eventually, the flame tilt and flame drag in case of wind.
A. Flame Height
Thomas, Ref. /7-2/ has developed the following correlation for the mean visible flame
height or length for the calm wind situation:
0.61
L=42D mr Eq. (7.9)

[
(po./go) ]
where:
Scandpower
L = average visible flame height1>(m)

D = equivalent air diameter from Eq. (7.8) (m)
m = burning rate per unit pool area (kg/m 2s)
Po = the density of the ambient air (kg/m3)
Although this correlation was originally developed on the basis of laboratory scaled
experiments with wood crib fires, it has been shown that this correlation represents
flame length data for large hydrocarbon pool fires as well.
Thomas, Ref. /7-2/, has also developed a similar correlation for the wind blown situa-
tion :
[ l
0.67
L = 550 m, . (u•ro.21 Eq. (7.10)
Po ✓QD
• _ Uw _ Uw
u - - - Eq. (7.11)
" 1/3
Ucr [g m1DI Pvl
where:
u = non-dimensional velocity(-)
Uw = wind speed (mis)
Ucr = critical velocity (m/s)
Pv = the fuel vapour density at normal boiling point (kg/m 3
) (see Eq. (7.14))
From Eq. (7.10) and Eq. (7.11) it appears that the flame length will decrease with in-
creasing wind velocities. This is explicable because a more efficient combustion in the
flame will result from the increased air entrainment due to the wind. Hence, the fuel
gases need not rise as high in order to get sufficient air for the combustion process.
Moorhouse et al., Ref. /7-3/, state that on the basis of the data from large LNG and
large liquid hydrocarbon pool fires it can be inferred that the Thomas equations are
unlikely to provide accurate results for LNG pool fires larger than 25 m and liquid pool
fires larger than 40 m.
8. Flame Tilt
Croce et al. Ref. /7-4/ maintain that for large scale fires, the following correlation
developed by the American Gas Association (AGA) on the basis of large scale LNG
fires appears to be most representative for prediction of the flame tilt a.:
1 foru'5; 1
cos a.= Eq. (7.12)
{ 11./J foru·~1
where u· is the non-dimensional velocity given by Eq. (7.11).
1) Due to the fact that the flame length fluctuates considerably during time, the mean or average flame length is
defined as the length at which the flame length is equal to or above 50 per cent of the time
Scandpower
~SINTEF
Risk Management AS
C. Flame Drag
Flame drag is the extension of the flame base downwind of the pool edge as shown in
Figure 7.1. The following correlation by Moorhouse, Ref. /7-5/ is correlated on the
basis of large scale LNG pool fires:
069
D' = 1.5(Fr10 )°" Dw Eq. (7.13)
where:
D' = the flame drag (i.e. the extent of the liquid pool in the direction of the
wind plus the extension of the flame base downwind of the pool edge.
See Figure 7.1) (m)
Dw = the extent of the liquid pool in the direction of the wind (m)
Fr,o = the Froude number given by Uw2/gD based on the wind velocity at a
height of 10 m
Caution must be exercised when using the Eq. (7.13) since it is based only on LNG
pool fire data.
By assuming the fuel vapour to be an ideal gas, the vapour density, Pv (kg/m 3), can be
predicted by the following relation (based on the ideal gas law) if the fuel molecular
weight Mw (kmol/kg) and the boiling point temperature Tb (K) of the fuel are known:
Pv"'12.18Mw Eq. (7.14)

Tb
It is of special importance to calculate the flame drag if the wind direction is more or
less coincident with the line from the flame centre to the target because the effective
radiative flame surface is then displaced a distance D' - Dw closer to the target. This
will result in significantly higher heat flux values onto the target.
Geometry of Trench Fires
A fire in a rectangular pool with high aspect ratios (i.e. length to width ratios) is usually
termed as a "trench fire". Large-scale tests with LNG-fires having aspect ratios of up
to 30.0 reported by Mudan et al., Ref. /7-6/, have shown that the flame geometry of
such fires is more sensitive to wind conditions than conventional pool fires discussed
so far. Flame heights were found to be a strong function of trench width rather than
length. For very large aspect ratios, the trench fire seemed to break up into small
flames having typical base dimensions equal to the trench width.
Mudan et al., Ref. /7-6/, suggest that the trench fire geometry may be represented by
a modified Froude number which is defined in the following way:
Scandpower
l' I
r I I I
I ·I
lI ,. !
Figure 7.1: A Sketch showing the Flame Drag Phenomenon which usually occurs
when high Wind Velocities are present (L is the total Flame Length, L1 is
the Length of the lower Part of the Flame not shielded by Smoke, Dw is
the Extent of the Liquid Pool in the Wind Direction, D' is the total Extent
of the Base of the Flame in the Wind Direction including the Flame Drag
and 0 is the Flame Tilt from the vertical Axis)
F'r = Uw Eq. (7.15)

2✓gW
where:
Uw = wind speed (mis)

w = width of trench (i.e. in the downwind direction) (m)
g = acceleration due to gravity (m/s2 )
The flame length, flame drag and flame tilt of trench fires can be expressed by the
following correlations:
Flame length L:
2.2 F'r ~ 0.25

0.88(F'r r 0 55
· 0.1 :$; F'r < 0.25 Eq. (7.16)
4.0 F'r :$; 0.1
~ SINTEF Scandpower
Risk Management AS
Flame drag W:
3.5 F'r <! 0.25

23.3(F'r)1·37 0.1,,; F'r<0.25 Eq. (7.17)
1.0 F'r,,;0.1
Flame tilt a.:
0.56 F'r <! 0.25

0.36(F'r )"0 ·32 0.25,,; F'r < 0.042 Eq. (7.18)
1.0 F'r,,; 0.042
Caution must be exercised when using the correlations above since they are based
only on LNG pool fire data.
The Burning Velocity
In most of the correlations so far, the burning velocity in terms of either the regression
rate, Vt (in m/s), or the burning rate per unit pool surface area, m (in kg/m 2 s) is inclu-
ded in the expression. Burgess et al., Ref. /7-7/, give the following expression for the
regression rate of "infinite large" pools2:
Vrro = 0.076[ ii He ] Eq. (7.19)

iiHv+cp(Tb-To)
where:
Vrro = regression rate of infinite large pools (mm/min)3

iiHc = heat of combustion (kJ/kg)
iiHv = heat of vaporization (kJ/kg)
Cp = specific heat (kJ/kgK)
Tb = boiling temperature (K)
Tr = actual temperature of the fuel (K)
For pool fires with a pool diameter less than the "infinite large" pool diameter, the re-
gression rate vr is given by the following relation:
Vr = Vrro[1- exp(-0.670)] Eq. (7.20)
The relationship between vr (mis) and m (kg/m 2s) is given by:
mt= Pt Vt Eq. (7.21)
2) An "Infinite large" pool is a pool which causes an optical thick flame plume. That is, the pool fire emits heat
radiation with a emissivity equal to unity. The regression rate of 'infinite large' pools is constant independent of the
pool diameter. For most hydrocarbon fuels the infinite regression rate is attained at an equivalent diameter above
1-3m
3) The regression rate v, will be calculated in mis if the constant 0.076 is replaced by 1.27·1/J6
Scandpower
where:
m = burning rate per unit pool area (kg/m2 s)

Pt = fuel density (kg/m 3)
7.2.4 Flame and Smoke Gas Temperatures

Since a pool fire in the open air is a buoyancy controlled diffusion flame, the correla-
tions for the centreline temperature of the rising flame and gas plume proposed in
Appendix G may be used in this case. The rate of heat release which is included in
these correlations, may be expressed by the following expression pertaining to pool
fires in general:
Eq. (7.22)
where:
TJ = Combustion efficiency (-)

m = Burning rate per unit pool area (kg/m 2s)
Ap = Pool area (m2 )
~He = Heat of combustion of the fuel (kJ/kg)

Heat Transfer to Objects Engulfed by a Pool Fire
The principle mechanisms of heat transfer to objects located inside the flame plume or
swept by the pool fire flame plume do not differ from other buoyancy controlled diffu-
sion flames like for instance gas flare fires. Hence, the same expressions presented
in Appendix H may also be used when calculating radiative and convective heat trans-
fer to objects more or less engulfed by a pool fire.
Heat Transfer onto External Objects from a Pool Fire
When it comes to targets which are located outside the flame plume and, thus, are
exposed to radiative heat transfer only, the situation is somewhat different from other
fire situations because hydrocarbon pool fires may generate large amounts of smoke
which cover large parts of the flame plume.
The "solid flame model" presented in connection with gas flare fires in Appendix H,
may also be applied for pool fires in the open air. However, an area averaged emis-
sive power as used in Eq. (H.11), can not be used in the case of large scale pool fires
with fuels which generate large amounts of soot and black smoke and for targets in
the near field of the pool fire. Figure 7.2 shows the measured average emissive
power as a function of the flame depth (here equal to the pool diameter) in a series of
petroleum pool fire experiments carried out by Hagglund, Ref. /7-8/.
Scandpower
The reduced averaged surface emissive power with increasing diameter from a certain
pool size is due to the increasing generation of black and thick smokei. This forma-
tion of smoke starts from a certain height above the pool. Hence, the lower part is still
strongly radiating with a heat flux in the range 130 - 150 kW/m 2• This area is be-
coming smaller for increasing pool diameters. The upper smoky part radiates with an
emissive power of about 20 kW/m 2 , which corresponds, to black body radiating smoke
of a temperature of 500 °c.
xx x = measured by raciometer
o o ,o = measured by spectrophotometer
i
e<.> 1 5 ; 1 - - - - - - - - 1 - - - - - - - - - ~ - - - - - - - - l
~
.?.-
~
~10•1------~lr-----'r--+--------l
C
'"
~.,,
E Sl---~'-----1-------1----------1
fl_
0.1 50 100
Flame Depth (m)
Figure 7.2: Averaged measured Surface Emissive Power as a Function of the Flame
Depth for Petroleum Fires (Hagglund, Ref. /7-8/)
The height above the liquid pool level at which the smoke generation starts, H1, is
given by the following equation developed on the basis of experimental data (Mudan
et al., Ref. /7-6/):
Eq. (7.23)
Targets located on the ground and in the near field (i.e. distances less than one or two
pool diameters from the fire), will receive significantly more radiation from the strongly
radiating lower part of the pool fire than from the upper part which is covered by
smoke. The incident radiation from each part according to Eq. (H.11) must be calcu-
lated. The sum of the results will give a higher and more correct value of incident heat
radiation to the target compared to when using an area averaged emissive power over
the entire flame surface. Hence, the following relation is valid:
Eq. (7.24)
4) It has been shown that for large scale hydrocarbon pool fires with a carbon to hydrogen ratio greater than about
0.3, a substantial part of the of the fire is obscured by thick and black smoke in the outer periphery from a certain
height above the level of the liquid pool. This smoke layer absorbs a significant part of the emitted flame radiation
and results in very little heat emission to the surroundings. For fuels with a carbon to hydrogen ratio less than 0.3,
one can disregard from the smoke screening effect even though the transition at a carbon to hydrogen of 0.3 is not
sharp. Hence, the entire flame plume may then be assumed to have an average emissive power of 130 - 150
kW/m 2
Scandpower
Risk Management AS {Gl)SINTEF
where:
= atmospheric transmissivity (-) (Eq. (H.13))

= the view factor for the target with respect to the lower and upper
part of the flame plume, respectively
E p1,Epu = The emissive power of the lower and upper part of the flame plume,
respectively (e.g. Epu ~ 20 kW/m 2 , Ep1 ~ 150 kW/m 2)
In the far field (i.e. distances greater than 1 - 2 pool diameters from the fire), an area
averaged emissive power and Eq. (H.11) may be used when calculating the incident
radiative heat flux to the target. The area averaged emissive power, Epa (kW/m 2) , is
given by the following equation (Ref. /7-1/):
Eq. (7.25)
7.2.6 Smoke
The correlations for prediction of the mass flow in the fire plume and visibility through
the smoke which are proposed in Appendi x J, will also apply when predicting the
smoke production of pool fires in the open air. Eq. (7.22) must be applied for the rate
of heat release from a pool fire.
7.2.7 Duration of Fire

Continuous Spills
In case of a pool fire due to a continuous release of a combustible liquid, the duration
of the fire will be equal to the duration of the release plus the time necessary for con-
suming the remainder of the oil after the release has terminated . Usually, the latter
time will be significantly smaller than the duration of the spill and the fire duration can
in this case be set equal to the duration of the spill.
Instantaneous Spills
For instantaneous spills the duration of the fire is equal to the time to reach the maxi-
mum diameter given by Eq. (7.4).
Physical Barriers Forming a Dike
If the area of the dike is smaller than the resulting area of an instantaneous or con-
tinuous release when no barriers are present, the duration of the pool fire in the dike is
dependent on the maximum liquid depth in the dike, h,. It is given by the total amount
3
of fuel (m ) divided by the area of the dike (m2). If this depth is larger than the height of
the barriers forming the dike, the height of the dike has to be used instead. The dura-
tion of the fire, ~ . is given by the following equation:
td =_!l =hi Pt Eq. (7.26)

Vt mt
~ SINTEF Scandpower
Risk Managem ent AS
7.2.8 Example
Continuous release from a storage tank containing 500 m3 of gasoline. The release
rate can for simplicity be considered as constant 1O kg/s until the tank is empty. The
tank is located in the centre of a rectangular shaped dike with dimensions 1O m x 20 m
x 0.15 m (downwind length x crosswind length x depth of dike). No walls are located
adjacent to the dike.
The maximum incident heat flux onto a point 1 m above ground of a pressurised stor-
age tank located 10 m from the dike shall be predicted for both the calm wind situation
and the windblown situation. The wind direction coincides with the straight line from
centre of dike to the tank centre. In order to address the dependence of a modelling
approach on the predicted results, both the one-zone and two-zone radiation model as
well as the point source model are to be used.
Solution
Molecular weight of gasoline: Mw = 100 kg/kmol

Heat of combustion of gasoline: AHc = 43,700 kJ/kg
Combustion efficiency: 1] = 0.7
Density of gasoline: pf = 870 kg/m 3
Regression rate of a "infinite large"
burning pool: vf = 0.8-10-4 mis
= 4.8 mm/min
Burning rate: m = 0.07 kg/sm 2
"Effective" vapour density at normal
boiling point: pv = 3.49 kg/m 3s
Ambient temperature: To = 20°C = 293 K
Ambient density: po = 1.2 kg/m 3s
Wind speed: uw = 5 mis
Relative humidity: r = 50%
2. Prediction of flame geometry:
Equilibrium pool diameter without physical restrictions: Eq. (7.6) and Eq. (7.7) predict
an equilibrium diameter and time to reach this diameter of 13.5 m and 35 sec, respec-
tively when inserting the release rate in m3/s into the equation (1 O kg/s 10/870 = =
0.0115 m3/s). The pool area of this circular shaped pool is 144 m2 .
Since the length of the dike is smaller in the downwind direction than the predicted
equilibrium diameter, the extent of the pool in this direction will be 10 m and 144 m2/
10 m = 14.4 m in the crosswind direction. Hence, this roughly speaking rectangular
pool will not completely fill the dike in the crosswind direction. The liquid spread
ceases 2.8 m from each short wall of the dike.
Equivalent pool diameter can be predicted

by Eq.(7.8) to: Deq. = 11.8 m
Flame heights: Calm wind situation (Eq. (7.9)): Lo = 20.5m

Wind blown situation (Eq. (7.10) & (7.11)): Lw = 14.8 m
Scandpower
Risk Management AS (Gl}SINTEF
Flame tilt (Eq. (7.12)): a. = 60 deg.

Net flame drag: (Eq. (7.13)): D'- D = 3.5 m
Hence, the radiative flame surface is displaced a distance of 3.5 m closer to the fire
exposed tank. It is more convenient in this case to assume the opposite, i.e. the tank
is moved 3.5 m closer to the dike. The effective distance between "pool centre" and
target:
x' = x + Dx/2 - (D' - D) = 10 m + 10 m/2 - 3.5 m = 11 .5 m
3. Prediction of atmospheric transmissivity:
For a distance of 10 m from radiative surface of the flame to the target (i.e. the calm
wind situation) and relative humidity of r = 50 %, Eq. (H.13) predicts a transmissivity:
-r = 0.885
For the wind blown situation this distance is reduced to 6.5 m. due to the flame drag.
Eq. (H .13) predicts a transmissivity when x = 6.5 m and r = 50 %:
-r = 0.91
4. Prediction of incident radiation onto the tank:
a. The one-zone model:
Average emissive power:
For an equivalent diameter of 11.8 m, an average emissive power of 52 kW/m 2 can be

predicted from Eq. (7.25). This is somewhat less (5 - 10 kW/m 2) than what can be
predicted from Figure 7.2, which shows measured averaged surface emissive powers
for a gasoline pool fire as a function of pool diameter.
View factor calculation :
Since the point of the tank at which the incident heat flux shall be predicted is located
1 m above the ground, the flame plume has to be divided into two parts; one lower
part of height 1 m and one upper part of height 20.5 m - 1 m = 19.5 m (in principle
shown in Figure 1.4). The radiation heat flux from each part has to be predicted sepa-
rately and then the results have to be summed.
Calm wind situation:
Lower part: a= 2L/D = 2-1/14.4 = 0.14, b = 2X/D = 2-17.2/14.4 = 2.4 inserted into Eqs.
(1.2) and (1.3) yield Fv = 0.024 and Fh = 0.0011. Eq. (1.6) yields Fmax = 0 .024.
Upper part: a = 2L/D = 2-19.5/14.4 = 2.7, b = 2X/D = 2-17.2/14.4 = 2.4 inserted into
Eqs. (1.2) and (1.3) yield Fv = 0.2 and Fh = 0.1. Eq. (1.6) yields Fmax = 0.223.
The resultant maximum view factor is achieved when the view factor of the lower and
upper part is summed up. A resultant maximum view factor Fmax = 0.25 is achieved .
Scandpower
The incident radiative heat flux is given by Eq. (H.11):
q= 52- 0.25- 0.885 = 11.5 kW/m2
The windblown situation:
Lower part: a= 2UD = 2-1/14.4 = 0.14, b = 2X/D = 2-13.7/14.4 = 1.9 inserted into Eqs.
(1.4) and (1.5) yield Fv = 0.03 and Fh = 0.001. Eq. (1.6) yields Fmax = 0.03.
Upper part: a= 2UD = 2-13.8/14.4 = 1.9, b = 2X/D = 2-13.7/14.4 = 1.9 inserted into
Eqs. (1.2) and (1.3) yield Fv = 0.6 and Fh = 0.3. Eq. (1.6) yields Fmax = 0. 7.
The resultant maximum view factor is: Fmax = 0.03 + 0.7 = 0.73.
The incident radiative heat flux is given by Eq. (H.11):
q = 52- 0.73• 0.91 = 34.5 kW/m 2.
Conclusion:
A wind speed of 5 m/s will cause the pool fire to tilt 60 degrees (measured from the
vertical axis through the pool centre) towards the tank and the radiating flame surface
is moved 3.5 m closer to the tank. These wind effects will cause three times the
amount of the heat flux compared to the calm wind situation, i.e. from 11.5 to
34.5 kW/m 2.
b. Two-zone model:
For an equivalent pool fire diameter of 11.8 m and flame height of 20.5 m, Eq. (7.23)
predicts a height of the lower strongly radiative part of the flame plume, H1, to approxi-
mately 5 m.
View factor calculations:
In case of a two-zone model the radiating surface has to be divided into three parts.
That is, a lower strongly radiating part of height 1 m, an intermediate strongly radiating
part of 4 m, and an upper sooty part of height 15.5 m. The view factors of the two
former parts have to be summed up in order to attain the total view factor of the lower
part with respect to the tank.
Lower part: a= 2UD = 2-1/14.4 = 0.14, b = 2X/D = 2-17.2/14.4 = 2.4 inserted into Eqs.
(1.4) and (1.5) yield Fv = 0.024 and Fh = 0.0011. Eq. (1.6) yields F1max = 0.024.
Intermediate part: a= 2UD = 2-4/14.4 = 0.56, b = 2X/D = 2-17.2/14.4 = 2.4 inserted

into Eqs. (1.2) and (1.3) yield Fv = 0.09 and Fh = 0.016. Eq. (1.6) yields F2max = 0.091.
The resultant maximum view factor for the lower strongly radiating part is:
Fmax = F1max + F2max = 0.024 + 0.091 = 0.115
Scandpower
The view factor of the upper sooty part, F3max, of length 15.5 m, is the view factor of
the upper 19.5 part of the flame plume (calculated to 0.223 in the one-zone model
= =
calculations) minus F2max, i.e. F3max 0.223 - 0.091 0.132.
The incident radiative heat flux is given by Eq. (7.24):
q = 0.884· [0.115-150 + 0.132-20) = 0.884• [17.3 + 2.64) = 17.6 kW/m 2
Conclusion:
The incident radiation heat flux onto the tank is predicted to 17.6 kW/m2 with the two-
zone model, while it was predicted to 11 .5 kW/m2 for the one-zone model. Hence, for
this tank which is located slightly less than one pool diameter from the fire, the one-
zone model underpredicts the incident radiation heat flux onto the tank by more than
35 % compared with the two-zone model. The two-zone model is more reliable for
close targets near the ground due to the fact that it takes into account the much
stronger importance of the lower strongly radiative zone of the flame plume.
c. The Point-Source Model:
The radiative heat flux based on the point source model can be predicted by Eq. (H.9).
This equation requires that the radiative fraction of the total rate of heat release is
known. This parameter can be predicted by Eq. (J.5) when knowing the average
emissive power (predicted by Eq. (7.25) to 52 kW/m 2), burning rate (0.07 kg/sm2) and
heat of combustion of gasoline (43,700 kJ/kg) as well as the flame length to diameter
of the pool fire (20.5/11 .5). These numbers inserted into Eq. (J.5) yield a radiative
fraction of fR = 0.14. The radiative fraction of the total rate of heat release can be pre-
dicted by Eq. (H.14) to 42,625 kW. Thus, the incident radiative heat flux given by the
point source model can be predicted by Eq. (H .13) to 11 .1 kW/m 2.
Conclusion :
The one-zone model predicts an incident radiation heat flux of 11 .5 kW/m2, the two-
zone model 17.6 kW/m 2 and the point source model 11.1 kW/m2. Hence, the one
zone model and the point source model predict roughly speaking identical results
which are more then 35 % lower than the heat flux predicted by the two-zone model.
Hence, there is nothing to gain by using the one-zone model instead of the point
source model in this case. The two-zone model is definitely most close to the real
value.
7.3 Pool Fire on the Sea Surface

7.3.1 Scenario
This fire scenario may take place if there is a major spill of a liquid fuel from offshore
installations or tankers and the liquid fuel is forming an "oil slick" on the sea surface
followed by an ignition of the oil. As in case of the "Pool Fire in the Open Air" scenario
(see Chapter 7.2), the liquid release may be spilled continuously at a constant or time
dependent release rate, or it may be released instantaneously where all the quantity of
oil is released within a very short time.
Scandpower
The oil slick burning process may be divided into three phases:
(1) Ignition and acceleration phase in which vigorous burning occurs (a typical
regression rate of 2 - 4 mm/min is reported Ref. /7-9/)
(2) A second phase which exhibits a slowdown in the rate of fire growth mainly due
to the cooling of the fuel by the sea water beneath the slick
(3) Extinction phase.
The oil slick may be ignited immediately or the ignition may be more or less delayed.
For a delayed ignition the oil will spread on the sea surface resulting in increased pool
area and decreased slick thickness for instantaneously released oil spills. The oil slick
must have a minimum thickness in order to be ignited. Thompson, Ref. /7-10/, reports
a minimum thickness of 1.25 mm if oil on the sea surface is to be ignited. If the igni-
tion source is present after the oil slick has reached its minimum thickness, the oil slick
can not be ignited.
The temperature of the oil at the time of ignition must be higher than the flash point5l of
the oil if there is no external heat source to heat the oil to the flash point. Ignition pos-
sibilities are strongly limited by the wind.
Elam et al., Ref. /7-11/, have found that ignition times for crude oils are approximately
inversely proportional to the imposed heat flux squared. The following theoretical
model is established:
Eq. (7.27)
where:
t;g = time to ignition (i.e. the time to reach the ignition temperature T;9 of
the crude oil) (s)
(A.pCp)f = product of thermal conductivity, density and specific heat capacity
of the crude oil (i.e. the thermal inertia of the oil) (W2s/m4 K2)
T;g = ignition temperature (K)
To = initial temperature of the crude oil at time t=O (K)
q" = imposed heat flux (incident to the surface of the oil slick) from an
external heat source (kW/m 2)
Elam et al., Ref. /7-11/ report that the effects of exposure (i.e. spark igniter or pilot
flame) and oil thickness are negligible when the oil is imposed with a radiant heat flux
of 1O kW/m2 . For less radiant heat fluxes, longer times are needed to ignite oil with a
spark than with a pilot flame.
After ignition the oil slick continues to burn until its thickness reaches a minimum value
at which the heat loss to the water consumes most of the heat feedback from the
flames above. Bruist et al., Ref. /7-12/, report an experimental value for this minimum
slick thickness of 0.8 mm resulting in a combustion efficiency•l in the range 85-95 %
depending on the released quantity. In general, as the initial oil volume increases so
does the combustion efficiency for instantaneously ignited oil slicks. Further, Bruist et
al., Ref. /7-12/, report that the sooner the oil slick is ignited the higher the burning effi-
5) The flash point of a liquid fuel is defined as the lowest temperature at which the fuel will emit vapours sufficient to
fonn a flammable mixture with air that will flash across the surface when ignition temperature is applied
6) Defined as the difference between the volume of oil spilled and the volume of the unburned residue which has
attained the minimum slick thickness divided by the volume of oil spilled
Scandpower
NoJWeQian Fire Research Laboratory
ciency. A rough approximation of the combustion efficiency for an instantaneously

ignited slick is:
1 1
llcomb = (1- - v ;s)x100% Eq. (7.28)
3
where
Vs = volume of instantaneously oil release (m 3)
In the first phase the fire will not differ very much from a pool fire on a solid substrate
of the same size. In the second phase the single large scale fire from the first phase
is replaced by several smaller fires in regions that still are thick enough to burn. Thus,
the total flame height of this fire is significantly reduced . This can explain the pheno-
menon of dramatically reduced flame length to diameter ratio often observed with oil
slick fires. The first signs of flame quenching are evident in this burning stage. The
burning rate drops dramatically in this phase of the fire, which also result in a smaller
flame height.
The third regime has quenching in the centre of the slick and a further drop in the total
burning rate, thus forming a ring fire consisting of many smaller flames.

Immediate Ignition
For an immediate ignition of liquid spill onto the sea surface the same relations as
proposed for spills onto a smooth solid substrate may be used (i.e. Eq. (7.3) including
Eq. (7.6)). The only deviation is that an effective acceleration due to gravity "g" should
be used instead of the real acceleration of gravity (i.e. g ~ 9.8 m/s2) .
The effective acceleration due to gravity is given by the following equation:
Eq. (7.29)
where:
PL = density of the liquid spill (kg/m 3)

3 3
Pw = density of sea water (kg/m ) (Pw ~ 1,025 kg/m )
Eq. (7.29) will result in an effective acceleration of gravity much less than the real
acceleration of gravity. From Eq. (7.3) it appears that a lower value of the acceleration
of gravity leads to a smaller spreading diameter over time and a smaller maximum
spreading diameter compared to a corresponding instantaneous spill on a smooth sur-
face on land. However, from Eq. (7.4) and (7.7) it appears that the time to reach the
maximum and equilibrium diameter of an instantaneous and continuous release, res-
pectively, will both be longer. From Eq. (7.6) it appears that the equilibrium diameter
is independent of the acceleration of gravity.
Scandpower
Delayed Ignition - Continuous Spills
In case of a delayed ignition of a continuous release in which the diameter of the oil
slick has become larger than the equilibrium diameter at the time of ignition, the
growing oil slick will soon attain the equilibrium diameter after the ignition of the fuel
has occurred. This diameter is sustained as long as the constant release rate is sus-
tained. Hence, in practice all continuous spills may be treated roughly as if they were
ignited immediately.
Delayed Ignition - Instantaneous Spills
In case of a delayed ignition of an instantaneous release of a comparatively large

quantity of oil onto the sea surface, the diameter of the oil slick as a function of time
during the period prior to ignition has to be calculated first. This is necessary in order
to check when the minimum thickness of the oil slick is reached and where no ignition
of the slick can take place. Stolzenbach et al., Ref. /7-13/, give spreading laws for
successive regimes in which the governing spreading and retarding force changes
between the regimes.
Phase 1
In the first regime in which the gravitational and inertial force are the prevailing
spreading and retarding force, respectively, the spreading diameter D1 is given by the
following relation:
Eq. (7.30)
where:
k, = a constant (k, = 1.14 is reported to be a good value, Ref. /7-13/)

Pw = density of water (kg/m 3 )
p, = density of liquid fuel (kg/m 3)
VL = volume of liquid fuel released instantaneously (m 3)
t = time from start of the release (sec)
Phase 2
A second phase of spreading occurs for slicks that are sufficiently large and have
been spreading for a sufficiently long time that the viscous shear on the bottom of the
slick becomes the primary force balancing the buoyant gravitational force. The
spreading diameter D2 may then be expressed by the following equation:
Eq. (7.31)
where:
k2 = =
a constant (k2 0.98 is reported to be a reasonable value, Ref. /7-13/)
vw = =
viscosity of sea water (m 2/s)" 1.31-10-e m2/s
Scandpower ~
Risk Management AS u SINTEF
Phase 3
The final spreading stage occurs when the slick has thinned sufficiently so that surface
tension is the dominant spreading mechanism. The balance of the surface tension
and the viscous force leads to the following equation for the spreading diameter 0 3 in
the third regime:
Eq. (7.32)
where:
a constant (k3 = 1.6 is reported to be a good value, Ref. /7-13/)

interphasial tension (N/m) (is in the range 0.005 - 0.02 N/m)
Eq. (7.30) of Phase 1 applies as long as 0 1 < 0 2 while Eq. (7.31 ) of Phase 2 is used
when 0 2 < 0 1 and 0 2 > 0 3 . Eq. (7.32) of Phase 3 applies as long as 0 2 < 0 3 . Imple-
menting Eq. (7.30) - Eq. (7.32) in a spreadsheet will be convenient when predicting the
oil slick spreading.
Based on the total amount of liquid fuel released, the predicted diameter of the
burning oil slick as a function of time and a minimum oil slick thickness of 0.8 mm, the
maximum time delay for ignition of the crude oil can easily be predicted. Buist et al.,
Ref. /7-12/, have developed the following equation for the maximum ignition delay,
ti.max, of an oil slick due to an instantaneous release of oil of volume Vs (m 3) based on
Eq. (7.30) - (7.32):
td max = 0.0975 V~.4S Eq. (7.33)
For an 100 m 3 spill of oil (p1 = 850 kg/m3 ) released instantaneously onto the sea sur-
face (Pw = 1,025 kg/m 3 and Vw = 1.31-10·5 m 2/s) and a interphasial tension of
0 .005 N/m, the transition from phase 1 to 2 occurs after 3 minutes when the spreading
diameter of the oil slick is 120 m. The thickness of the oil slick is then slightly less
than 10 mm. The transition from phase 2 to phase 3 occurs after 6 hours when the
diameter of the oil slick is approximately 380 m. The thickness of the oil slick is then
slight ly less than 1 mm. Hence, if the oil slick needs a thickness of at least 1O mm to
be ignited and burn, there can not be any fire after the oil spread on the sea surface
has finished the first regime of radial spreading on the sea surface.
7.3.3 Flame Shape, Flame temperature, Heat Transfer Assessments,

Smoke Production, Duration of Fire and the Effects of Wind
All these important fire characteristics are treated together in one section due to the
fact that calculation of these characteristics will not deviate from those calculation
methods proposed in the section dealing with the "Pool Fire In the Open Air Scenario",
i.e. Chapter 7.3. The flame shape (i.e. flame height, flame drag and flame tilt), flame
temperatures, heat transfer to objects outside and inside the flame plume, smoke pro-
duction and duration of the fire can be calculated by using the same formulae as pro-
posed in Chapter 7.2.
Scandpower
Effects of Wind on the Fire
The effects of wind on the fire from an oil slick will be the same as a corresponding
pool fire in the open air sited on a solid substrate with respect to flame tilt and the
flame drag. However, oil slicks tend to drift/turn to the right with respect to the direc-
tion of the wind. A good empirical rule of thumb that agrees well with large scale
experiments and numerical calculations of oil slick behaviour is that the drift velocity of
a slick is 3 % of the wind velocity and that it will turn 12-15° to the right with respect to
the direction of the wind.
7.3.4 Hazard Assessment and Thermal Loads from Crude Oil Fires
on the Sea Surface
7 .3.4.1 General
This section comprises a synopsis of the main results from the SINTEF research
project "Fire on the Sea Surface". It is based on the reports STF25 F95063 "Thermal
Loads from Crude Oil Spill Fires on the Sea Surface", Ref. /7-14/ and STF25 F95064
"Fire on the Sea Surface - Hazard Assessment 1995", Ref. /7-15/.
The hazard assessment of the scenarios leading to a fire on the sea surface may be
divided into "probabilistic factors" and "consequence factors". In the following chap-
ters these factors are commented separately.
7 .3.4.2 Probabilistic Factors
Wind Effects
SINTEF suggests as a first approach, that with a remote heat source and/or an igni-
tion source producing a radiation less than 5 kW/m 2 over an area of 1 m2 , the limiting
wind speed for ignition and sustained burning of fresh crude oil is 12 m/s. The ignition
source in question here is one that does not add substantial heat to the oil slick, i.e.
like a spark. However, ignition of an oil slick may take place in case of wind velocities
above 12 m/s if the heat flux from a remote source and/or an ignition source is suffi-
ciently high.
In order to take into account varying states of the oil spill (e.g. varying temperature,
content of light fractions of the oil) the range for the limiting wind speed for ignition and
sustained burning is suggested to be 12 + 3 mis (i.e. 15 m/s).
Wave Conditions
Due to the lack of experimental documentation of the effects of waves on the ignition
of oil slicks, SINTEF suggests that wave action does not have any influence on the
probability of ignition or sustained combustion within the first half-hour after a release
of fresh crude oil. However, further research is needed to make this statement less
conservative.
Scandpower
Slick Movement
Driven by the surface current and the wind the oil will move on the sea surface. The
basic rule of thumb for the speed and direction of travel is a vector addition of 3 % of
wind velocity and 100 % of surface current.
Based on experimental findings, a drift velocity of 1.5 % of wind velocity measured

1O m above the sea surface is suggested to be a better estimate valid for the initial
phase of a burning oil slick, Ref. /7-14/ and /7-15/. This statement will, in most situa-
tions, lead to a conservative prediction of the fire hazard. To cover an uncertainty of
what happens in the initial phase of an oil spill, a drift velocity induced by surface cur-
rent ranging from 70 - 100 % of the current velocity is proposed. Care must be taken
when using weather and sea current data, so that the wind effect is not added twice to
a surface current. In weather and current databases these two factors, surface cur-
rent based on tide and large-scale sea current and wind induced current are normally
presented separately.
Sustained Combustion
In the literature there are three main characteristics, which have shown to be impor-
tant regarding the possibility of sustained fire; namely: the flash point of the oil, the
thickness of the oil slick, and the wind velocity. The use of flash point or the fire point
(20 - 40 °C above the flash point) of a fuel as a temperature to be a limiting condition
for a sustained combustion, may be misleading. The flash point and the fire point are
fuel characteristics useful for classification regarding ignitability and safe handling of
the fuels , but it is determined in small scale test apparatus (cup burner tests etc.). In
realistic fire scenarios these temperatures may only be used as guidance of the fuel
characteristics.
If a sufficiently large ignition source is deployed to a fuel of high flash point, ignition
and sustained combustion may occur as a result of the dynamic process of flame pro-
pagation and radiative feedback to the fuel surface, irrespective of the initial fuel tem-
perature. The condition , in which sustained combustion may occur, is depending on
the possibility to establish a certain temperature in the oil layer.
Flash Evaporation
It is often supposed that a substantial fraction of a fresh crude oil is evaporated by

flash evaporation , e.g . when falling from topsides of a platform to the sea surface.
There is no evidence that the evaporation rate of light end fractions (CS, C6 ,C7, C8)
of the crude oil will be crucial in the first minutes after a spill or release. No direct
measurements of the rate of evaporation of the different fractions have been found in
this study.
The evaporation process of crude oil exposed to different environments may be found
more accurately by laboratory studies. It is recommended that the oil , which is falling
from a platform to the sea surface, i.e. the well stream, is considered as fresh crude oil
with its original chemical composition based on process state, Ref. /7-14/ and /7-15/.
However, flashing of oil fractions or oil components, which have their boiling point
below ambient temperature at atmospheric pressure, have to be taken into account.
This latter applies to both the well stream of highly unstabilised and stabilised crude
oil.
Scandpower
Limiting Thickness of the Oil Slick
The "rules-of-thumb" for the limiting oil slick thickness for ignition and extinguishment
in relatively calm, quiescent conditions are, Ref. /7-14/ and /7-15/:
Minimum ignitable thickness for fresh, volatile crude oil is about 1 mm

Minimum ignitable thickness for aged, not emulsified crude oil and diesel fuels is
about 3 to 5 mm
Minimum ignitable thickness of residual fuel oils, such as Bunker "C" or no.6 fuel
oil, is about 10 mm
2
Once 1 m of burning slick has been established, ignition can be considered
accomplished
Extinguishment at a 0.5 mm thick oil residue seems to be a reasonable estimate
The thickness of an oil slick is not uniform over the whole slick, and the tendency
to form patches with thickness above 1 mm must be taken into account re-
garding the possibility of sustained fire. A rule-of-thumb is that 90 % of an oil
spill is contained in 1O % of its area.
Ignition
Given that burning is possible, the probability of ignition can be estimated, taking into
account, Ref. /7-14/ and /7-15/:
If burning oil can reach the sea (running or falling), then the probability of ignition
can be assumed to be 1
If there is a major topsides fire then a probability of the pool being ignited by
burning debris should be estimated. It has not been found any references or
guidance on this, but fairly obviously, ignition from light, flaming debris is likely
while ignition from falling "red hot ", heavy objects is not
Without a credible, flaming ignition source, crude oil burning is unlikely
The situation where ignition is possible without a flaming ignition source is with a
remote fire or heat source, which may increase the evaporation rate of oil and
form a flammable mixture. Then a falling object of the type "red hot falling
object" may ignite the oil slick as well.
A first estimate of a heat flux sufficient to produce flammable mixture above the oil
slick is: A heat flux above 5 kW/m 2 over an area of 1 m2 during a time period above
5 minutes may induce a flammable mixture above a crude oil slick. This statement is
not verified, but with the combination of indications from different small scale and in
situ burning experiments, it is a lower limit with some experimental verification.
Further research may reduce the conservatism of this statement.
7.3.4.3 Dynamics
Flow to the Sea Surface
Oil can reach the sea surface in two ways, by "Plunging" or "Running":
Oil will "Plunge" if it falls freely from the platform

Oil may "Run" down the jacket, risers, etc. and may still be burning when it
reaches the sea surface
Oil released sub-sea can be treated like "Plunged" oil
Scandpower
Oil released near the surface, e.g. as a result of collision damage to a riser can
be treated like "Run" oil.
The time history of rapid flashing and equilibrium flashing evaporation of spilled oil is
not known in sufficient detail to predict the fate of the oil from the leakage to it is for-
ming a slick at the sea surface. Plunging and running oil from a platform to the sea
surface is not considered as significantly different processes with respect to evapora-
tion and emulsification, Ref. /7-14/ and /7-15/. Hence, the condition of the oil after
plunging or running is to be considered as if it was fresh from the release point. How-
ever, further studies are needed to reduce this conservatism.
Pool Diameter
There are many factors affecting the spread of oil on the sea surface. It is sufficient to
say that formulas reported in this chapter are sufficient to model the following cases,
Ref. /7-14/ and /7-15/:
Continuous spills, no ignition, delayed ignition or immediate ignition

Instantaneous spills, no ignition, delayed ignition or immediate ignition.
To calculate these cases, one needs to know the mass leakage rate and duration
(continuous) or the volume (instantaneous) spilled, and the fuel density.
The three phases of oil slick spreading; gravity driven, viscous force limited and sur-
face force limited spread are well defined and verified by Fay, Ref. /7-16/. However, it
should be emphasised that the given spreading laws are only order of magnitude
estimates of the oil slick diameter and they are only valid in case of calm sea. How-
ever, there is evidence that the final stage of surface tension limited spread is not uni-
form. The formation of a sheen of oil over a large area and the existence of thicker
flakes (thickness above 1 mm) in between is to be noted here. Emulsified oil tends to
stay in thicker layers more than oils, which are not emulsified.
Inventory available to burn
If the pool of oil is ignited then it should be assumed that the total inventory of running
and plunging oil would burn (except for the 0.5 mm "residue". Some references state
that water-oil emulsions can burn up to 70 or 80 % water.
7.3.4.4 Consequence Factors
Burning Rates
The burning rate of an oil slick increases with diameter up to a maximum of about
4 mm/min for crude oil pools at 10 m diameter, then drops, probably due to oxygen
deprivation at the centre of the fire for larger diameters.
Recommended values are:
4 mm/min (0.055 kg/sm2) for crude oils, Ref. /7-17/ and /7-18/
9 mm/ min (0.1 kg/sm2) for condensate, Ref. /7-17/.
The combination of fire load and duration is the measure of the consequences of a
fire. A lower burn rate may lead to a longer fire duration and a greater extent or dia-
Scandpower
G}) SINTEF
Risk M anagement AS
meter of the fire on the sea surface. (However, a greater diameter of the oil slick will
also result in an increased total burning rate due to the increased extent or area of the
fire, and thus, necessarily not an increased duration of the fire). This may, in some
situation, be more critical than a high burning rate, leading to high flame height and
smaller extent of the burning oil slick.
To take into account the consequence of a possible longer duration of a fire, a lower
burning rate than the supposed should also be checked in the consequence assess-
ment. Burn rates of 2 and 4 mm/min for crude oils and of 4 and 9 mm/min for conden-
sate are proposed, Ref. /7-14/ and /7-15/.
Fire Dimensions
The recommended calculation models or correlation for flame height, flame deflection
and flame drag in Chapter 6 and 7 is in line with current knowledge. The correlation of
flame height versus pool diameter for hydrocarbon fires developed by Heskestad, Eq.
6.51 in Chapter 6, may be used as a first assessment. This correlation predicts flame
heights with accuracy, which is reasonable for some fire hazard assessments. The
main problems of identifying a flame height are the pulsation of the flames (typically
with a frequency related to the square root of the pool diameter, Ref. /7-14/). Further
the difficulty of observing flames due to smoke obscuration and the difficulty of de-
fining a lower temperature of which flaming combustion occurs.
The correlation for the flame height in Eq. 6.51 yields an average or "effective" flame
height over time, i.e. the flame height that is equal to or above the pulsating flame
height in 50 % of the time. When using this flame height in calculations of incident
heat radiation onto external targets, a value close to an average or effective heat
radiation exposure is predicted.
The flame tilt or deflection due to wind may be predicted by the correlation of Mudan
and Croce, Ref. /7-19/ valid for circular hydrocarbon pool fires:
cos 6 = 1.1 · D116 I u112
where:
6 = angle of deflection, from vertical

D = pool diameter (m)
u = wind velocity (mis) at 10 m level
Fire Duration
The duration depends partly on the pool dimensions and weather the spill is con-
tinuous or instantaneous. Fires from continuous spills are likely to last as long as the
fuel is supplied. In both cases, the fire will only last until the oil thickness is 0.5 mm.
The efficiency of combustion drops with the film thickness. This statement is relevant
for the hazard assessment as long as the efficiency of combustion is taken as the
fraction of the original spilt oil, which is burnt.
Flame Spread, Flame Temperature, Heat Fluxes and Smoke Production Rates
Upwind flame spread may be limited by strong wind. The prediction methods referred
to in Chapter 6 and 7 are usually based on small-scale experiments. They predict a
quite low spread velocity, in the order of 2-5 cm/s. However, even in moderate winds
Scandpower
both the prediction methods and the larger scale experiments show that flames spread
occur upwind. The crosswind flame spread velocity is also of the same order as the
upwind velocity.
The main conclusion with respect to flame temperature and heat fluxes is that when
the size of the pool fire increases, the heat loss from the central parts of the flames
are prevented. Hence, the temperatures of certain regions inside the flames may be a
high as 1,300 °C or more. The very hot regions are located where the turbulent
eddies of the flames entrain sufficient air to complete combustion, and where the
flames are optically thick. The fire load to objects inside a fire plume is of the same
order as those found by experiments on land.
Smoke production rates are varying with pool diameter and fuel properties. The main
tendency is that crude oil fires on the sea surface are producing a high rate of smoke,
typically in the order of 10-15 % of the fuel burning rate. (Smoke in this context is de-
fined as particulate, solid soot and liquid hydrocarbon fractions). This indicates that
combustion is not complete inside the fire plume, and the fire plume is typically also
covered with thick, black smoke. This effect makes the radiation to remote objects far
less than it would be without the smoke screening effect.
Thermal Loads and Burning Rates
An oil slick fire on the sea surface will produce a thermal environment, which is
varying in time and space. A target enveloped in the flames will experience a fluctua-
ting heat load, with peak values of approximately 400 kW/m2 , Ref. /7-14/ and /7-15/.
To assess the heat flux density in the flames as part of an engineering calculation, the
peak values are not suitable. One has to take into account variations in time, i.e. fluc-
tuations of the combustion process. It may be possible to take into account the posi-
tion of the target relative to the sea surface. The typical heat flux density in the ver-
tical profile, for a fire with pool diameter of 15 m, shows that the maximum fluxes occur
about 1/3 to 1/2 of the continuous flame height (6 - 9 m) above the sea surface. The
position of maximum heat fluxes and temperatures laterally is not easy to assess.
One should consider the possibility of exhibiting peak fluxes anywhere within the flame
envelope at each level.
Based on experiments, reflecting the scale, the environment and the fuel type used,
the measured imposed heat flux density onto a cold steel target with untreated surface
will be in the order of 260 kW/m2 in time periods of half a minute. The value of a
1-minute exposure will be in the order of 220 kW/m2 • The SINTEF experiments at
Spitzbergen were lasting some 10-20 minutes, and the average heat flux densities
measured over time intervals more than 1 minute are not necessarily representative of
a longer lasting fire. As a value to be used as an engineering assessment, the
1-minute average value of 220 kW/m2 , is proposed to be valid.
A burning rate of 0.03-0.05 kg/m 2s is typically measured in large open crude oil fires
on the sea surface.
Conclusions with respect to Consequence Factors
The following consequence factors regarding flame temperatures, heat flux densities,
smoke production and burning rates may be used in engineering assessments of the
fire hazard in connection with large crude oil fires on the sea surface, Ref. /7-14/ and
/7-15/:
Scandpower
G) SINTEF Risk Management AS
No,wegian Fire Research Laboratory
Maximum temperature exposure (over short time): 1,370 °C

Average or effective maximum flame exposure temperature
(over 1 minute): 1,130 °C
Maximum heat flux density over½ minute: 260 kW/m 2
Average or effective maximum heat flux density: 220"
Fraction of burned fuel converted to smoke (i.e. soot particles): 15%
Burning rate: 0.05 kg/m 2s
7.3.5 Example
100 m3 of crude oil is spilled from an offshore platform onto the sea surface forming an
oil slick under the platform. This may cause an extremely hazardous situation for the
entire platform in case of ignition of the oil slick. No wind is present at the time of igni-
tion. The height from the sea surface to the cellar deck is 27 m. The oil slick is expo-
sed by a heat flux of 1O kW/m 2 due to the flare fire from the flare boom. Calculate the
spreading of this oil slick in case of no ignition and the ignition possibilities of the oil
due to the heat flux against the slick surface due to the flare fire. In case of ignition,
the average and maximum extent of the fire (diameter and flame height) and fire dura-
tion as well as the average heat exposure (temperature and heat flux) to the cellar
deck and the smoke production of the fire at weather deck level are to be predicted.
Solution
Volume of crude oil spilled onto the

sea surface: VL = 100 m3
Molecular weight of gasoline: Mw = 100 kg/kmol
Heat of combustion of gasoline: ~He = 43,700 kJ/kg
Combustion efficiency: TJ = 0.7
Density of crude oil: pf = 850 kg/m 3
Regression rate of a "infinite large"
burning pool: Vt = 0.8-104 mis
= 4.8mm/min
Burning rate: m = 0.07 kg/sm 2
"Effective" vapour density at normal
boiling point: pv = 3.49 kg/m 3s
Average thermal conductivity of crude oil: = 1.5-104 kW/mK
Specific heat of crude oil: = 0.879 kJ/kgK
Ignition temperature of crude oil: 220 °c = 493 K
Ambient temperature: To = 1O 0 c = 283 K
Ambient density: po = 1.24 kg/m 3
Density of sea water: pw = 1,025 kg/m 3
Viscosity of sea water: vw = 1.31-10·0 m2/s
lnterphasial tension: fr = 0.005 N/m
2. Prediction of oil slick spreading
Eqs. (7.30) - (7.32) yield the spreading diameter of the oil slick as a function of time
before ignition as shown in Figure 7 .3. The calculation shows that the transition from
phase 1 to phase 2 occurs after 4 min when the oil slick diameter is 120 m. The slick
thickness is then slightly less than 1O mm. The transition from phase 2 to phase 3
Scandpower
occurs after 6 hours and 23 minutes when the diameter of the oil slick is approximately
380 m and the slick thickness is slightly less than 1 mm.
OIL SLICK SPREADING ON THE SEA SURFACE

Voh,1tn11 of oll c,pl l l at sa- · 100 tn2
450
◄ 00
350
'E? 300
V
i
1 II
250
v
.ll 2 00
~
-m
1 :,0
0
1 00
50
a
0 2 ◄ 6
T lrre (hour sJ
Figure 7.3: The calculated Spreading of an Oil Slick due to an instantaneous Spill of
100 m 3 of Crude Oil on the Sea Surface provided no Ignition of the Oil
3. Prediction of ignition possibilities of the oil slick
Eq. (7.33) predicts a maximum ignition delay of slightly less than 50 min. From calcu-
lations of the oil thickness (Figure 7.4) it appears that the oil thickness at this time is
2.4 mm. Consequently, two times larger than the minimum oil thickness reported by
Thompson, Ref. /7-10/.
Scandpower
OIL SLICK SPREADING ON THE SEA SUAFACE

Vn lu,,,. of or J apl I I at •aa• 100 rn2
32
30 -
29 -
2S -
:!4 -
iu 22 -
. 20 -
Y.
I 19 -
.
~
.c
l!
16
14
-
-
ai
12 -
10 -
0
B -
6 -
-
"'
2 -
0 I
0 2 ' I
'
"' 6
Tim (hours)
Figure 7.4: Calculated Oil Slick Thickness due to an instantaneous Release of

100 m3 of Oil on the Sea Surface provided no Ignition of the Oil
Eq. (7.27) predicts for an imposed heat flux of 1O kW/m 2 a maximum time to ignition of
the oil of approximately 40 sec if one assumes that the initial temperature of the oil
slick is equal to the ambient temperature. The ignition time drops sharply for higher
initial temperatures of the slick because from Eq. (7.27) it appears that this time is pro-
portional to the difference between the ignition temperature and the initial temperature
of the oil squared. Hence, since the predicted time to ignition due to the heat ex-
posure from the flare boom fire is significantly less than the maximum ignition delay
time, this oil slick will be easily ignited.
4. The maximum extent of the fire
Eq. (7.3) and Eq. (7.4) yield a maximum oil slick diameter and a fire duration of
126.8 m and 3½ min, respectively. Eq. (7.5) predicts an average oil slick diameter of
86.6 m. Eq. (7.9) predicts an average flame height of 80.5 m.
5. Heat exposure to the cellar deck
Eq. (G.4) predicts a distance to the virtual source of -30.9 m. The negative sign indi-
cates that the virtual source is located 30.9 m below the pool surface. The distance
termed 'z' in (G.1) is: 30.9 m + 27 m = 57.9 m.
Since the ratio Q215/D = (12612558.2 kWt5t86,6 = 8 < 10.2 and z/Q215 = 0.08, Eq.
(G.1) predicts a flame temperature 27 m above the pool surface of 990 °c. Eq. (G.8)
predicts a fire gas velocity of 52.0 m/s. Eq. (H. 7) predicts a heat transfer coefficient of
Sc andpower
0.064 kW/m 2 • Eq. (H.3) predicts then a heat transfer by convection : qc = 0.069-(990-
10) = 62.4 kW/m2 . Ultimately Eq. (H .2) predicts a radiative heat transfer by radiation:
qr = 56.7-10-12 ·(990+273)4 = 144.3 kW/m2 .
The average total heat transfer to the main support frame of cellar deck is the sum of
radiative and convective heat transfer, i.e. 62.4 kW/m2 + 144.3 kW/m2 = 206.7 kW/m2
where 70 % is radiative heat transfer and 30 % is convective heat transfer. This is a
serious heat load onto the cellar deck, but a duration of 3 ½ min is definitely too short
to cause any structural damage to the platform construction. Due to the extreme
average flame height of 80.5 m, the platform personnel on the weather deck will be
exposed to lethal fire exposure. This fire will definitely escalate and cause other fires
on the platform, which most probably will lead to a loss of the whole platform.
7.4 Pool Fire in an enclosed Area

7.4.1 Scenario
Principally, this fire scenario does not deviate much from pool fires in the open air with
respect to the lateral spreading of the liquid fuel on the floor of the enclosed area. The
most apparent deviation from the open air situation is that the vertical spreading of the
fire gases may be restricted by the ceiling.
As shown in Figure 7 .5 the hot plume of fire gases (either reacting fire gases, i.e. fuel
vapour and air forming a flame, or hot combustion gases) impinges on the ceiling and
spreads radially along the ceiling forming a relatively thin radial ceiling jet. This jet of
fire gases entrains air from the room along the length of the plume in addition to re-
circulated smoke generated from the combustion process. As the jet moves under the
ceiling surface, it entrains further ambient air from below. The ceiling jet transfers
energy by convection to the cooler adjacent ceiling . Further, the spreading velocity of
the fire gases is retarded by friction and less buoyant force due to decreased tempe-
rature along the ceiling jet.
The ceiling jet continues to move radially outward under the ceiling surface until it
eventually reaches the walls of the enclosure and then turns downward along the
walls. Since the downward wall jet is of higher temperature than the ambient air, and
is, thus, retarded by buoyancy, the downward wall jet stops at a certain distance below
the ceiling. In this way a hot gas layer consisting of more or less well mixed fire gases
is stratified under the ceiling above a comparatively cool layer of almost clear air by
the floor at a height, h9 , above the floor as shown in Figure 7.5.
In case of openings in the fire enclosure, the fire gases tend to flow out through the
upper part of the opening, while air from the outside flows into the enclosure through
the lower part of the opening. Hence, the fire plume, which needs air for the com-
bustion, acts as a 'pump' that provides the air flow into the enclosure. The ventilation
opening acts as a 'throttle' in this schematised flow picture.
Scandpower
~ SINT EF Risk M anag em ent AS
Figure 7.5: A simplified Drawing of the enclosed Pool Fire Situation (m, = total
Burning or Evaporation Rate of Fuel, ma = Fire induced Air Supply Rate,
m9 = Mass Rate of Combustion Products, unburned Fuel and Excess Air
leaving the Fire Enclosure, h9 = the Height of the hot Gas Layer above
the Floor and H is the Ceiling Height)
In enclosed fires there generally exists a complex interaction between the three
physical processes "fluid flow" (by the outflow and inflow of hot fire gases and air),
"combustion" (by burning of the fuel) and "heat transfer" (by radiative and convective
heat transfer from the fire gases and heated walls and by heat conduction in the walls
and liquid fuel). These processes are closely linked together in such a way that each
process occurs in response to each other in a very complex way. A mutual interaction
between all these physical processes is taking place. The severity of the enclosed
pool fire situation is a direct result of the equilibrium state of these three physical
processes.

The pool fire diameter in case of continuous or instantaneous spills can be calculated
according to the models proposed for pool fires in the open air in Chapter 7.2.2.
However, one has to be aware of the fact that the spread of the liquid fuel may be re-
stricted by the surrounding walls for large liquid releases.
7.4.3 Effects of Ventilation

As accounted for in the scenario description the fluid flow processes in terms of air in-
flow into the enclosure and the outflow of hot combustion gases as well as the gas
flow within the fire enclosure, may control the severity of the fire condition. Enclosed
pool fires may be divided into "well ventilated pool fires" and "underventilated pool
fires" dependent on the rate of air supply to the fire enclosure. However, there is no
sharp division between these two stages of the enclosed fire situation and therefore
Scandpower
there is no single, clean cut definition for these stages of the fire. The ratio between
the rate of air supply to the fire enclosure and the evaporation rate of fuel vapour from
the pool surface is decisive for whether the fire will be underventilated or well ventila-
ted. The size of the compartment in relation to the pool size is also a key parameter
determining the severity of the fire.
In well ventilated fires the supplied air to the fire enclosure is in excess and, thus, the
fire enclosure is significantly cooled. The rate of burning is controlled by the pool area
and the properties of the fuel. Hence, such fires are often termed as "fuel controlled
fires". Generally, the fire severity of well ventilated fires is not very great. The visibility
through the smoke and the degree of smoke filling the fire room (i.e. the downward
extent of the hot gas layer from the ceiling) as well as the temperature of the hot gas
layer are usually significantly less critical in well ventilated fires compared with under-
ventilated fires. The very early phase of any enclosed pool fire will be at this stage.
Underventilated pool fires are characterized by a hot and thick layer of smoke beneath
the ceiling of the fire room and flames emerging out from the openings of the fire
enclosure. This layer extends normally two-thirds to three-fourths of the room height
from the ceiling. The rate of burning inside the fire enclosure is controlled by the
restricted rate of air supply. Pool fires which are slightly underventilated are normally
most severe in terms of burning rate, gas temperatures concentration of toxic gases
and visibility through the smoke in the fire enclosure. Excess fuel, which can not be
burned inside the fire enclosure due to oxygen depletion, will burn from the opening
where the fuel vapour comes in contact with sufficient air.
The transition between underventilated and well-ventilated fires is not clearly defined.
One definition is that the transition occurs when flames emerge from the openings of a
fire enclosure. This may occur long before the fire has become understoichiometric.
A definition, which will be used here, is that underventilated conditions are achieved
when the hot gas layer has descended to a level beneath two-thirds of the height. In
Chapter 7.4.7 correlations are given for prediction of the hot gas layer height above
the floor.
Fire Induced Air Supply Rate
The fire induced air supply rate, ma (kg/s) , to a fire enclosure through a single opening
with height Ho and width Wo is given by the following expression (Ref. /7-20/):
ma= k WoH;' 2 Eq. (7.34)
where k is given by the following expression :
~ CdP
3 °
2g(1- TTo)
Eq. (7.35)
where:
opening flow coefficient ~ 0.7

average temperature of the hot gas layer
Scandpower
For normal indoor conditions (i.e. T0 = 293 Kand p0 = 1.2 kg/m 3 ), k is approximately
equal to 0.5 when T9 and the ratio mdma vary within wide ranges. Table 7.1 should
show the constancy of k for widely different fire conditions. Hence, a constant value of
k = 0.5 is selected and Eq. (7.34) and Eq. (7.35) are simplified as follows:
ma= 0.5 WoH~12 Eq. (7.36)
However, Kanury, Ref. /7-21/, maintains that for large openings the factor is signifi-
cantly less than 0.5. k = 0.13 is given by Kanury, Ref. /7-21/, for large openings.
If the room has n openings, then the effective or resultant opening height H. may be
calculated from the following expression (from Ref. /7-22/):
n
12
I<woH~ );
-}He =-'i=1c,_n_ __ Eq. (7.37)
L(WoHo);
i=1
Table 7.1: Values of the Coefficient kin Eq. (7.34) for varying hot Gas Layer Tem-
oeratures, T"• and Ratios of the burnim Rate to Air Sunnlv Rate, mlma
.
TJTa (K) 800 1,000 1,300 1,500
mlma
1/10 0.50 0.49 0.47 0.46
1/15 0.51 0.50 0.48 0.47
1/20 0.51 0.51 0.49 0.48
The factor W0 H/12 of Eq. (7.37) is often termed as the "opening factor". This factor
must be replaced by an "effective" or "resultant" opening factor , F, given by the ex-
pression:
n
F = L(WoHo); · -}He Eq. (7.38)
i=1
The fire induced air supply rate is for a fire compartment with n openings given by the
expression:
n
ma= k I<woHo);. -}He Eq. (7.39)
i=1
where k = 0.5 for normal sized openings (e.g. windows, doors etc.) or k = 0.13 for
large openings (i.e. with a typical dimension or extent larger than 5 - 10 m).
7.4.4 Flame Shape

The shape of the flame will for large pools in relation to the fire compartment and ven-
tilation rate be determined by the ceiling height and the location of the vent openings.
Scandpower
Risk Management AS (!])SINTEF
For ceiling heights more than 1 - 2 times the pool diameter, the rising flame plume will
probably not impinge on the ceiling. Thus, the flame height may be predicted in the
same way as for pool fires in the open air (i.e. according to Eq. (7.9) in Chapter 7.2.3).
For low ceiling heights in relation to pool diameter (H < D), the flame plume will im-
pinge on the ceiling and there will be an axisymmetric radial flame plume impingement
under the ceiling of length Hr as shown in Figure 7.6.
1 'I
Figure 7.6: A Sketch showing the Radial Flame Plume Impingement under the
Ceiling (D is the Diameter of the burning Pool, H is the Ceiling Height and
Hr is the Radial Flame Plume Impingement on the Ceiling)
Radial Flame Plume Impingement under the ceiling.
The radial flame plume impingement under the ceiling , Hr, is given by the following ex-
pression by You et al., Ref. /7-23/:
Ceiling with no side walls (L > H)
0 95
Hr /D = 0.5 [(L - H) ID] · Eq. (7.40)
Ceiling with side walls (L > H)
0 89
Hr /D = 0.69 [(L - H) ID] · Eq. (7.41)
where D is the equivalent diameter of the pool fire, L is the flame height in the open
and H is the ceiling height.
Fire Expelling from Openings
For underventilated fires the excess fuel vapour will burn outside the compartment
where the fuel comes in contact with sufficient air. A diffusive flare fire will emerge
from the opening with a width equal to the width of the opening and a depth equal to
approximately two-thirds of the height of the opening as accounted for in Chapter
6.3.2. The height of the flames emerging from the opening is given by Eq. (6.48)
Scandpower
where the effective diameter of the rising flame plume is predicted according to Eq.
(6.49) in order to take into account the effect of a wall above the opening.
A criterion for predicting the amount of excess fuel burning from the openings as a
diffusive flare fire may be the stoichiometric air to fuel ratio. That is, the theoretical
maximum amount of fuel burning inside the fire enclosure, mfi, is dependent on the
rate of air supply, m., and the stoichiometric fuel ratio, r, according to the following
equation:
rh _ rha Eq. (7.42)

n--r-
The remaining part of the fuel which will burn from the opening as a diffusive flare fire,
is then given by the following equation:
rhto =rhr - rhfi =rht - rhar Eq. (7.43)
where m1 is the total amount of fuel evaporating from the fuel surface.
In the equation above it is assumed that all available oxygen in the fire enclosure is
consumed by the fire. That is, an ideal combustion reactor is assumed in which no
excess air is supplied and all the oxygen in the air is being consumed by the combus-
tion process. Such a reactor provides for good mixing between fuel vapour molecules
and the available oxygen. A necessary prerequisite for such a theoretical reactor is
generally high reactor temperatures.
In the early stage of the enclosed fire in which the temperatures are generally low, a
part of the fuel may not even react with the oxygen in spite of excess air supply due to
the incomplete mixing of fuel and oxygen molecules. Eq. (7.43) applies primarily for
fully developed, ventilation controlled fires in which the fire temperature is generally
high. Eq. (7.40) will usually underestimate the amount of fuel vapour burning from the
ventilation openings of a fire enclosure. In the early stage of the fire, there may
appear fire expelling from openings long before the fire has become understoichio-
metric (i.e. m. < m,-r). In this situation Eq. (7.43) will give no excess fuel and thus no
external flames from the openings. However, since the fire severity of fully developed
enclosed pool fires is the most interesting, Eq. (7.43) may be an approximation to the
amount of excess fuel burning from the openings.
Burning Rate of enclosed Pool Fires
Due to the intense heat release of most ventilation controlled pool fires which cause a
complex radiative exchange between the flames, hot combustion gases and internal
surfaces of the fire enclosure, it is often anticipated that enclosed pool fires have a
higher burning rate compared to an equally sized pool fire in the open. The enhanced
burning rate is due to radiative feed back from the hot gas layer and the surrounding
solid boundaries of the fire compartment.
Experiments (Ref. /7-241) with small scale enclosed pool fires have shown that the
burning rate may increase by a factor of 6 - 7 compared to corresponding burning
rates in the open. However, the flame plume in these experiments was not optically
thick due to the small scale and the fuel, which did not generate much soot (i.e. an
alcohol fuel). Consequently, most of the radiative feedback did reach the pool surface
causing increased fuel evaporation.
Scandpower
In case of large-scale crude oil pool fires, which generate large amounts of soot and
black smoke, the flame plume covering the pool surface will be optically thick with res-
pect to thermal radiation. In this case, the radiative feedback will not reach the pool
surface. Hence, the burning rate of enclosed large-scale hydrocarbon pool fires will
hardly be much higher than corresponding pools in the open. The burning rate may
be slightly increased due to increased flame temperature. The burning rate may be
predicted as for open pool fires as accounted for in Chapter 7.2.3. It may, however,
be advisable to increase the open burning rates by 10 - 20 % in order to take into
account the effects of increased flame temperatures.
7.4.5 Flame and Gas Temperatures

Well Ventilated Pool Fires
As a fire plume intercepts the ceiling, the flow is turned, resulting in an axisymmetric
ceiling jet. If there are nearby walls, the flow will be transformed from a radial flow to a
downward wall flow. Alpert et al. , Ref. /7-25/, give the following expressions for the
temperature rise of steady state unbounded smooth ceiling flows:
For r < 0.2 H:
213
~T= 22.2Q To Eq. (7.44)
H513 293
where r is the radial distance along the ceiling from the centre of the area of the ceiling
at which the plume is impinging (m), TO is the ambient temperature (K), Q is the rate of
heat release by the fire (kW) and His the ceiling height of the fire enclosure (m).
Alpert et al., Ref. /7-25/, state that the temperature is fairly uniform at the value given
by Eq. (7.44) for a radius, r, along the ceiling above the fire less than 0.2H. Beyond
this area of the ceiling the following expression is valid (from Ref. /7-25/):
For r > 0.2H:
Q)2/3
~T =
6.8 -
r
( To Eq. (7.45)
H513 293
Further, Alpert et al., Ref. /7-25/, maintain that these equations are empirical and lose
precision as the distance between the pool and the ceiling becomes either very small
or very large. Accuracy can also be lost in small rooms where the heat is confined. It
is, therefore, wise to depend on results from these equations only when the smallest
dimension of the ceiling is larger than 2H. A more exact expression is the following
expression for the minimum permissible ceiling area, Amin, given by Alpert et al., Ref.
/7-25/:
A . = 0.048 M,~Hc Eq. (7.46)

min (~)2/3
Scandpower
Page7-36 Handbook for Fire Calculations and Fire Risk Assessment in the Process Industry
where M1 is the total amount of liquid fuel (in kg) involved in the fire. The Eqs. (7.44)
and (7.45) should only be applied when the ceiling area is larger than Amin•
When assessing the hazards when the ceiling flow is limited between two beams, the
use of the unbounded ceiling correlation will underestimate the thermal exposure.
Seyler, Ref. 17-26/, recommends that where beamed ceiling or corridor configuration
confines the flow, the following ceiling jet temperature correlation developed by Deli-
chatsios, Ref. /7-27/, should be used for r > lb'> (see Figure 7.7):
Eq. (7.47)
where .1.T is the temperature rise above ambient after the flow has impinged on a
beam or corridor wall, i.e. when the distance r down the corridor is greater than half
the corridor width or half the distance between two beams (i.e. r > lb)- .1.T is the tem-
perature rise near the ceiling over the source fire and can be predicted by Eq. (7.44).
Further, Eq. (7.47) should only be used when the beams are deep enough to fully con-
fine the flow (i.e. when: hJH > 0.1 (IJH)"113).
Underventilated Pool Fires
Stensaas, Ref. 17-28/, has on the basis of an extensive experimental programme with
enclosed pool fires in a room with the dimensions 1 m x 2 m x 1 m, developed the
following expression for the temperature rise as a function of time:
Eq. (7.48)
where:
mfi = the part of the total burning or evaporation rate of fuel per pool surface
area at which there is sufficient air according to stoichiometric condi-
tions•> (kg/sm 2)
Ap = pool area (m2)
.1.Hc = heat of combustion (kJ/kg)
t = time (sec)
ma = rate of air supply rate (forced or fire induced) (kg/s)
Cpg = specific heat of the fire gases•> (kJ/kgK)
p = density of the bounding walls (kg/m 3)
Cp = specific heat of the boundaries (kJ/kgK)
A, = thermal conductivity of the boundaries (kW/mK)
Aw = inner surface area of the fire enclosure (m 2) (exclusive the opening
area)
7) '• is half the distance between two beams or half the corridor width and h• is the vertical depth of the beams (see
Figure 7.5)
8) For understoichiometric conditions m"fi is given by the following relation:
m"fi = ma/rAp
where r is the stoichiometric air to fuel ratio for the actual fuel (given by Eq . (6.47))
9) Can be set equal to specific heat of air because fire gases consist mostly of air
Scandpower
:m: - Channel Flow

( Critical Flo w }
t
14----21b----
H _,__ _ Buoyan t ]I - Flow Transition

Plume (Density Jump)
l_ I
I
/
Fire Source /
I
'----t-Turnino
I
Reqion
,
,Fire Radius r0
I
\ ,_, Source
Fr o nt View \ I- Ax isymmtt ric //
'\, Rodial Cei linQ /
, Jet
..... _- _,,,,,..
/
Top Vit w
Figure 7. 7: Ceiling Flow induced by a Fire Source centrally located between two long
Ceiling Beams (from Ref. /7-27/)
Eq. (7.40) is capable of predicting the average hot gas layer temperature development
as a function of time of the transient phase of the fire when the burning rate per pool
surface area, the pool area, the air supply rate to the enclosure, the inner surface area
of the fire room, the heat of combustion of the fuel and the density, thermal conduc-
tivity and heat capacity of the solid boundaries are known.
For the steady state phase of the fire the following expression for the average hot gas
layer temperature is developed:
Eq. (7.49)
where o is the thickness of the enclosing boundaries of the fire enclosure. A serious
drawback of the expression of Eq. (7.49) is that it has not yet been verified experi-
mentally as is the case of Eq. (7.48). That is, the constant and the exponents of Eq.
(7.49) are found on the basis of more or less highly transient experiments and is, thus,
valid only for this phase of the fire. However, Eq (7.49) seems to yield reasonable
values for the steady-state value of the average hot gas temperature.
The steady-state temperature is attained when the transient temperature correlation of

Eq. (7.48) yields a value of T9 which is equal to the value given by the steady-state
expression of T9 , i.e. Eq. (7.49). Hence, the average hot gas layer temperature T9 is
always equal to the smaller of the values given by the Eqs. (7.48) and (7.49).
Scandpower
~ SINTEF
Ris k M anagement AS
From Eq. (7.48) and Eq. (7.49) one can deduce the following expressions which re-
present the effective heat transfer coefficient of the bounding walls for the transient
phase and the steady state phase of the fire, respectively (from Mccaffrey et al., Ref.
/7-29/):
h = ✓(PCpA-)w Eq. (7.50)
Eq. (7.51)
If the enclosing walls of the fire enclosure consist of different constructions, they must
be considered separately and then summed according to the following equation, Ref.
/7-29/:
Eq. (7.52)
If a construction consists of several material layers, the following expression applies to

the effective heat loss coefficient of this construction 10>:
Eq. (7.53)

1. Heat Transfer on Object engulfed by Flames or Hot Gases
The heat transfer to objects engulfed by flames or hot gases (e.g. in the hot gas layer)
in connection with an enclosed pool fire will not in principle deviate from the heat
transfer in the open air. This subject is covered by Appendix H.
2. Ceiling Heat Transfer - Well Ventilated Pool Fires
The general convective heat transfer correlations presented in Appendix H, may not
be valid for ceiling jets. This correlation shows a strong dependence of heat transfer
coefficient on the object size or jet scale. In case of ceiling jet heat transfer where
large scale turbulent eddies are constantly being generated by buoyancy forces, the
heat transfer depends more on these large scale eddies than on the local turbulence
typical of the forced convection heat transfer correlation presented in Appendix B.
Hence, the convective ceiling heat transfer must be treated separately.
The ceiling heat transfer correlations of Cooper, Ref. /7-30/, are a result of a review of
other investigators' data and other sources. As such these correlations are recom-
mended for general use. The general convective heat transfer correlation is given by:
10) Since Eq. (7.48) is correlated on the basis of enclosed pool fire experiments in which the walls of the fire enclosure
were made of fibreboards with a ve,y low thennal conductivity compared with that of steel (i.e. 0.01 kW/mK vs. 50
kW/mK for steel), this correlation will underestimate the average hot gas layer temperature in case of walls with a
high thermal conductivity, also in cases where steel is protected by a fire insulation material. Consequently, in
cases with wall materials of high thennal conductivity (i.e. ;... > 0. 1 kWlmK), the thennal conductivity have to be set
equal to 0. 01 kW/mK. This can be justified since it is supposed that the enclosing boundaries of the fire enclosure
do not affect the fire severity ve,y much
Scandpower
q" =h(T - Ts) Eq. (7.54)
where h is the ceiling heat transfer coefficient by convection, T is ceiling exposure

temperature predicted by the previously proposed equations and Ts the ceiling surface
temperature. The following correlations for the ceiling heat transfer coefficient is given
by Cooper, Ref. /7-30/:
-0.65
0.213 Re·0 ·3 (
~) when 0.2 < ~ < 1.03 Eq. (7.55)
-1 .2
0.217Re·0 ·3 (
~) when1 .03<~
where:
• (Q)1/3
h =0.365 H Eq. (7.56)
and
Re= 0.304Q113H213 v·1 Eq. (7.57)
where r is the radial distance along the ceiling from the centre of the area at which the
ceiling jet impinges the ceiling, H is the enclosure height, Q is the rate of heat release
by the fire (predicted from Eq. (7.22)) and the kinematic viscosity v is given by the
following expression:
T 512
V =4.128 -10- 9 -
9
- -- Eq. (7.58)
T9 + 110.4
3. Ceiling Heat Transfer - Underventilated Pool Fires
Underventilated fires are, as already described, characterized by a thick and hot gas
layer under the ceiling which often extends from two-thirds to three-fourths of the
ceiling height. The presence of this hot gas layer may increase the temperatures in
plumes and ceiling jets immersed in the hot layer. Therefore, Cooper, Ref. /7-30/, pro-
poses the following alternative set of equations for calculating the ceiling heat transfer in
case a hot gas layer in the fire room:
q"=h'(Tad - Ts} Eq. (7.59)
Scandpower
-0.65
- ·3
0
0.213 R'e ( ~. ) when 0.2 < ~. < 1.03 Eq. (7.60)
-1.2
0.217 R'e -0-3( ~. ) when 1.03 < ~.
where:
•
h =0.365 H'
(Q')113 Eq. (7.61)
and:
T Q' 213 r
(Tad - T g) -- 28 •1ToH'513e
g -1-77-
H'
Eq. (7.62)
r
for 0 ,;; H' ,;; 0. 75
or:
_ Tg Q' 213 ( r )-0.00

(Tad - T 9 )-5.77 To H' 513 H'
Eq. (7.63)
for 0.75,;; ~.
R'e = 0.304Q' 113 H' 213 v-1 Eq. (7.64)
H' and Q' are given by the following expressions where h is the hot gas layer height
above floor (sE!!e next section), H is the ceiling height, T9 is the average hot gas layer
temperature and T 0 is the ambient temperature:
,_ [ (T9)3'5(Q'- )•"5]
H -H - h9 1 - -
To Q
Eq. (7.65)
Eq. (7.66)
7.4.7 Smoke
The Height of the hot Gas Layer above Floor
The following equation represents an expression for the hot gas layer height above
the floor for a room height H, burning rate m1, ventilation rate ma and gas temperature
T9 (Ref. /7-21/):
Scandpower
Ventilated compartments
H
Eq. (7.67)
hg = ( 213( )112
1 + 1 + ~:) ;:
When the average hot gas layer temperature, T9 , is predicted from Eq. (7.48), the fire
induced air supply rate, ma, is predicted from Eq. (7.36) and the burning rate, m1, is
given an appropriate value, then the hot gas layer height above the floor of ventilated
compartments may be predicted by Eq. (7.67).
Completely closed Compartments
If, on the other hand, the room is completely closed without any ventilation opening
(i.e. in case of forced ventilation and shutdown of the ventilation system when the fire
is detected), the hot gas layer height above the floor h9 as a function of the time t is
given by the following expression for normal indoor conditions (Ref. /7-31/):
Eq. (7.68)
where:
H = the internal height of the enclosed space (m)

At = floor area (m2)
Q = the rate of heat release (kW)
t = the time from ignition (sec)
The Rate of Smoke Production of enclosed Fires
For calculation of the smoke production, m9 , within an enclosed area in the appea-
rance of a hot gas layer of height h9 above the floor, the following expression of
Bitchen et al. , Ref. /7-32/, is often used:
Eq. (7.69)
where:
P = perimeter of the fire (m)

h9 = smoke layer height above the floor (predicted by Eq. (7.61) or Eq.
(7.62)) (m)
Po = =
density of the ambient air (kg/m3 ) 352.17/T0 where T0 is the ambient
temperature (K)
T9 = temperature of the hot gases in the plume at the ceiling (K) (can be
predicted by Eq. (7.48) or Eq. (7.49) dependent on whether the fire is
underventilated or well ventilated)
Eq. (7.69 ) assumes that the rate of smoke production is primarily determined by the
rate of air entrainment in the plume of flames and combustion gases rising above the
Scandpower
fire. Further, this equation assumes constant and upward velocity of the flame zone
and that the following dissimilarities must be met:
JAt > hg and L > h9 Eq. (7.70)

2
where Ar is the floor area, h9 is the smoke layer height above the floor and L is the
flame length predicted by Eq. (7.9).
For normal indoor conditions (i.e. T0 =

293 K, p0 =
1.2 kg/m 3 , and T 9 = 1,073 K =
800°C), Eq. (7 .69) reduces to the following simplified expression:
Eq. (7.71)
The Optical Density of enclosed Fires
The optical density per meter path length, OD, can be expressed by the following rela-
tion:
OD=ax[ A;] Eq. (7.72)
where:
OD = (log(lofl))/L where 10 r,N/m 2) is the intensity of the light source and I

(W/m2) is the intensity of the light beam after passing through the
smoke over a distance L (m- 1)
L\m= total mass of fuel burnt and contained in the enclosure at time t (is
equal to the total mass of material burnt if the enclosure is closed) (kg)
V = is the volume in which the smoke is contained (m 3)
ax = "mass optical density" (m 2/kg) where:
(1 = "particulate optical density" (m 2/kg). It is for flaming combustion often
assumed to be a constant "' 3,300 m2/kg
X = the mass fraction converted to smoke or soot
If the "particulate optical density", X, (varies in the range 0.05 - 0.15 for liquid hydro-
carbon fuels) and the total mass burnt of fuel Am are given (the latter as function of
time from the outbreak of fire), the optical density per meter path length may be pre-
dicted by Eq. (7.72).
Visibility in the Gas Layer of enclosed Pool Fires/Criteria for Smoke Blockage
The visibility is given by Eq. (J.9) when the optical density is known.
Benjamin, Ref. /7-33/ reports that an OD of 0.25 m·1 is approximately the threshold of
diminished performance and that this is the smoke level that one should have in mind
when designing building systems. An OD of 0.25 m·1 corresponds to a visibility of
about 4 m which is the dimension of a typical living room. Further, it has been estab-
lished that a visibility of about 1 m (OD = 1 m·'), i.e. a visibility corresponding to the
length of an arm, will be of no help at all when escaping from an enclosed fire atmos-
phere. Figure 14 shows the visibility in a smoke laden atmosphere as a function of the
optical thickness per meter length.
Scandpower
Benjamin, Ref. /7-33/, maintains that for even major changes in the properties of the
burning fuel as for instance its mass optical density, there will be small effects on the
available safe egression time which generally will be low almost regardless of the
properties of the burning fuel. Benjamin, Ref. /7-33/, ascertains that the escape time
can be more effectively influenced by decreasing the burning rate than by trying to
alter the smoke producing properties of the burning substance. Further, that the size
of the fire will effectively control the amount of smoke generated and therefore have
an indirect effect on smoke spread control. Hence, the use of automatic sprinklers will
be an effective means for smoke spread control.
~
10 I .A. I
•
5
---
E
>-
2
■
~
..0 1
u;
• Focused lamp in yellow smoke?9
> VFocused lamp in black smoke29
□ Illuminated slgn' 9
0.5
0 Illuminated slgn29
■ Diffuse light2&
~ Focused lamp in yellow smoke,29 0
ABacidighted slgn 5
0.2 A Illuminated p lacard5
~ Illuminated sign'
0.05 0.1 0.2 0.5 1 2 5 10

0/L, optical 1length of smoke (m-1 )
Figure 7.8: The Visibility in a Smoke Laden Atmosphere as a Function of the Mass
optical Density per M path Length (from Quintiere, Ref. /7-34/)
7.4.8 Duration of Fire

With respect to the calculation of the duration of an enclosed pool fire, it will not
deviate from pool fires in the open air, which is dealt with in Chapter 7.2.7.
7.4.9 Overview of Findings from Fire Tests with Confined Pool Fires
SINTEF Energy - Norwegian Fire Research Laboratory (NBL) has carried out a series
of fire tests with confined pool fires, according to the "Test Programme F3 of Blast and
Fire Engineering for Topside Structures", Ref. /7-35/. The final test programme con-
sisted of 7 pool fires, in two compartment sizes, 135 m3 and 415 m3 . In all tests the
fuel was a Statoil S/eipner Condensate.
Scandpower
{r;J) SINTEF Risk M anagement AS
Page 7--44 Handbook for Fire Calculations and Fire Risk Assessment in the Process Industry
The test results are compared with results from a previous test series in the 135 m3
compartment. The conclusions from the tests including confined pool fires are given
in the following sections.
7 .4.9.1 Effect of Confinement on the Behaviour of Pool Fires
The following conclusions can be drawn with respect to the effect of confinement on
the behaviour of pool fires.
During the initial stages of fire development, confined pool fires behave as they
would in the open
After a short period, ranging from a few seconds to a few minutes, the develop-
ment of the fire depends on the degree of ventilation control, specifically on the
value of global stoichiometry. (Definition: .i, = Air/fuel mass rate into the com-
partment divided by the air mass requirement for stoichiometric combustion)
A well-defined horizontal interface between an upper hot gas/smoke layer and

lower cool air layer forms. Depending on the relation of the vent flow and the
size and position of the fire source, the conditions for ventilation or fuel con-
trolled burning are established. When the burning approach ventilation controlled
conditions, it is possible for combustion at the interface between these layers to
be highly oscillatory and unstable. This may lead to rapid vigorous combustion
and high heat fluxes and temperatures above 1,350 °c due to soot oxidation
when temperatures rise above approximately 1,200 °c
Before steady state conditions are achieved, incident heat fluxes and tempera-
ture rise rates can diminish if the fire enters a ventilation controlled regime.
Copious amounts of soot are produced from incomplete combustion, particularly
when the temperature of the smoke layer is greater than 900 •c. If the soot is
deposited in an area where the temperature is below 900 °c, the soot can act
like a heat shield insulating the surface of the walls, roof and objects from the
radiative flames. However, if the soot is deposited in an area where temperature
is high enough (greater than 1,200 °C) the burning soot may contribute by in-
creasing the heat load to the walls, roof and objects
How severe this effect is and the exact temperature at which this becomes im-
portant, can only be estimated. An estimate based on the experimental condi-
tions in this project, a "fireball" of diameter greater than 4-6 m is needed to en-
sure extensive soot oxidation, leading to the increased heat load
When soot oxidation occurs within a large combustion zone inside the compart-
ment, the radiation from this hot zone makes the surroundings hot as well. The
heat loss from the core zone is subsequently reduced, as the radiation is
"trapped", and the temperature increases further. The "radiation trapping" may
occur in larger flames, making the influence of the thermal properties of the com-
partment walls and ceiling insignificant
The diameter of a "fireball" which may be critical for reaching a regime of com-
bustion with heat fluxes in the order of 350-400 kW/m 2 and gas temperatures
above 1,370 °c, is between 4 - 6 m. This will then coincide with the critical resi-
dence time. of soot involved in the reaction, and with "radiation trapping"
Scandpower
Risk Management AS lt))SINTEF
It seems easier to obtain higher heat flux values and temperatures to the roof
and wall surfaces located near the corners. The geometry of a corner may pro-
mote higher turbulence, and it is also possible that interchange of radiation bet-
ween hot surfaces creates higher heat fluxes than flame radiation only. An
example of this effect is the steel plates of the west wall were totally damaged
and to some extent melted
In general the CO level increases with decreasing <j, for all large scale fires, but
the temperature and residence times are also important as these parameters
determine the dominant combustion reaction kinetics. For the small scaled fires
3
(135 m ) the CO level seems to be more constant when <j, varies. In the tests
with temperatures above 1,200 °c in large volumes, the concentration of CO is
also significantly higher than in other tests. This indicates more extensive soot
oxidation resulting in more CO under these conditions
Confined condensate pool fires never reach extremely ventilation controlled con-
ditions (<j, is never smaller than 0.8). The overall burning rate of condensate pool
fires enters a self-limiting regime as ventilation controlled conditions are approa-
ched such that <j, is always greater than approximately 0.8. The final burning
rate is lower than expected when comparing with the burning rate of an open
pool fire of the same size
In ventilation controlled pool fires, soot ignition on exiting the vent can produce
high levels of external radiation. Inside the compartment heat fluxes and tempe-
ratures increase when the soot burning. The temperature and chemical compo-
sition of the smoke layer at the vent determine whether or not sustainable com-
bustion is possible
The general trends observed in these tests are expected to be applicable to

uninsulated compartments. However, the surface and gas temperatures will be
lower for uninsulated compartments of similar size, which will affect the che-
mistry and soot production in areas close to these surfaces. When enclosures
are large, the influence of thermal properties of walls and ceiling (for instance
insulation) on fire development is reduced
In the test condition with the vent opening in one wall, the region of maximum
combustion intensity shifts from the pool towards the vent as <j, decreases
The results show that there is no difference in the pool fire burning rate between
a pool fire on water or a pool fire on a steel substrate
The start conditions for a pool fire are important with respect to the rapidity of
establishing high heat loads inside the compartment during the fire. The forma-
tion of a large "fireball" initially may trigger the process of soot oxidation, which
would not occur if the fire size slowly increased
The equivalent ratio (<j,) seems to be very much effected by the vent configura-
tion. This is easily seen in the tests with the split vent opening. It is also obser-
ved that variation in the fuel flow rate has only minor influence on the equiva-
lence ratio (<j,), while <j, is much more sensitive to changes in ventilation opening
size and location. These observations are valid for the test configurations inves-
tigated
Scandpower
Page 7-46 Handbook for Fire Calculations and Fire Risk Assessment In the Process Industry
The radiation from the external flames seems to be shielded by soot in all the
condensate fires. For the corresponding confined propane fire tests, the radia-
tion from the external flames to the surroundings were 10 times higher than the
enclosed condensate pool fires due to less soot shielding.
7.4.9.2 The Thermal Load onto Process Vessels, Pipework, Module Walls and
Module Decks from Pool Fires
The following conclusions can be drawn with respect to the thermal load onto process
vessels, pipework, module walls and module decks from enclosed pool fires:
The general finding in the tests is that a heat flux density onto a target (total
incident heat flux, both radiative and convective) is up to 200 kW/m 2 • However,
at certain conditions considerably higher fluxes may be seen, e.g. in the order of
350 - 400 kW/m 2• These fluxes occur simultaneously with high temperatures,
above the saturation level of the data logging system 1,370 °c, extensively
throughout the compartment
At steady state conditions, incident heat fluxes to the surrounding walls, ceiling
and impinged objects are comparable in magnitude to those found for impinging
pool fires in the open, but can be higher under certain conditions
For confined pool fires, the final equilibrium temperatures of the walls, ceiling
and objects generally increase with decreasing ,i,, due to greater soot radiation.
Average combustion gas temperatures do not depend strongly on ,i, suggesting
that the increase in heat flux to the walls is attributable to an increase in optical
thickness in the smoke layer. The increased optical thickness is due to higher
concentration of soot particles
Generally, confined pool fires are not extinguished by water deluge, but the fire
is controlled and burns at a much reduced rate.
An extensive set of quality data for model validation and development of fire hazard
prediction techniques is presented in reference, Ref. /7-35/, in the technical reports
(one individual report for each test) and in video recordings from all tests. The accu-
racy and usefulness of data are presented in this reference.
7.4.10 Example
Unstabilised crude oil is leaking from the liquid outlet of the test separator at a con-
stant rate of 10 kg/s. The oil is filling a dike of dimensions 8.5 n x 13.5 m x 0.1 m.
The test separator is. located in an offshore module of gross dimensions (width x
length x height): 20 m x 40 m x 15 m. There is a ventilation opening of dimensions
20 m x 7.5 m on each of the short walls. All boundaries consist of fire insulated steel
plates. Due to the fact that the oil is unstabilised, 25 % of the burning rate per pool
area has to be added to the burning intensity of stabilised crude oil.
The fire development, flame shape, hot gas layer height above the floor and maximum
ceiling temperatures as well as the maximum ceiling heat transfer shall be predicted
for this enclosed pool fire within two hours from ignition.
Scandpower
Solution
Burning rate per pool area of stabilised crude oil: 0.04 kg/sm 2
+ 25 % due to gas flashing: 0.01 kg/sm 2
burning rate per unit pool area of unstabilised crude oil: 0.05 kg/sm 2
Regression rate of crude oil: 3.6 mm/min
Burning efficiency: 70%
Density of crude oil: 830 kg/m 3
Initial room temperature: 20 °C
Effective properties of the boundaries:
* Thickness: 0.03 m
* Density: 400 kg/m 3
* Specific heat: 1.0 kJ/kg
* Thermal conductivity: 0.12 W/mK
2. Prediction of flame geometry:
Eq. (7.8) predicts an equivalent diameter of the rectangular pool of 10.4 m. Since Eq.
(7.6) predicts an equilibrium diameter of 16.0 m which is larger than the extent of the
dike, the released liquid fuel will be contained in the dike. However, because the leak-
age rate is larger than the burning rate of the pool fire, the dike will be completely filled
with crude oil and the oil will spread beyond the extents of the dike and attain a cir-
cular equilibrium pool of diameter 16.0 m.
The net filling rate of oil will be the filling rate minus the burning rate =
(10 • 60,000/830)/115- 3.6 = 2.7 mm/min. The overflow time is: 100/2.7 = 37 min.
Eq. (7.9) predicts a flame height of 15.4 m until 37 min. of the fire and a flame height
of 20.7 m from 37 min when the pool area is increased due to flow of oil beyond the
boundaries of the dike.
Likewise Eq. (7.44) predicts the extent of the flame plume impingement under the
ceiling of 8.5 m and 10.7 m, respectively before and after the overflow.
3. Prediction of the hot gas layer temperature:
11
Eq. (7.39) > predicts a fire induced air supply or ventilation rate of 410.8 kg/s.
Eq. (7.49) predicts a stationary temperature of hot gases of 663 °c after 50 min. Eq.
(7.48) predicts the following expressions before and after the overflow of the dike:
0-37 min.: Tg,1 = 20 + 179 . t 118 (°C)
37 - 120 min.: T 9 , 2 = 20 + 236. t 118 (°C)
Both the expressions above predict that the stationary temperature of 643 °C is
attained after approximately 50 min regardless of the size of the pool area. Figure 7.9
shows the development of the average hot gas layer temperatures within two hours.
11) A value of 0. 5 is assumed for the factor kin Eq. (7.39)
~ SIN TEF Scandpower

Risk M anagement AS
4. Prediction of the hot gas layer height:
With a total burning rate of 5.8 and 1O kg/s before and after the 37th minute (i.e.
before and after the overflow of oil), Eq. (7.67) predicts a hot gas layer height above
floor of 5.7 m and 5.3 m, respectively.
5. Maximum ceiling temperature and heat transfer:
Eq. (7.22) predicts a total rate of energy release of 175,510 kW which is increased to
305,900 kW after the overflow. Eq. (7.41) predicts a maximum ceiling flame tempe-
rature of 783 and 1,135 °c before and after the overflow of the dike. Eq. (7.59) pre-
dicts a maximum convective heat transfer to the ceiling of 42.9 kW/m 2 and
51.6 kW/m2 • Likewise, Eq. (H.2) predicts an increase of radiative heat transfer from
65.3 kW/m 2 to 216.6 kW/m 2 • Thus, the total maximum ceiling heat transfer to the
ceiling which occurs over a circular area of diameter 6 m with a centre just above the
pool centre, is increased from 108 kW/m 2 and 268 kW/m 2 at the time of overflow.
700
800
\i'u
•
~
'""
i! 400
300
l
!
J 1QC
0
0 20 40 eo 80 120
TrJM 1.-011 ron1t1on (lft1n,)
Figure 7.9: Development of the Average Temperature of the hot Gas Layer for the
first two Hours of the Fire
Conclusions:
This enclosed pool fire exhibits a fire growth after 37 min due to the the enlargement
of the pool area from a rectangular pool of dimensions 8.5 m x 13.5 m with an area of
115 m2 to a circular pool with a diameter of approximately 16 m and a pool area of
200 m2 due to the overflow of oil beyond the extents of the dike. In both cases the
enclosed pool fire will reach the ceiling creating a radial 'flame plume impingement'
under the ceiling. The diameter of this flame impingement area is increased from
8.5 m to 10.7 m by the time of overflow.
Scandpower
Risk Management AS Gl SINTEF
The average hot gas layer temperature is stabilised after 50 min at 663 °c. The maxi-
mum ceiling temperature is increased from 783 °c to 1,135 °c at the time of overflow.
This results in a dramatic increase in the radiative heat transfer to the ceiling from
65.3 kW/m2 to 216.6 kW/m2 . Hence, the total ceiling heat transfer is increased from
108 kW/m2 to 268 kW/m2 at 37 min after ignition.
It has to be pointed out that the average hot gas layer temperature as indicated in
Figure 7.9 as well as the other calculated characteristics, exhibit a more smooth tran-
sition between the values calculated here because it is assumed that the pool area is
increased immediately from 115 to 200 m2 •
7.5 Fireball
7.5.1 Scenario
The preceding event for a fireball may be a pool fire enveloping a pressurised storage
tank containing liquefied gas or a jet fire impinging on the storage tank. After some
time, the tank will rupture due to the intense heating. A BLEVE (Boiling Liquid Expan-
ding Vapour Explosion) with overpressure at the source in the order of 0.05 bar will
occur.
An other preceding accident causing this fire scenario is a major structural damage of
the tank leading to a more or less full rupture of the tank. Volatile fuels are being
transported in rapidly increasing volumes both by highway vehicles and trains, which
may be exposed to accidents (collision and derailment). A spark ignition may easily
occur in such situations resulting in a BLEVE/fireball.
A fraction of the liquefied fuel subsequently released will evaporate immediately and
take part in a more or less huge fireball which has the shape of a hemispherical
burning cloud or ball of fire which emits heat radiation over a relatively short period of
time. The strong buoyancy force of a hot combustion gases results in a highly turbu-
lent cloud in which rapid air entrainment occurs. A hemispherical shape is maintained
during most of the initial expansion until the fireball growth is exceeded by the buoy-
ancy, and then the spherical shape develops. Once completely formed, the fireball
will lift off entraining further air, which results in a cooling of the fireball. Normally the
complete process takes 5 - 30 sec. It is the high degree of turbulent mixing and rapid
air entrainment which allow large quantities of fuel to be consumed in such a short
period of time.
The major hazards from BLEVEs/fireballs are blast effects, projectiles and thermal
radiation. Some BLEVE incidents have shown that close to the container the physical
blast during the rupture is equivalent to a wind speed of about 100 km/h. Further,
BLEVE incidents have caused projectiles to be thrown up to 600 m. Perhaps the most
serious aspect of fireballs is their effect on people who may suffer severe burns some
distance away from the fireball. Owing to the short duration of a fireball and the
thermal inertia of buildings and process equipment, it is unlikely that such construc-
tions will be damaged due to the thermal heat loads even by 100 ton fireballs.
Scandpower
7 .5.2 Diameter and Height of Fireball

Numerous empirical models are available for calculating the fire characteristics of fire-
balls. Most of the models are based on the following basic assumptions:
The fireball is homogeneous, isothermal and spherical

All the fuel within the expanding vapour cloud is consumed
The fireball radiates as a black body (i.e. the flame emissivity is unity)
The wind has no effects on the fireball characteristics.
Most of the empirical models available for predicting both the maximum, Dmax, and
average diameter, Ds, of a fireball are of the form:
D=k•m" Eq. (7.73)
where m is the total mass of fuel involved in the fireball, k is a constant and n is an
exponent greater than zero. The range of mass of fuel in the experimental series from
which each respective model is developed, varies greatly (i.e. from milligrams to
1,000 tons). In spite of this fact the constant k and the exponent n of the empirical
correlations for the maximum diameter of a fireball developed on the basis of these
experimental series do not vary very much. The value of the exponent is generally
about 1/3 and the constant varies in the range from 5 to slightly above 6. The general
expression of Eq. (7.73) does, however, not take into account the fuel properties.
Lihou et al., Ref. /7-36/, have, on the basis of the experimental work of Maurer et al.,
12
Ref. /7-37/, developed the following general relationship for the average l diameter Ds
of a fireball:
Ds = 3.44[44.8 :J 113
Eq. (7.74)
where m is the total amount of fuel participating in the fireball d (in kg) and Mt is its
molecular weight of the fuel (kmol/kg).
Fay et al., Ref. /7-16/, present the following relationship for the maximum diameter
Dmax which is independent of fuel:
1
Dm=6.36 · ms Eq. (7.75)
The elevation height of the fireball centre above ground is given by the same general
expression as shown in Eq (7.73). However, there is very little documentation
12) It is more convenient to use the average diameter of the fireball than the maximum diameter when calculating the
incident heat flux to targets some distance away from the fireball because it will give more realistic heat loads over
the duration of the fire than using the maximum diameter
13) As already mentioned most models assume that all the fuel within the expanding vapour cloud is participating in
the fireball. However, this amount need not to be the total amount of fuel in the tank container because all this fuel
need not flash off to vapour. The total amount of fuel in the tank must be multiplied with a flash fraction in order to
get the total amount offuel participating in the fireball (termed as m).
Mudan, Ref. 17. 81 suggest the following rule of thumb for mass of fuel for fireball calculations:
1. If the flash fraction exceeds 30 percent, it should be assumed that the entire mass of fuel is contained in the
vapour cloud
2. If the flash fraction is less than 15 percent, it may be assumed that the remaining liquid will bum in the form of
a pool fire
3. If the flash fraction ranges from 15 to 30 percent, a linear interpolation is assumed for the liquid fraction
Scandpower
Risk Management AS
G)SINTEF
Norwegian Fire Research Laboratoiy
showing models of fireball height above ground. Various authors report that the fire-
ball height is from 0.5 - 3 times the diameter of the fireball.
Lihou et al. Ref. /7-36/ report that the average ascension rate of large-scale fireballs
(independent of fuel) is approximately 1O mis. In the Chapter 7.5.6, which deals with
the duration of fireballs, there is a similar expression for the fireball duration as for the
fireball diameter. The only difference is that the constant is equal to 0.31 . The height
of the fireball may also be expressed by the same general expression of Eq. (7.73)
with an exponent equal to 1/3.
The height of the fireball is given by the average rate of rise (10 mis) times the dura-
tion of the fireball. Hence, the following expression may be deduced for the average
elevation height of the fireball centre above the ignition point:
H =3.1 [ 44.8 :t ]
1/3
Eq. (7.76)
7.5.3 Temperature of Fireball

While peak temperatures during the initial expansion phase as high as 2,100 °c have
been reported, the effective fireball temperature is significantly less due to the air
entrainment. Crawley, Ref. /7-38/, reports that the effective fireball temperature will lie
in the range 1,100 - 1,200 °c. Lihou et al. , Ref. /7-36/ reports that the temperature of
a fireball will be fuel dependent and report the following effective radiation tempera-
tures:
Methane: 1,230 ± 20 °C
Propane: 1,020 ± 20 °C
Butane: 1,330 ± 20 °C

Since the duration of a fireball is from 5 - 30 sec, the structural response will not be
considered in this section. Only incident radiative heat transfer to human beings some
distance away from the fireball will be considered.
Crawley, Ref. /7-38/, maintains that the fireball flame will be optically dense behaving
as black body radiators with emitted surface heat fluxes in the order of 200 -
2
270 kW/m . According to the fuel dependent radiative temperatures listed in Chapter
7.5.3, the corresponding surface emissive powers of a fireball will be as follows when
assuming black body radiation according to the Stefan Bolzmann's law of radiation:
Methane: 290 ± 15 kW/m2

Propane: 160 ± 10 kW/m 2
Butane: 375 ± 20 kW/m2
The incident radiation to a target at a distance x from the fireball centre is given by:
q; = -rFEp Eq. (7.77)
~ SINTEF Scandpow er
Risk M anagement AS
where,: is the atmospheric transmissivity (given by Eq. (H.13) where xis equal to the
radial distance x' in Figure 7 .10) and F is the view factor"> of the target with respect to
the fireball. For a spherical fireball of diameter D, and height of fireball centre above
ground H, the value of the view factor F at a radial distance x' from the radiating sur-
face of the fireball to the target (see Figure 7.10), is given by the following equation
(from Lihou et al., Ref. /7-36/):
(H - Ds /2) + _
05
F= Ds Eq. (7.78)
4{ [ (H -□~• /2) + 0.5] 2+ [;.] 2r,2
When the average fireball diameter D,, height H and radiative temperature are pre-
dicted, the incident radiative heat flux onto a target outside the fireball may be pre-
dicted by Eq. (7.77), Eq. (7.78) and Eq. (H.13).
If the target is elevated a height h above the ground, this height has to be subtracted
from the fireball height H in Eq. (7.78).
MEAN DIAMETER D
...
l\~~i ~.
'-··-..=-~
r2 = d 2 + (z-h) 2
·- .......
........ / LOCATION OF
INTEREST
GROUND LEVEL
Figure 7.10: Schematic Sketch of the Fireball Radiation Model (Z is the Height of the
Fireball Centre, h is the Height of the Target above Ground Level, d is
the horizontal Distance of the Target from the Fireball Centre and D is
the Diameter of the Fireball)
14) The fraction of radiation emitted from the fireball, which would shine directly on a ground level target (i.e. the same
level as the pressurised storage tank)
Scandpower
Risk Management AS (Qi)SINTEF
7.5.5 Duration of Fireball

As already stated several times (!), the duration of fireballs is very short (5 - 30 sec.).
Lihou et al. , Ref. /7-36/, have derived the following similar relationship for the duration
of a fireball, ti, as for the fireball diameter:
Eq. (7.79)
7 .5.6 Example
A pressurised storage vessel containing 10,000 kg of butane is fully enveloped by a
large pool fire. Calculate fireball dimensions (i.e. diameter and height) and duration as
well as the incident heat flux onto the face of a human being 100 m from the tank 2 (!)
m above the ground. It is assumed that 80 % of the fuel flashes off to vapour imme-
diately at the time of tank rupture.
Solution
Fuel : Butane (C4H10)

Molecular weight: 58
Fuel mass: 10,000 kg
Flash fraction of the mass released: 80%
Effective radiation temperature: 1,350 °C
Emissivity of fireball: 1
Convective fraction of total rate of heat release: 0.8
Relative humidity: 30%
Location of fire exposed target: - Distance from tank: 100 m
- Height above ground: 2m
Since more than 30 % of the fuel flashes off to vapour, the rule of thumb presented in
the footnote on page U--says that all the fuel contained in the tank will take part in the
combustion process, i.e in the formation of the fireball.
Maximal diameter of fireball:
Eq. (7.75) yields form= 10,000 kg and Mr = 58: Dm= 137.0 m
Average or effective diameter and height of fireball:
Eq. (7.74) yields: Average or effective diameter: Ds = 68.0 m

Eq. (7.73) yields: Average or effective height: H = 61.3 m
Fireball duration:
Eq. (7.77) yields: Duration of the fireball: ti = 6.1 sec
Scandpower
(,J}SINTEF Risk M anagement AS
Page7-54 Handbook for Fire Calculations and Fire Risk Assessment in the Process Industry
Incident radiation onto the target:
The distance x, which is the distance between the hot core of the fireball and the fire
exposed surface, may be easily calculated by means of Pythagoras to: 82.3 m
For a distance x = 82.3 m and a relative humidity of 30 %, Eq. (H.13) predicts an

atmospheric transmissivity of: ~ = 0.8
For an effective radiation temperature of 1,350 °c = 1,623 K and fireball emissivity of

unity, Eq. (H.12) yields: Ep = 393.4 kW/m 2
Eq. (7.78) yields for a fireball height and diameter of 61.3 m and 68 m, respectively, a
view factor of: F = 0.044
Ultimately, Eq. (H.11) yields an incident radiation heat on to the face of the human
2
being of: QR= 393.4-0.044-0.8 = 14 kW/m
According to API RP 521 Guide given in Table 8.1 in Chapter 8.2, this level of thermal
radiation is permissible for escape for only a few seconds. Eq. (8.1) predicts for this
radiation level a time to pain on dry skin of: Ip "' 3 ½ sec
Sc and power
/7-1/ K.S. Mudan: "Thermal Radiaton Hazard from Hydrocarbon Pool Fires", Prog.
Energy Combust. Sci. , Vol. 10, 1984.
/7-2/ P.H. Thomas: "The Size of Flames from Natural Fires", pp.844-859, 9th Int.
Combustion Symposium, Comb. Inst., Pittsburgh, PA, 1963.
/7-3/ J. Moorhouse and M.J. Pritchard: "Thermal Radiaton Hazard from large Pool
Fires and Fireballs - A Literature Review", Inst. of Chem. Eng. Sym., Man-
chester, 1982.
/7-4/ P.A. Croce and K.S. Mudan: "Calculating Impacts for large Hydrocarbon Pool
Fires", Fire Safety Journal 11 , pp. 99-112, 1986.
17-51 J. Moorhouse: "Scaling Criteria for Pool Fires derived from large Scale Experi-
ments", Inst. of Chem. Eng. Symp. 71 , pp. 165-179, 1982.
/7-6/ K.S. Mudan and P.A. Croce: "Thermal Radiation for LNG Trench Fires", ASME
Winter Annual Meeting, New Orleans, LA, 1984.
17-71 D. Burgess and M. Herzberg: "Radiaton from Pool Flames, Heat Transfer in
Flames", Afgan and Beer, 1974.
/7-8/ Hagglund, B.: "The Heat Radiation from Petroleum Fires", FoU-Brand, pp.
18-24, 1977.
/7-9/ D. Evans et al.: "Burning, Smoke Production and Smoke Dispersion from Oil
Spill Combustion", NIST Report No. NISTIR 89-4091 , U.S. Department of
Commerce, National Institute of Standards and Technology, Gaithersburg,
May 1988.
/7-10/ C.H. Thompson et al.: "An Oil Mitigation Tool", U.S. Department of Energy
Report EV-1830-1 , Washington D.C.
/7-11/ S.K. Elam et al.: "Cone Heater Ignition Tests of Liquid Fuels", Fire Safety Jour-
nal, Vol. 16, No. 1, 1990.
/7-12/ I.A. Buist and E.M. Twardus: "Burning uncontained Oil Slicks: Large Scale
Tests and Modelling", S.L. Ross Environmental Research Limited, Ottawa,
Ontario.
/7-13/ K.D. Stolzenback et al.: "A Review and Evaluation of Basic Techniques for
predicting the Behaviour of Surface Oil Slicks", Ralph M. Parsons Laboratory,
Report No. 22, National Oceanic and Atmospheric Administration , U.S. Depart-
ment of Commerce, 1977.
/7-14/ Wighus, R., U.mvik, L.E. and Drangsholt, G.: "Thermal Loads from Crude Oil
Spill Fires on the Sea Surface"; SINTEF Report STF25 F95063, SINTEF NBL -
The Norwegian Fire Research Laboratory, 1996 (will be published together
with this report).
/7-15/ Wighus, R. , NBL and Guenette, Applied Chemistry: "Hazard Assessment

1995", SINTEF Report STF 25 F95064, SINTEF NBL - Norwegian Fire
Research Laboratory, 1996.
Scandpower
/7-16/ Fay, J.A.: "The Spread of Oil Slicks on a Calm Sea", Dep. of Mechanical Eng.,
Massachuttes Inst. of Technology, May 16, 1969.
/7-17/ Stensaas, J.P.: "Fire on the Sea Surface, State of the Art and the Need for
further Research", SINTEF Report STF25 A92035, December 1992.
/7-18/ M. Fingas and N. Laroche: "An Introduction to the In-situ Burning of Oil Spills",
Spill Technology News Letter, 15 August 1991.
/7-19/ Mudan, K. and Croce, P.: "Fire Hazard Calculations for large open hydrocar-
bon Fires", pp. 2.45 - 2.87, SFPE Handbook of Fire Protection Engineering,
National Fire Protection Association, 1988.
/7-20/ M.L. Bullen: "A combined overall and Surface Energy Balance for fully deve-
loped ventilated controlled Liquid Fuel Fires in Compartments", Fire Research,
1977/78.
/7-21/ A.M. Kanury: "Scaling Correlations of Flashover Experiments", National

Bureau of Standards, Report No. NBS-GCR-83-448, October 1983.
/7-22/ American Iron and Steel Institute: "Fire Safe Structural Steel, A Design Guide",
1979.
/7-23/ H.Z. You and G.M. Faeth: "Fire Mater.", 3, 1979.
/7-24/ P.H. Thomas, 9th Symposium (Int.) on Combustion, 844, 1963.
/7-25/ R.L. Alpert and E.J. Ward: "FMRC", J.I. No. 01836.20, Factory Mutual Corpo-
ration, 1982.
/7-26/ C.L. Seyler: "Fire Plumes and Ceiling Jets", Fire Safety Journal 11, 1986.
/7-27/ M.A. Delichatsios: "The Flow of Fire Gases under a beamed Ceiling", Combus-
tion and Flame 43, pp. 1 - 10, 1981.
/7-28/ J.P. Stensaas: "Physical Modelling of enclosed Pool Fires", SINTEF Report
No. STF25 A87006, The Norwegian Fire Research Laboratory, Trondheim,
1987.
/7-29/ B.J. McCaffrey, J.G. Quintiere and M.F. Harkleroad: "Estimating Room Tem-
peratures and the Likelihood of Flashover using Fire Data Correlations", Fire
Technology 17, pp. 98-120, 1981.
/7-30/ L.Y. Cooper: "Journal of Heat Transfer", 104, 1982.
/7-31/ J.R. Lawson and J.G. Quintiere: "Slide Rule Estimates of Fire Growth", Fire
Technology, Vol. 21, No. 4, 1985.
/7-32/ E.G. Bitchen and A.C. Parnell: "Smoke Control in Fire Safety Design", E. &
F.N. SPON Ltd., 1979.
/7-33/ Benjamin/Clarke Associates, Inc.: "Smoke and Life Safety", prepared for the
Society of Plastic Industry, August 1983.
Scandpower
Risk Management AS (J) SINTEF
/7-34/ J.G. Quintiere: "Measurements: An Assessment of Correlation between Labo-

ratory and Full-Scale Experiments", Fire and Materials, Vol. 6, Nos. 3 and 4,
1982.
/7-35/ G.A. Chamerlain and M.A. Persaud, Shell Research Thornton and R. Wighus
and G. Drangsholt, SINTEF - Norwegian Fire Research Laboratory: "Blast and
Fire Engineering for Topside Structures, Test Programme F3, Confined Jet
and Pool Fires - Final Report", SINTEF Report No. STF25 F95028, Norwegian
Fire Research Laboratory, 21 May 1997.
/7-36/ D.A. Lihou and J.K. Maund: "Thermal Radiation from Fireballs", I. Chem. E.
Symposium Series No. 71 , The Institute of Chemical Engineers, North
Western Branch, 1982.
/7-37/ Maurer et al. : 2nd Int. Symposium on Loss Prevention, EFCE, Heidelberg,
1977.
/7-38/ F.K. Crawley: "The Effects of the Ignition of a major Fuel Spillage", I. Chem. E.
Symposium Series No. 71 , Manchester, 1982.
Scandpower
8. Response to Fire
8.1 General
The fire exposure or fire load which will effect personnel, structures and equipment will
be found through the previous mentioned calculation methods, or it will be defined by
the internationally accepted standardised fire exposure models.
Where the fuel or burning material is mainly of cellulosic type, the Standard Fire Curve
according to ISO 834 is often adopted when no further analysis of the fire is carried
out. This exposure model is meant to cover most fires in buildings and residential
houses. When the fuel mainly consists of hydrocarbons (oil, gas), the so-called hydro-
carbon-curve (or the "HG-curve", Ref. /B.55/ and /B.56/) is adopted'1. This exposure
model is meant to cover fires in areas on offshore platforms and processing plants
where hydrocarbon fluids are stored and/or processed. These two standardised time/
temperature-curves are the internationally accepted exposure curves for classification
of the fire resistance time of partitions and elements of construction.
The different methods of design and assessment of fire resistance time for a structure
may be visualised by the matrix shown in Figure 8.1. The structure may be defined
either by a single element of construction, a simple frame or the construction as a
whole. Likewise the fire exposure may be defined with different degrees of accuracy.
The most simple fire models are defined by the standardised time/temperature-curves.
Only a simple judgement, based on the actual burning fuel/material is done in order to
choose either the ISO-curve or the HG-curve. Exposure model H2 indicates that a fire
analysis is made, and the actual fire exposure is transformed to an equivalent
exposure time - t. according to one of the standardised time/temperature-curves.
When a complete fire analysis is carried out, the result will be the characteristics of the
actual fire. The fire may be defined either by a time/temperature-curve (H 3), a time/
heat flux-curve or by both. The characteristics of the fire itself are also important; such
as the radiative and convective part of the heat flux, and the emissivity of the flames.
1) The characterisffcs of fires in hydrocarbon fuels may vary to a great extent, and different hydrocarbon fires are de-
scribed in Chapter 6 and Chapter 7. Fires in hydrocarbon fuels may vary over a much greater range than fires in
cel/ulosic materials. The standardised "HG-fire" simulates an enclosed pool-fire, and reaches a maximum gas tempe-
rature of 1, 100 'C
Scandpower
G)SINTEF Risk Management AS
MODEL FOR s, S2 S3
STRUCTURE COl1PLETE
ELEHENT SUBSTRUCTURE
STRUCTURE
1.. I •
AA
*
MODEL FOR
THERMAL I•I 1
EXPOSURE
H1 T- TEST OR
CALCULATION
fDETERl1-
NISTIC)
CALCULAT ION
EXCEP -
TIONALLY
TESTING
~DETER11I -
ISTIC)
V
T-
ttd
TES1' OR CALCULATION CALCULATION

CALCULATION EXCEP- (P ROBABILIS
'rlONALLY TIC)SllOULD
(PROBABI-
H2 LISTICJ
T ESTI NG
(PROBABI -
BE AVOIDED
Ti:,
LISTIC)
t.
CALCULAT!O N CALCULATIO N CALCULATION
(PROBABILIS
p>rtOBABI- (PROBABI - TICt I N
H3 ISTIC ) LIST!C) SPE I AL
CASES AIID
FOR
RESEARCH
t
Figure 8.1: Different Methods of Structural Design for Fire Safety. The Columns de-
scribe three different Element Models, from the simple Element Type
Models (S1) to the complete Structure (S3 ) . The Rows describe different
Models for thermal Exposure, from the standardised Time-Temperature
Curves (H1) to a real Fire Exposure Curve (H 3). The Figure indicates the
suitable Verification Methods for Combination of Structure- and Exposure
Model
Most fire testing for classification is a combination of the simple elements of construc-
tion (S1) and standardised exposure models {H 1). However, sophisticated calculations
with other exposure models and on more complex structures may be carried out on
computers.
In terms of fire classification, structures, partitions and equipment (with passive fire
protection) must meet certain requirements. The primary criteria of stability, integrity
and insulation must be satisfied throughout a prescribed time of heat exposure.
The criterion of stability is that the structure shall fulfil its loadbearing capacity through-
out the fire exposure period.
The criterion of integrity is that partitions shall prevent spread of flames and hot fumes
throughout the fire exposure period.
The criterion of insulation is that the unexposed side of partitions shall not reach sur-
face temperatures in excess of a certain level throughout the fire exposure period.
Scandpower
8.2 Exposure Criteria for Personnel

8.2.1 Introduction
The hazard to human beings exposed to a fire is often considered in terms of the
following effects:
Thermal effects
Obscuration of vision
Toxic effects.
Thermal effects may be achieved either by direct contact with fire gases21 or at some
distance away from the fire. In the first case human beings will be exposed to both
thermal radiation and convection, while in the latter case the heat load consists only of
thermal radiation. Direct flame exposure to human beings is not taken under discus-
sion as inevitably this would result in serious injury and/or ultimate death.
The smoke effluents from a fire may cause obscuration of vision leading to a serious
impairment of escape possibilities. Obscuration of vision will in most cases be hazar-
dous when considering enclosed fires often far away from the fire source.
However, obscuration of vision can also be serious in open fires with strong winds or
in cases were the exposed human beings are located on a level above the fire source
(e.g. in case of a fire on the sea surface under an offshore oil production platform).
Toxic effects are hazardous in cases of prolonged exposure of fire gases usually in or
very close to the fire, often due to impairment of escape possibilities as a result of
smoke obscuration.
8.2.2 Thermal Effects

Incident Radiation Heat Flux
The resulting injury due to thermal radiation from a fire onto the bare skin of human
individuals is mainly a strong function of the magnitude of the incident heat flux and
the exposure time. For steady incident heat fluxes, as from pool fires or jet fires, im-
pact criteria have been determined in terms of steady heat flux level and associated
duration of exposure.
Examples of such criteria are according to Croce et. al., Ref. /8-1/, that a steady heat
2
flux of 5 kW/m causes extreme pain after 15 seconds. Further, that a steady heat flux
of 10 kW causes blistering within 15 seconds and severe blistering within 30 seconds.
On the basis of such data, Mudan, Ref. /8-2/, presents the following correlation for
time required for pain, tp (sec), correlated with the radiation intensity, q" (kW/m 2):
33
=[35]'· Eq. (8.1)
tp <'.J"
2) Fire gases in general can be considered as either reacting fires gases which constitute the flame plume, or com-
bustion products mixed with entrained excess air (i.e. fire effluents)
Scandpower
Figure 8.2 indicates exposure times for the threshold of pain, threshold of blister, and
full blister.
.400
kW/m2
15
.3 50
.3 00
u
G)
·!!- .2 50
E
g
IU
t>
.200
I
><
::,
;
....
IU .150
Cl)
:c
.100
.050 - - ~- - - - -- - - - ~- - - 0
0 5 10 15 20 25 30
Time- s ec
Figure 8.2: Exposure Times for the Threshold of Pain, Threshold of Blister and
Blister (from Veghte, Ref. /8-3/)
Table 8.1 gives the API RP 521 , Ref. /8-4/, recommended design flare radiation levels
for personnel when assessing exposure at the plant heat-release conditions.
Scandpower
Table 8.1: Recommended Design Flare Radiation Levels of the API RP 521 Guide
for Pressure-relievin and de ressurin S stems, Ref. /8-4/
e~f~in""is~'~~ib""'1""e;:-:e,c,":'r:""'2.-r:;,-=
'".p""·
qesi~~flare ' ?2li':t"f0:2;:riF~1Cr:./f.,>••2cc•itf.'/·

radia\i~r:i level•·
<~\?Jtm>
15.77 Heat intensity on structures and in areas where operators are not
likely to be performing duties and where shelter from radiant heat
is available
9.46 Value of the permissible flare radiation level at design flare release
at any location to which people have access. Exposure must be
limited to a few seconds, sufficient for escape only
6.31 Heat intensity in area where emergency actions lasting up to 1
minute may be required by personnel without shielding, but with
appropriate clothing
4.73 Heat intensity in area where emergency actions lasting several
minutes may be required by personnel without shielding, but with
appropriate clothing
1.58 Value of the permissible flare radiation level at design flare release
at any location where personnel are continuously exposed
In Lees, Ref. /8-5/, there are described so-called "probit methods" for predicting the
degree of injury due to thermal radiation from fires. A probit relation is a function,
which relates the probit to a causive variable, which is an expression for the hazar-
dous condition. An example of a causive variable might be the incident radiation
intensity combined with exposure time. In this case the probit function has the
following form:
Y = 14.9 + 2.56 - In· [t. ·(q")4 13

] Eq. (8.2)
where t. is the exposure time (in sec.), and q" is the incident radiation intensity (in
kW/m 2).
The following expressions can be derived for the necessary incident radiation intensity
to human beings, q " (kW/m 2), from fires in general expressed in terms of the expo-
sure time t. (sec.) for causing 1 %, 50 % and 99 % probability of death as well as blis-
tering:
a) 1 % lethality: q" --172 t .
- 0·14 Eq. (8.3)
b) 50 % lethality: q" = 341 t. - 0·75 Eq. (8.4)
c) 99 % lethality: q" -- 674 t .

- 0·15 Eq. (8.5)
d) Blistering: q" = 56. 7 t. -0.7l Eq. (8.6)
The expressions above are given to be valid for transient fires as well, like for example
flash fires. However, in the case of fires which yield highly transient heat fluxes (e.g.
Scandpower
Gi) SINTEF Risk Management AS
fireballs and flash fires), absorbed energy is considered rather than a steady heat flux
criterion. Table 8.2 gives critical energy absorbed for transient heat fluxes after a skin
surface temperature of 55 °C is reached. According to Mudan, Ref. /8-2/, blistering
occurs when the skin surface temperature reaches about 55 °c.
Table 8.2: Critical Energy for Transient Heat Loads (Energy absorbed after the Skin
Surface Temperature reaches 55 °C) (from Ref. /8-1/)
Human response
q" (kJ/m2)
40-50 Pain or mild second-degree burns
80 - 100 Blistering or severe second-degree burns
150 - 170 Severe third degree burns and skin tissue will be permanently damaged
Direct Exposure to Fire Effluents
Table 8.3 indicates some physiological responses of human individuals for several
elevated temperature levels of the surrounding air.
Spieth et al. , Ref. /8-6/, have examined the thermal hazard of aircraft occupants. It
was indicated that a maximum temperature of about 200 °c can be tolerated by the
human respiratory tract. An air temperature of about 31 O °C is reported to be the tem-
perature at which individuals suffer third-degree burns with 20 sec. exposure. Escape
is, thus, assumed to be impossible. Figure 8.3 presents the time to incapacitation as a
function of the air temperature.
Table 8.3: Elevated Temperature Responses of Human Individuals (from Pryor,

Ref. /8-7/
Temperature Physiological response
(OC)
125 Difficult breathing
140 5 min. tolerance time
150 Mouth breathing difficult, temperature limit for escape
160 Rapid , unbearable pain with dry skin
180 Irreversible injury in 30 seconds
205 Respiratory system tolerance time less than four
minutes with wet skin
Scandpower
Air Temperature ·c
50 1qo 1 ;o 2q0
300 .:
250
,e:::s
.5 200
Ti • 5.33 X 10"
E , ' Rlll1 - [("F-32}5/9]3.66
g
s 150 \. / '---r
·1
B 100 ' \
.E
.s., 50
\_.
'"': •)r.,._,....
~ ~
• . ..
0'.
100 150 200 250 300 350
-;:
4 0
Air Temperature "F
Figure 8.3: Time to Incapacitation as a Function of Air Temperature for Aircraft

Occupants (from Spieth et al., Ref. /8-6/)
Heselden, Ref. /8-8/, indicates a tenability limit of 160 °C corresponding to an incident

2
heat flux of 2 kW/m for safe escape from fires in road tunnels. National Research
Council of Canada, Ref. /8-9/, reports that a temperature of 150 °c is the upper limit
for the air temperature in which human individuals are capable of breathing.
Kaplan et al., Ref. /8-10/, state that a second effect of heat stresses is that the toxicity
of carbon monoxide may be modified. Experiments on test animals under heat stress
conditions showed that the COHb (carboxyhemoglobin) concentration at the time of
death was much lower compared with animals not stressed with heat.
8.2.3 Obscuration of Vision

The absence of vision may delay or prevent escape from fires and cause people to be
exposed to the fire gases for an unacceptable long period of time. While the exposure
to high concentrations of toxic and hot gases usually will be significant only in the
vicinity of the fire, the effect of reduced visibility may also be significant far away from
the fire source. For example, in multi-compartment buildings, the smoke blocking
effect may be significant in rooms far away from the room of fire origin.
Moreover, the smoke blocking effect is reported to be the first condition becoming
critical of the three hazardous conditions of fires listed in the introduction (i.e. heat
stresses, obscuration of vision, toxic effects).
Measure of light Obscuration
Light obscuration is often expressed by the concept of "optical density" per unit path
length, D/L (m· 1), which is determined by monitoring the attenuation, D (-), of a beam
Scandpower
of light passing through the smoke over a distance L. The optical density per unit path
length (D/L) can be expressed by the following expression31:
D 1 lo
-=-log- Eq. (8.7)
L L IL
where lo is the light intensity of a standardised light source and IL is the light intensity
after passing through the smoke over the distance L.
Limits of Smoke Tolerance
Several authors (e.g. Jin Ref. /8-11/) have carried out experiments designed to test
the ability of human beings to function effectively in a smoky environment. Jin, Ref.
/8-11/, draws the conclusion that an optical density of 0.25 m-1 (corresponding to a
visibility of 4-5 m) is about the threshold of diminished performance, and that this is
the smoke level that one should have in mind when designing smoke ventilation sys-
tems.
Quintiere, Ref. /8-12/, has compared data from several authors with the corresponding
visibility through the smoke. In spite of some variation in the data depending on the
experimental approach, the following relation could be established between the optical
density D and the visibility O through the smoke:
V = 00-1 Eq. (8.8)
From this plot it was established that the visibility is reduced below the dimensions of
an average residential room (i.e. 1 - 4 m) at D/L levels between 0.25 and 1 m-1 .
A report from the Benjamin/Clarke Associates Inc., Ref. /8-13/, states that the visibility
of less than one arm length will be of no help at all when escaping from a fire environ-
ment. If one assumes that the length of one arm is about 1 m, it can be found from
Eq. (8.8) that a visibility of 1 m corresponds to an optical density of 1 m-1 •
According to the report of Benjamin/Clarke Associates Inc., Ref. /8-13/, smoke detec-
tors usually respond to an optical density per unit path length in the interval of 0.005 to
0.05 m-1 . The time available for safe escape is the time it takes for the smoke level to
increase from detection point to the hazard point either due to smoke blocking, toxicity
or heat stresses. For each of the three critical conditions (i.e. heat stresses, smoke
obscuration and toxicity) , Friedman, Ref. /8-14/, picked a level, which when reached in
a fire, represents a serious threat to life safety. Several types of materials (e.g. filter
paper, several types of PU-foam, nylon. polystyrene, PVC, and several types of bran-
ched polyethylene) were burned and all three effects were compared. The result of
this extensive investigation was that obscuration of vision was the first threat to reach
critical levels for all materials except from one (i.e. filter paper) .
The time to reach the level of total absence of vision (defined at D/L = 0.5 m-1) was
estimated to be in the time interval of 1½ - 6 min by means of full-scale experiments
with full-scale fires in passenger rail cars.
3) In the literature covering this field of smoke obscuration there exists several definitions of the parameter OIL. One
definition multiplies with a factor of ten on the right side of Eq. (8. 3). Another definition uses the In-function instead of
the log-function. In this latter case the light attenuation is termed as 'the smoke extinction factor' k (m-1) rather than
"optical density". All definitions vary from each other by a constant
Scandpower
Risk Management AS (t;»SINTEF
Handbook for Fire Calculations and Fire Risk Assessment in the Process Industry Page lt-9
Friedman, Ref. /8-14/, maintains that in exponentially growing fires, escape time is
rather independent of smoke properties of the burning material. Flame spread or
burning rate characteristics of the material are much more important.
8.2.4 Toxic Effects

General
Kaplan et al., Ref. /8-10/, define smoke as "a complex mixture of the airborne solid
and liquid particulates and gases evolved when a material undergoes pyrolysis or
combustion". Further, it is ascertained that over 100 chemical species have been
identified from the burning of some materials. Although fire gases may include a wide
variety of toxic gases, the concept of toxicity is usually grouped into narcotic gases
and irritant gases. Narcotic gases may cause narcosis and death, while irritants may
cause incapacitation mainly by effects on the eyes and the upper respiratory tract
which may impair escape capability and sometimes cause delayed death due to lung
damage (i.e. so-called "delayed toxicity").
The main narcotic gases known to appear in fires are CO, CO2, and HCN. The gases
causing irritation are mainly HCI, NO,, NH 3, SO2, HF, HBr and Acrolein. Table 8.4
gives a survey of the main sources of these fire gases.
In addition to the main toxic products listed in Table 8.4, the lack of oxygen may cause
serious incapacitation and, at worst, death. Reduced levels of oxygen will appear
more or less in all fire atmospheres in which the air supply is restricted.
Most of the combustible materials used in the process industry are made from both
natural and synthetic materials, which may generate the toxicants listed in Table 8.4.
In the subsequent sections there will be a review of the main toxic gases, their effects
on people involved in fires, and tenability limits of each toxicant including tenability
limits with respect to insufficient oxygen. However, first there will be a brief review of
different measures of toxicity used in assessing toxic effects of fire effluents.
Measures of Toxicity
Along with so-called clinical data (i.e. analysis of autopsies of fire victims), the res-
ponse of toxic gases has been quantified by measuring the response of animals (nor-
mally rodents) exposed over a fixed time to different concentrations of fire gases. This
latter method requires a combustion system, an animal exposure system, and a
chemical gas analysis system. A great number of different toxicity test methods exist
which intend to simulate the smoke characteristics of real fires. Hence, there may
exist deviations between tenability limits reported in the literature depending on the
experimental method applied.
Scandpower
..
T a bl e 84 . T ox1can
The S ources ofth emam . ts appearing .m Fire Gases
Toxicant Sources
CO (carbon monoxide)
All materials, liquids and gases containing carbon
CO2 (carbon dioxide)
HCN (hydrogen cyanide) From combustion of wool, silk, polyacrylonitrile, nylon,
polyurethane, PU-foam, etc.
NOx (nitrogen oxides) Produced in small quantities from fabrics and in larger
quantities from cellulosic nitrate, celluloid, etc.
NH 3 (ammonia) Produced in combustion of wool, silk, nylon and meamine,
concentrations generally low in ordinary building fires
HCI (hydrogen chloride From combustion of materials containing chlorine, e.g.
polyvinyl chloride (PVC), and some fire retardant (treated)
materials
SO2 (sulfur dioxide) From combustion of materials containing sulphur (rubber)
HF (hydrogen fluoride) From combustion of fluorinated resins or
Hbr (hydrogen bromide) films and some fire retardant materials containing bromine
Acrolein From pyrolysis of polyolefins and cellulosic at low tempe-
ratures {~ 400 °C) (fats and oils)
Behavioural and physiological response data and chemical gas analysis from a toxicity
test method are compared in order to assess the toxic effects at different concentra-
tions of fire gases. The number of animals showing a response in a toxicity test
method, either as lethality or as incapacitation, will increase as the exposure concen-
tration and time of exposure are increased. When the percent of animals showing an
effect at a given gas concentration and within a specified time is graphed as a function
of the logarithm of the gas concentration , a straight line may be approximated. From
this plot the concentration which will produce an effect on 50 % of the animals within a
specified time interval (preferably 30 min or less) can be obtained by interpolation. If
the effect is incapacitation, the concentration is often termed as IC50 . When lethality is
the observed response, the term LCso is used to denote the concentration causing
death in 50 % of the animals for a specified exposure time. LC50 is a more reliable
measure than ICso because death is the most clearly defined and easily measurable
response to fire effluents. IC50 and LC50 are usually expressed in ppm, but the units %
and mg/I are also frequently used.
The time of exposure is very important in inhalation toxicology. As the ex posure time
is increased , the necessary concentration of toxic gases in order to produce a certain
effect is decreased. The reverse is also true, i.e. as the exposure concentration is
increased, the time to produce the effect is accordingly decreased. This relationship 4J
can be expressed by Haber's law, which states that within certain time and concentra-
tion ranges, the product of concentration, C , and time to effect, t, is a constant which is
characteristic of the toxicant, or expressed mathematically:
C • t = constant Eq. (8.9)
4) An exception to the generalisation above is sensory irritation which is primarily related to concentration rather than to
dose
Scandpower
Risk M anagem ent AS Gl SINTEF
Thus, this product is a quantitative expression of the "exposure dose" to which a sub-
ject is exposed. Hence, if the toxicological effect is incapacitation, the exposure dose
is termed ICTso- Correspondingly, LCT50 may express the exposure dose, which is
causing death of 50 % of the animals. ICT50 and LCT50 are usually expressed in
ppm-min. For certain toxic gases (e.g. HCN) there is a tendency that progressively
lower doses (C-t values) can be tolerated as the concentration is increased.
As already pointed out the measures for assessing the toxicity of fire effluent men-
tioned so far, are determined by using rodents, which are exposed to fire effluents for
a specific period of time. An important question in this context would be whether the
rodent is a reasonable model for human exposure to fire atmospheres.
According to Hilado et al., Ref. /8-15/, "mouse-rat comparisons are a potential value
from the viewpoint of comparative toxicology, because they provide the possibility of
extrapolating laboratory animal data to humans by means of parameters such as
metabolic rate". It was established that for most narcotic gases the use of toxicologi-
cal data based on animal response yields reasonable values for human response. On
the other hand, concerning toxic gases where irritation is the predominant mechanism,
the correlation between human and animal response appears not to be so good.
Carbon Monoxide
Extensive investigations of human fire fatalities have shown that carbon monoxide
(CO) represents the major threat in most fire atmospheres even though this toxicant is
not the most toxic of fire gases. CO is usually one of the most abundant. The toxic
effect of CO is tissue hypoxia caused by the inability of the blood to carry sufficient
oxygen to critical body organs. This arises from the fact that the affinity of haemoglo-
bin for carbon monoxide is 300 times greater than that for oxygen. Thus, CO ties up
the haemoglobin to produce carboxyhaemoglobin (COHb).
Reinke et al., Ref. /8-16/, summaries the responses of different concentrations of car-
bon monoxide given in Table 8.5. From the table it appears that the carbon monoxide
becomes extremely dangerous when the concentration of this gas is in the range of
some thousands ppm (i.e. in the range 3,000 - 5,000 ppm). Such concentration levels
are quite commonly encountered in enclosed fire atmospheres with restricted ventila-
tion.
Table 8.5: Carbon Monoxide Responses (from Reinke et al., Ref. /8-16/\
.· . .
Concentration· of Response
' .
carbon monoxide
.I
. (ppm). I
100 Allowable exposure for several hours
400 • 500 No appreciable effect after 1 hour
600- 700 Just appreciable after 1 hour
1000-1200 Unpleasant after 1 hour
1500- 2000 Dangerous when inhaled for 1 hour
4000 Fatal when inhaled for less than 1 hour
10000 Fatal when inhaled for 1 min
Scandpower
According to Kaplan et al., Ref. /8-10/, the threshold of COHb content in blood for
escape impairment of human individuals appears to be in the range of 30 to 40 %.
From concentration-time curves for a man with a respiratory capacity of 20 I/min (light
physical activity) for different COHb-concentrations, it was concluded that a COHb-
concentration in the range of 30 - 40 % corresponds to an accumulated C-t dose of
CO of approximately 35.000 - 45.000 ppm-min . This order of magnitude is in accor-
dance with the ICTsa-value obtained from experiments with baboons.
Carbon Dioxide
Carbon dioxide is not particularly toxic at levels normally observed in fires, even
though moderate concentrations do stimulate the rate of breathing. This condition
may contribute to the overall hazard of a fire gas environment by causing accelerated
uptake of other toxicants.
According to Kaplan et al. , Ref. /8-10/, the rate and depth of breathing is increased
with 50 % by 2 % carbon dioxide and doubled by 3 % carbon dioxide in air. At 5 %,
breathing becomes difficult for some individuals, although this concentration of carbon
dioxide has been inhaled for up to 1 hour without serious after-effects. In Table 8.6
Bryan, Ref. /8-17/, indicates the human responses of higher CO2 exposures. It
appears that the CO2 contents in air have to exceed 1O % in order to threaten the
exposed person's ability to escape under his own power. Concentrations of this order
of magnitude are usually encountered in enclosed fire atmospheres.
Table 8.6: Carbon Dioxide Responses (from Bryan, Ref. /8-17/)

Concentration of Response
carbon dioxide
(%)
10 Approaches threshold of unconsciousness in 30 min
12 Threshold of unconsciousness reached in 5 min
15 Exposure limit 1 min
20 Unconsciousness occurs in less than 1 min
Hydrogen Cyanide
Hydrogen cyanide (HCN) is according to G.E. Hartzell, Ref. /8-18/, a very rapidly
acting toxicant which is more than 20 times more toxic than carbon monoxide. While
the main toxic effect of CO is that it disturbs the ability of the blood to carry oxygen to
the cells, the main toxic effect of HCN is that it disturbs the ability of the cells to utilise
oxygen. It does not combine appreciable with haemoglobin, but does bind with the
enzyme in body cells. The result is inhibition of the utilisation of oxygen by the cells.
The role of HCN in fire casualties is less clear than that of CO. It is known that some
HCN is produced when polymeric materials containing nitrogen are being burned (cf.
Table 8.4). From the data of Kimmerle, Ref. /8-19/, shown in Table 8.7 , one can con-
clude that 50 ppm may be tolerated by human individuals for 30 - 60 min without diffi-
culty, while 100 ppm for the same period is likely to be fatal. 135 ppm may be fatal
after 30 min and 181 ppm may be fatal after 10 min. However, according to Kaplan et
al., Ref. /8-10/, documented cases in which HCN alone can be considered to be the
primary toxicant are rare. However, in fire situations where for example upholstered
Scandpower
furniture consisting of large quantities of polyurethane, the toxicity of HCN may accor-
ding to H0yland, Ref. /8-20/, exceed the toxicity of toxic gases as for example CO.
Table8.7: Hvdroaen Cvanide Resoonses /from Kimmerle, Ref. /8-19/l

·.. ,, __ .
Confamtra,tibn of·. Resppns~ ..

hydrogen cyanide. ,·
..
I ,
,: . : ·- .
·. (ppm) .·· .. ·
'
' . '
0.2 • 5.1 Threshold of odour

18-36 Slight symptoms (headache) after several hours
45. 54 Tolerated for½ to 1 hour without difficulty
100 Death after 1 hour
110-135 Fatal after½ to 1 hour or dangerous to life
135 Fatal after 30 min
181 Fatal after 1O min
280 Immediately fatal
Levin et al., Ref. /8-21/, report a 30-min LC 50 value of 160 ppm of HCN with 95 % con-
fidence limits of 140 to 180 ppm. Most of the post-exposure deaths occurred within 24
hours. If the 24 hours post-exposure deaths were included in the calculation, the LCso
value was reduced to 11 O ppm with 95 % confidence limits of 97 to 127 ppm. Further,
Kaplan et al., Ref. /8-10/, report that the C- t products of rats at incapacitation are in
the range 1,200 to 2,700 ppm-min, while the C· t product at incapacitation of active
rats is about 800 ppm min.
Insufficient Oxygen
Oxygen is consumed from the atmosphere when combustibles are being burned
during fires. Kaplan et al., Ref. /8-10/, state that when oxygen drops from its usual
concentration of about 21 % in atmospheric air to about 17 %, motor co-ordination is
impaired. When the level of oxygen drops further into the range 14 - 1O %, a person
will still be conscious, but may exercise faulty judgement and will be quickly fatigued.
If the contents of oxygen drops to 1O- 6 %, the person loses consciousness and must,
according to Reinke et al., Ref. /8-16/, be revived with fresh air or oxygen within few
minutes to prevent death. If the exposed person is in strong physical activity, in-
creased oxygen demands may result in such symptoms of oxygen deficiency as de-
scribed above at significantly higher concentrations.
Synergetic Effects
Levin et al., Ref. /8-21/, draw the conclusions that CO and HCN act in an additive
manner and that synergetic effects have been found when the animals are exposed to
certain combinations of CO and CO,. 5 % CO, raised the threshold of deaths due to
hypoxia and decreased the LCso value of HCN. By decreasing the 02 concentration in
the presence of other major fire gases, the toxicity increased even further. In some
few cases minor combustion products have also been found to play a predominant
role. Feill Fant ikke referansekilden. shows both lethal concentrations of CO, 02
and HCN, and with various combinations of these gases with and without 5 % CO,.
From the figure one may see that the non-lethal area becomes much smaller
compared to when CO, o, and HCN act separately.
Scandpower
«JI SINTEF Risk Management AS
Nomegian Fire Research LaboralOJY
Figure 8.4: Lethal Concentrations of CO, 0 2 and HCN and various Concentrations of
the Gases with and without 5 % CO2 (from Levin et al., Ref. /8-21/
Prediction Methods for Estimation of the Toxicity of Narcotic Gases
Purser, Ref. /8-22/, gives the following expressions for the fractional incapacitation
doses for CO, HCN, CO 2 and insufficient oxygen:
k (CO1.03s ) ~t
Fco Eq. (8.10)
0
where:
Fco= fraction of incapacitating dose

~t = exposure time (min.) (= 1 in this case)
k = 0.00082925 for 25 I/min respiration rate
co = concentration of CO (ppm)
D = COHb (carboxyhemoglobin) concentration at incapacitation (D = 30 %
for light activity)
1
FHCN = e<S.39s-o.023. HCN> - 1 Eq. (8.11)
where:
FHcN = fraction of incapacitation dose of HCN

HCN = concentration of HCN (ppm)
1
F02 = ----
e1.9s-o.s2s. - ->
c20.9-o Eq. (8.12)
2
Scandpower
Risk Management AS (G}) SINTEF
where:
Fo2 = fraction of incapacitation dose of insufficient 0 2

02 = concentration of 02 (%)
1
Fco2 = s.1s23-o.s1sexco 2 Eq. (8.13)
8
where:
Fc02 = fraction of incapacitation dose of CO2

CO2 = concentration of CO2(%)
The increased respiration rate due to CO 2 exposure can be taken into account by
means of the following multiplication factor, Vco2:
e0.2496XCO2+1.9086
Vea,= _ Eq. (8.14)
68
Purser, Ref. /8.22/, has developed the following expression for the "total fractional
incapacitation dose" F1N:
Eq. (8.15)
As a basis for this expression the following is assumed:
1. CO and HCN are considered to be directly additative

2. CO2 increases the rate of uptake of CO and HCN in proportion to its effect on
respiration rate
3. The narcotic effect of low oxygen hypoxia is considered to be directly additative
to the combined effects of CO and HCN, but not increased by CO 2 induced
hyperventilation
4. The narcotic effect of CO2 is considered to act independently of the effect of
other gases.
Eq. (8.15) has to be calculated for each minute after the start of the fire and the values
are summed. When the running total sum exceeds unity, this indicates the onset of
incapacitation.
Irritants
According to Birky, Ref. /8-23/, it is not always the light attenuation due to smoke
obscuration which is the major problem for people escaping from the fire and for fire
fighters. The impairment of vision caused by irritants represents also a serious threat
in escape or fire fighting situations. The initial painful effects of irritants (sensory irri-
tation) are mainly on the eyes and upper respiratory tract. These pains do not get
worse with prolonged exposure. The pains may even lessen as the exposure time.
proceeds. The toxic effects on the lungs increase with prolonged exposure. They are
often most serious some hours after the exposure and they may cause death.
The main irritants of fire effluents are NOx, NH3, Hcl, SO2, HF, Hbr, and Acrolein. Very
little is known about the irritating effects from different concentrations of these toxi-
cants. Irritant gases may also frequently result in both immediate death and post-
Scandpower
Gl SINTEF Risk Management AS
exposure deaths due to pulmonary complications. For example, NOx is a strong pul-
monary irritant capable of causing immediate death as well as delayed injury. Hilado
et al., Ref. /8-15/, indicate predicted human LC 50 values51> (5-min and 30-min expo-
sure) for some irritants in Table 8.8.
Table 8.8: Predicted Lethal Concentrations for Human Individuals and published
Values (from Hilado et al., Ref. /8-151)
Toxicant LCso (ppm) LCso (ppm)
5min 30min
NH3 55000
SO2 17000 8000
HCI 100000 7700
HF 44000 4600
NOx 410 180
As it evidently appears from Table 8.8, NOx is definitely most toxic of the irritant gases
presented. In fact, Hilado et al., Ref. /8-15/, conclude that NOx along with HCN are
the most toxic gases in the smoke from fires. However, the LC50 value of Acrolein is
reported to lie in the range 50-136 ppm, Ref. /8-15/.
Tenability Limits of Fire Effluents - Concluding Remarks
A conclusion of this review of physiological effects and the tenability criteria of human
individuals exposed to fire effluents is that the toxic effects of CO by far exceed those
of other toxic gases present. CO is, as can be seen from Table 8.9, not the most toxic
of the gases appearing in fire effluents, but it exists usually in high concentrations
compared to other toxicants. It appears that HCN, NOx, and Acrolein are far more
toxic than CO. In most real full-scale fires a good first approximation may be that the
tenability limits can be predicted solely on the basis of the CO-concentration. How-
ever, one must be aware of the fact that other toxicants also may be important (e.g. in
enclosed fires involving upholstered furniture, the toxicity of HCN may exceed the toxi-
city of even CO).
The observed toxicity of fire effluents can also be explained by a interaction between
the primary toxic gases and, to a certain extent, to minor combustion products.
Especially, CO2 and reduced levels of 0 2 in combination with CO may have synergetic
effects, while CO and HCN act more in an additative manner. Toxicological data re-
ceived from tests using animals (usually rodents) seem to yield reasonable tenability
criteria for human individuals. In Table 8.9 there is a summary of the different ten-
ability criteria of the main toxicants of fire effluent.
5) Predicted from toxicological tests with mice and rats by means of metabolic rate
Scandpower
Handbook for Fire Calculations and Fire Risk Assessment in the Process Industry PageS-17
Table 8.9: Tenability Criteria of Fire Effluents in Terms of 30-min. IC50 , LC50 , ICT50
and LCT50 Values
,.i:•·, ' :
'JiSxil:ant ' 1c,l.;·· •· .. 1:v··· LC:Jf'' ..-.;:-;' ···t.cT···. .•
·...·. • . . :·• ' • ,!;O
. (ppm)
..
(ppt)l) . > < ·• .ICTs'o ··•··
' •·· (pprh,min) · .· (ppm'min) •
•·
co 1400-1700 4500 35000 - 45000 70000 - 135000
CO2 100000 146000
HCN 100- 200 160 750-2500 1500- 7500
02 75000
NHs 55000
(5 min. exposure)
NOx 180
HCI 7700
SO2 8000
HF 4600
Acrolein 50-135
8.3 Fire Response of Steel Structures

8.3.1 Introduction
When a steel structure is exposed to fire, either through radiation or convection, or
both radiation and convection, the steel temperature will increase. The mechanical
properties of steel are highly dependent on the steel temperature, and it is often
necessary to protect the steel from reaching a critical temperature. The critical steel
temperature Ts.er of a loaded steel member is the temperature at which the loaded
member cannot carry the design effects of actions (load) for the fire design situation.
The critical steel temperature depends on the utilisation level I-lo (see Eq. (8.16)):
E%
_ fi d
µ o- 0
R
fi,d,t=O
Eq. (8.16)
where
Efi,d = the design effects of actions for the fire design situation
Rfi,d,t= = the. corresponding design resistance of the steel member for the fire
situation at time t=0 (i.e. normal temperature)
and to a certain degree on the steel quality.
The critical steel temperature will normally be within the temperature range 400 °c -
650 °C.
Note:
Fire design of loaded steel structures from the utilisation level /lo is relevant for certain
cross-sections (class 1, 2 and 3). For class 4 - sections special design rules comply.
See Ref. 18-28/ and 18-241.
Scandpower
«) SINTEF Risk Management AS
NotWegian Fire Research Laboratory
Further, in some situations when the structure is subjected to certain stress- and res-
traint conditions the critical steel temperature may be considerably lower than this
range, Ref. 18-251. In other situations the redistribution of loads may lead to higher
acceptable steel temperatures.
Hence, the critical steel temperature should be determined carefully.
Structural steel may be protected by a heat shield (radiative panels or deluge sys-
tems), by passive fire protection materials (boards or spray-applied materials) or by
internal heat sinks, such as water filling of hollow steel sections.
8.3.2 Properties of Steel at Elevated Temperatures

General
There are numerous different steel qualities both regarding mechanical, physical and
thermal properties. Mechanical and thermal properties may be highly dependent on
the steel temperature, and care should be taken when choosing material data for use
in temperature analyses.
The test method from which the material data are derived may be of great importance.
The method should correspond to the actual load situation, which is to be analysed.
E.g. the temperature dependent mechanical properties for steel are highly dependent
on the stress-strain conditions and heating rate, Ref. /8-26/, and one should seek to
find data derived from a test set-up similar to what the structure to be analysed will ex-
perience during the fire exposure.
Mechanical Properties
Data on temperature dependent mechanical properties can be taken from Ref. /8-27/
and /8-28/.
Density:
Density is considered to be constant in the relevant temperature range. The unit

mass of steel is given by Eq. (8.17).
Ps = 7850 kg/ m3 Eq. (8.17)
Strength and Deformation:
The strength and deformation properties of steel at elevated temperatures are, for
heating rates between 2 and 50 °C/min, characterised by a set of stress-strain rela-
tionships with a linear-elliptical shape. A stress-strain curve can be written for each
level of steel temperature. A mathematical model and tabulated data for this relation-
ship are given in Ref. /8-28/.
Reduction factors for stress-strain relationship of steel at elevated temperatures

{effective yield strength and modulus of elasticity) are given in ECCS and Euro-
code 3-ENV 1993-1-2 (Ref. /8-27/ and /8-28/) for all grades of steel (Fe 310-Fe 510).
Normally the effective yield stress at elevated temperatures has been defined for a
maximum accepted yield strain, e.g. elastic strain 0.2 %.
Scandpower
Within the European harmonisation of codes it has been suggested to define the
effective yield strain as high as 2.0 % for steel at elevated temperatures. This gives
totally different effective yield strength as a function of temperature.
According to Eurocode 3-ENV 1993-1-2 (Ref. /8.28) the following reduction factors
can be used for the stress-strain relationship of steel at elevated temperatures when
the yield strength is defined at a strain ey. T = 2.0 %:
Table 8.1 0: Reduction Factors for the Stress-Strain Relationship of Steel at elevated
Tern Jeratures /Yield Strain Ev T = 2.0 %), Ref. /8-27/ and /8-28/
SteeLtemperature · Reduction fac:tors. at temperatareT, relati.ve to the value of.a, ore; at 20 •c
Reductio1Hactor for ReductionJactor for, . ·· ···.. R~uctionJactortql ·
effective yiel~ strength proportional limit slope of linear elastic range
··Ts
. (relativetof,l ·(rel~\iveto. f,) (relativ~ to E,)
. . k y.T =.a.,.ta, k p.T.=.apo/ay 1<!ET~Esr/Es
20°c 1.000 1.000 1.000
100 •c 1.000 1.000 1.000
200°c 1.000 0.807 0.900
300°c 1.000 0.613 0.800
400°c 1.000 0.420 0.700
500°c 0.780 0.360 0.600
600°c 0.470 0.180 0.310
700°c 0.230 0.075 0.130
800°c 0.110 0.050 0.090
000°c 0.060 0.0375 0.0675
1,000 •c 0.040 0.0250 0.0450
1,100°c 0.020 0.0125 0.0225
1,200°c 0.000 0.0000 0.0000
Annex A of Eurocode 3-ENV 1993-1-2 gives the stress-strain relationship at elevated

temperatures for different steel grades and strain levels. (The reduction factor for the
cry, T/cry ratio at a strain level of e = 2.0 % is the same for all relevant steel grades and
temperatures levels).
The thermal elongation valid for all grades (Fe 310- Fe 510) of steel may be defined
by Eq. (8.18) - (8.20), Ref. /8-28/:
T. s 750 °c:
Al= -2.416-10-4 +1.2-10.s • T, +0.4·10·8 • T; Eq. (8.18)

I
750 °Cs Ts s 860 °C:
Al= 11-10.., Eq. (8.19)

I
860 °cs T. s 1,200 •c:
Al _, .s
-=-6.2-10 +2-10 -T, Eq. (8.20)
I
Scandpower
(t,SINTEF Risk Management AS
Norwegian Fire Re!earch LaboratoJY
The coefficient of elongation may in simplified calculations be considered as indepen-

dent of the steel temperature, and Eq. (8.21) can be used:
Lil =1.4-10-5 • (Ts -20 °c) Eq. (8.21)

I
where:
I = length at room temperature (m)

Lil = temperature induced expansion (m)
Ts = steel temperature (°C)
Thermal Properties
Specific Heat:
The specific heat of steel is a function of temperature, and may, for grades Fe 310 -
Fe 510, be approximated by the following equations (J/kg-0 C), Ref. /8-28/:
20 °C ~Ts~ 600 °C:
Eq. (8.22)
600 °C ~Ts~ 735 °C:
=666- 13002 Eq. (8.23)

Cps 738-Ts
735 °C ~Ts ~ 900 °C:
= 545 + 17820
Cps Ts - 731
Eq. (8.24)
900 °C ~Ts ~ 1,200 °C:
Cps= 650 Eq. (8.25)
where:
Cps = specific heat of steel (J/kg- 0 C)

The specific heat may in simplified calculations be considered as independent of the

steel temperature, and then the following value should be taken:
Cps = 520 J/kg°C Eq. (8.26)
Thermal Conductivity:
The thermal conductivity of steel is a function of temperature, and may, for grades
Fe 310 - Fe 510, be approximated by the following equations (W/m-°C), Ref. /8-28/:
Scandpower
Risk M anagement AS G) S I NTEF
Handbook for Fire Calculations and Fire Risk Assessment in the Process Industry Paga 8-21
T, !> 800 °C:
:\.s = 54-3.33 -10-a · Ts Eq. (8.27)
800 •c !, T, !, 1,200 •c:
:>..,=27.3 Eq. (8.28)
where:
:>... = thermal conductivity of steel 0JV!m-•C)

Eq. (8.16) - (8.28) are plotted as curves in Appendix B.
8.3.3 Methods to calculate Temperature Response of Fire exposed

Steel Structures
General
There are a number of different methods to predict the temperature response of fire
exposed steel structures. The simplest methods are directly based on results from fire
tests, while the most sophisticated computer models calculate the temperature in-
crease of a fire exposed cross section based on a given exposure curve and tem-
perature dependent thermal properties of the materials. The verification methods will
often use a combination of results from fire tests and calculation procedures.
Uninsulated Steel
When unprotected steel is exposed to fire, the temperature will in most cases increase
to a critical level within minutes. However, if the steel section is heavy and the tem-
perature level is moderate, the structure may have sufficient fire resistance without
further fire protection.
The temperature of an unprotected steel section may be predicted through an iterative

process using Eq. (8.29), Ref. /8-28/:
1 A ·
II.Ts = - - · _s · hnet,d · 11.t Eq. (8.29)
CpsPs Vs
where:
II.Ts = temperature increase in steel section (°C)

h net,d = the design value of the net heat flux per unit area 0JV!m 20C)
Cps = specific heat of steel (J/kg°C)
Ps = unit mass of steel (kg/m 3)
AJV. = section factor of steel section (m.1), where
As = area of the inner surface of the insulation material per unit length of
the member (m2/m)
v. = volume of the member per unit length (m2)
11.t = time interval (sec)
Scandpower
This equation is based on the assumption of quasi stationary, one dimensional heat
transfer with the steel considered as a heat sink, in which the heat supply is instan-
taneously distributed to give an uniform temperature due to the great thermal conduc-
tivity of steel.
When the steel section is large, there may be considerable temperature gradients in
the cross sections. In such cases there should be carried out calculations using a
suitable computer code in order to predict the temperature response more correctly.
In Norwegian Standard NS 3472 (Ref. /8-24/) simplified equations for steel tempera-
ture response in unprotected steel sections subjected to a standard cellulosic time/
temperature exposure according to ISO 834 are given as follows:
T,,,, = 310[( :J' - l0,83 for R15 Eq. (8.30)
T,,, = 153[( :J'+ l 2,38 for R30 Eq. (8.31)
Insulated Steel
There are few hand methods for the calculation of temperature response of steel with
passive fire protection, and the available methods are not applicable for all types of
fire protection materials. These hand methods are simple iterative equations where
the material properties of both steel and insulation materials are given as constants.
However, the thermal properties of both steel and insulation materials may vary con-
siderably with temperature, and the results from such calculations should be carefully
checked with test results.
Direct use of Test Results:
When the fire exposure is described by one of the standardised time/temperature-

cuNes (ISO-cuNe or HC-cuNe), the simplest, but perhaps most used, manual
methods are the direct use of data from fire tests. Most materials have been subjec-
ted to a large number of fire tests when used as passive fire protection on steel struc-
tures, and interpolation between different test results may in many situations give the
probable temperature response.
Indirect use of Test Results:
Due to the vast number of test results that normally are available for each fire pro-
tection material, the results relevant to one specific material are often organised in
tables or graphs to simplify interpolation.
Results from fire tests of insulated structural steel may be organised in many ways,
since there are four test variables:
Scandpower
Risk Management AS {G)) SIN TEF
Fire resistance time FR (min)

Section factor61 AN. (m"')
Insulation thickness i5; (m)
Critical steel temperature (Tc,) (OC)
Critical steel temperature is defined as the steel temperature at which the steel mem-
ber cannot carry the design load in the fire situation.
Tables and graphs may be constructed in different ways:
Hand drawn lines based on test results

Linear regression analysis of test results
Backward calculation of test results (NT Fire 021 ).
Linear regression is a commonly used method of curve-fitting to organise a large

number of test results. Such a method is described in Ref. /8-29/, and the analysis
can be carried out in a spreadsheet computer programme, i.e. Excel.
This method is strictly based on test results, and it is assumed that there is a mathe-
matical relation between the different test results.
Details of each specimen in the fire test programme (and simulation series) are used
as input data in the mathematical model which is used to relate the fire resistance, F/A
value and protection thickness.
The multiple linear numerical regression analysis is conducted using the following
equation:
Eq. (8.32)
Hence:
t Eq. (8.33)
6) The section factor may be expressed in different ways. H,IA, ANs and FIA are the most common expressions. In this
paper the section factor is expressed by the equation ANs where
A; = area of the inner surface of the insulation material per unit length of the member [rrrlm]
Vs = volume of the member per unit length [m3lm]
7) For uniform members ANs equals to FIA where F =the internal perimeter of the insulation [m], and A =area of the
steel cross section (rrr).
The section factor for different steel sections are given in Appendix C
Scandpower
Norwegian Fire Research Laboratocy
where
FR = time in minutes to a defined steel temperature level of the tested

section (30, 60, 90 minutes) - (fire resistance)
t = thickness of fire protection material (mm)
Ts = critical steel temperature (°C)
F/A = section factor (m-1)
Ao-A7 = constants (applicable to the specific analysis only)
Based on (Eq.8. **) a linear regression analysis is carried out on the parameters FR, t,
t/(F/A), Ts, txTs, txTJ(F/A), TJ(F/A) and 1/(F/A) calculated from the test and simulation
data. The constants A0 -A7 are determined using the regression analysis function of
Microsoft Excel. The multiple correlation factor is calculated. For acceptance of the
suitability of the input data, the square of this value (coefficient of determination or R
squared") should be in excess of 0.9.
The output from the analysis is shown in Figure 8.5. The constants A 0-A7 and R-
squared are marked.
R-squared
SUMMARY OF OUTPUT
Regresionsstatislikk
Mullippel R O 99851993
R-kvadrat I o'.99704205I
Justert R-kvadrat 65535
Standardfeil o
Observasjoner 8
Variansanalyse
fq SK GK F Signifkans-F
Regresjon 7 2425,67867 346,525524 O #NUM!
Residualer 0 7, 19633012 65535
Totalt 7 2432,875
Koefflsienter Standarr:Jfeif t Stat P•verr:Ji Nederste 95% 0 verste 95% Nedre 95 0% 0 verste 95 0%
Skjm<ingspunkt 36,976635 0 65535 #NUM! 36,97663502 36,97663502 36,97663502 36,97663502
X•variabel 1 5,67215582 0 65535 #NUM! 5,672155817 5,672155817 5,672155817 5,672155817
X•variabel 2 8,4113E-14 0 65535 #NUM! 8,4113E-14 8,4113E-14 8,4113E•1 4 8,4113E-14
X•variabel 3 -0 11876688 0 65535 #NUM! -0,118766878 -0,118766878 -0,118766878 -0, 118766878
X•variabel 4 -0 00565896 0 65535 #NUM! -0,005658963 -0,005658963 -0,005658963 -0,005658963
X-variabel 5 0 2489466A 0 65535 #NUM! 0,248946677 0,248946677 0,248946677 0,248946677
18,8906901
~
X-variabel 6 0 65535 #NUM! 18,89069013 18,8906901 3 18,89069013 18,89069013
X-variabel 7 -5150,11765 0 65535 #NUM! -5150,117647 -5150,117647 -5150,117647 -5 150,117647
~
I Constants Ao - A7 I
Figure 8.5: Output from Regression Analysis in Excel
Based on Eq. (8.33) and the analysis output (constants A 0-A7) the thickness of the
specified passive fire protection material necessary to protect steel sections from fire
can be calculated for combinations of section factor (F/A), critical steel temperature
(Ts) and fire endurance time (FR). Eq. (8.33) is easily programmed in Excel, and the
thickness can be calculated in a table.
Scandpower
With many fire protection materials it is possible to achieve a coefficient of determi-

nationn in excess of 0.95, which indicates a high level of agreement between test data
and predicted data.
If a sufficient level of agreement is achieved, Eq. (8.31) is used to calculate necessary

thicknesses for combinations of FR and A/V•.
The results may be presented either in a tabular manner as shown in Table 8.11, or by
graphs as shown in Figure 8.6.
Table 8.11: Predicted Insulation Thicknesses for Material X based on a linear Re-
aression Analvsis of Data from Fire Tests given in a tabular Manner
AJV.··•·. •··.. 1ri~ulatio11th!Ckl)l:l~s (mm) a,tgiv~1 FR-tir:ne (min) • ·. <
tm·1) Critical- .- steeltemperafure:
. . . . .,. 450 °c Jh' steel ',se.ction
Mean . . ', ,, ,, .,, ,,
.
.· . 30min, 60.min 90rnin 1.2Q.min 180min 240 mi.n
30 3 5 7 9 13 17
40 4 6 9 11 16 21
50 5 8 10 13 19 25
60 5 8 12 15 21 28
70 6 9 13 16 23 30
80 6 10 14 18 25 33
90 7 11 15 19 27 35
100 7 11 15 20 28 37
NECESSAAY INSULATION THICKNESS
"
.
!•
•~
~
~
%
"'
L----
L---
L.----
L--
--- L---
e
"
,
<
~
---- !.----
c--
" ---------
0
" " 70
SECTION l"ACTCJ:l Al/Ye C-1]

""
0 30 "''"
-+ 60 111rri 0 90 11U1
.I> 120 ..,1,. ,c 110 "''" V 240 /Oln
Figure 8.6: Predicted Insulation Thickness based on linear Regression Analysis of

Data from Fire Tests given as Graphs. Alternative Presentation of Data
from Table 8.10
8) A factor that evaluates the level of agreement between test data and predicted data
Scandpower
O)SINTEF Risk Management AS
Note:
Normally more than five test specimens are necessary to define the validity range of
the calculation. The calculations are only valid for interpolation, and not valid for fire
resistance times, section factors and thicknesses outside the tested range.
It should also be emphasised that the calculated data may only be valid for sections of
the same shape as tested. This has to be considered especially for intumescents,
since this type of fire protection material behaves differently on different shapes of
steel sections. For intumescents it is often necessary to carry out two linear regres-
sion analyses: one for I-shaped sections and one for hollow sections.
It has been more common to have different design tables valid for different heat flux
levels (ranging from e.g. 50 kW/m 2 - 300 kW/m 2 ) . It is in these cases, very essential
that the design tables are supported by relevant test data at each heat flux level, and
not by calculations alone. Specially for intumescents this is important, since the be-
haviour in fire of these materials is dependent on the heat flux. The results from a
200 kW/m2 fire test cannot be used in predicting the behaviour of the intumescent at
other heat fluxes!
The tables and graphs shown in Table 8.11 and Figure 8.6 are either used directly to
find the right thickness for a combination of FR and section factor, or to find the end
temperature for a combination of FR, section factor and thickness. If tables are avail-
able for a range of different critical temperatures T er, it is possible to find more points
on the temperature curve.
Simple Calculation Methods:
Simple manual methods for the calculation of temperatures in an exposed insulated

steel structure are limited to a one-dimensional approach. The equations for this prob-
lem vary very little in different references.
For a uniform temperature distribution in a steel cross-section the temperature in-

crease ,1.Ts of an insulated steel member during a time interval ,1.t may be obtained
from Eq. (8.34), Ref. /8-28/:
Ll
A Ts= - oi · -Ai · [ -1-] . <Tg-Ts >·,1t - <e1111 10 - 1>· LlA Tg
'J...J- Eq. (8.34)
CpsPs V s 1+0/ 3
with :
Ci'J=
c pl· · lll
s:· •p.I •AI
Eq. (8.35)
Cps · Ps · Vs
where
A.j = thermal conductivity for insulation (W/m°C)

Oi = thickness of insulation (m)
Cpi = specific heat of insulation (J/kg0 C)
Ps = density of steel (kg/m 3)
Cps = specific heat of steel (J/kg 0 C)
Pi = density of insulation (kg/m3)
Scandpow er
Risk M anagement AS ~ SINT EF
A;Ns = section factor for steel member (m- 1)

T9 = ambient gas temperature at time t (°C)
Ts = steel temperature at time t (°C)
AT9 = increase of the ambient temperature during the time interval At (°C)
At = time interval (sec)
Eq. (8.34) • (8.35) are valid for steel members insulated with dry materials. For moist
materials the calculation of the steel temperature increase AT. is based on the same
equations with the following modifications:
a) Before reaching a steel temperature of 100 °C, Eq. (8.34) is used

b) At a steel temperature of 100 °C, a delay in the steel temperature rise, Atv is
accounted for. Different methods for predicting the time delay are presented in
Ref. /8-28/. Where other exposure models than the Standard Fire is used, Atv•
values from fire tests should be used.
Note:
When using manual methods for the calculation of temperature in fire exposed insula-
ted steel structures, care should be taken when choosing material data for the insula-
tion material. First, the thennal conductivity varies with temperature. If the thennal
conductivity is described with one single value, a relevant average value should be
used. Second, all materials have an upper validity temperature, and calculations
should not be drawn outside this level. Third, it is a good custom to validate calcu-
lations with test results.
In Norwegian Standard NS 3472 (Ref. /8-24/) simplified equations for steel tempera-
ture response in protected steel sections subjected to a standard cellulosic time/tem-
perature exposure according to ISO 834 are given as follows:
Eq. (8.36)
T._30
A "-·
=218[( v: · d:
)o. -2,79
]
forR30 Eq. (8.37)
02
T.,160 =318[( A
v: · "-·)
d: ,
-2,54] forR60 Eq. (8.38)
02
T~ 90 =324[( A "')
v:·d: , -2,15] forR90 Eq. (8.39)
Backward Calculation of Test Data:
The test method which is used for the verification of passive fire protection materials
for steel beams and columns in the Nordic countries is the "Nordtest Method NT Fire
021 ", Ref. /8-30/. This method specifies a test method and a method for the calcula-
tion of the effective thermal conductivity or thermal resistance for an insulation system
Scandpower
(t,StNTEF Risk Management AS
for steel structures. The thermal conductivity, calculated from the test results, is
strictly related to the simplified equations used, and to the method by which the insu-
lation is attached to the steel members, and should therefore not be considered as a
true material parameter.
The test and calculation methods described are intended for insulation systems appli-
cable to steel beams and columns. The insulation may be intumescent paint, mineral
wool slabs, sprayed mineral wool, vermiculite or calcium silicate slabs, fibre slabs, re-
inforced gypsum slabs etc. The calculation method is not immediately applicable to
very dense materials, which protect by heat absorption rather than by insulation, or
combustible materials or deteriorating materials - materials that are "consumed" at a
certain rate when subjected to fire exposure.
The fire resistance test is performed according to ISO 834 with an exposure equiva-
lent to the standard fire or the hydrocarbon fire. The specimens are selected so that
they cover the range of section factors used in practise, and the insulation is tested
with maximum and minimum thickness, respectively. The specimens are equipped
with thermocouples, and both the furnace temperatures and steel temperatures are
recorded during the test period.
The temperature dependent heat flow through the insulation, here denoted as the
"effective" thermal conductivity, is determined through a "backward calculation" of the
measured steel temperatures. Eq. (8.34) is solved for thermal conductivity Ai, and the
iterative calculation method is carried out with a computer programme.
By using this thermal conductivity as an input to Eq. (8.34) the steel temperature can
be calculated for any combination of section factor and insulation thickness within the
tested range. However, since the computer programmes are not always available, the
most common use of this temperature calculation is to use the calculated tempera-
tures in the design of curves similar to Figure 8.6. The design process is normally
reduced to a graphical method using the design curves.
Computer Codes:
There are several non-linear computer codes for temperature analysis of elements of
construction. Most computer codes can solve 1- and 2-dimensional problems. A few
codes are capable of solving 3-dimensional problems.
SUPER TEMPCALC, Ref. /8-31/, and TASEF-2, Ref. /8-32/, are examples of two com-
puter codes for personal computers.
Both codes are based on the finite element method (FEM), and cover 1- and 2-dimen-
sional heat flow. The computer codes can analyse a cross section of a structure ex-
posed to an arbitrary time/temperature-curve. The cross section can consist of diffe-
rent materials, and the thermal properties (A, cp) can be given as non-linear tempera-
ture-value curves. The heat exposure is normally modelled as a time/temperature-
curve with a given convection- and emissivity-value at the boundaries.
The geometric structure is modelled as a plane cross-section with an element mesh

configuration, and the governing transient heat conduction equation is solved at the
boundaries of each element. The equations to be solved are given in Ref. /8-31/ and
Ref. /8-32/.
Scandpower
Risk Management AS ({;)) SINTEF
Note:
These two FEM-codes cover only the energy balance in each finite element and the
element model as a whole; the mass balance is not included. The thermal conducti-
vity (,t-values) used in the models to describe heat transfer in solids are "effective ;,.
values", these cover heat transfer by convection, conduction and radiation inside the
material.
Figure 8.7 shows an element mesh for an insulated I-section steel column. The
column is subjected to uniform heat exposure, and due to the two symmetry lines, only
a ¼ of the cross section has to be analysed.
Finite element d1v1s1on

111.n, ,,1n
/ ( Q, Ill
1110 +-+-+-+--+--+-+-+--+
I 1,1 I.I Q,I 1 1,J t,1 ••• :,

Tl- I " J
Figure 8.7: Element Mesh for an insulated I-Section steel Column. Results from a
Temperature Analysis with either of the two FEM Computer Codes may
be presented as Time/Temperature Curves for the individual nodal
Points, or as isothermal Charts at given times
Note:
Results from a calculation should not be accepted unless the validity of the input-
parameters are documented, and the calculations are compared with test results.
As for manual methods, care should be taken when choosing the thermal parameters
,i and Cp for the materials. When calculating outside the validity range for the material
(eg. gypsum, mineral wool), this should be taken into account in the ,i. and Cp•curves.
Scandpower
it,SINTEF Risk Management AS
Example: When gypsum has reached a temperature of 600-650 'C the chemically
bound water is released and evaporated, and a gypsum board will loose its strength
and if possible fall off the protected structure. In calculations it is suitable to simulate
this by a rapid increase in the conductivity-curve at e.g. 600 'C.
In all temperature analyses of insulated structures it is essential to know the reaction

of the individual materials in the model when exposed to the actual fire situation:
Will the material endure the temperatures? Will there be any melting, sintering
of mineral wool, charring etc. that can deteriorate the materials? Any reactions
have to be taken into account
What is the operating temperature range for the materials?
Will there be any fall off, spalling etc.?
Will the boundary conditions change during the fire? A shining surface with a
very low emissivity may change due to high temperatures and depositions of
soot
Is there room for expansion of intumescents? An intumescent material expands

when exposed to fire and forms an insulating char.
It is also essential to have a range of fire test results in order to validate the material
properties. One should be very careful to carry out calculations for structures that
strongly differ from the tested sections.
8.3.4 Simple Design of Steel Structures

Fire design of steel structures may be carried out according to Ref. /8-28/ or /8-24/.
Except when considering deformation criteria , the critical steel temperature Ts.er at a
time t for a uniform temperature distribution may be determined for any degree of
utilisation I.lo at time t=0 using
1
Ts.er = 39.19 In[ 3 _833 - 1] + 482 Eq. (8.40)
0 .9674µ0
For members Class 1, Class 2 and Class 3 cross sections and for all tension
members, the degree of utilisation I.lo at time t= 0 may be obtained from Eq. (8.40).
The design procedure is then
determine the cross-section class

determine the degree of utilisation I.lo according to Ref. /8-28/ of /8-24/
calculate the critical steel temperature Ts.er from Eq. (8.40)
calculate the uniform temperature distribution Ts,d in the heated steel member
stability control Ts,d < Ts,cr•
Scandpower
8.4 Fire Response of Concrete Structures

8.4.1 Introduction
Concrete structures are in general considered to be fire resistant, and there are many
test reports, which demonstrate the fire resistance of reinforced concrete. However,
nearly all available successful fire endurance tests have been carried out under "cellu-
losic" type fire conditions, whereas fires on offshore oil production platforms are nearly
all of the "hydrocarbon" type, which is much more severe. A hydrocarbon fire, (HC-
fire), may be characterised by a very fast temperature rise and by high stationary
temperatures. Knowledge about the behaviour of concrete structures, especially high
strength concrete used in offshore platforms, under such conditions has been limited.
8.4.2 Behaviour of Concrete at elevated Temperatures

Spalling
The concrete used in offshore structures is characterised by high compressive

strength, very low permeability and relatively high moisture contents. The low perme-
ability and high moisture contents are known as two of the main parameters influen-
cing the risk of extensive spalling of concrete during fire exposure. The use of pre--
stressed reinforcement, which induces compressive stresses in the concrete, will also
have a negative influence, i.e. higher risk of spalling. A number of tests carried out at
SINTEF NBL - Norwegian Fire Research Laboratory have proved that severe spalling
will occur from unprotected offshore-type concrete when exposed to a high intensity
fire simulating real hydrocarbon fires. The spalling starts after only a few minutes, and
is characterised by a continuous cracking of small aggregate- and cement paste parts
from the surface of the exposed concrete element. The most extensive spalling is,
contrary to common interpretation, observed for the tested light weight aggregate
(LWA) concrete with compressive stresses introduced through prestressed reinfor-
cement. This concrete quality has a very low permeability. A 7 year old beam of
LWA-concrete experienced the same extensive degree of spalling as the younger
ones, Ref. /8-33/. Rate of spalling up to 3-4 mm/min was recorded for unprotected
LWA-beams. A local peak value of 8 mm/min was recorded for a LWA-beam where
the passive fire protection failed.
Different protection methods have been tested in order to prevent or reduce the des-
tructive spalling. Constructional details, such as the use of reinforcement net near the
surface of the concrete, did not stop or limit the degree of spalling.
Fire tests have proved that the application of passive fire protection may prevent
spalling. Several materials, both spray applied and boards, and different application
methods have been investigated. Fire tests have shown that 10-20 mm of e.g. a
siliceous board is sufficient to reduce the temperature gradient over the outer concrete
layer, and, thus, prevent the spalling.
Any passive fire protection applied to a concrete surface to prevent spalling must stay
put during the entire period of fire exposure.
The fire endurance tests have shown that spalling during HC-fire may be a problem
which seriously has to be considered in design, Ref. /8-33/. For certain concrete
qualities spalling may be a severe problem at fires with even lower intensities.
Scandpower
Mechanical Properties
Mechanical properties of concrete which are highly dependent on the temperature are
stress-strain relationships, modulus of elasticity, compressive- and tensile strength,
creep and thermal elongation.
Due to the vast number of different concrete types and qualities, it is difficult to give
general equations for these properties. For high strength concrete qualities (often with
light weight aggregates) there are limited data with respect to temperature dependent
properties.
However, in Ref. /8-27/, /8-34/ and /8-35/ information about mechanical properties of
normal strength concrete qualities can be found.
The quality of the reinforcement steel may also vary, and the temperature dependent
mechanical properties must be determined. Data can be found in Ref. /8-26/ and Ref.
/8-34/ for some types of reinforcement steel.
Thermal Properties
Normal weight concrete:
The specific heat Cpc (J/kg°C) of normal weight concrete may be adopted according
to Eq. (8.41) , Ref. /8-27/ and /8-34/:
20 °C :s; Tc :s; 1,200 °C:
= 900 + 80-TC - 4 . (-1.s.. )2 Eq. (8.41)

Cpc 120 120
where
Cpc = specific heat of concrete (J/kg°C)

Tc = concrete temperature (°C)
In order to take into account the moisture contents the specific heat may be completed
by a peak value situated between 100 °c and 200 °c:
Peak value cpc· = 1,875 J/kg-°C for 2 % moisture by concrete weight

IN simple calculation models the specific heat may be considered to be independent

of the concrete temperature. In this case the following value should be taken:'
= 1,000 J/kg°C
The thermal conductivity Ac (W/m-°C) of normal weight concrete in the range 20 °C ~

Tc :s; 1,200 °c may be adopted according to Eq. (8.42) Ref. /8-27/ and /8-34/:
2
Ac= 2-0.24 • Tc +0.012 Tc ) Eq. (8.42)
120 ( 120
Scandpower
---------- ------ -----------~· --------- ------· - - -
Light Weight Concrete:
The specific heat Cpc (J/kg°C) of light weight concrete may be considered to be inde-
pendent of the concrete temperature
Cpc = 840 J/kg°C

The thermal conductivity for light weight concrete may be determined from the
following
A.c=1.o • ...l£.. Eq. (8.43)

1600
T 0 ~800°C
A.c = 0.5 Eq. (8.44)

where:
).c= thermal conductivity of light weight concrete (W/m°C)

Tc = concrete temperature (°C)
Eq. (8.41- 8.44) are plotted as curves in Appendix B.
8.4.3 Methods to Calculate Temperature Response of Fire exposed

Concrete Structures
Manual Methods
Most countries have national standards for the design of concrete structures exposed
to fire, and design manuals for fire exposed concrete elements of construction have
been made. Based on a great number of fire tests of elements with different shapes,
dimensions, concrete mixes, positions of reinforcement etc., design tables have been
made. Variables in these tables are often type of aggregate, dimensions (thickness of
partitions and width of beams), concrete cover and fire resistance time. With a certain
combination of prescribed fire resistance time and dimensions, the designer can find
the necessary concrete cover to protect the reinforcement steels from reaching a
critical temperature.
For load-bearing structures the critical steel temperature in the reinforcement is often
set to 500 °C. For pre-stressed structures the critical steel temperature is often set to
350 °C.
The simplest method of calculating the temperature response of a fire exposed con-
crete structure is to use the design tables as a basis. It can be assumed that the tabu-
lated concrete cover will prevent the steel reinforcement from reaching 500 °C/350 °C
for the given fire endurance time. The approximate temperature will differ by ± 50 °c
for concrete cover differing ± 5 mm from the tabulated values. This is valid only for
concrete exposed to the standard fire.
Scandpower
ais1NTEF Risk Management AS
A simple way of calculating the load bearing capacity of a fire exposed concrete mem-
ber is to find the 500 C isotherm in the cross section of the concrete element at diffe-
rent stages of the fire exposure (both heating and cooling). It can then be assumed
that the concrete outside the 500 °c isotherm is not contributing (or has a limited con-
tribution) to the total load-bearing capacity of the section. Further, it can be assumed
that the concrete subjected to lower temperatures has its full capacity. The capacity of
the reinforcement is a function of the steel temperature, which can be assumed to be
approximately equal to the concrete temperature in the same position.
The 500 °c isotherm can be found from the above-mentioned simple method, or by
pre-calculated temperature profiles of prototype cross-sections which are often given
as an annex to the national design standards, Ref. /8-27/ and /8-34/.
Note:
National Standards for the design of fire exposed structures mainly focuses on
structures exposed to the standard fire. No standards are made for hydrocarbon fire
exposure or general fire exposure curves. Care should be shown in using tabulated
values when the fire exposure differs from the standard fire.
Insulated Concrete
It is possible to insulate concrete structures in order to compensate for insufficient

concrete cover. There are a number of different insulation materials available, and
their insulating capacity on concrete is often given as an equivalent concrete thick-
ness.
Example: A high density mineral wool board of 10 mm is equivalent to 25 mm con-

crete when regarding the necessary fire insulating cover for the reinfor-
cement. If additional 25 mm of concrete cover is needed to meet a pre-
scribed fire resistance, this may be replaced by 10 mm of high density
mineral wool.
Computer Codes
In order to carry out reliable calculations of the fire resistance of concrete structures, a
full temperature analysis of the exposed cross-section is necessary. There are a
number of computer codes available, which can be used for such an analysis.
SUPER TEMPCALC, Ref. /8-31/, and TASEF-2, Ref. /8-32/, are examples of com-
puter codes which can be run on a personal computer.
Both codes are based on the finite element method (FEM), and covers 1- and 2-
dimensional heat flow. The computer codes can analyse a cross-section of a struc-
ture exposed to an arbitrary time/temperature-curve. The cross section can consist of
different materials, and the thermal properties ().., Cp) can be given as a non-linear tem-
perature-value curve. The heat exposure is modelled either as a time/temperature-
curve with a given convective and radiative heat transfer coefficient at the boundaries,
or as a prescribed time/total heat flux - relationship. Hollow sections can also be
analysed. Ref. /8-31/ and /8-32/.
Figure 8.8 shows an element mesh for an insulated I-section concrete beam. The
column is subjected to uniform heat exposure, and due to the symmetry line, only half
of the cross section has to be analysed.
Scandpower
- -
~
}-
,.,., .
··- t")
.
Element model Exposure model
10o1••'"" ,,o~ u I ( "l

uuu, ... , ...... u , ( "'
After 1 hour After 2 hours
Figure 8.8: Element Mesh for an insulated I-Section Concrete Beam, and Positions
of Isothermal Lines at a given Stage in the Calculation
In such a temperature analysis it is not necessary to model the reinforcement; the

whole cross-section is assumed to be concrete. Based on the temperature analysis it
is possible to carry out an accurate calculation of the load bearing capacity of the con-
crete member. The temperature of reinforcement steel is assumed to be equivalent to
the concrete temperature at the same position, and the loadbearing capacity of each
steel can be calculated individually.
Note:
As when carrying out all temperature analyses it is essential that care is shown when
choosing the thermal properties of the materials. Since there are a number of different
concrete qualities, the essential factors are type of aggregate, density and moisture
content. These factors can help choose the adequate thermal properties.
The next important issue is whether the concrete quality will take the heat exposure
without spa/ling, (See Chapter 8.4.2). When analysing high strength concrete subjec-
ted to high intensity fires, such as fires in hydrocarbons, special attention should be
given to the spa/ling problem.
When analysing prestressed concrete structures the critical temperature of the steel
strands is often set to about 350 'C, depending of the stress level and on the proper-
Scandpower
ties of the steel. However, the bond strength between the steel strands and the con-
crete may fail long before the critical temperature is reached.
When analysing composite structures (e.g. insulated concrete structures) it must be

documented that the insulation will stay put to the concrete surface during the fire
exposure. If not, this has to be taken into account in the calculations.
8.5 Glass Fibre Reinforced Plastics (GRP)

8.5.1 Introduction
Glass fibre Reinforced Plastics (GRP) is a group of polymer based composite mate-
rials that have found wide acceptance within industrial and consumer oriented appli-
cations over the last two decades. This is due to the special material properties they
possess. Such properties as high strength combined with corrosion resistance and
light weight, are of great interest to the offshore industry and to the process industry
onshore.
A composite is a material built up of two or more different components. In a fibre com-

posite at least one of the components is in the form of fibres. The other ones form the
matrix which fix the fibres and transfer load between them.
Since GRP generally is a less known material in the industry compared to steel and
concrete, a general description of GRP and its physical properties has been included
in the handbook. Ref. /8-36/ gives further information regarding properties and appli-
cations of GRP materials.
Matrix Materials in GRP Composites
There is a great variety of plastic materials that can be chosen as matrix materials in a
composite. The first distinction to be made is between the two basic types of plastics:
Thermoplastic and thermosetting materials.
A thermoplastic is characterized by the fact that it softens and melts when the tempe-
rature is raised and hardens again when the temperature is lowered. GRP materials
with thermoplastic matrix materials will not be discussed further in this handbook.
Thermosetting plastics are characterized by being formed and cured under the appli-
cation of heat or chemical means into crosslinked and substantially infusible and in-
soluble materials. The materials exhibit a drop in mechanical properties at the glass
transition temperature (T9) of the resin, but retain, due to the fibre reinforcement, a
significant level of properties up to the temperature of disintegration. Hence, they re-
tain to a large extent their physical properties up to the point of disintegration. It is the
thermosetting plastics that are most commonly used as matrix materials for con-
tinuous fibre reinforced plastic composites.
Physical properties of thermosetting materials are superior to those of low priced

thermoplastics, and they have proven very good in transmitting forces between the
fibres in composites. Competitive prices and manufacturing processes have in addi-
tion made thermosetting plastics the predominant matrix material in composites.
Scandpower
Polyester Resin
Polyester resin is the most widely used resin type for reinforcement with glass fibres.
The most common categories of polyester resins include orthophtalic, isophtalic, bis-
phenol A, chlorendics (or HET acid polyesters) and vinylester. Vinylester resins are
often classified resistance of the thermosetting plastics. The chemical structure has
similarities with the epoxies, but the curing process is the same as for the polyesters.
They are also processed into products by the same methods used for common
polyesters.
An advantage compared to previously used thermosetting materials is the ability of the

material to cure at room temperature and without any compression. This opened the
possibility to contact mould curable reinforced laminates with virtually no size limit.
This meant simple and cheap moulds, and a manual lay-up process requiring no ex-
pensive equipment.
Since their introduction, polyesters have been adapted to more advanced production
processes such as compression moulding, filament winding, centrifugal casting etc.
The basic material has also been developed into a variety of specialities with em-
phasis on properties such as fire resistance, chemical and corrosion resistance,
weatherability, physical behaviour etc. Specialities from which one can benefit when
designing products for offshore and process industry.
Epoxy Resin
Epoxy resins are among the most versatile of the commercially available matrices.
They cover a broad range of physical and mechanical properties and processing re-
quirements. They have in common that they all contain epoxide groups.
Curing may be effected at room or at elevated temperature depending on the resin

and hardener used. Amines, polyamides and acid anhydrides are commonly used
hardeners or curing agents in resins intended for structural applications.
Characteristics of epoxy resins, which make them especially suitable as matrix mate-
rials in fibre composites, include:
Epoxy resins adhere well to a wide variety of fillers, reinforcing agents and sub-
strates
The wide variety of available epoxy resins and curing agents can be formulated
to give a broad range of properties after cure and to meet a diverse spectrum of
processing conditions
The chemical reaction between epoxy and curing agent does not release any
volatiles
The shrinkage during cure is usually lower than for polyester, vinylester and
phenolic resins.
Epoxies have similar properties and production features to the ones found in poly-
esters although epoxies represent a somewhat greater health hazard during proces-
sing. Due to their somewhat better bonding and physical characteristics, and better
ability to transfer loads throughout the laminate they are predominantly used as a
matrix in more advanced composites. The same advantage is found when used with
glass fibres.
Scandpower
The chemical resistance of epoxies is of importance especially in high performance

products such as pressure vessels and pipes, applications often found in offshore
installations and process plants.
The higher price is partly due to more expensive raw materials and more complicated
fabrication processes and partly due to a possible health hazard in connection with the
hardener systems. The most frequent processes used for pipes and pressure vessels
are filament winding and centrifugal casting. Filament winding gives better physical
properties whereas centrifugal castings are used for highly corrosive fluids.
Phenolic Resins
Phenolic resins are among the very oldest plastic materials known. Traditionally they
have been processed at elevated temperatures in matched metal dies under high
pressure. Properties such as dimensional stability, high heat distortion temperature,
good thermal and electrical insulation properties, water resistance and low price have
lead to a wide use of phenolics in appliances and laminated sheets. Their use has,
however, been restricted to fairly small components because they are more difficult to
process. Recently new versions of phenolic resins, which can be cured at room tem-
perature, have been developed. The fact that they can also be processed with no re-
quirement for compression, brings them into the same category as polyesters and
epoxies production wise. There appears to be remaining problems that need to be
solved in the area of processability, but the outstanding properties of phenolics in
terms of flammability and smoke emission, will enhance the demand and most prob-
ably lead to increased consumption of these materials. Their properties should point
them out as being of special interest to the industry.
Reinforcing Materials
Glass fibres are the dominating reinforcing material. Glass exhibits outstanding pro-
perties when processed in the form of fibres. The diameter of fibres used as rein-
forcement is in the range of 9-24 µm, typically 11 µm. They are produced in con-
tinuous filaments from molten glass. The filaments are assembled parallel in bundles,
called rovings, which form the basis for the manufacture of different types of glass re-
inforcement. Some of the most common types are chopped strand mats, woven
rovings or weaves made from yarns (spun filaments).
A very important factor in the manufacture of glass fibres for reinforcing purposes is
the sizing. Sizing is a thin multifunctional coating applied on each fibre during manu-
facture. The main purpose is to secure good bonding to the matrix material, but also
to ease the further processing of rovings into their length, and spraying them onto a
conveyor together with a binder.
Glass fibre weaves are made in the same way as ordinary textiles, and are charac-
terized by type and thickness of the yarn used, weave pattern, and by the amount of
threads in the warp and weft directions.
GRP Composites
The objective of fibre composites is to obtain materials with properties superseding

those of the individual components. The reinforcing fibres have high stiffness and
strength and good creep and temperature resistance, but they are too brittle and
easily damaged to be used alone as engineering materials. Unreinforced plastics
Scandpower
have, compared to metals, low stiffness and strength, and they creep under load. An
addition of reinforcing fibres will greatly enhance the properties of the matrix materials.
The purpose of the matrix material is to fix the structure of the fibres, keep them in
position and transfer the loads taken up by each individual fibre. The objective is that
the fibres will normally exhibit strength and moduli 10 to 100 times that of the matrix.
Kept together and forced to act together by the matrix, the fibres contribute to a com-
bined material with physical properties approaching those of the fibre material. On the
other hand the matrix is of vital importance for corrosion resistance and weatherability
of the composite. The fact that the matrix encompasses and protects each fibre and
also forms the surface of the end product, underlines the importance of correct choice
of plastic material for the matrix.
The properties of the GRP materials can be varied within the ranges by variation of
matrix material, fibre type, fibre orientation, fibre reinforcement geometry, fibre volume
fraction, etc. This gives the designer freedom to tailor the mechanical properties of
glass fibre reinforced plastics for specific applications and according to required loads.
The strength of a finished part is strongly dependent on the orientation and the
amount of glass fibres. The glass fibres are available in three main groups or types:
1) Chopped strand mats (CSM), which are short fibres with random orientation in
plane. The amount (by weight) of fibres in a product when using CSM is usually
in the range of 10-40 wt%
2) Woven roving and fabrics in which the glass fibres are orientated in two main
directions, 0° and 90°, giving the highest strength and stiffness in these direc-
tions. The fibre contents in products are normally 40 - 70 wt%
3) The use of unidirectional fibres gives high strength in the direction of the fibres.
The amount of fibres in a product based on unidirectional fibres may be in
excess of 80 wt%.
Figure 8.9 illustrates typical ranges for tensile strength and tensile modules (in prin-
cipal fibre direction) for the three main categories of reinforcement in GRP products.
The values are based on a collection of empirical data.
The coefficient of thermal expansion varies from about 5 to about 40 (-10-6/°C) and
for most of the GRP materials in the same range as for steel 8 - 16 (·10-6/°C) and
aluminium 22 - 44 (-10-6/°C).
The coefficient of thermal conductivity varies from 0.2 - 0.35 (W/m, °C) while the cor-
responding values for steel and aluminium are 10 - 36 (W/m, °C) and 120 - 129 (W/m,
°C) respectively.
The variation of properties within each category is affected by the use of different
qualities of glass fibres, by using different matrix materials and also by variation in pro-
duct quality.
Mechanical and physical properties for the categories of fibres and matrix materials in
addition to GRP materials are given in the Appendix F.
Scandpower
(l)SINTEF Risk Management AS
TENSILE STRENGTH TENSILE MODULUS
(MPa)
1000 -
----------=:-------, ~ ,~ - 60
(GPat
- Unidirectional
llbrea (UDJ
"" so
-
I e-
..
40
- I j Ii
500 - ... 30
.
. ,
Awoven rovfnga (WRI
,- 20
-
,_ 10
.
~ h < > 1 1 ~ d st, _ mat (CSM)
0 I I I I 0
I I t I
0 so 100
WI ,c. Class
Figure 8.9: Typical Values for the Mechanical Properties in Principal Fibre Direction
of GRP as a Function of Weight Percent Glass for the three Main Types
of Reinforcements. The Values are based on a Collection of Empirical
Data, Ref. /8-37/
GRP Products in the Process and Offshore Industry
In recent years a lot of different GRP products have been taken into use in the pro-
cess and offshore industry. Amongst these are pipes, tanks, life boats, gratings, hand
rails, stairways, ladders, walls/panels, boxes, impellers, profiles, cable trays and cable
ladders.
In the future it is likely to believe that composite materials will be used in more critical
applications such as pipes containing hydrocarbon liquids (subsea or in the process
modules), high pressure pipes, pumps, primary and secondary structural elements,
subsea structures, risers, etc.
To use composite materials in such applications a lot of qualifying work prior to in-
stallation has to be carried out based on risk assessment analysis for specific com-
posites. This includes for instance documentation of mechanical properties, environ-
mental resistance, chemical resistance, fire performance, etc.
8.5.2 General Fire Properties of GRP Materials

Composites of commercial thermoset polymers represent generally flammability
hazards. There are, however, methods to make the resins more fire retardant. This
includes use of fire retardant additives or modification of the resin with fire retardant
groups. Fire retardants act generally by decreasing the efficiency of combustion and
they tend to contribute markedly to smoke evolution. These fire retardants are basi-
cally effective in the very early stages in a fire to reduce ignitability and prevent flame
Scandpower
Risk M anagement AS ('G» SINTEF
spread. In a fully developed fire they have little effect and will contribute significantly
to an increase in smoke evolution and in some cases also to evolution of toxic gases.
There is a tendency to avoid the use of flame retardant resins in offshore applications
of GRP materials.
Different test methods or test standards have been used to characterise a certain fire
property. This is often due to different standards in different countries, and can make
it difficult to compare fire properties for different materials.
Below are some important fire properties presented for different types of GRP mate-
rials, and some general comments are drawn.
Oxygen Index
The 01 (Oxygen Index) test has become widely accepted in the flammability testing of
polymeric materials. The 01 is defined as the minimum concentration of oxygen, ex-
pressed as volume percent, in a mixture of oxygen and nitrogen that will just support
flaming combustion.
The Oxygen Index for GRP materials varies within each type of matrix material depen-
dent on type of additives, amount of additives, etc. In Table 8.12 the 01 is indicated
for GRP materials with different types of matrix material.
Table 8.12: Oxvaen Index

GRP with pti~n:Olip C3RP":,yith epoxy as· . GRP witffstaridarct GRF'.wi!h flame
as matrix materiaf I •• rriafrix material ·.· pc,lyest~ras rnatril( reta"rdaot polyester
· · .· · . · · material · · as matriXinaterial
40-80 25-40 15-20 25-50
Materials with an Oxygen Index higher than 21 are self-extinguishing under normal
atmospheric conditions. Under other conditions this is not necessarily the case and a
higher 01 may be needed for the material to be self-extinguishing. Hence, Table 8.12
shows that phenolic resins have better self extinguishing properties than epoxy and
standard polyester resins, while flame retardant polyesters can gain 01 similar to some
phenolic resins.
Resistance to Ignition
Regarding GRP materials' resistance to ignition some testing has been done at Nor-
wegian Fire Research Laboratory (NBL) according to the British Standard (BS 4790:
1972 (hot metal nut method). Flaming times were recorded for four different laminates
made of glass fibre reinforced polyester, fire retardant polyester, epoxy and phenolic
resins.
Flaming times for glass fibre reinforced polyester laminates with and without retardant
additives were typically 40 and 30 seconds respectively. For glass fibre reinforced
epoxy laminates the flaming times ranged from 40 to 80 seconds. No flaming was
observed with a glass fibre reinforced phenolic laminate. Later, flames were not ob-
served on any of the laminates. The results indicated little effect of fire retardants on
polyester resin in this test.
Scandpower
The same laminates were tested with respect to ignition and burn through upon expo-
sure to welding flames. Five different types of welding flames were applied, pointing
directly onto the laminates at a distance of 10 mm. The burner was removed after 15
seconds of exposure. The maximum temperature of an oxygen/acetylene welding
flame is 3,100 °C. Time until burn through and after flame time was recorded.
This test showed significant differences for polyester laminates with and without fire
retardants. Typical after flame times were about 2 minutes for the samples without fire
retardant and for some the flames were increasing until they were extinguished after 3
minutes. For samples with flame retardant polyester the after flame time was typically
3 seconds or less. In all tests with epoxy laminates the flames were increasing until
they were extinguished after an after flame time of 3 minutes. For the phenolic lami-
nates the after flame time was typically 2 seconds or less. Some of each category of
laminates was burned through during the 15 seconds exposure, dependent on the
type of flame. There was no significant difference between the two polyester lami-
nates and the epoxy laminates. The phenolic laminates seemed to be more resistant
to burn through.
Spread of Flame
The spread of flame property expresses the GRP material's ability to limit the fire-
affected area after exposure of flaming for a given period of time. As a general rule
GRP products of phenolic matrix perform better than GRP products of epoxy of stan-
dard polyester resin when speaking in terms of flame spread. Again products of
epoxy matrix perform better than standard polyesters matrices. GRP products with
polyester matrix filled with flame retardants tend to spread the flame less than pro-
ducts of standard polyester matrices, and increasing the amount of flame retardant
fillers seems to lead to increasing reduction of flame spread.
Experiments with GRP pipes carried out at Statens Provningsanstalt in Sweden,

Ref. /8-38/, showed that water would keep the pipe cool and tend to reduce the flame
spread, and also that fire retardants seem to have an effect of reducing the flame
spread.
Regarding GRP materials with vinylester as matrix material, it seems that these have
more or less the same flame spread properties as GRP materials of epoxy materials.
Smoke Generation
The smoke generation property tends to express the combustion characteristics of

GRP materials. Usually the tests explore the smoke production potential of a small
sample of material under specified conditions of pyrolysis and combustion. It does not
necessarily indicate the smoke production under other thermal exposure conditions. A
common smoke generation test method used to determine the smoke generation
property of GRP products is the ASTM E662 (smoke chamber test). This method is
widely used as a test for characterisation of different combustible materials and deter-
mines smoke evolution as maximum specific optical density, Dm.
In Table 8.13 typical figures of smoke generation properties for GRP products of diffe-
rent matrix materials are shown, tested according to ASTM E662. As for all other fire
properties also the smoke generation property varies within each type of matrix mate-
rial, but the figures given can be used to characterise the smoke generation in general
terms.
Scandpower
Risk Managem ent AS G) SINTEF
Table 8.13: Smoke Generation Properties (flaming) of Test Speciments exposed to a

Combination of an Electrical Heater and a Propane Burner
5- 75 230-900 500· 340 - 1200
• Based on only one test
Comparison of GRP materials of phenolic, vinylester, epoxy or polyester resins in

terms of smoke generation show that the phenolic laminates perform best. Thereafter
follow the vinylester, epoxy and unfilled polyester. The flame retardant filled polyester
composites generate more smoke than the others. This is due to the fact that the fire
retardants generally act by decreasing the efficiency of combustion, thereby contri-
buting markedly to smoke generation.
The phenolic resin's low smoke generation properties, is especially important in the
case of a fire situation where a person's ability to see is of vital importance.
A comparison of smoke generation curves for similar samples of GRP produced using·
phenolic, polyester and epoxy resins are shown in Figure 8.10.
700
-- -- - - -
600 ,/
Epoxy.,.,,,.
-
C
>.
t i)
500
400 I
I
/
<!)
-0 I
300
0 I
u I,
=+=
(l_
200
0
100
Phenolic
.....................................
0 2 4 6 8 10 12 14 16 18 20 22
Time (minutes)
Figure 8.10: Comparative Smoke Curves - NSB Smoke Chamber (smouldering), Ref.
/8-39
Scandpower
General
The test methods and standards given for determining the combustion characteristics
of materials are useful for the quality control of the materials and for research and
development. A direct comparison between different materials tested according to
different methods/standards is not recommended. When comparing different mate-
rials tested according to identical methods/standards, it is necessary to be aware of
the test conditions, i.e. for instance test temperature, substrate curing time, glass fibre
content, type and amount of additives, etc.
The toxic fume production of phenolic resins is low. This is illustrated in Table 8.14
where typical emission from phenolic GRP in NBS smoke chamber test is given.
Table 8.14: Gaseous Emission from Phenolic GRP in NBS Smoke Chamber Test,
Ref. /8-39/
Gas Smouldering Flaming
(ppm) (ppm)
CO2 300 5,000
co 50 100
HC1 0 0100
SO2 25 00
NO2 0 0
NO 0 0
NH3 0 0
HCN 0 0
Formaldehyde 0 0
Phenol 0 0
Phenolic composites perform significantly better than composites based on polyester,

vinylester and epoxy with respect to the general fire properties included in this hand-
book. Composites based on polyester, vinylester or epoxy resin have more similar fire
properties and differences may depend on variations in compositions of the individual
resins. Use of fire retardants improves oxygen index and flame spread properties but
has a negative effect on smoke evolution and in some cases on the evolution of toxic
gases. Results from the literature on general fire properties give indications of the
performance of the different types of composites. The development in the North Sea
is towards avoiding the use of flame retardants. In some applications of GRP off-
shore, use of fire protective coatings will be required and these will influence espe-
cially flame spread, ignitability and smoke evolution.
Fire protective coatings can be used when an increase in the fire resistance of the
GRP materials is required. The fire protective coatings' aim is to delay a temperature
rise of the material beneath the coating. Several different types are available including
intumescent paints, intumescent epoxy based coatings, intumescent highly filled cera-
mic forming thermoplastic systems, phenolic foams, non expanding organic systems,
inorganic systems based on vermiculite, magnesium oxychloride, mineral fibres and
ceramic fibres, etc.
It should be noted that real fires can not be exactly simulated by standardised labo-
ratory tests. For this reason supplementary large scale tests may be necessary in
Scandpower
cases where new and critical applications of GRP pipes are being considered, to
demonstrate fire safety and to establish satisfactory test and acceptance criteria.
8.5.3 Performance of selected GRP Products at elevated Tempera-

tures
In this handbook two relevant products are selected as examples of how GRP pro-
ducts typically perform at elevated temperatures. These two selected products are
tanks and pipes.
Tanks
Under fire testing of GRP tanks the fire temperature exposure has in most cases been
in the range of 800 - 1000 °C. These tests have mainly been pit fire tests where the
tanks have been placed above a burning pool of naphtha, petrol, heating oil or other
readily combustible hydrocarbons. The content of the tanks have included liquids
such as naphtha, petrol, heating oil and water, and also gases, e.g. engine gas and
natural gas. In some investigations the performance of GRP tanks has been compa-
red to corresponding tanks in steel and aluminium. The main goal of these investi-
gations has been to document the fire performance of GRP tanks from a safety and
functionality point of view. Thon and Stokke, Ref. /8-40/ and /8-41/, summarise seve-
ral different fire tests of tanks in GRP, steel and aluminium.
Fire testing of GRP tanks containing flammable liquids (e.g. naphtha) has shown im-
proved performance compared to steel and aluminium. They are generally considered
safer for this application. This is mainly due to the low heat conductivity of GRP,
which are 40 and 700 times less than for steel and aluminium respectively. The heat
conductivity of the GRP tanks is further reduced upon fire exposure due to the for-
mation of a carbonised outer layer. GRP tanks are for this reason also considered
representing reduced risks for contamination of the surrounding environment by loss
of contents and for personnel involved in extinguishing the fire. Due to GRP's low
thermal conductivity one does not find the two scenarios one usually finds with a steel
tank exposed to fire where boiling of the liquid leads to a pressure build-up and
eventually exceeds the ultimate strength of the tank, or a partly filled tank where parts
of the tank will drastically reduce its mechanical properties leading to a potential of
explosion. Neither will a fire exposed GRP tank melt as is the possibility for an alumi-
nium tank. As an example, the characteristics of a failure scenario of water filled GRP
tanks are typically as follows:
Water drops on the outer surface of the laminate

A thin water jet up to one meter from the laminate. The water pressure of the
water jet is so low that you can meet it with the hand without any damage
A fitting or manhole cover may blow off. It could be blown a couple of meters
from the tank. The fracture is signalled a long time before by audible acoustic
emission. Also whitening of the failure area will occur some time before fracture
(cracking)
Fracture of the tank occurs with flowout of the liquid. The water pressure is
eliminated immediately.
Hence, GRP tanks and vessels are considered to have a safer behaviour than steel
tanks in cases where the internal pressure may exceed the ultimate strength of the
tanks or where the dry parts of the tank is heated leading to reduction in mechanical
properties. Liquid filled GRP tanks in particular will fail in a non dramatic way, while
Scandpower
{!)SINTEF Risk Management AS
for steel tanks there is a risk of explosion causing a great deal of damage to the sur-
roundings.
The support system for GRP tanks is an important factor for the performance in fire
situations. With an inadequate support system the tank may collapse.
In two fire tests where the GRP tanks were supported on two horizontal steel bars the
tanks collapsed after 12 and 16 minutes. These supports were not according to any
design standard for GRP tanks. In fact all standards emphasise very strongly that
horizontal GRP tanks had to have saddle support covering at least 120 °C of the cir-
cumference and with even load distribution over the saddle.
There has been concern about the possible hazards regarding electrostatic charging
of GRP tanks. According to the literature, among these the British Standard for con-
trol of undesirable static electricity (BS 59958-1 ,2) GRP tanks and pipes should not
represent any increased risk for static electricity build up compared to metallic tanks
and pipes when containing sea water if the sea water is properly earthen at some
points, Ref. /8-42/. The same considerations as for metal tanks and pipes have to be
taken regarding filling procedures and earthening. If possible hazards should arise
regarding electrostatic build up, the technology exists today, which can make the
whole, or a part of a tank or pipe electrically conductive.
Little information of fire behaviour of thick-walled laminates is reported in the literature.

Based on this fact, Senter for lndustriforskning (SI) in Oslo, Norway, carried out their
own fire testing of thick-walled laminates, Ref. /8-43/. The aim of Si's fire testing was
to see how thick-walled laminates behaved in a fire situation and to measure the
temperature profile through the wall thickness. The tests are relevant for thick-walled
GRP products such as tanks and pipe flanges and to describe how the temperature
through the wall may develop. The test parameters that were varied were wall thick-
ness, cooling of the unexposed side, type of reinforcement and the use of fire protec-
tive coating.
The results from the tests can be combined with measurements of strength and stiff-
ness as a function of temperature to give a basis for calculation of residual strength of
a thick-walled GRP product at elevated temperature.
Laminates were mounted as the "roof' of a small propane heat furnace. The cooling
of the unexposed side was either by convective air or water and in one of the tests the
laminate was insulated with ceramic wool on the unexposed side.
The tests, which included water-cooling of the unexposed side of the laminates, illu-
strated that water cooling is effective in keeping the temperature towards the un-
exposed side of the laminates at a low level. GRP tanks that contain water or other
liquids start to boil. Degradation will then develop in vapour filled areas.
There are differences in the behaviour of laminates reinforced with Combimat (WR +
CSM) and those reinforced with chopped strand mat (CSM). The CSMs seem to give
an improved resistance during the first stages of a fire especially in the parts close to
the unexposed side. This is probably due to a higher expansion of these laminates
during fire exposure with a corresponding increase in insulation effect. This opening
of the structure may, however, also lead to more rapid burn away of the resin at later
stages. The laminates made of Combimats do not have any significant volume in-
crease during heat exposure and the continuous fibres keep the laminate tighter to-
gether when the outer resin has burned away.
Scandpower
The expansion or opening of the CSM laminates seems to give these laminates higher
resistance against inward propagation of melting of the glass fibres. The tighter fibre
bundles of WR laminates will be more exposed in thin walled laminates. The only
laminate tested at SI that suffered a burn through was a thin walled laminate (4.7 mm)
made of Combimat where the fibre bundles in the WP part melted progressively.
It has been reported that water filled unprotected GRP tanks have been tested for
more than 30 minutes without leakage in open hydrocarbon pit fires. In addition, if
small amounts of fire protective coating or intumescent paint are added, the fire resis-
tance of the tanks will be significantly increased and the tanks can withstand even
significantly longer times without any leakage occurring.
SI also investigated the effect of applying external fire protection on the thick-walled
laminates. The effect of applying a thin layer (approx. 2 - 2.5 mm) of highly intumes-
cent paint of the exposed side of the laminates was clearly seen. The temperature
development through the wall of a laminate is much slower for the laminates with
protective coatings. The protective coating is highly intumescent paint (which expands
more than 30 times) reinforced with chopped strand mat. The results showed for the
tested laminates that the coating provided good protection for more than 120 minutes.
The erosion resistance of these coatings, for instance when exposed to jet fires, has
to be investigated further. The test results also indicate that after longer times of fire
exposure the differences tend to be reduced.
Addition of a fire retardant filler may improve the fire resistance and slow down the
temperature increase for a period of time. The tested filler did not have the same
effect as a fire protection coating. Use of fillers has the advantage that they can be
added during the production process.
The fire tests also demonstrated that the ability of a laminate to resist a fire depends
on the wall thickness. An increase of the thickness of GRP laminates increases the
fire resistance considerably. The glass fibres in thick laminates remain in place, and
these glass fibres, together with charred resin further into the laminates, function as a
thermal barrier between the rest of the laminate and the fire outside. These results
correspond well with other test results reported, such as the 6 hours fire in the engine
room of a minehunter made of GRP, where no burn through of the approximately
20 mm thick laminates occurred.
The measured temperature profile through the laminates, combined with measure-
ments of strength as a function of temperature, provides a good basis for a computer
simulation calculating the residual strength of a structure. The difficult part in this cal-
culation is to obtain reliable mechanical data as a function of temperature relevant for
the products in question.
Pipes
One of the main obstacles for the use of GRP pipes on offshore oil installations has
been the concern about a possible fire hazard. GRP pipes have, however, been sub-
ject to extensive fire testing in order to document their performance in fire situations.
In this handbook a summary of several independent fire test programmes are inclu-
ded.
The results document that unprotected plain and jointed GRP pipes with flowing water
inside remain leakfree and maintain their function for long times under fire exposure,
indicating that for offshore sea water pipe systems normally maintained in this con-
Scandpower
dition, GRP might be applied without particular fire protection. The same reasoning
should be valid also for pipe systems that might be exposed for short periods in stag-
nant water conditions before water begins to flow, which is the situation for the wet
parts of the fire water system. The results indicate that common joints (e.g. cemented,
muff cemented and laminated) do not seem to weaken the performance of the pipes in
this condition. Additional tests might, however, be required for further documentation.
Tests with unprotected empty pipes have indicated burn through times for pipes with
5 mm wall thickness of 2-6 minutes dependent on test conditions. Results indicated
that the endurance time was mainly influenced by the pipe wall thickness and rela-
tively independent of the matrix material.
An experience gained from observing these tests was that the smoke development
was fairly low and mainly limited to the initial period where the outer pipe surface de-
grades or carbonises. After the tests the ends of the pipes which had not been
directly exposed in the fire remained unaffected and thus indicating insignificant flame
spread along the pipes during and after the test.
The mechanisms that lead to this good performance of water filled GRP pipes in stag-
nant water condition and especially in flowing water condition include:
1) Degradation (carbonisation) of the outer surface forming a layer with conside-

rably reduced heat conductivity compared to the pipe itself
2) In later stages microcracks may develop leading to slight weeping through the
pipe wall giving a cooling effect (especially in stagnant water conditions)
3) Development of an equilibrium situation between heat input and heat transfer

which stops or significantly reduces further degradation
4) The slight weeping does not influence the functioning of the pipe and full opera-
ting pressure may be retained for long periods in this condition .
A simple small scale fire test with a small diameter pipe 1.5 inch, relevant for offshore
sprinkler systems, has documented the fire resistance of small unprotected pipes with
streaming water inside, Ref. /8-44/. The test was performed with a propane burner
and the flame temperature for the first 8 hours was 850 to 870 °C and increased up to
1,070 °C for the last 2 hours. A water flow of 0.05 mis inside provided enough cooling .
These tests showed that the pipe remained leakfree for more than 1O hours and that
only about half the wall thickness was degraded both in the 3 hours' and the 1O hours'
test, thus, confirming the establishment of an equilibrium situation.
This indicates that for offshore sea water pipe systems normally maintained in this
condition, GRP might be used without particular fire protection. Non-critical sea water
pipe systems are today generally installed without fire protection.
The fire water system represents a more critical application for GRP pipes where the
use of GRP is especially advantageous if sufficient safety and reliability can be docu-
mented. There is a corrosion problem with the metal alloys used today, which affects
the systems' reliability. There is a risk that internal corrosion products can plug the
sprinkler nozzles when the system is put into operation.
The pipes in the fire water system are either water filled or empty. On the basis of
safety studies and probability calculations, the following requirement for the fire water
Scandpower
Norwegian Fira Research Laboratory
system on a North Sea platform has been indicated by the NPD (the Norwegian
Petroleum Directorate). Pipes in the empty and stagnant water condition shall be able
to resist a fire for 5 minutes before the water starts to flow inside the pipes. The whole
assembly shall then remain functional and supply water at full capacity for at least 2
hours.
This means that fire protection will be required at least for pipes and joints of the dry
parts in the water system. For pipes with stagnant water the need for fire protection is
more uncertain. It should, however, be recommended to protect the joints in such
parts of the system.
It is important to optimise the use of fire protective coatings in order to maintain the
weight and cost benefit offered by the use of GRP. Important requirements of such
coatings are long-term environmental resistance and the maintenance of properties of
fire protection after long exposure to the offshore environment. This is especially im-
portant for intumescent coatings.
Results of tests with the use of fire protective coatings show that by proper design of
the fire protection it should be possible to satisfy the most severe requirements to fire
performance, including GRP pipes being either empty or in an airfilled condition. The
results indicate that for pipes that might be exposed to a fire in the empty condition,
even if it is only for limited time periods, the use of fire protective coatings seems to be
necessary, especially for systems intended for service in emergency situations as for
instance the dry parts of the fire water system. For pipes that will be exposed to a fire
for only short periods in the empty condition, another solution might be to increase the
wall thickness sufficiently.
Use of fire protective coatings will also improve general fire properties such as flame
spread, smoke emission, ignitability, etc. They do, however, introduce new problems
such as requirements to their long term environmental resistance, maintenance, etc.,
which will increase weight and cost.
As mentioned in Chapter 8.5.2 there is a lot of types of fire protective systems.

Systems like Chartek have been used in offshore applications for years and have
shown to be a good alternative for fire protection of different structures. In recent
years other promising alternatives have been developed. One of these is the intumes--
cent paints. Tests of several different coating alternatives (for instance Nullifire and
Uniterm) with a thickness of 1-2 mm applied on 4" GRP pipes, may be able to offer
sufficient protection of the GRP pipes surface to above 300°C. No degradation of the
matrix materials in the pipe will then take place. The intumescent paint can favourably
be reinforced with a chopped strand mat (CSM). It is, however, important to investi-
gate the long term environmental resistance of these intumescent paint systems to
qualify these for use outdoors. Another type of recently developed fire protective sys-
tem is a non-expanding rubber based system from Viking Mj,:mdalen A/S. A 5-8 mm
thick layer of this system was able to protect a 4" GRP pipe for 8-1 O minutes without
an increase in the GRP pipe surface temperature to above 300 °C. This coating sys-
tem is said to have good long-term environmental properties, but fire properties like
flame spread, smoke generation and toxicity have to be investigated before use.
Scandpower
«)SINTEF Risk Management AS
The failure criteria are important when considering the performance of GRP pipes in
fire situations. Two different criteria are often used in evaluation of test results:
1) Time until first weeping or leakage

2) The time the pipes are able to maintain operating pressure and their intended
service.
These two criteria will in some cases lead to significantly different results. From a
safety point of view, the second criterion should be of prime interest.
The properties of GRP pipes regarding flame spread, smoke evaluation and ignitability
should not restrict the use of these materials in sea water pipe systems offshore.
Available documentation on the performance of various joint designs has been limited.
Hence, Senter for lndustriforskning (SI), initiated a limited test programme to investi-
gate how various GRP joint designs perform in a fire, Ref. /8-45/. In this handbook the
main conclusions from the testing of jointed pipes of 100 mm of diameter are presen-
ted. All these pipes were tested in the stagnant water (7 bar) condition until failure.
The various joint design tested are shown in Figure 8.11 . Because the pipes are
freely supported, the joint is exposed to axial loads due to the water pressure of about
470 kp during the fire test.
W
l#/$#/#lll!Wl#/!Wllll/lll/11//Pl~-j///jf#.@'//#,I/Q#Rfflh'l##/ll ,/41
con/con cemented
----~~~ con/con muff cemented
fW&&D@,Wf-&wz;,wiiW/4if%1kW-i&'•4
conrcyl cemented
W##&#/&1$$/Wff@.Wl?},:~s;, "i . i ~#//1~

laminated (plain, con/con)
-a=•-- rubber seal lock joint
Figure 8.11: Sketch of various Joint Designs included in own Fire Testing, Ref. /8-45/
The tests revealed that most of the joints remained intact until termination of the tests
when the internal water temperature reached the boiling point (approximately 168 °C)
after test periods of approximately 30-40 minutes. Joint separation occurred in two
cases, for an epoxy pipe with con/cylindrical joint after about 12.5 minutes and for a
vinylester pipe with a muff cemented joint after 34 minutes. The con/cyl joint design
may be more critical from a fire technical point of view compared to the con/con con-
figuration because as degradation progresses form the outer surface, the entire
cemented area will be exposed simultaneously. The thickness of the wall outside the
Scandpower
Risk Management AS {G)) SINTEF
cemented area seems to be a critical factor. A thicker epoxy pipe with con/cylindrical
cemented joint withstood a test period of 35 minutes.
Unprotected rubber seal lock joints with thermoplastic locking strips showed as expec-
ted poor performance in this test and failed after exposure times of only 2 and 4
minutes respectively. Testing of both plain pipes and jointed pipes in the stagnant
water condition proved similar to other tests in that slight weeping started after
approximately 15 to 25 minutes and that this weeping usually did not increase notice-
ably upon continuation of the test indicating establishment of a quasi steady state con-
dition.
An experience from observing these tests was also that the smoke development was
fairly low and mainly limited to the initial period where the outer pipe surface degrades
or carbonises. Empty pipes were tested for up to 7 minutes. After the test the ends of
the pipes which had not been directly exposed in the fire remained unaffected indica-
ting insignificant flame spread along the pipes during and after the test.
It is important to be aware of the fact that even if it is not required to protect the pipe
itself in a fire, it might be necessary to protect other parts of the pipe system. Flanges,
blind flanges, valves, joints and test connections are typical examples of items that
require special attention with respect to fire protection.
A commercial finite element code, ABAQUS, has been used at SI to develop a model
for simulation of fire testing of GRP pipes, Ref. /8-46/. It is, however, possible to use
other commercial finite element codes, which can solve the heat transfer equation with
temperature dependent material properties and time dependent boundary conditions.
The model can calculate the temperature history of a GRP pipe, tank, etc. with and
without external protection exposed to different levels of fire loading. The temperature
history is simulated as a typical solid body heat transfer problem with general, tempe-
rature dependent conductivity, internal energy and general convection and radiation
boundary conditions. Input to the model is temperature history of the fire. The initial
temperature and the heat transfer coefficient between the different materials must also
be given. The model is intended for use as a tool prior to fire testing in order to
minimise the costly experimental work required to develop optimum solutions.
Comparison between computer simulations and experimental results shows that the
model gives results that correspond reasonably well with experimental results.
A sensitive analysis shows that the calculated temperature development on the outer
surface of the GRP pipe is sensitive to a variation of the thermal material properties.
The temperature development on the outer surface of the GRP pipe is also sensitive
to a variation of the heat transfer coefficient between the GRP pipe and the fluid inside
the pipe. This indicates that the results from the model should be used with some
care, since it is hard to find the real material properties experimentally.
An example of the possibilities of using such a model is given in Appendix 2 for a GRP
pipe insulated with 12 mm Kaowool (ceramic fibre).
Another application of the model is to combine a heat load applied to a structure with
the structure's mechanical properties as a function of temperature. Hence, the resi-
dual strength of the structure at a given time can be calculated.
Scandpower
/8-1/ P.A. Croce, K.S. Mudan: "Calculating Impacts of large open Hydrocarbon
Fires", Fire Safety Journal, 11 , 1986.
/8-2/ K.S. Mudan: "Fire Hazard Calculations for large open Hydrocarbon Fires", The
SFPE Handbook of Fire Protection Engineering, ISBN: 0-87765-353-4, Sep-
tember 1988.
/8-3/ J.H. Veghte: "Designing protective Clothing, Part 2, Fire Service Today", 49
(4), 1982.
/8-4/ American Petroleum Institute: "Guide for Pressure-relieving and Depressuring

Systems", API RP 521, Washington DC, USA, September 1982.
/8-5/ F.P. Lees: "Loss Prevention in the Process Industries", Vol. 1, Butterworth &
Co. (Publishers) Ltd., 1980.
/8-6/ Spieth et al.: "A combined Hazard Index Fire Test Methodology for Aircraft
Cabin Materials", Vol. I and II, Dep. of Transportation , Atlantic City, April 1982.
/8-7/ A.J. Pryor: "Full-Scale Evaluation of the Fire Hazard of interior Wall Finishes",
Southwest Research Institute, San Antonio, Texas, 1968.
/8-8/ A.J.M. Heselden: "Studies of Fire and Smoke Behaviour relevant to Tunnels",
Fire Research Station, Borehamwood, England, November 1978.
/8-9/ National Fire Protection Association: "Fire Protection Handbook", 14th Edition,
Boston, Massachuttes, USA, 1976.
/8-10/ H.L. Kaplan, A.F. Grand and G.E. Hartzell: "Toxicity and the Smoke Problem",
Fire Safety Journal, 7, 1984.
/8-11/ T. Jin: "Human Behaviour in Fire Smoke", page, 324, Preliminary Report,
Proceedings of the 4th Joint Panel Meeting, UJJNR Panel on Fire Research
and Safety, Tokyo, 1979.
/8-12/ J.G. Quintiere: "Smoke Measurements: An Assessment of Correlations bet-

ween Laboratory and Full-Scale Experiments, Fire and Materials", Vol. 6, No. 3
and 4, 1982.
/8-13/ Benjamin/Clarke Associates, Inc.: "Smoke and Life Safety", prepared for the
Society of Plastic Industry, August 1983.
/8-14/ R. Friedman: "Quantification of Threat fromm a rapidly growing Fire in Terms

of relative Material Properties, Fire Materials", 2, p. 27, 1978.
/8-15/ C.J. Hilado, H.J. Cumming: "A Review of available LCso Data", Fire Safety
Centre, University of San Francisco, California, 1977.
/8-16/ R.E. Reinke, C.F. Reinhardt: "Fires, Toxicity and Plastics, Modern Plastics",
February 1973.
/8-17/ J.L. Bryan: "Damageability of Buildings, Content and Personnel from Exposure
to Fire", Fire Safety Journal, 11 , 1986.
Scandpower
/8-18/ G.E. Hartzell: "Evaluation of Toxic Hazards in Fires", Southwest Research

Institute, San Antonio, Texas, 1986.
/8-19/ G. Kimmerle: "Aspects and Methodology for the Evaluation of Toxicological

Parameters during Fire Exposure", The Journal of Fire and Flammability Com-
bustion Toxicology Supplement, Vol. 1, February 1974.
/8-20/ Personnel communications with 0. H0yland who has carried out full scale
experiments with furnitures, SINTEF NBL - Norwegian Fire Research Labora-
tory, Trondheim, Norway.
/8-21/ B.C. Levin, M. Paabo, J.L. Gurman, S.E. Harris: "Effects of Exposure to single
or multiple Combinations of the Predominant Toxic Gases and low Oxygen
Atmospheres produced in Fires, Fundamentals and applied Toxicology", 9,
1987.
/8-22/ D.A. Purser. "Modelling Toxic and Physical Hazard in Fires", 2nd Int. Symp. on
Fire Safety Science, Tokyo, Japan, 1988.
/8-23/ M.M. Birky: "Hazard Characteristics of Combustion Products in Fires: The

State-of-the-Art-Review", Centre of Fire Research, National Bureau of Stan-
dards, Washington, 1977.
/8-24/ Norwegian Standard NS 3472: Steel Structures. Design Rules. 3rd • Edition,
September 2001.
/8-25/ Martin McD. Grant: "Quantitative Risk Assessment: Likely Error Ranges and
the Need for a Multi Diciplinary Approach", Management and Engineering of
Fire Safety and Loss Prevention, Onshore and Offshore, 1990.
/8-26/ Y. Anderberg: "Behaviour of Steel at high Temperatures", Rilem-Committee

44-PHT, GIB W14/82/44 (S).
/8-27/ "European Recommendations for the Fire Safety of Steel Structures. Calcula-
tion of the Fire Resistance of Load Bearing Elements and Structural Assem-
blies exposed to the Standard Fire", EGGS-Technical Committee 3 - Fire
Safety of Steel Structures, European Convention of Constructional Steelwork,
1983.
/8-28/ Eurocode No. 3: "Design of Steel Structures. Part 1.2: Structural Fire
Design", Draft pr ENV 1993-1-2, May 1993.
/8-29/ DRAFT ENV YYY5 part 4: "1995 Fire Tests on Elements of Building Con-
struction. Test Method for determining the Contribution to the Fire Resistance
of Structural Members: by applied Protection to Steel Structural Members",
Document GEN/TC 127 - N949.
/8-30/ Nordtest Method NT Fire 021: "Insulation of Steel Structures: Fire Protection".
/8-31/ SUPER TEMPCALC: "User's Manual for TCD 3.0 with Tempcalc®", Fire
Safety Design, Lund, Sweden, December 1990.
/8-32/ TASEF-2: "TASEF Ver. 3.0 User's Manual", Swedish National Testing Institute,
Boras, Sweden, February 1988.
Scandpower
I!) SINTEF Risk Management AS
/8-33/ U. Danielsen, T.A. Hammer, H. Justnes, S. Smeplass: "Marine concrete Struc-

tures exposed to Hydrocarbon Fires", SINTEF NBL - Norwegian Fire Research
Laboratory, STF25 A88064, Trondheim, December 1988.
/8-34/ Eurocode No. 2: "Design of concrete Structures. Part 1.2: Structural Fire
Design", Commission of the European Communities, pr ENV 1992-1-2,
October 1993.
/8-35/ U. Schneider: "Properties of Materials at high Temperatures: Concrete",

Gesamthochschule Kassel Universitat, Kassel, 1985.
/8-36/ H. Thon, R. Stokke: "Performance of GRP Pipes in Fire Situations", SI-Report

No. 850202-2, Senter for lndustriforskning, Oslo, April 1987.
/8-37/ H. Thon, R. Stokke: "Technical Booklet - GRP Offshore", SI-Report No.

850202-6, Senter for lndustriforskning, Oslo, September 1988.
/8-38/ B. Sundstr0m: "Plastic Pipes - Testing for Fire Endurance and Flame Spread
Properties", SP-RAPP 1985:31 , Statens Provningsanstalt, Baras, 1985.
/8-39/ J. Hunter, K.L. Forsdyke: "Phenolic GRP and its recent Applications", BP
Chemicals Int., Sully, England.
/8-40/ H. Thon, R. Stokke: "Behaviour of GRP Tanks in Fire Situations, SI-Report No.
850202-5, Senter for lndustriforskning, Oslo, May 1988.
/8-41/ R. Stokke: "Use of GRP in Sea Water Pipe Systems Offshore", OTC 5744,
Houston, Texas, May 2-5, 1988.
/8-42/ H. Thon, R. Stokke: "Static Electricity in Association with Glass Fibres reinfor-
ced Plastic (GRP) Pipes and TanksNessels", SI Report No. 850202-4, Senter
for lndustriforskning, Oslo, May 1988.
/8-43/ H. Haanes, H. Thon: "Properties related to Fire Performance of Glass Fibre

reinforced Plastic (GRP) Materials, SI Report No. 890210-4, Senter for lndu-
striforskning, Oslo, October 1988.
/8-44/ H. Thon, H. Haanes, R. Stokke: "Glass Fibre reinforced Plastic (GRP) Pipes in
the Fire Water System Offshore", SI Report No. 890210-3, Senter for lndustri-
forskning, Oslo, October 1988.
/8-45/ R. Stokke: "Glass Fibre reinforced Plastic (GRP) Offshore - Final Summary
Report", SI Report No. 850202-7, Senter for lndustriforskning, Oslo, Septem-
ber 1988.
/8-46/ S.H. H0ydal: "Computer Simulation Model for predicting the Temperature
Development of Fire exposed GRP Pipes", SI Report No. 890210-3, Senter for
lndustriforskning, Oslo, October 1988.
Scandpower
9. Fire Protection
9.1 Passive Fire Protection
9.1.1 Introduction and Philosophy
Passive fire protection is any built-in action in order to improve the fire resistance of
partitions, loadbearing structures and equipment. The most common type of passive
fire protection is an insulation to limit temperature rise in structures and equipment in
case of fire.
Fire Protection Philosophy
The requirements for sufficient protection against fire on an offshore installation or in

an industrial plant are directly related to the actual fire potential in the area. On any
installation or industrial plant, passive fire protection must be arranged in such a way
that fire will be prevented from spreading to other areas (confinement of fire) and to
the loadbearing structure.
Separation of areas by fire resisting divisions should be done in order to
limit spread of fire

reduce size of fire area which in turn limits requirements to the capacity of the
active fire protection system
protect adjacent structures
protect escape routes.
In general, fire resisting divisions shall not be used as a justification for reducing dis-
tance between areas which normally should be well separated, but shall be a means
to maintain the total safety when such separation of areas is inadequate.
Fire Classes
Partitions are divided into different classes described by a letter (B, A or H) and a
number (0, 15, 30, 60, 90, 120 etc.)
The letter describes the type of construction and according to which fire exposure
curve the element has been classified.
The number denotes the fire endurance time in minutes for which the element has
satisfied the criteria of stability, integrity or insulation when exposed to the specified
fire exposure curve. The least endurance time to satisfy one of these criteria is
the fire endurance time for the element. However, the minimum endurance time
according to the criteria of stability and integrity of partitions is equivalent to the pre-
scribed duration of the fire tests.
In terms of fire classification, structures, partitions and equipment (with passive fire
protection) must meet certain requirements. The primary criteria of stability, integrity
and insulation must be satisfied throughout a prescribed time of heat exposure (see
Chapter 8.1 ).
Scandpower
9.1.2 Separation by Distance

Separation by distance is in many ways the most simple method of fire protection.
The main principle is to separate all critical process equipment by specifying sufficient
distances to any possible fire caused by failure of a process equipment so that the
incident radiation flux onto relevant process equipment is below a critical value. For
example, Institute of Petroleum, Ref. /9-1/, reports maximum radiation heat flux of 44
and 8 kW/m2 for thermally protected and unprotected pressurized storage vessels,
respectively. Hence, a fire due to a leakage of hydrocarbon fuel from any process
equipment in the process area must not cause higher heat fluxes onto the nearest
process equipment for a given assumed safe separation distance.
The spread of a liquid hydrocarbon from a storage vessel or process train is usually
limited by sufficiently high curbs surrounding the vessel or process train. Alternatively,
the storage vessel or process train may be located within a dike. In this way the maxi-
mum lateral size of the fire is known and the maximum radiation heat flux as a function
of distance from the fire is easy to find from calculation methods presented in Chapter
7 combined with Appendix Hand I. In case of gas releases, Chapter 6 combined with
Appendix H and I, present methods for prediction of the heat flux as a function of
distance from the resulting gas fire.
If thermal radiation on human beings exposed for several minutes is considered, an

incident heat flux of 1.5 - 6 kW/m 2 (dependent on degree of protective clothing) is
often assumed to be the maximum permissible radiation intensity (see Chapter 8.2.1).
When considering safe separation distances for process equipment or human beings,
only the radiative heat transfer has to be taken into account. In Appendix H two
methods are presented for prediction of the radiative heat transfer to external targets.
First we have the "Point Source Model" which is rather simple to use. The incident
heat flux to a target located a distance x from the fire (virtually from the centre of the
fire plume) is given by the equations (H.9) and (H.10). For a given maximum accep-
table heat flux for a storage vessel q,max, the safe separation distance, Dsate, can be
predicted from Eq. (H.9) when setting: Dsate = x. The following expression will result:
_ fR mt £\He
Dsafe- Eq. (9.1)
41tq"r,max
where:
fR = radiative fraction which is given by Eq. (J.5) or experimental data for

buoyancy controlled diffusion flames (e.g. pool fires and diffusive gas
fires) and by Eq. (6.45) or Table 6.2 (-)
m, = burning rate when assuming a combustion efficiency Tlett = 1. Other-
wise, the burning rate has to be multiplied with an TJett < 1 (kg/s)
£\He = heat of combustion for the burning fuel (kJ/kg)
As accounted for in Appendix H, this method tends to overestimate the heat flux in the
near field . Hence, for a given heat flux, q,sate, the safe distance squared, Date, will be
correspondingly underestimated and also Dsate, although to a much less extent. This
is the reason for the warnings put forward in Appendix H when considering safe dis-
tances for process equipment. Too short separation distances may be predicted
which yield higher incident heat fluxes than presupposed. Consequently, this method
is not recommended for use on offshore installations where safe separation
Scandpower
Risk Management AS (GJ} SINTEF
Norwegian Fire ResearchLaboratory
distances for process equipment have to be predicted due to the small areas on
such installations.
When considering safe radiation distances for human beings, which have to be much
larger than in the case of process equipment, Eq. (9) is assumed to be satisfactory. In
this case the atmospheric transmissivity, , , may be multiplied with the nominator of
Eq. (9) in order to take into account atmospheric absorption of radiation by the humi-
dity in the air. , is given by Eq. (H.13). However, the distance, x, from the fire is inclu-
ded in this equation. First the atmospheric transmissivity has to be predicted by
assuming a reasonable separation distance. If the error is too large, a different dis-
tance has to be selected. A trial and error procedure has to be followed until a rough
agreement is achieved. The atmospheric transmissivity is not particularly sensitive to
the distance from the fire.
The other method for calculating the heat flux onto external targets presented in
Appendix H, is the so-called "Solid Flame Model" which is given by Eq. (H.11). This
method is reasonably accurate also in the near field of the fire as well as in the far
field. Here the atmospheric transmissivity can be set equal to unity.
Due to the rather complicated expressions given in Appendix I for prediction of view
factor F in Eq. (H.11), it is impossible to achieve an analytic expression for the safe
distance, Dsare, such as given by Eq. (9.1). It is recommendable to use Eq. (H.11)
combined with the use of the charts given in Figures 1.2 and 1.3 in Appendix I, for pre-
dicting the view factors in order to find the minimum separation distance that yields the
maximum permissable heat flux.
When an acceptable heat flux at a given distance is found, a more accurate calcula-
tion can be carried out by using the expressions in Appendix I in order to find a more
accurate estimate of the view factor. The maximum view factor, which is the vectorial
sum of the horizontal and vertical view factor, is recommended. This will lead to a
conservative estimate.
9.1.3 Fire Partitions

Fire partitions are any structural divisions, e.g. vertical walls or bulkheads and horizon-
tal decks with a fire rating.
It is worth mentioning that there might be some differences between the national stan-
dards of acceptance criteria for fire rating of partitions in the building industry and the
international acceptance criteria for fire rated partitions for use at sea, e.g. on board
ships and on offshore platforms. The differences are mainly found for B-classed par-
titions and fire rated doors.
Fire Classes
Fire rated partitions are divided in three different classes:
B-Class: The partition shall maintain its stability, and prevent penetration of flames
and hot fumes for a period of 30 minutes when tested according to the
standard fire time/temperature-curve. The average and maximum tempe-
rature. rise on an unexposed face of the partition shall, within the classi-
fication time, not exceed 140 °c and 225 °C respectively
Scandpower
A-Class: The partition shall maintain its stability, and prevent penetration of flames
standard fire time/temperature-curve. The average and maximum tempe-
rature rise on an unexposed face of the partition shall, within the classi-
fication time, not exceed 140 °c and 180 °c respectively
H-Class: The partition shall maintain its stability, and prevent penetration of flames
hydrocarbon fire time/temperature-curve. The average and maximum
temperature rise on unexposed face of the partition shall, within the classi-
fication time, not exceed 140 °c and 180 °C respectively
Example: A H-60 partition has been tested for 120 minutes according to the hydro-
carbon fire time/temperature-curve, and during the period of testing there
were no penetration of flames or hot fumes. During the first 60 minutes of
the testing the average and maximum temperature rise on the unexposed
face of the partition did not exceed 140 °c and 180 °C respectively.
A H-0 partition has been tested for 120 minutes according to the hydro-
carbon fire time/temperature-curve, and during the period of testing there
was no penetration of flames or hot fumes. The number O denotes that
there are no insulation requirements, but in order to satisfy the stability
criterion throughout the period of testing, the structural core has to be
protected. The necessary fire protection will depend on the structural load
and the critical temperature (Tcr) of the core.
There may be given additional requirements concerning the contents of combustible

materials and the materials used in the structural core of the partitions.
The relevant test methods are:
Combustibility test: IMO Res. A472 (XII), Ref. /9-2/, which corresponds to ISO
1182 Ref. /9-3/
Fire endurance test: IMO Res. 517 (XIII), Ref. /9-4/, which corresponds to ISO
834, Ref. /9-5/.
Fire partitions may be constructed in different ways, from a dense concrete wall or
deck to complicated system walls or decks with different insulation materials.
When passive fire protection is provided, the material is normally applied to the side of
the partition, which will be facing the fire. The partitions may be designed for a one-
sided exposure (fire only onto the specified side) or a two-sided exposure (fire onto an
arbitrary side of the partition).
Spray applied fire protection normally requires a retention net of galvanised or PVC-
coated steel mesh, secured to the surface by different types of welded studs or pins.
The retention net is fixed to the primed surface prior to the fire proofing, and is nor-
mally positioned at mid-depth in the insulation. A spray applied fire protection should
be regarded as a system, since the system consists of a priming procedure, a reten-
tion net and fixing device, the passive fire protection material and finally a top coat
system to seal off the surface.
Scandpower
B-class partitions are normally unloaded walls made of a steel-sheet cassette system
with mineral wool between the sheets.
A-class and H-class partitions are loaded or unloaded walls or decks with a structural
core. The core may be protected by insulation board or blankets, by a spray-applied
insulation or by a combination of different insulation materials.
A sufficient fire resistance may also be achieved by e.g. waterfilling of a hollow steel
structure.
Examples of different fire rated partitions are given in lists of classified products from
the approving authorities and classification companies. Examples of such lists are the
"List of equipment approved by the Norwegian Maritime Directorate", Ref. /9-6/, and
"Type Approved Products" from Det Norske Veritas, Ref. /9-7/.
As a general requirement, penetrations (cables, ducts or pipes) through fire rated par-
titions shall not deteriorate the partition. The same requirement is valid where the
partitions are penetrated by doors and windows. That is, cable- and pipe-penetra-
tions, door- and shutter-assemblies and windows have to be classified if intended for
use in fire partitions. The criteria for doors and windows are the same as for fire par-
titions. For cable- and pipe-penetrations the insulation criteria on an unexposed side
of the penetration is limited to the maximum temperature rise.
Examples of different fire rated doors, windows, cable- and pipe-penetrations are
given in lists of classified products from the approving authorities and classification
companies, Ref. /9-6/ and /9-7/.
Note:
It has to be documented that the protection materials will endure the gas temperatures
that are probable in the fire.
In some situations a fire partition may be subjected to relatively low temperatures due
to a limited or remote fire. For some intumescent materials it is in such situations
important to know the activation temperature for the material, and what will happen if
the material is subjected to these low temperatures in terms of expansion, possible
melting, smoke generation etc. A slight fire exposure may be followed by an intense
fire, and the material must not be deteriorated during the first stage of the fire.
9.1.4 Structural Fire Insulation

Here fire insulation is defined as passive fire protection materials applied to different
types of structures to limit structural and other consequence damage by fire.
Fire Classes
Fire insulation alone cannot be given a fire rating. All classifications of fire insulation
materials concerning fire endurance are strictly connected to the use of the material.
The fire rating is given to the insulated element, which consists of a structural core
(e.g. steel), and the fire insulation. Fire rated loadcarrying elements are divided into
two different classes:
Scandpower
(G)iSINTEF Risk Management AS
A-Class: The element shall maintain its stability when tested according to the stan-
dard fire time/temperature-curve. The stability criterion is defined by a
critical temperature for the structural core, which the tested element shall
not exceed during the specified period of testing time. For insulated steel
structures this critical temperature is normally in the range 350-550 °c
H-Class: The element shall maintain its stability when tested according to the
hydrocarbon fire time/temperature-curve. The stability criterion is defined
by a critical temperature for the structural core, which the tested element
shall not exceed during the specified period of testing time.
There is no specified period of time for the fire test. The test may last for 30 minutes
or 4 hours, depending on the fire rating which is aimed for. Normally the fire ratings
are A-30, A-60, H60 and H120.
Example: A H-120 steel structure has been tested for 120 minutes according to the
hydrocarbon fire time/temperature-curve, and during the period of testing
the steel temperature did not exceed the critical temperature.
Different methods of presenting classified insulated steel structures are given in

Chapter 8.3.3.
Additional requirements may be defined regarding the combustibility of the materials.
The relevant test methods are:
Combustibility test: IMO Res. A472 (XII), Ref. /9-2/, which corresponds to ISO
1182, Ref. /9-3/
Fire endurance test: NT Fire 021, Ref. /9-8/, which corresponds to ISO 834,
Ref. /9-5/.
Structural fire protection may be applied to the member in a variety of ways using
spray applied materials, boards or blankets. The material may be formed as a profile
insulation, a box cladding or a solid insulation. See Figure 9.1.
Scandpower
Risk Management AS (t",J) SINTEF
Profile Box Solid

(with orwltlout
9ar>over~angos)
Protection techniques for three sided protection
Proflle Box Sold

(wllhorwllhoot
gap o~••nse•)
Protection techniques for four sided protection
Figure 9.1: Protection Techniques for Structural Fire Insulation
Profile insulation is most common for the different spray applied systems, due to the
increased durability of the material when directly applied to the steel surface.
9.1.5 Materials for Passive Fire Protection

There are a number of different types of fire protecting materials. Some materials
base their insulating capacity on low thermal conductivity alone, others on a combina-
tion of low thermal conductivity and high specific heat and yet others on high specific
heat alone.
One group of materials (intumescent epoxies) is dense coatings at normal tempera-

tures, but react and expand to form a protective cellular char during fire exposure.
Note:
Instrument epoxies may behave differently on different steel shapes. I.e. thicknesses
for different ratings are different for RMS sections and /-sections.
All passive fire protection materials have to go through certain classification proce-
dures in order to achieve approvals from the authorities. Such procedures may be
different furnace tests, jet fire tests and/or calculations. Passive fire protection may
have a number of fields of operation, and must be tested as such in each case.
Passive fire protection materials are either combustible or non-combustible, accor-

ding to a specified test procedure.
NPD are now developing acceptance criteria for use of combustible materials off-
shore.
Scanclpower
itj)SINTEF Risk Management AS
Norwegian Fire Research Laboratoty
Insulation materials for fire protection may be divided into different types of insulation,
based on method of application.
Table 9.1 gives some examples of the most common insulation types and some trade
names within each group:
From "Regulations concerning explosion and fire protection of installations in the

petroleum activities" issued by Norwegian Petroleum Directorate (NPD), 1992 (Ref.
/9-9./) the following requirements relating to passive fire protection are given:
IV-3 REQUIREMENTS RELATING TO PASSIVE FIRE

PROTECTION
Section 19
General Requirements relating to Passive Fire Protection
The passive fire protection shall be such as to ensure that relevant
structures or equipment components have adequate fire resistance
with regard to loadbearing properties, integrity and insulation proper-
ties during a dimensioning fire, and such as to contribute to reducing
the consequences of fire in general.
The fire loads that may o=ir shall be determined by calculations.
Section 20
Fire Technical Requirements relating to Materials
Materials on the installation shall as a rule be non-combustible.
If in a total evaluation it is deemed justifiable from a safety point of
view to make use of materials that do not meet the requirements to
incombustibility, such materials shall have limited flame spread
properties, low smoke development and low heat generation.
An assessment shall be made of the toxicity of gases emitted in the
event of a fire.
Documentation shall be available to support the basis for the decision
regarding choice of materials.
Table 9. 1: Examples of some commonly used Insulation Types and some Trade
Names w1"th·1n eac h Group
Main insulation group Types of insulation Trade names
1
Intumescents CHARTEK Ill, CHARTEK IV
PITTCHAR XP
THERMO-LAG 440
Cementicious Dossalack HCD
Spray applied materials Mandolite 550
Pyrocrete 201
Fibre insulation Blaze Shield D C/F
Contra Flam
Kerlane 45 BAP 620
Fibre boards Rockwool Boards
Conlit
Keranap 40
Boards Dense boards Skamolex
Promatect H,L
Monolux
Vermiculux
Mineral fibre blankets RockwoolBatts
Ceramic fibre blankets Durablanket
Blankets TRITON Kaowool
Kerlane 45
FIREMASTER Blanket X607
1) lntumescent materials are materials that expand when exposed to fire. There are various types of intumescents.
Here only epoxies are mentioned. A number of different paint types are omitted. CHARTEK and PITT CHAR are
intumescents, whereas TERMO-LAG 440 is a subliming material
Scandpower
Risk Management AS (r,J) SINTEF
Examples of different fire rated structural fire insulation are given in lists of classified
products from the approving authorities and classification companies. Examples of
such lists are the "List of equipment approved by the Norwegian Maritime Directorate",
Ref. /9-6/ and "Type Approved Products" from Del Norske Veritas, Ref. /9-7/. Different
methods of application may be found in the data sheets given in Ref. /9-10/.
9.1.6 Environmental Effects/Explosion Effects on Passive Fire Pro-

tection
General
Passive fire protection materials, both structural and for partitions, are fire-tested
according to internationally accepted methods, and classifications are given based on
the test certificates and in some cases supplementary calculations.
However, fire is only one type of load the materials probably will be subjected to
during their lifetime. All materials will be subjected to mechanical wear and tear, and
materials for exterior use will in addition be exposed to environmental effects.
Other types of accidental load, like jet fires and explosions, may also be relevant.
Environmental Effects
The materials are normally tested under optimal conditions both regarding bond
strength, quality of the retention net and curing conditions of spray applied materials.
In addition the materials are new and undamaged.
For exterior applications, especially in marine climate (offshore conditions), unfavour-

able curing conditions for spray applied materials, environmental effects, mechanical
wear and tear and poor durability of the material may impair both the mechanical
strength and insulation properties of the passive fire protection materials.
A survey of different offshore installations in the North Sea has shown poor conditions
for some originally applied exterior passive fire protection materials.
The main types of damage to the insulation caused by environmental effects are:
Damage to the top-coat followed by wetting of the insulation and erosion

Corrosion of unprotected steel retention net and fixing points to the steel sur-
face.
The corrosion damages are found only in cementicious type spray applied materials.
This may be a serious problem in the case of fire, since the corrosion may have de-
stroyed the net completely. The material may thus fall off long before the fire rating
time.
Fire tests carried out on different spray applied cementicious materials exposed to
years of North Sea climate show no impaired insulation capacity of the material itself.
Note:
When choosing fire protection materials for exterior use, the environmental effects on
the material should be paid as much attention as the fire characteristics.
Scandpower
Actions should be taken to minimise the negative effects on the material from the
environment.
Damage in the top-coat on exterior spray applied materials should be repaired as

soon as possible to prevent ingress of water (sea water) which may cause corrosion.
Where the retention net and/or studs are damaged by corrosion, the fire protection
system should be renewed or secured by means of bolts, nets, panels etc.
Explosion Effects
Statistics based on accidents, which involve fires and/or fires in offshore installations,
show that fires and explosions often appear together. A typical situation, in case of a
critical fire, is an explosion and a subsequent fire.
A number of test series have been carried out by Haverstad, Ref. /9-11/, in order to
find out
how gas explosion loads affect different types of fire protection

what types of damage are typical
how the damage affect the thermal insulation of the structure.
The explosion tests showed that the structural response was the most important para-
meter regarding damage to the protection. Only secondary responses to the direct
effects of the explosion (explosion wind, over- and underpressure, etc.) were obser-
ved. Typical damage included cracking, crushing and bond failure, Ref. /9-11/.
Later, protected beams and plates were tested for a simulated explosion load (struc-
tural response forced on by a hydraulic jack) and a subsequent HC-fire. These tests
showed that the simulated explosion load caused damage to the fire protection in the
same way as in the gas explosion tests. The fact that this method is simpler, easier to
control and cheaper to carry out compared to gas explosion tests, led to the recom-
mendation to use this method as much as possible. The subsequent fire tests showed
the effects of the damage to the thermal insulation.
The design philosophy and the experimental work carried out by Haverstad, Ref.
/9-11/ formed the basis of the following results:
Specifications for passive fire protected structures subjected to gas explosions

and subsequent fire loads
Test methods for passive fire protected structures subjected to gas explosions
and subsequent fire loads
Methods of calculation to prevent considerable damage to the fire protection in
case of an explosion load.
The specifications consist of
a functional requirement
documentary requirements
approval requirements.
The specifications give supplementary requirements to the already existing regulations

for fire and fire testing. In principle they are independent of the structural material and
the type of fire protection.
Scandpower
Risk M anagement AS (tj) SINTEF
The functional requirement is the superior one, which states that the fire protection is
to resist the actual explosion load and still satisfy the functional requirement as fire
protection.
Compliance with the functional requirement is to be documented. This documentation

is to include a report from gas or simulated explosion tests and subsequent fire tests.
If it is impossible to represent all the details of a protective system on the recommen-
ded test specimens, it is allowed to document the serviceability of the omitted parts by
calculations.
The approval requirements for types of passive fire protection which have been sub-
jected to explosion and fire loads, are the criteria for the temperature rise for the actual
fire class given in official regulations. The approval is valid for the class of fire, which
is tested, and within the tested level of the structural response and/or blast load. A
condition for the approval is that the damage to the protection is not to be increased
because of the direct effects of the explosion if simulated explosion tests are used.
The test methods include both explosion and fire tests. The experimental work
showed that a simulated explosion load is representative for some types of protection
(spray applied protection materials). If not, gas explosion tests have to be done. The
subsequent fire test is to follow the procedure of the standardised fire test.
9.2 Active Fire Protection

9.2.1 Introduction
The mechanisms of extinction of flames in a fire situation are several, and a success-
ful fire fighting often depends on more than one mechanism. The main mechanisms
of extinction may be grouped in six different physical processes, Ref. /9-12/:
Cooling of the flames to a temperature where the chemical reactions can not be
sustained
Reduction of oxygen and/or fuel concentration to a level where no combustion
can be sustained
Speeding up flow velocity to a level where the residence time of fuel and oxygen
in combustible mixture is shorter than the time scale of the chemical reactions
Adding components in the combustion zone, which break the chain of chemical
reactions by substitution with endothermal reactions.
Cooling the combustible item to a temperature where gasification rate is below
the limit for sustainable combustion
Adding components (liquid or solid) to the surface of the combustible item that
increases the heat of gasification. Wetting by water is one example.
All mechanisms may be related to the so-called fire triangle, where one of the three
sides represents one main requirement for sustained fire. This is shown in Figure 9.2.
A sufficient supply of fuel, a sufficient amount of heat, either produced by the fire itself
or supplied from a source and an amount of oxygen are required.
Scandpower
HEAT
Figure 9.2: The "Fire Triangle"
If one of the requirements is missing, the fire will be extinguished. Table 9.2 describes
some of the most frequently used fire suppression methods, and gives indications of
the active mechanisms. Some of the most important physical phenomena, which
should be described in attempts to model the suppression, are also listed. The table
is mainly a reproduction from Pietrzack and Ball, Ref. /9-13/, with small alterations
done by the author.
Table 9 2 Fire Suppression Methods and important related Physical Phenomena

Extinguishing agent Fuel Important phenomena Comments
Charring and non- Water spray: Used for compartment
charring solid mate- - Wetting of combustibles, pre- fire, pre- and post flash-
rials venting fire spread over
- Induced flow and gas entrain- Flaming fires
ment Deep seated fires
- Drop dynamics Outside fires
- 'Water curtain", partitioning
Water spray Evaporative cooling of:
- Hot gases
- Fire plume
- Burning and non-burning
surfaces
Steam generation:
- lnerting effect
- Blocking air flow
Scandpower
.
~hguishingiigenf .·. ,,:!, ,!;' Fiiel:\:,. ;/,< :,:; • · 1mportanf.pbeiiomena:,: •··:;;/ ¢omments• .. . .
Flammable liquids Same as solid materials, except Used for pre- and post
from char combustion. flashover flaming com-
In addition for partment fires and
1. Water-miscible liquids: outside fires
- Diluting of liquid surface to
bring below flash point
2. Water-immiscible liquids:
- Drop penetration into liquid
surface
- Heat transfer to water drops
while in liquid to cool below
flash point
3. Water immiscible liquid with
low flash point
- Reduction of fuel evapora-
tion by gas/plume cooling
only
Flammable liquids or Cooling of hot gases by evapo- Used as volume pro-
gases. ration of water. tection in compartment
Used for solid lnerting effect by steam produc- fires.
Water mist materials protection lion, diluting oxygen and fuel Used as local applica-
when water is limited vapour. tion fire suppression
or water damage is Radiation blocking by small systems, mainly in
critical droplets enclosures
Flammable liquid Fixed systems: Used for pre-flashover
fuels - Induced flow compartment fires and
- Foam partitioning outside fires
Surface distribution and

spreading
Cooling of fuel surface below

vaporization temperature
Foams
Barrier to:
- Vapour transport
- Heat transfer
Steam generation:
- inerting effect
- Blocking of air flow
Charring and non- Same as flammable liquid fuels,
charring solid but in addition char combustion
materials cooling
Ail fuels Gas injection: Used for pre-flashover
- induced flow compartment fires
- Mixing involving flaming
combustion.
Vent sealing:
- Impact on flow Total flooding systems
- Gas/air leakage only.
Inert gases
Combustion inhibition by Local application of CO2
oxygen dilution.
Natural cooling of fuel and other

surfaces.
Agent cooling of flame and fuel

surfaces
Scandpower
Extinguishing agent Fuel Important phenomena Comments

All fuels Gas injection: Used for pre-flashover
- Induced flow compartment fires.
- Mixing
Total flooding systems
Vent sealing: only.
Chemically reacting gases - Impact on flow
(halons) - Gas/air leakage With gas inhibiting
flames
Combustion chain reaction
inhibition.
Natural cooling of fuel surfaces

Flammable liquids Powder injection
- Particle dynamics
- Induced flow
- Partitioning
Vent sealing and effect on com-

partment ftowfield.
Chemically inhibiting Thermal radiation attenuation

powders and shielding.
Combustion chain reaction

inhibition.
Natural cooling of surfaces

Charring solid Sarne as above, but in addition: Melting powders
materials
Fuel coating acting as air barrier
9.2.2 Properties of Water and Common Fire Extinguishing Media

Water is one of the most commonly used substances in fire suppression. The reason
is partly the physical properties, the availability and the environmental friendliness.
Compared to almost all other substances, water has a high specific heat capacity and
heat of vaporisation (per mass unit). The heat of vaporisation is about 6 times higher
than that of CO 2 . Specific heat capacity is almost twice as high as that of CO 2 and
higher than other inert gases and Halon alternatives used in fire suppression. This is
shown in Figure 9.3, Ref. /9-14 /.
Scandpower
2500 ~ - - - - - - - - - - - - - - - - - -~ 2,0
• I■ Hvap • cp vapour (20 C) I 1,8
2000 1,6
1,4
1500 1,2 ~
~
~
• 1,0 ~
a
1000 • • 0,8 ~
• • • • 0,6
500 • 0,4
0,2
0 L.-- -L.J_ 0,0
Figure 9.3: Heat of Evaporation and specific Heat capacity of some Fire Extin-
guishing Agents, Ref. /9-15 and /9-16/
Water is not an inert gas at ambient pressure at normal temperature, but becomes an
inert gas (water vapour, steam) by heat transfer from a heat source after release. This
is the main difference of water and other frequently used fire suppressants: All other
alternatives occur as gases after release into an atmosphere without a heat source.
This is shown by Figure 9.3, listing the boiling point of some frequently used fire sup-
pressants at atmospheric pressure.
Water, however, may exist as vapour even at ambient temperature, but at a maximum
volume fraction represented by its saturation pressure at the actual temperature. The
maximum saturation pressure of water as a fraction of the atmospheric pressure, at
temperatures between 0 - 100 °C, is shown in Figure 9.4. When water mist is relea-
sed into the atmosphere with or without a fire, it will evaporate and saturate the atmos-
phere if heat is supplied, but only to its saturation pressure. First when the atmos-
phere reaches the boiling point of water, 100 °C, water can exist as vapour (steam) at
100 % concentration. In a typical fire environment, the water mist (the droplets) will
evaporate in contact with the flames and the hot fire plume, and will re-condense and
form minor droplets in zones where the temperature decreases below 100 °c. These
droplets will collide and grow in size as they will in clouds, and after a certain time they
may rain out as droplets or be re-circulated into the hot zone.
Scandpower
G) SINTEF Risk M anagement AS
Saturation pressure of water in air at

atmospheric pressure
100
-- 80
I
I
~
60 /
s
ca V
-
D.
"C
D.
40
20 /
/
~
0
0 10 20 30
----
40
i..-----
50 60 70 80 90 100
Temperature (C)
Figure 9.4: Saturation Pressure of Water in Air at ambient Atmospheric Pressure

(1.013 bar)
Ta bl e 9..
3 B0I1ng . t of some exf1nguIs
T PQin . h.ing A,gens
t at Atmosp henc
· Pressure
Substance Boiling point at atmospheric pressure (°C)
IG-541 (lnergen) -196
CO2 -79
Halon 1301 -58
HFC-227ea (FM 200) -16
Halon 1211 -4
Water +100
9.2.3 Deluge/Sprinkler Systems

Deluge and sprinkler systems are used in process areas to suppress and extinguish
fires. The purpose of a sprinkler system is "to ensure early response to a fire and dis-
charge a specified amount of water over a designed area for a relevant time in order
to control/extinguish the fire". A deluge system, often called water spray installation,
has the purpose of releasing a predetermined pattern and amount of water over a
predetermined area for the purpose of control and/or to extinguish the fire, or to pro-
vide cooling to an installation which, if affected by heat (usually but not exclusively
radiated heat) would aggravate the situation by explosion, collapse, release of fuels or
otherwise. A third purpose of a waterspray system may be to provide a barrier to
inhibit the spread of fire.
The main difference between a sprinkler and a deluge system is the spray nozzle
working principle. The sprinkler nozzle is a combined detector, actuator and spray-
forming device which is activated by elevated temperature at the position of the
nozzle. A glass bulb or a fusible link is predetermined to break at a certain tempe-
Scandpower
rature level, and will spray water by the activated nozzles only. A deluge nozzle is
normally fully open all the time, and water supply is provided based on a detector
signal or manual action. The other possibility is that the system is pressurised or
water-filled, and the nozzles are opened over a predetermined area by a mechanism
at each nozzle. The activation of the system may then be based on smoke, flame or
other fire detection criteria, not only temperature.
The dimensioning criteria for sprinkler and deluge systems are based on detailed
guidelines, and the only part of the dimensioning which is really based on calculations
is the pipe work, pumps and other parts of the water supply system. There are no
commonly accepted formulas or correlations describing quantitatively the interaction
between the water spray and the fire, although some computer models have been
developed for this purpose. The basis for the dimensioning is partly tests and partly
experience from practice. Some calculations of parts of the spray/fire interaction may
be carried out, although a complete computer model is not available.
9.2.4 Sprinkler Installation Rules

There are standard specifications for installation and for components of sprinkler- and
water spray systems. A classification system for typical buildings and activities is nor-
mally part of the standard. A new harmonised European Standard for Fixed Fire-
fighting Systems is worked out by GEN (European Committee for Standardisation),
Ref. /9-17/. This standard is at present at the stage of Formal Vote. The sprinkler
standard also consists of a number of component standards, still in the process of
being finalised.
Process industry and offshore activity have traditionally used National Fire Protection
Association's rules. Similar procedures are described in NFPA 15, Ref. /9-18/.
A general procedure for dimensioning a sprinkler system can be summarised:
Decide on the building classification and activity. This will give a hazard class,
and a category of stored material
Determine required water application density, litres per minute per square meter,
(l/m 2 min), also denoted in millimetre per minute, (mm/min). For a given class
and category, a minimum "density of discharge" is defined. For sprinklers the
density of discharge varies between 2.25 mm/min for the lowest hazard class up
to 30.0 mm/min. NFPA 15 advises water application rates in the same order of
magnitude
Determine area of operation. For sprinkler systems this is the area which is
supposed to receive the minimum density of discharge with the supply system
and the pipes installed. This corresponds to a fixed number of sprinkler nozzles,
which are supposed to open in case of a fire. If the number of sprinkler nozzles,
which are activated in a fire, exceeds this number, the water discharge density
will be lower than the design value
Specify operation time. Different hazard classes require a minimum effective

time of operation, from 30 minutes for the lowest class to 90 minutes for high-
risk areas. This will put requirements to the water reservoir capacity, and there
are places where extra water tanks are necessary to achieve either pressure or
capacity
Scandpower
Determine required sprinkler spray pattern. The type of storage or objects to be

protected requires different type of nozzles: conventional type sprinkler, side wall
sprinkler, flat spray sprinkler, ceiling pattern sprinkler or a deluge type nozzle
Determine water supply requirements, and check the actual water supply capa-
city
Dimension pipes, connections, pumps and other components required to deliver

the required water flow rate. This may be done by hydraulic calculations.
9.2.5 Deluge Systems - Water Sprays

9.2.5.1 Tests by VTT, Finland
In a test series carried out by VTT, Finland, Ref. /9-19/, several liquid pool fire
scenarios subjected to different sprinklers and water sprayers were investigated. The
test set-up is shown in Figure B.5.
All together 109 tests were conducted . The extinguishing system consisted of four
nozzles located symmetrically above a pool mounted on a weighing platform. The test
system was set up in an 18 m high, 14 m wide and 27 m long test hall. Tests were
conducted with ten different liquids with a flash point in the range from -6 °c to 234 °c
using seven different nozzles. The level of the nozzles varied from 3 m to 8 m above
the fuel surface. The pool size was varied from 0.4 m2 to 12 m2 . Most of the tests
were carried out with a 1.6 m2 circular pool.
t t
Figure 9.5: Schematic of the Test Arrangement in the extinguishing Tests at VTT,
Ref. /9-19/
Scandpower
Risk Management AS ~ S I NTEF
A criterion for assessing the suppression of a liquid pool fire is presented in Ref.
/9-19/. The surface temperature of the pool is used as a measure of the burning rate
of a specific fuel. When water is applied, the surface temperature is reduced, reflec-
ting a reduced burning rate. A ratio Rr of the difference between surface temperature
and liquid flash point at free burning and with water applied, respectively, is defined:
Eq. (9.2)
where:
Ts = Surface temperature with water applied

Ta = Surface temperature without water applied (free burning)
T FL = Flash point temperature of the liquid (Closed cup value)
This non-dimensional number RT is used to characterise the performance of some

water sprays against liquid pool fires. Figure 9.6 shows how different pool sizes in-
fluences the time history of the RT. Small pools react quickly to the water application,
and the surface temperature is reduced considerably. Larger pools have a less pro-
nounced reduction of surface temperature.
!
~ SPR2/8 m/3mx3m/3,4bar
•..J
.8
1-""
I
1-"' .6 4.0 m2
~
"
~
f-l'"
I
..J .4 ~::::: r0,4 m2
1-'"
.2 I I
I
'------
~
0
LIAV230 I •
-.2
0 1 2 3
TIME (min)
Figure 9.6: Non-dimensional Surface Temperature of LIAV 230 (Liquid Hydrocarbon

Solvent with < 3 % Aromatics, Density 790 kg/m 3, closed cup Flash Point
58 - 60 °C), in Pools of different Size as a Function of Time in Tests with
Sprinkler SPR2 in a 3 m x 3 m Square at a Height of 8 m. Water
Pressure is 3.4 bar corresponding to a nominal Application Rate of
16 mm/min, Ref. /9-19/
Scandpower
(Gl)SINTEF Risk Management AS
This non-dimensional surface temperature is used to determine the critical flash point
of liquids that can be extinguished by cooling with water sprays. The basis of this is
shown in Figure 9.7. A fuel with flash point below about 60 °c is difficult to extinguish
by surface cooling in a pool. If water is supplied at a temperature above the flash
point of the fuel, the water may contribute to evaporation and enhance burning rather
than reducing or controlling the fire.
This correlation may be useful as a criterion for extinguishing of liquids, even in re-
lease configurations other than pools.
1
,-... L:::.. SN 1 /3 m / 2 . 1 mx 2 . 1 m/9.4 bar
...,
•8 A SN2 / 5m/2.1mx2.1m / 7.3ba
~
I "v SPR1/5m/3mx3m/ 5.3bar
!-CD •6 0 SPR2/8m/3mx3m / 1 . 1 bar
'--'
'-
,....._
e SPR2/5m/3mx3m/1,1bar
..., .4 ■ SPR3/5m/3mx3m / 7.2bar
~
I
.2
~
'--'
-.2
-50 0 50 100 150 200 250
FLASH POINT (°C)
Figure 9.7: Non-dimensional Surface Temperature 1 min after Start of Water Appli-
cation as a Function of closed Cup Flash Point of the Liquid. The
Legend of the Measurements reflects the different Nozzle Configuration,
Height and Water Pressure, Ref. /9-19/
9.2.5.2 Tests by Shell/BG UK
Shell and British Gas (BG) carried out a test series on jet fires and deluge system per-
formance in the 1990's, Ref. /9-20/. The test set-up was a 3 kg/s horizontal natural
gas sonic jet fire impinging a steel vessel 2 m diameter, 8 m long, with 12.5 mm wall
thickness at a distance of 9 m. The vessel was protected by a local deluge system,
producing a uniform coverage rate of 10.2 litres/min m2 using 24 medium velocity
nozzles. The set-up is shown in Figure 9.8.
Scandpower
Risk Managem ent AS ~ SI NTEF
<1.
I
1480 1475 14~5 1475 1480
I· I I·
. . . • .
5 9
.
10/16
17
18/24
•
1
25
2$2
33
•
34140
41
•
42/48
49 53
I
• • • • l • •
43/47
•
50/52
•
54/56
214 6/8 11/15 19/23 27(31 35139
12/14 20122 2sbo 36138 44/46
3
•
13
•
•
21
•
. •
29[
•
.
37 45
.• 55
•
I
i
2460
I· !
•32 2s•
Direction of flame
..,.0
<D
C\l
•31 21•
• 30 2s •
i
<1.
,ft =Spray nozzle

• =Thermocouple position
All dimensions =mm. ·
Figure 9.8: Test Set-up and Position of Thermocouples in the Large-Scale Jet Fire
Experiments by Shell/BG, Ref. /9-20/
The measured thermal response of the vessel is shown in the graphs in Figure 9.9
and Figure 9.10. The test numbers indicated at the graphs are as follows:
Test 12: No deluge

Test 4: Deluge running when jet was ignited
Test 6: Deluge initiated 60 s after the jet was ignited
Test 8: Water monitor (located behind the jet release point) and deluge operating
Test 10: Water monitor only on from ignition of jet
Test 11: Water monitor on 30 s after the jet was ignited.
(j)SINTEF Scandpower
Risk Management AS
700 -
700 Q = Test No.
0 =TestNo. @
@
600 -
S00
200
Tlmc (minutes) Time (minutes)
Figure 9.9: Temperature Development of Steel Tank. Left: Front Centre Line (Pos
27). Right: Rear Centre Line (Pos 31), Ref. /9-20/
700
0 = Test No.
700
@ Q = Test No.
600
600 - @
500
500
ug_, ®
e
400
© 0-
~ 400
::,
~ e
::,
1i.
E
e.,
a.
~ E
.,
I-
200
100
100
100
(@)
0 0
0 2 4 2
0
Time (minutes) Time (minutes)
Figure 9.10: Temperature Development of Steel Tank. Left: Top Centre Line (Pos
25). Right: Bottom Centre Line (Pos 29), Ref. /9-20/
Scandpower !Fa\\
The conclusion drawn by the authors was that a standard offshore deluge system was
unable to protect vessels directly impinged by a jet fire. The use of monitors alone
(flow rate not available) directed along the jet axis (Test 10 and 11) was effective.
When flow from monitors together with deluge was tested (Test 8), this had less effect,
probably due to less water to the monitor.
9.2.5.3 Tests by NIST, USA
Large-scale simulation tests were carried out in the mid 1980's out to demonstrate the
effectiveness of water spray systems to control and extinguish gas-well blow-out fires,
Ref. /9-21/. Water was injected into the gas jet by two different systems, one internal
and one external. The application mode, flow rate of water and fire size was varied.
The vertical natural gas jet fire was from 144 - 222 MW. The external water flow from
four nozzles surrounding the gas jet could extinguish the gas jet at a mass ratio
(water/gas) of 2.17, and left in burning at a rate of 1.56. Tests in small scale with diffe-
rent application modes were carried out prior to the large-scale testing. The results of
this work are summarised in Figure 9.11.
. ·~· ,,
Spray geometries, effectiveness
MW/Mg= 4.2
.Mw/Mg=S.lr .t
00,
Figure 9.11: The Effectiveness of different Spray Geometries in extinguishing Gas Jet
Fires, Ref. /9-21/
Scandpower
In the laboratory tests, two nozzles were used for injection at the base of the fire,
which led to a critical water/fuel mass ratio of 4.2. With different application modes, as
shown in Figure 9.11, the critical mass ratio was as high as 9.5 when the two nozzles
were directed horizontally towards each other at the base of the fire. In large scale,
however, the arrangement with four nozzles made the water requirement even less,
as low as 2.17 was tested successfully.
9.2.5.4 Blast and Fire Engineering for Topsides Structures
22 large-scale pool and jet fire experiments were carried out by Norwegian Fire
Research Laboratory, SINTEF, in 1997-98, Ref. /9-22/. These included 15 jet fires (4
horizontal jets and 11 vertical jets) and 7 pool fires, in two insulated compartments,
135 m3 and 415 m3 • A sketch of the 415 m3 test compartment is shown in Figure 9.12.
b - -
0 - ·..
"'
~ _ ss0o
- - - -______ _ _ _ _ _ Origin
y ___.__--,,-
Figure 9.12: Schematic of the 415 m3 Test Rig showing Location of the horizontal and
vertical Targets, Ref. /9-22/
A range of parameters was varied, including ventilation opening size and location, fuel
type, substrate, release height and pressure. Four vent sizes were investigated which
included the effect of splitting the vent area. In all except one test with gaseous pro-
pane, the fuel was a Statoil Sleipner-condensate. The fuel flow rate was about
0.95 kg/sec (about 40 MW). In some tests cylindrical steel targets were introduced
inside the test rig. Gas temperatures, steel temperatures (of walls, ceiling and tar-
gets), heat fluxes (to walls, ceiling and targets), gas compositions and fuel release
rates were measured. In 11 of the tests the effect of water deluge was investigated.
The layout of the deluge system is shown in Figure 9. 13. The nozzle used in these
tests is a modified Wormald MV57 with 12.6 mm orifice diameter (similar to the nozzle
denoted N5 in the medium-scale test programme). The K-factor was measured to
99.5 (litres/min bar112) . The water application rate was aimed at 14 (litres/m2 min), but
varied between 10.4 and 28.7 (litres/m2 min).
Scandpower
Conclusions from the confined jet and pool fire test programme:
During the deluge experiments the well ventilated condensate jet fires were not
extinguished by typical offshore water deluge. The jet fires continued to burn at
the same rate but there was a substantial reduction in fire intensity. However,
well ventilated propane gas jet fires have been seen extinguished during
smaller-scale experiments conducted prior to the Phase 2 of the project con-
ducted by Shell Research, HSE and SINTEF, Ref. /9-23/
Fuel controlled (underventilated) condensate jet fires were controlled, but not ex-
tinguished when deluge was activated soon after ignition
Fuel controlled (underventilated) condensate jet fires were extinguished when

deluge was activated 10 to 12 minutes after ignition and the compartment was
"hot"
There were no significant differences between the effect of water deluge on ver-
tical and horizontal condensate jet releases, for the conditions studied
It is possible for the fire to re-ignite after the water deluge is terminated due to
the presence of hot gases or surfaces coming into contact with fuel
Extinguished jet fires represent a potential explosion hazard if the fuel continues
to be released
Generally, confined pool fires are not extinguished by water deluge, but the fire
is controlled and burns at a much reduced rate.
Scandpower
lt,SINTEF Risk Management AS
Water supply - - - -- - -- - ~
from pump 4"
.----------- -+-----
West wall
-------~
- .+-
" E 6000 mm
~,
El
Q I
~,
2.0" 2.0"
i-
1585 mm 2750 mm 1665 mm
E
E
: o
I ~ Distribution pipes
I N to nozzle were 0.75"
I
I
El
E
~ f
r--
N
E C
Roof
E ·.;
8 E
LO en
N C
a:
Ring main is located on

E
E top of the roof.
0 Distribution pipes are run
0 through t he ceiling and
"'
N the nozzles are mounted
vertically downwards
just below the ceiling.
~----~----------- - ----~---~~
10 10 mm 3450 mm 1040 mm
< ><- - - - ---~---►
Figure 9.13: The Deluge System installed on the Roof of the 415 m3 Test Rig at
SINTEF, Ref. /9- 16/
Scandpower
The distribution of droplets of different sizes in a spray is governed by the geometry of

the spray nozzle and the water pressure.
For two geometrically similar nozzles the droplet size is also a function of orifice
diameter. A basic correlation for the relative droplet diameter is described by:
d, = {Ll.p/Ll.po) ·113 (D/D0)213 Eq. (9.3)
where:
d, = relative median droplet diameter, d/d 0

d = actual droplet diameter
do = reference droplet diameter
Ll.p = actual pressure immediately upstream the nozzle
LI.po = reference pressure
D = orifice diameter
D0 = reference orifice diameter
The median droplet diameter is defined as the diameter where 50 % of the mass of
water is made up of droplets, which are equally large or larger than the median droplet
diameter and 50 % smaller than this diameter.
A smaller nozzle diameter leads to smaller droplets, and a high nozzle pressure leads
to smaller droplets.
The droplet size distribution is normally a Gaussian distribution of droplet sizes varying
around a median droplet diameter. Special designed nozzles may, however, give
other distributions. For instance, a special nozzle is developed to give a certain
amount of very fine droplets together with relatively large droplets, in order to obtain
both penetration through flames and effective cooling of smoke, Ref. /9-24/.
The influence of droplet size on cooling and extinguishing of gas fires has been
studied experimentally by Wighus, Ref. /9-12/. The tests were carried out in a 2.5 m
wide, 2.5 m high and 5 m long enclosure, with a propane burner at floor level. Results
from experiments with a 1 MW fire are shown in Figure 9.14.
Three different nozzles were tested, at similar experimental conditions. The enclosure
was naturally ventilated through an air inlet opening at the floor level, spanning over
one long side. The smoke outlet was located at ceiling level, at the opposite long wall
of the enclosure. The test procedure was to ignite the propane burner, let it burn
freely for 5 minutes, and then activate the water spray. Heat transfer rates from the
fire to the enclosure, to the ventilation air and to the water were registered by a com-
prehensive measurement technique. Tests were run with increasing water application
rates, until instantaneous extinguishment occurred. The ratio of heat absorbed by the
water spray to the total heat production was measured, and the results for each test
condition are given in Figure 9.14 as a "SHAR" value, i.e. Spray Heat Absorption
Ratio.
Scandpower
(Gl)SINTEF Risk Management AS
SINTEF-NSL
5~ - - - - - - - - - - - - - ~ - - - - - - - ~
4 E)rtlngulshed
,a
+ 11J Shar = 0.15
+ Shar = 0.2
Not extingui.stHJd
0 ~ ~ ~ ~ 1~ 1B 1~ 1~ 1~ ~
Mean droplet diameter (miototneter)
DH7 + H4 ◊ G22
Figure 9.14: Results of Extinguishment Tests for a Propane Fire about 1 (MW) in a
30 (m 3) Compartment, with Fire induced Ventilation
SHAR is defined as:
SHAR = Heat absorbed by water spray

Eq. (9.4)
Heat released by fire
The tests clearly indicate that if a water spray can absorb more than 30 % of the heat
release rate of a fire, there is a possibility for instantaneous extinguishment. The
smaller the droplets, the lesser were the water application rate needed to obtain this
SHAR value.
9.2.6 Pressure - Water Flow Rate

Bernoulli's equation for flow in pipes and through orifices gives a correlation for water
flow rate as a function of nozzle pressure:
V = K .1p1/2 Eq. (9.5)
where:
'ii = Water volume flow rate

K = Pressure drop coefficient
.1p = Pressure drop
Scandpower
The pressure drop coefficient is commonly presented for typical sprinkler- and deluge
nozzles, named the k-factor. The k-factor is normally presented as litres/min bar112 •
The procedure to measure the k-factor is described by ISO 6182-1: 1993(E), Ref.
/9-25/. Variations in test set-up and instrumentation may influence the measured
value.
In addition to the uncertainty of the k-factor, the water application density may vary
significantly over a floor area. Different nozzles produce different spray patterns, often
characterised by a spray angle and a characteristic flow pattern (hollow cone, full
cone, flat spray etc).
If one looks in three dimensions, the spatial distribution of water droplets is even less
quantified. Interaction between a water spray and flames and the interaction with a
flame front of a gas explosion is definitely a function of spatial distribution of droplets,
and this feature of water sprays, denoted water volume fraction, should be focused in
future research.
9.2.7 Cooling by Water Droplets

Heat transfer to a droplet in a hot atmosphere is a highly transient phenomenon. The
track of the droplet from the nozzle through a gas with variable temperature, first
heated up to its boiling temperature, thereafter reducing volume by evaporation may
be described by a set of equations, and needs to be solved by a numerical scheme.
The motion of a droplet can be described by the following equation, taking into
account the drag force from the surrounding gas flow and the gravity force, Ref. /9-26/:
Eq. (9.6)
where:
Eq. (9.7)
Eq. (9.8)
Here Vd ,md, Ad and Dd are the velocity, mass, cross section area and diameter, res-
pectively, of the water droplet. V9 , p9 and ).lg represent the gas phase velocity, density
and viscosity, respectively. g is the acceleration of gravity.
Fundamental equations of heat transfer to a droplet can be described by Ref. /9.16/:
Eq. (9.9)
where md , cd ,A, and Td are the mass, specific heat capacity, surface area and tem-
perature of the droplet, respectively. Further, h1g is the latent heat of water evapo-
ration and ht is the convective heat transfer coefficient, which can be evaluated using
the following relationship: (Nusselt Number of a sphere in natural convection):
Scandpower
Nu= h, · Dd = 2.0 + 0,6 • Re 112 • Pr 113 Eq. (9.10)

kg
where kg is the thermal conductivity of gas phase, Re is Reynolds Number, Pr is

Prandtl Number.
The evaporation rate is the sum of diffusive mass transport by vapour pressure diffe-
rence and boiling rate of the droplet:
Eq. (9.11)
where
(Tdew<Td<Tbp) Eq. (9.12)
The Sherwood Number, describing the concentration gradient at the surface, is quan-
tified by:
Sh= hm Dd = 2.0 + 0.6. Re 11 2 • Sc 113 Eq. (9.13)

Dvapour
Sc= - -µ_g__ Eq. (9.14)

Pg· Dvapour
where hm is the mass transfer coefficient. Psat and Pg are the saturated vapour
pressure and the static pressure in gas phase. Cvapour is the vapour mole fraction and
Dvapour the mass diffusion coefficient of water in the gas phase. Tvapour and Tbp are the
vapour dew point and boiling point, respectively, of water at atmospheric pressure.
Finally, the boiling rate of a water droplet in a warm environment can be estimated by
the following equation:
(Td>Tbp) Eq. (9.15)
Eq. (9.16)
where Cg is the specific heat of gas phase and Pd the density of water.
In addition, the velocity difference between the droplet and the gas phase should be
calculated for a droplet with a finite exit velocity from the nozzle, subjected to gravity
and drag force, into a flow field of varying velocity.
From these equations, a droplet lifetime may be computed.
A simplified model for droplet lifetime is described by the D2 law for droplet evapo-
ration, Ref. /9-27/:
Scandpower
Norwegian Fire Research l aboratory
dD~ 8-k 9
-= - - - -ln(B +1) Eq. (9.17)
dt Pd·C
9
q
where Bq is the Spalding Number or the transfer number,
B - Cg -(Tg -Td) Eq. (9.18)

q - h
fg
This model holds after an initial transient period associated with the heating of the
droplet to near the boiling point.
Letting the diameter of the droplet be zero, representing complete evaporation, the
correlation of droplet lifetime is:
Eq. (9.19)
where
8-k
K =- -9- - ln(Bq + 1) Eq. (9.20)
Pd ·Cg
2
Figure 9.15 shows examples of calculation of droplet lifetime using the simplified 0 -
law. The results represent the time after the droplet has reached the boiling tempe-
rature of water.
Time to evaporation
1000
100 ~... .....r--,..__

:i 10
~I'-......."'---
r--,..r---
G)
E ~ I"'---..
-
"'-
i= 1
""'r--.
0, 1
0,01
0 100 200 300 400 500 600 700 800 900 1000 1100
Temperature difference rCJ
1- 1000 - 500 - 250 - 100 J micrometer
Figure 9 .15: Droplet Lifetime as a Function of Temperature Difference between Gas

and Boiling Water, and Droplet Diameter, calculated by the 0 2-Law, Ref.
/9-27/
Scandpower
(t;J)SINTEF Risk Management AS
9.2.7.1 Droplet - Fire Plume Interaction
Figure 9.16 shows some of the trajectories a droplet may follow in a fire plume. Track
I show a droplet that is large enough to survive the travel through the hot smoke gases
and through the flames finally to hit objects or the fire base before it is evaporated.
This droplet has the ability to cool or wet an object or to interact with a liquid pool
surface.
If the temperature of this droplet is below the flash point of the liquid when it reaches
the surface of a pool, it may cool the liquid to a temperature below the flash point and
subsequently lower the evaporation to a rate at which the fire is extinguished .
•
•
III.
••
II
• •
• I
(jJ
~
~
Flame ©
~
•
• • IV
Entrained gas
Object
Fire base
Figure 9.16: Schematic of different possible Trajectories of Water Droplets in a Fire

Plume
If the temperature of the droplet is above the flash point but below the boiling point of
the liquid, it may limit the evaporation of the fuel and subsequently decrease the
burning rate. If the temperature is above the temperature of the liquid layer, it may en-
hance the evaporation of the fuel and increase the fire.
Track II shows a droplet that is heated to the boiling point of water when it has entered
the flames. This droplet size may reduce the combustion efficiency both by cooling
the flames and by the dilution of reactants.
Track Ill shows a droplet pattern where the boiling point is reached in the hot smoke
plume. The effect of this type of droplets is mainly to cool the smoke gases.
Scandpower
Handbook for Fire Calculations and Fire Risk Assessment in the Process Industry Page 9.33
Track IV shows a pattern where droplets may enter the combustion zone by the en-
trained air. This category of droplets may interact with combustion and may also
effectively cool objects and fire base, because they have not been heated by passing
a long track of hot gases before entering the fire zone.
To evaluate the behaviour of different categories of droplets, a model containing the

most important features of transport and heat transfer of an evaporating droplet, into a
transient temperature field is needed.
9.2.8 Water Mist

The water mist technology has developed rapidly the last ten years, with the phase-
out of Halon for fire suppression as one of the main driving forces, Ref. /9-18/. As a
result of the development of water mist fire suppression technology, new fields of
application have also emerged. In addition to replacement of Halon gases, water mist
is now applied as an alternative to inert gas systems and as replacement for sprinkler
systems. At present, the general trend is that a new application of water mist needs to
be tested in a set-up similar to the practical situation where it is to be used. In parallel
to the manufacturers testing efforts, test institutions and researchers world-wide focus
on more general questions regarding the effect of water mist in interaction with fire,
and standardisation committees work with new standards and test methods for diffe-
rent applications. The general trend in fire science, with focus on scenario based
design and fire safety engineering, leads to the need for documented calculation
methods and engineering practice.
9.2.8.1 Suppression Mechanisms of Water Mist
When water is added as a mist to the atmosphere in a fire scenario its effects may be
of different kind:
It may cool the atmosphere by convective heat transfer from water droplets to
the atmosphere
It may dilute and possibly inert the atmosphere by adding water vapour
It may act upon the flames by enhancing or decreasing the rate of entrainment
It may separate the flaming zone from the source of fuel vapour
It may interact with reactants in combustion
It may block radiant heat
It may interact with the fuel directly, by cooling and wetting surfaces of solid or
liquid combustibles.
The two main modes of engineering the water mist systems are either local application
or total volume protection. With local protection the interaction of water mist with fuel
flow and evaporation and the displacement of flames, either by blow-off or separation,
are the dominant suppression mechanisms. In total volume protection the recirculation
of fire products and evaporated water into the combustion zone is of importance. In
many practical applications the water mist systems are in a grey zone between local
application and volume protection, for instance in large compartments. This is illustra-
ted in Figure 9.17.
Scandpower
IG))SINTEF Risk Management AS
Local application Grey zone: Volume protection

Large
comoartments
Interaction with
combustion,
fuel-flow and --------- ---------►
evaporation
Flame displacement
•blow-off --------- ---------►
•separation
Re-circulation of
+------ combustion products
and water vapour
Figure 9.17: Different Mechanisms of Water/Fire Interaction in local Application and

Volume Protection
The challenge for water mist system producers is to produce and deliver an optimal
spray that can be transported into the combustion zone and interact with the combus-
tion, for different locations of a fire, in various protected volumes and with different fire
types and sizes.
9.2.8.2 Physical versus Chemical Effect
Combustion is a chain reaction with a large number of chemical reactions taking place
simultaneously at different rates and speed. In common fires, the main reactants are
the oxygen from the air that reacts with carbon and hydrogen from the fuel, forming
carbon dioxide and water. When oxygen supply is limited, the combustion becomes
incomplete and carbon monoxide is formed at the expense of carbon dioxide. Bet-
ween the start- and end products there are hundreds of different reactions.
Figure 9.18 shows the importance of some of the different factors, based on mathe-
matical modelling of the combustion process, Ref. /9.28/. The graph illustrates calcu-
lated laminar flame speed as a function of water vapour concentration at 100 °c am-
bient atmosphere. Laminar flame speed is a measure of reactivity. The effect of
cooling by evaporation of water is not taken into account. Calculations are carried out
with the computer program CHEMKIN - which calculates chemical kinetics in reac-
tions, Ref. /9-29/. The mixture of fuel and air in the shown example is at 10 % above
stoichiometric fuel amount. The fuel is a mixture of methane (87.79 %) and some
ethane (9.04 %). The results are based on computation of 108 different reactions.
The figure shows that the effect of water mist is mostly physical, which includes dilu-
tion and thermal effects. The thermal effect is about twice as high as the effect of dilu-
tion (inerting). Direct influence on the combustion process (chemical, catalytic) is
minimal.
Scandpower
70 perimental points
j
~ ea"""-----------------------
~
~
~ 50
·u
0
~ 40 Physical
g> suppression
.E
::, 30
.0
~
"'t:
·E 20
"'
....1 Fuel equivalence ratio ( (I)) = 1.1
10 Temperature of cold premixed gases (7) = 100°C
0 5 10 15 20 25 30 35 40 45 50
Content of water vapour in premixed gases (XH_ 0), %
Figure 9.18: Laminar Burning Velocity of Methane/Ethane in Air with Water Vapour
added at 100 °C, illustrating the Effect of cooling, inerting and Chemical
Suppression, Ref. /9-28/
9.2.8.3 The Paradox of Water Mist
The big paradox in the effect of water mist used in compartment fire protection is that
the most difficult fires to extinguish are the smaller ones. If a fire grows big enough,
the experience with water mist is that these fires are quite easily extinguished. The
smaller fires, however, either diminish or stabilise at a constant burning rate, Ref.
/9-30/. Most of the experience with this is obtained with liquid spray or pool fires. If
higher water flow rates are introduced, the survival of smaller fires was even more pro-
nounced, but reducing the water application rate some times leads to fire extinguish-
ing. A possible explanation to this phenomenon is in the high boiling point of water
compared to other extinguishing agents, as shown in the former chapter. At ambient
temperatures below the boiling point of water, water vapour can exist at a maximum
concentration dependent of temperatures illustrated in Figure 9.4, as the saturation
pressure of water as a fraction of atmospheric pressure.
If water mist is introduced as a volume protection agent with nozzles arranged as in

typical gas extinguishing systems, there will typically be zones where a fire may be
hidden from direct impingement by water droplets. Flame displacement and direct
surface cooling effects will then be minor or non-existing, and all the effect of water
has to be due to droplets of water following the air entrained into the fire plume or re-
circulated fire products. It seems quite unrealistic that a concentration of water drop-
lets that are able to float with the air can be produced by present technology, if it is
physically obtainable at all. However, well-engineered water mist systems with
nozzles designed and localised to cover obstructed spaces will overcome these prob-
lems.
Scandpower
(» SINTEF Risk Management AS
9.2.9 Water Concentration and Droplet Sizes for Extinguishment

9.2.9.1 Small scale
In small scale tests, using the so-called cup burner test and the counterflow burner
test the effect of water mist and Halon 1301 is compared, Ref. /9-31 /. By counterflow
burner tests the conclusion was that water mist with a droplet diameter of 20 µm,
water has the same affect as Halon 1302 (on mass basis). Then the heat of evapo-
ration of water was maximally utilised in the flame of the burner.
At the University of Lund, Sweden, several basic research programs of the effect of
extinguishing media have been carried out, among these a substantial amount of
water mist research. A test apparatus basically developed to study solid particle ex-
tinguishing media has been used to study water mist, Ref. /9-32/. The REMP-value,
Required Extinguishing Media Portion, is found by an apparatus.
From theoretical considerations and calculations, it has been shown that a critical
concentration for extinguishing a fire is about 30 %. In practical tests with the REMP-
apparatus, the flame was extinguished at a water concentration of 100 - 180 grams/
m3 , depending of droplet diameter, Ref. /9-33/.
9.2.9.2 Medium Scale
In this category, there are rather limited publications regarding experiments with water
mist. One typical medium-scale test series was carried out by SINTEF NBL in a 31 m3
test rig during the FIREDASS research programme, Ref. /9-34/. FIREDASS (FIRE
Detection and Suppression Simulation), financed by the European Commission, had
the aim to develop a simulation tool to study the effect of water mist in small enclo-
sures. Tests were carried out in enclosures similar to aircraft baggage compartments,
and a detection system linked to a water mist system was tested in both a 31 m3
model compartment and in a realistic full-scale baggage compartment. The SINTEF
task was a series of tests with gaseous, liquid and solid materials as fuel, carried out
in a compartment with a ventilation rate similar to a realistic rate of an aircraft. Based
on the extensive measurements of temperatures, concentrations of oxygen, CO2 , CO
an water vapour, an empirical correlation for the conditions where water mist will
extinguish a fire was developed, Ref. /9-35/. This model is based on the conditions of
the gas that is entrained into the base of a fire plume, and shows a combination of
oxygen concentration and a temperature where flames are extinguished. The correla-
tion is:
0 2extlimit = 20.9 - k . Tn Eq. (9.21)
where:
k = 0,000045
n = 2
T = Representative temperature of the gas entrained into the combustion
zone (K)
The correlation is valid for T > 35 °C.
The same correlation is shown graphically in Figure 9.19. It shows that if the tempe-
rature of the supply air to the combustion zone increases, the oxygen concentration
Scandpower
needed to obtain extinguishment is lower. This correlation is in accordance with the

knowledge about flammability limits of gases; higher temperature corresponds with
lower flammable limit of flammable gases. The correlation shown here is only using
the other participating reactant, namely oxygen.
Some of the points shown in Figure 9.19 indicate that extinguishment occurs even at
higher oxygen concentration than given by the correlation. The reason is that the cor-
relation is truly empirical, and the limit is set lower than all the experiments shows.
Some of these points were obtained by fires that were directly hit by the droplet spray
from the water mist nozzle. In the tests enveloped by the limit of the correlation, most
of the results were obtained with fires located at a longer distance from the spray
nozzle to the fire base. In some fires obstacles were also present.
In the FIREDASS tests, it was not possible to extinguish flames only by the droplets
that are floating in the air; some help from the vaporised water was also needed. In
the lower temperature regime, the low saturation pressure of water vapour in air at the
low temperature limits the water vapour concentration and hence the direct inerting
effect. For small flames, the re-circulation of combustion products is normally also
diminished due to reduced buoyancy. At a temperature as low as about 35 °C, the
maximum water vapour concentration is about 5.5 %. The effect on oxygen concen-
tration is then maximum 1.1 %, reducing the ambient oxygen concentration to 19.8 %.
The extinction limit at 35 °c based on the above-mentioned correlation, Ref. /9-35/, is
about 16.6 %. The gap between the limit and the maximum saturation pressure of the
water in air is to be filled by oxygen consumption by the fire and evaporation of water
droplets in the hot flame zone. The direct action seems to be limited to the difference
at about 35 °C in the FIREDASS tests.
This correlation is as mentioned empirical and based on experiments in medium scale,

with water mist produced by a twin-fluid nozzle, and is valid for these conditions.
However, it has shown to be representative for other water mist systems and at diffe-
rent scales as well.
Critical oxygen concentration for

extinguishment with water mist
C 20
;
b
0
1G
15
••
C
0 -
u~
C ~
0
u >
C
0
Q
-
0 10
5
~
0
0
0 100 200 300 400 500
Temperature (Celcius)
• Extinguished • Not extinguished - Extinguishing limit I
Figure 9.19: Test Results from medium-Scale Water Mist Experiments, with an
empirical Correlation for the extinguishing Limit shown, Ref. /9-35/
Scandpower
(,i)SINTEF Risk Management AS
At the University of Lund some experiments has been carried out in the so-called 1/3
scale room. The reference room is here the test room of the room corner test, ISO
9705. A critical concentration of 70 - 140 g/m 3 was found. This was valid for droplet
diameters of 30 - 40 µm. Critical concentration varied with the distance from the water
mist nozzle to the flame and the elevation of the nozzle, Ref. /9-33/. The water mist
was supplied in a way to make it follow the airflow into the flames, and the purpose
was to study how water mist acts as a total flooding extinguishing medium.
9.2.9.3 Full Scale
At the International Conference on Water Mist Fire Suppression Systems in Boras,

Sweden in 1993, researchers from Canada, Sweden and Norway presented results
from full scale tests with water mist systems, Ref. /9-36/. National Research Council
of Canada, National Fire Laboratory, had managed to get the test data from a variety
of water mist producers, and presented design criteria on water mist systems for use
in ships, for the Canadian Navy. A practical upper limit for room size was then set to
200 m3 , Ref. /9-37/.
SINTEF had already carried out test series with water mist in offshore turbine hoods,
and the client allowed publication of results from the project "Halon Replacement by
Fine Water Spray Technology - Turbine Hood Application", Ref. /9-38/. This work
documented that water mist was suitable to extinguish gas- and liquid fires in rooms
with limited air supply, that it was easier to extinguish larger fires than small, that
sequential supply of water mist could be optimal under certain circumstances, and that
obstructed fires were difficult to extinguish, but still possible where air supply was
limited. Typical design concentrations of water for closed rooms with large fires were
set to 60 - 70 grams/m 3 room volume, but 400 - 600 grams/m3 was needed to ex-
tinguish small fires. The location of water mist nozzles was of great importance for the
result, especially regarding hidden fires. Nozzles should be located with the aim to
spray towards the base of hidden fires. Ceiling mounted nozzles spraying downwards
were not necessarily the best choose. The SINTEF test programme with turbine
hoods formed the background for standardisation work, and was referred to in the very
first standard of water mist systems, NFPA 750, Ref. /9-39/. The above mentioned
design criteria used by the Canadian Navy were also based on these tests.
In December 1994 IMO approved regulations on how to test and install alternatives to
Halon in machinery spaces and pump rooms onboard ships. This opened up for the
use of water mist, Ref. /9-40/. Different classes of machinery spaces were introduced,
based on room volume, see Table 9.4. Classification authorities very soon introduced
the practice that no producer got a certificate to install a water mist system in room
sizes larger than the maximum it has been tested to. The IMO 668 got an addendum
in 1996, MSC Circ.728.
T abl e 94 Cl asses of Mach'merv s,paces .m IMO 668, Ref. /9-40/

Class Volume (m 3)
1 up to 500
2 up to 3000
3 above 3000
Scandpower
In 1995 IMO approved Revised guidelines for approval of sprinkler systems equivalent
to that referred in SOLAS regulation 11-2/12, Ref. /9-41/. These were based on the
principle that an alternative system might be approved if it was shown by experiments
that the performance was equivalent or better. To get a reference to sprinkler systems,
the Swedish and the Finnish fire laboratories carried out a series of tests with a con-
ventional sprinkler installed, to establish the acceptance criteria for a system. The first
tests were carried out in a cabin- and corridor configuration, later in public spaces like
restaurants and in tax-free shopping area.
In May 1999 IMO then approved guidelines for the approval of fixed water-based local
application fire-fighting systems for use in category A machinery spaces, and opened
up for the use of water mist in spaces with a high risk of releasing combustible liquids,
Ref. /9-42/.
These IMO regulations have given most of the full-scale test results, especially in
machinery spaces, and to some extent also with public areas and cabins. Many
manufacturers have carried out tests to get approval for installation in machinery
spaces, both for volume protection and for local application. The biggest machinery
space any manufacturer has got approval for seems to be about 2,000 m3 , but in 2000
some manufacturers also have tried 3000 m3 •
Experience so far is that water mist systems manage to extinguish both spray and
pool fires quite rapidly, but it takes longer time to extinguish the smallest hidden fires
of the test programme. The 0.5 m2 concealed heptane pool fire and a low-pressure
concealed light diesel oil spray fire of 0.03 kg/s are the fires normally lead to the
longest time to extinguishment. The limiting time of the IMO 668 test is 15 minutes.
This time to extinguishment is exceeded when the room sizes approaches about 2000
m3• The temperature of the test room is rapidly reduced to a very low level for most
tests, far from hazardous to people.
At present, this author does not know any manufacturer that has passed the test
criteria of IMO 668/728 for room sizes above 2000 m3 • This problem is presently
addressed by experts around the world, with the aim to improve the IMO 668/728 test
procedures and acceptance criteria in a way that is safe but takes into account the
"paradox of water mist", Ref. /9-43/.
9.2.10 Foam Media

Foam media are added to water and are used in both fixed and manually based sys-
tems. The main mechanism is to form a layer on top of a fuel surface to prevent mass
and heat transfer. The foam media must be suitable for the actual fuel type. Some
fuels will dissolve certain types of foams.
The expansion rate of the foam is also different. The application system will also
determine the expansion rate. The expansion rate is defined as the volume multiplica-
tion factor an amount of foam/water solution is enlarged by a foam applicator.
Scandpower
There are three classes of foam expansion rates:
Foam class Expansion factor

Heavy foam 6-20
Medium foam 20-500
Light foam 500-1250
. f t Total volume of foam

ExpansIon ac or= - - - -- - - - - - - -- Eq. (9.22)
Volume of water + foam concentrate
Foams are sorted in three types:
Detergent and protein foam

AFFF and fluor-protein foam
Alcohol resistant foam.
AFFF is the abbreviation for Aqueous Film Forming Foam.
The three different types of foams are suitable for various fuel types. All foams are
most applicable for liquid fires as they are designed to form a layer upon a horizontal
surface. Foam concentrates may also be useful in other types of fires as they may
affect the surface properties of water and promote wetting. A special observation is
that foam concentrate added to water promotes the extinguishment of liquid spray
fires. This may be explained by the wetting conditions of water and oil droplets. An oil
droplet hit by a water droplet will normally not be wetted. However, by the action of a
film forming foam concentrate, the water may form a layer around the droplet preven-
ting evaporation and cooling the oil more effectively. This mechanism is, however, not
verified.
9.2.11 Carbon Dioxide, CO2 • (Inert Gas Systems)

Principally, all non-oxygen gases may be used as inert gases.CO2 and to some extent
Nitrogen, are traditionally utilised as fire suppressants. In later years, other inert gases
as Argon and some mixtures of different inert gases (lnergen, IG-541) have been
launced. Some of the hazards of use in manned areas have then been addressed.
Carbon dioxide is used in fixed systems as total flooding suppression systems. The
principle is to supply gaseous CO2 to a space under pressure, and seal vents to pre-
vent leakage of gas. The air in the space is diluted, and the oxygen concentration is
reduced until extinction of the fire occurs. This mechanism is called inerting. The
CO2 gas is very cold at the entrance, due to evaporation from the liquid phase in the
pressurised tank or bottle. This leads to additional cooling of the flames and the en-
closure. A critical volume concentration of about 30 % CO2 is needed to obtain extin-
guishment in most fuels and a design concentration is normally 40 - 50 % of the en-
closure volume. The expansion factor of CO2 from liquid to gaseous phase is about
450.
An important factor regarding CO2 systems is that activation of the suppression sys-
tem leads to low oxygen concentration in the enclosure, lower than human tenability
Scandpower
Risk M anagement AS ~ S I NTEF
limits. A safety measure preventing activation when people are in the protected enclo-
sure is necessary.
9.2.12 Halon Systems

Halons, Halogenated Hydrocarbons, have a special character among the fire sup-
pression media. The mechanism of extinguishment is of chemical character, as the
combustion process is influenced dramatically by the presence of Halons. In the chain
of reactions from complex to more simple hydrocarbons and further to the oxidation of
carbon with oxygen, some chains are replaced by the Halons. The needed concen-
tration of gaseous Halon in an enclosure is significantly less than in the case of
inerting gases.
Table 9 s· Halons and their Chemical Formulas

--;--;,, ,'•.,,' ~ - ,
Aalon ... · 'ch~inicalfo,1111ula ''
1301 C F3 Br
1211 C F2 Cl Br
2402 C2 F• Br2
Halon 1301 is frequently used in fixed systems, and the main advantage of Halon
1301 is that the required concentration for extinguishment is much less than inert
gases, i.e. in the order of 3 %. This means that a quick response activator can be
connected to the Halon system, regardless of whether people are present at the loca-
tion of the release. The Halons themselves are not directly toxic and the oxygen con-
centration in an enclosure after release of the extinguishing system is sufficiently high
for people to escape.
However, the Halons have other disadvantages, which are connected to a more re-
cent concern. Their Ozone Depletion Potential (ODP) are among the highest, and
authorities all over the world have laid restrictions on the use of Halons. In many
countries Halons are on the list of components which are forbidden to use. A pro-
gramme for phasing out Halons and destructing the present amounts has been star-
ted.
A replacement for Halons in existing systems is approached by so-called drop-in

gases. These are gases, which can be a direct replacement for Halons. None of the
drop-in gases known has yet been approved by authorities, because the long term
effect on atmosphere, safety requirements and toxic effect of reaction products after
they have been used in a fire, have not been proved. Some replacement gases for
Halons have been· launched and are listed as fire suppressing agents with low ODP,
and may be used under special conditions.
However, in many places where Halons have been used, other conventional fire
fighting systems may be used. A promising replacement is water mist sprays with
very small droplet sizes applied in enclosures by nozzles located near potential fire
sources. Combined with the shut-down of electrical equipment and ventilation, this
may be a good and safe replacement for Halons.
Scandpower
9.2.13 Dry Chemical Powders

Dry chemical powders have been used for fire extinguishment for several years, and
they are frequently used in portable systems. The powders are classified as A,B,C
and D type of powder.
Powder type Effective against fires in

A Cellulosic materials, flammable solid materials
B Liquid fires
C Electrical equipment
D Metal fires (Each metal may require a special powder)
The main effect of dry chemicals seems to be the cooling of flames and fuel, even
though some effects of inhabitation of combustion reactions are claimed by suppliers.
The melting of the powder, heating up to boiling point and sometimes evaporation of
the melted component will absorb much heat from the base of the fire and may pro-
duce an inert atmosphere locally.
The powder is propelled by a pressurised gas, normally nitrogen, and the particle size
distribution is important in determining both the transport mechanisms and the melting/
evaporation of the powder.
Combined systems using water to transport the powder have been developed, and
these systems are effective in some types of liquid pool fires.
9.2.14 Pyrotechnically generated Aerosols

Pyrotechnically generated aerosols (abbreviated aerosols) for fire extinguishing is
introduced to the market in western Europe and USA the later years, but has a long
record in Eastern Europe, Ref. /9-44/. An aerosol consists of a mixture of gas and
particulate material, where the particles can be in a liquid or a solid form. The aerosol
is by definition this mixture of phases, but is often associated to the particulate phase
only. The characteristic properties are that the particles follow the gas flow, and are to
a small extent settling by gravity.
Primarily, pyrotechnically generated aerosols extinguish the fire chemically, by inter-

fering with the combustion chain reaction. Physical action is due to the lowering of the
flame temperature by heat absorption by particles that is heated and/or evaporated.
The technology is usually not suitable for use in occupied areas due to the health risk
of breathing the particles for a longer period. In addition, the visibility inside an en-
closure is severely reduced upon activation of an aerosol system. IMO has released
guidelines for approval of such systems, with quite similar requirement as for water
mist systems, Ref. /9-45/. Added tests of health risk by the product are needed.
Scandpower
9.2.15 Effect of Active Fire Protection Systems on Smoke Generation

and Escaping/Evacuation Possibilities
Some of the aspects of safety for personnel in fires are affected by the active fire pro-
tection systems. The already mentioned possibility of suffocation due to lack of
oxygen in fixed CO2 systems, is a problem in both fire situation and if the system is
activated unintentionally.
Another problem, which has been addressed, is the use of intumescent passive fire
protection on structure and components in enclosed spaces. The reaction of some of
the intumescent materials to fire is that the char formation is combined with smoke
production. The smoke may be irritating or even toxic.
Soot/fuel rat:10
H?-1-2
0.01
D,009
0.008
Q.007
D.005
o.oas
0.00◄
D.003
D,002
0.001
0
0 2 ◄ 6 8 10 12 ,. 16 18 20
Tlfl'lrf (mlnJ
Figure 9.20: Soot/Fuel Ratio as Function of Time in a typical SINTEF Experiment,

Ref. /9-12/
The effect mentioned above is, however, not the main effect of the protection system.
The effect of prevention of fire spread and control of fire will mainly promote safe eva-
cuation and escape of personnel. An effect seen by using water sprays in hydrocar-
bon fires is a substantial reduction of soot contents in the smoke. A typical result of
the experiments with enclosed gas fires, reported by Wighus, Ref. /9-12/, where soot
concentration of the effluent gases was measured by an optical device, is shown in
Figure 9.20.
The soot production is calculated as a function of the fuel burning rate. The soot
production ratio is increasing as the fire develops in the enclosure, probably due to
reduced air/fuel ratio. When the water spray is activated, the soot production is either
Scandpower
reduced or some of the soot is washed out of the smoke. The spray is activated after
5 minutes.
Radiant heat is also to some extent absorbed by water spray and steam. This effect is
however much less than the effect of reduced flame temperature on the radiation out-
put as a result of water spray action.
Scandpower
/9-1/ Institute of Petroleum: "Liquified Petroleum Gas", Volume 1, "Large Bulk
Pressure Storage and Refrigerated LPG", John Wiley & Sons, February 1987.
/9-2/ IMO Res. A472 (XII): "Improved Recommendation on Test Method for
assessing Construction Materials for Marine Applications as non-combustible",
International Maritime Organisation, November 1981.
/9-3/ International Standard ISO 1182: "Fire Tests - Building Materials - Non-com-
bustibility Test", Second Edition, International Organisaton for Standardisation,
1983.
/9-4/ IMO Res. A517 (XIII): "Recommendation on Fire Test Procedures for "A", "B"
and "F" Class Divisions", International Maritime Organisation, February 1984.
/9-5/ International Standard ISO 834: "Fire Resistance Tests - Elements of Building
Construction", International Organisation for Standardisation, 1975.
/9-6/ Del norske Veritas: "Fire restricting Materials and Fire Technical Equipment.
Type approved Products", Lists of approved Manufacturers and Type
approved Products, No. 8, 1989.
/9-7/ "List of Equipment approved by the Norwegian Maritime Directorate", 1990.
/9-8/ NT Fire 021: "Insulation of Steel Structures. Fire Protection", Nordtest, 1985.
/9-9/ Norwegian Petroleum Directorate (NPD): "Guidelines to Regulations concer-

ning Explosion and Fire Protection of Installations in the Petroleum Activities",
1992.
/9-10/ Association of Structural Fire Protection Contractors and Manufacturers Limi-

ted: "Fire Protection for Structural Steel in Buildings", 2nd Edition, 1988.
/9-11/ T.A. Haverstad: "Passive Fire Protection subjected to Gas Explosion and Fire
Loads", Doctor Engineer Thesis 1989:24, Division of Structural Engineering,
The Norwegian Institute of Technology, Trondheim, 1989.
/9-12/ R. Wighus: "Active Fire Protection - Extinguishment of enclosed Gas Fires with
Water Sprays", SINTEF Report STF25 A91028, Trondheim, June 1991.
/9-13/ L.M. Pietrzak and J.A. Ball: "A physically based Fire Suppression Computer-
Simulation-Definition, Feasiability Assessment, Development Plan and Appli-
cations", Mission Research Corporation, Santa Barbara, USA, April 1983.
/9-14/ R. Wighus: "Water Mist Fire Suppression Technology - Status and Gaps in
Knowledge", 1st International Water Mist Conference 2001, Vienna, Austria, 4-
6 April 2001.
/9-15/ Dougal D. Drysdale: "An Introduction to Fire Dynamics", John Wiley and Sons
Ltd. 1985. ISBN O 471 90613 1.
/9-16/ Frank P. lncropera and David P. DeWitt: "Fundamentals of Heat and Mass
Transfer'', Second Edition, John Wiley & Sons, 1985, ISBN 0-471-82561-1.
Scandpower
/9-17/ Pr EN 12845: "Automatic Sprinkler Systems - Design, Installation and Mainten-

ance". Document is out for formal vote.
/9-18/ National Fire Protection Association: "NFPA 15: Standard for Water Spray
fixed Systems for Fire Protection", USA, 1985.
/9-19/ Matti Kokkala: "Extinguishment of liquid Fires with Sprinklers and Water
Sprays - Analysis of the Test Results", VTT Technical Research Centre of
Finland, Report 696, 1990.
/9-20/ L.C.Shirvill and G. White: "Effectiveness of Deluge Systems in protecting Plant

and Equipment impacted by High-Velocity natural Gas Jet Fires", Journal of
Hazardous Materials (Special issue).
/9-21/ David Evans and Dwight Pfenning: "Water Sprays Suppress Gas Well Blowout
Fires", Oil and Gas Journal, April 29, 1985.
/9-22/ C.A. Selby and B.A. Burgam: Blast and Fire Engineering for Topsides Struc-
tures - Phase 2. Final Summary Report", The Steel Construction Institute,
Silwood Park, Ascot, Berkshire SL5 7QN UK. SCI-P-253, ISBN 1 85942 078 8.
/9-23/ G.A. Chamberlain: "An experimental Study of Water Deluge on Compartment

Fires", International Conference and Workshop on Modelling and Mitigating the
Consequences of Accidental Releases of Hazardous Materials, pp 763-766,
Sept. 26-29. 1995.
/9-24/ C. Yao and A.S . Kalelkar: "Effect of Drop Size on Sprinkler Performance", Fire
Technology, Volume 6, Number 4, November 1970.
/9-25/ ISO 6182-1, First Edition: "Fire Protection - Automatic Sprinkler Systems - Part
1: Requirements and Test Methods for Sprinklers", Reference Number: ISO
6182-1 :1993(E), 1993-07-01 .
/9-26/ S. Kumar, G.M. Heywood and S.K.Liew: "Superdrop Modelling of a Sprinkler

Spray in a Two-phase CFO-Particle Tracking Model", Proceedings of the Fifth
International Symposium on Fire Safety Science, Melbourne, Australia, ISBN
4-9900625-5-5, 1997.
/9-27/ Steven R. Turns: "An Introduction to Combustion, Concepts and Application",

McGraw-Hill Book Co, ISBN 0-07-114783-7, 1996.
/9-28/ B.Z. Dlugogorski, R.H .Hickens, E.M.Kennedy and J.W.Bozzelli: "Water Vapour
as an inerting Agent", Halon Options Technical Working Conference, 6-8 May,
Albuquerque, New Mexico, USA, 1997.
/9-29/ R.J. Kee, F.M. Rupley, J.A. Miller: "Chemkin-11: A Fortran Chemical Kinetics
Package for the Analysis of Gas-Phase Chemical Kinetics", SAND89-8009B,
UC-706, Sandia National Laboratories, Albuquerque, New Mexico,
(1993a).http://www.sandia.gov/1100/CVDwww/chemkin.htm
/9-30/ Gerard G. Back, Craig L. Seyler, Phil J. DiNenno, and Richard Hansen: "Water
Mist Protection Requirements for very large Machinery Spaces", Report No.
CG-D-15-00, US Coast Guard Research and Development Center, March
2000.
Scandpower ~
Risk M anag ement AS ~ SINTEF
(
/9-31/ A.K. Lazzarini, R.H. Krauss, H.K. Chelliah and G.T. Linteris: "Extinction of
Counterflow Diffusion Flames with fine-Water Droplets", Halon Options Tech-
nical Working Conference, Albuquerque, New Mexico, USA, 2-4 May 2000.
/9-32/ NT FIRE 044, Fire Protection: "Fire extinguishing Media - Powder", Nordtest
method, 1991-11.
/9-33/ Petra Andersson and Goran Holmstedt: "Limitations of Water Mist as a total
flooding Agent In Lund University, Sweden", Report No, LUTVDG/(TVBB-
1015), 1997.
/9-34/ "FIREDASS - Fire Detection and Suppression Simulation - Final Technical

Report", Report No. 560-75919, Project No. BE95-1977, Project funded by the
European Commission under the BRITE/EuRam Programme, 14 April 1999.
/9-35/ Ragnar Wighus: "An empirical Model for Extinguishment of enclosed Fires with
Water Mist", Halon Options Technical Working Conference, Albuquerque, New
Mexico, USA, 12-14 May 1998.
/9-36/ Swedish National Testing and Research Institute, Brandteknik: "Proceedings

of the International Conference on Water Mist Fire Suppression Systems", SP
Report No. 1994:03, November 4-5, 1993.
/9-37/ Jack Mawhinney: "Design of Water Mist Fire Suppression Systems for Ship-
board Enclosures", SP Report No. 1994:03, Pages 16-44, Swedish National
Testing and Research Institute, Brandteknik, Proceedings of the International
Conference on Water Mist Fire Suppression Systems, Boras, Sweden,
November 4-5, 1993.
/9-38/ Ragnar Wighus, Petter Aune, Geir Drangsholt and Jan P Stensaas: "Full Scale
Water Mist Experiments", SP Report No. 1994:03, Pages 101-152, Swedish
National Testing and Research Institute, Brandteknik, Proceedings of the Inter-
national Conference on Water Mist Fire Suppression Systems, Boras,
Sweden, November 4-5, 1993.
/9-39/ NFPA 750: "Standard for the Installation of Water Mist Fire Suppression Sys-
tems", Draft Edition, 1994.
/9-40/ International Maritime Organisation (IMO): "Alternative Arrangements for Halon

Fire-extinguishing Systems in Machinery Spaces and Pump Rooms", MSC/
Circ.668, Adopted 30 December 1994, and MSC/Circ.728 (Revised) adopted 4
June 1996.
/9-41/ International Maritime Organisation (IMO): "Revised Guidelines for Approval of

Sprinkler Systems equivalent to that referred in SOLAS Regulation 11-2/12
(Resolution A.BOO (19)", 23 November 1995.
/9-42/ International Maritime Organisation (IMO): "Guidelines for the Approval of fixed
Water-based local Application Fire-fighting Systems for use in Category A
Machinery Spaces", MSC/Circ.913, 4 June 1999.
/9-43/ Gerard G. Back, Craig L. Seyler, Phil J. DiNenno, and Richard Hansen: "Water
Mist Protection Requirements for very large Machinery Spaces", Report No.
Scandpower
G)SINTEF
Risk Management AS
CG-D-15-00, US Coast Guard Research and Development Centre, March

2000.
/9-44/ Magnus Arvidsson: "Pyroteknisk genererade aerosoler for brandslacking", SP

Rapport 2001 :28, Swedish National Testing and Research Institute.
/9-45/ IMO MSC/Circ. 1007: "Guidelines for the Approval of fixed Aerosol Fire-
extinguishing Systems equivalent to fixed Gas Fire-extinguishing Systems", as
referred to in SOLAS 74, for machinery spaces.
Scandpower
10. Smoke
10.1 Methods for Prediction of Smoke Behaviour
Smoke is one of the major hazards in fires. The reduced visibility due to thick, black
smoke is the first threat to people who want to escape from or fight a fire. Knowledge
of smoke production, smoke flow and impact of smoke on people and constructions
are available from literature, from laboratory tests and from experience from real fires,
like the fire on the Piper Alpha platform in the North Sea, Ref. /10-1/. In this chapter a
link between the pieces of information is presented. An assessment of smoke hazard
is described by a sequential procedure, shown in Figure 10.1.
Wind Ventilalion
natlll"al
Mass rate Reduced s t

SMOKE Scot
SMOl<E FLOW 1--'T~e:::m=ra~tu:;::re~--I HUMAN
GENERATION Toxic as&s AND NASIL
Tempera!lll"e DISPERSION Heat flux LIMITS
Toxic gas
concentration/dose
Fire enclosure
geometry Geometry of oo]ecl
Figure 10.1: Frame Work for Smoke Hazard Assessment
Smoke generation is studied at SINTEF-NBL by small and medium scale experiments

with liquid hydrocarbon pool fires in enclosures, Ref. /10-2/, /10-3/ and /10-4/. Theory
and available literature are also sources for analysis of smoke generation. A com-
puter code simulating transient hydrocarbon fires in enclosures has been developed at
NTH/SINTEF by professor Magnussen and his group, Ref. 10-5/. The three-dimen-
sional computer code KAMELEON FIRE E-3D estimates fire development and smoke
production at specified geometry and ventilation conditions, by modelling of the inter-
action between combustion, fluid flow and heat transfer.
Smoke flow and dispersion around oil rigs or other buildings and constructions are
simulated by the 3-dimensional computer code, KAMELEON 11, Ref. /10-6/. The com-
puter code calculates the velocity and temperature distribution three-dimensionally
around an object. The code may also be used to calculate smoke flow and dispersion
inside complex structures, like hotels, ships and other buildings. KAMELEON II was
used to reconstruct the smoke flow and dispersion in the catastrophic fire of the
passenger ship Scandinavian Star, Ref. /10-7/
Production rates of soot and toxic gases are conversed to concentration at various
positions. Conversion factors between smoke temperature and length of vision and
carbon monoxide concentration respectively, may be developed specially for this pur-
pose.
IG/)SINTEF Scandpower
Risk Management AS
Impact on people is assessed by tenability limits for soot, temperature and toxic
gases. The tenability limits are taken from literature, reported separately by Stensaas,
Ref. /10-8/. This is also treated in detail in Chapter 8.2.
10.2 Smoke Generation

10.2.1 Introduction
The term "smoke" in this chapter means the total output of a fire, including fuel vapour,
combustion gases, excess air, liquid droplets and solid soot particles.
Smoke generation in enclosed hydrocarbon fires is mainly controlled by fuel properties

and ventilation conditions. The nominal ventilation factor a, the air-to-fuel ratio, de-
fines certain regimes of a fire, from well below stoichiometric, characterised by oxygen
starvation, through close to stoichiometric fires, to well ventilated fires. The smoke
generation data presented here are mainly based on experiments in a 2.5 x 2.5 x 5
(m3 ) model compartment, containing a liquid pool fire, Ref. /10-3/, /10-4/ and /10-9/.
The data are compared with literature data and theory.
A set of empirical correlations representing the production of toxic gases and the con-
centration of oxygen in an enclosed liquid pool fire is presented in this chapter, based
on the experiments referred. The correlations may be used as a first estimate of
smoke production. A more accurate estimate of smoke generation can be obtained by
using computer models.
The computer model KAMELEON FIRE E-3D is a 3-dimensional fluid dynamic code
including the production of soot by combustion modelling. This model is used for cal-
culations of smoke production on a real scale, and the results from this simulation may
be used as a reliable estimate for smoke production of hydrocarbon fires.
The example scenario described further in this chapter is a fire in a process module of
an offshore oil rig.
10.2.2 Mass Flow Rate

An important fire scenario in offshore platform fires is the enclosed fire. The estimate
of smoke production in this case is different from open fires, since the combustion
products can be recirculated into the flames. The enclosure itself may then be con-
sidered as a combustion chamber, and the mass- and heat balance for the enclosure
is the basis for smoke production rates, concentration of soot and gases and tempe-
rature estimates.
A rough estimate of smoke production rate in an enclosure is to take the fuel burning
rate multiplied with the amount of air needed for complete combustion. For most
hydrocarbon fuels this amount is about 15 times the fuel burning rate. This number
can be noted ka, the mass air-to-fuel stoichiometric ratio. In fires where the air supply
is sufficient for complete combustion the total mass rate of smoke can be calculated if
the fuel burning rate is known. The total smoke production rate is:
ITltot = ITlt + ka · ITlt =(1 + ka) · ITlt : : : 16 · ITlt Eq. (10.1)
Scandpower
I
I
where:
111101 = total mass flow of smoke from the outskirts of the flames
m, = fuel burning rate
ka = mass air-to-fuel stoichiometric ratio
This gives a first step estimate of smoke generation, to give an idea of the amount of
smoke at high temperature which is produced.
The smoke produced by the fire is then mixed with entrained air, a process which
goes on inside the fire module if access air is available, or outside the module when
the smoke escapes. This mixing process dilutes the smoke, leading to reduction of
temperature, soot concentration and concentration of toxic combustion products. The
oxygen concentration increases by mixing with fresh air. The prediction of mass pro-
duction rate of smoke is determined by the fuel burning rate, and the ventilation of the
fire enclosure.
The prediction of the fire development in terms of fuel burning rate and ventilation
conditions is a rather complex task. Methods for this type of prediction are advised in
the final report from the research project "Modelling of hydrocarbon fires offshore",
Ref. /10-10/.
10.2.3 Temperature
The temperature of the smoke leaving a fire enclosure is a result of the burning con-
ditions. If the assumptions of "worst case" for fire severity are taken, a maximum
smoke temperature will be up to "' 1,200 °c. In most cases a temperature significantly
lower than this can be expected. It is not obvious that "worst case" fire development
inside the fire enclosure leads to the worst hazard to people on board a platform,
regarding smoke. A very hot smoke plume developing from an opening in a fire com-
partment will have high buoyancy, and will tend to rise rapidly. In some cases this will
be favourable, since smoke can flow above the platform body. A cooler smoke plume
may easier follow the flow field of air around the platform, and may conflict with
escape routes and life boat stations at lower levels.
Computer simulation of fire in the actual modules will give more confident input to
modelling of smoke dispersion.
10.2.4 Soot Production

Soot production rates are given as soot yield, Y5 (g/g) , expressed by a ratio of soot
production versus fuel burning rate. The soot yield varies with fuel type, geometrical
configuration of the fuel within the fire enclosure and air-to-fuel ratio of the combus-
tion. The air-to-fuel ratio is denoted ex. Traditional building materials and furniture
have a soot yield typically in the order of 1-2 %. Liquid and gaseous hydrocarbon
fuels may have characteristic soot yields significantly above this. Soot yield of enclo-
sed liquid hydrocarbon pool fires is studied experimentally in this project, and results
are presented in a SINTEF report, Ref. /10-9/.
Some of the conclusions regarding the SINTEF experiments with enclosed hydrocar-
bon fires are given here:
Scandpower
The soot/fuel ratio, the soot yield, is typically 7 - 10 % (0.07 - 0.10 g/g) in enclo-
sed liquid hydrocarbon pool fires. The soot/fuel ratio is the production rate of
soot divided by the fuel evaporation rate
The soot yield is quite insensitive to air excess for a fuel with a boiling point
range 140 - 165 °c, tested in the experiments
A fuel with lower boiling point range, 62 - 82 °c, gives less soot/fuel ratios in well
ventilated conditions, typically in the order of 2 %. The soot/fuel ratio is increa-
sing significantly, up to 10 %, when the air-to-fuel ratio becomes closer to the
stoichiometric ratio, ct = ka = 15
Variation of fuel preheat temperature shows little effect on the soot/fuel ratio
Introduction of small objects, like a rack of pipes in the combustion zone, may
lead to less soot/fuel ratio. The reduction may be explained by increased tur-
bulence intensity, more effective mixing of air and fuel, and subsequently higher
temperature.
10.2.5 Carbon Monoxide

The concentration of carbon monoxide (CO) in the compartment and at the
outlet is typically less than 0.1 % for well-ventilated fires. This is characterised
by an air-to-fuel ratio a > 50
The concentration of CO in fires with limited air supply goes up to about -;::: 2 %
when the air-to-fuel ratio becomes close to stoichiometric, a = ka = 15. CO con-
centration above 0.5 % can be found in fires with air-to-fuel ratio up to three
times the stoichiometric ratio
For ventilation controlled fires and fires with oxygen starvation CO concentra-
tions of maximum 35 % can occur, from theoretical considerations. In practice,
a CO concentration of 5 %, at air-to-fuel ratios about half the stoichiometric ratio,
has been measured in diffusion flames in poorly ventilated enclosures. This
may be considered as a maximum CO concentration in offshore fires
Correlations for the concentration of CO versus the air-to-fuel ratio have been
developed. A correlation (correlation 1 in Figure 10.2), found by linear regres-
sion of the measured values is:
Ceo (vol%)= 429. (ar 2.4, Cl 2::: 8 Eq. (10.2)
where:
Ceo = volume fraction of carbon monoxide, %

ct = air-to-fuel ratio (mass rates)
A more conservative correlation (correlation 2 in Figure 10.2), representing the

maximum of the measured values is:
Ceo =1 .7- (kafa)2·5 , ct2::: 15 vol% Eq. (10.3)
where:
ka = stoichiometric air-to-fuel ratio, 15 for the fuels used in the experi-

ments
It is important to observe that these correlations are valid for a specific experimental
set-up and the types of fuels used here only. They are only verified for air-to-fuel
Scandpower
Risk M anagement AS ~ S I NTEF
,/
'
ratios from stoichiometric (a = 15), and above. For air-to-fuel ratios below 15, a con-
centration of about 5 % can be used as an approximation.
These correlations together with CO concentration measured in medium scale experi-

ments are presented in Figure 10.2.
co concentratron versus atr/~uel r~~ro
..... oo
3,50
3.00
2.50
:iiLOO
1 :so
1 00 0
0 :so
CJ
□
0 00
a 20 ◄o ISO ao soc
Atr/~u~I ratio C-J
0 COr-r•ta'C.1on 2
Figure 10.2: Correlations for Volume Concentration of Carbon Monoxide
10.2.6 Carbon Dioxide

Carbon dioxide concentration varies with air-to-fuel ratio, and a maximum concentra-
tion of 14 % has been measured at the outlet opening in the experiments by SINTEF.
The variation of the CO2 concentration with the air-to-fuel ratio has been characterised
by a correlation valid for the type of enclosed fire examined in the 1/4 scale module. It
is only valid for fires with an air-to-fuel ratio above stoichiometric ratio. The correlation
for the carbon dioxide concentration valid for these experiments is:
Cco2 [vol%)= 10.8 • (15/a) 0·88 , a ;,,.15 Eq. (10.4)
where:
Cco2 = volume fraction of carbon dioxide, vol%

ot = air-to-fuel ratio
This correlation is presented in Figure 10.3.
Scandpower
1:,.0
14.0
13,0 eaCb
12 0
11 0
~
a
!
C
10
8 .0
0
11 .a
~
~ 7.0
i
N
8 0
5 a
8 .. a
3.0
2 .0
1 ,0
o a
0 :io 40 80 80 1 00 1:10 1◄ 0
Ar r/~..-1 r •tlo C-J

0 .,....u,,..a va 1u•• - - Cc::arra I at ron
Figure 10.3: Carbon Dioxide Concentration versus Air/Fuel Ratio in Experiments in

1:4 Scale Offshore Module
10.2.7 Oxygen
Oxygen concentration at the outlet opening varies with air-to-fuel ratio, from close to
zero at stoichiometric ratio, up to the concentration of ambient air at large air excess
ratios. An empirical correlation valid for the types of fires examined in the experiments
is developed. A correlation for the oxygen concentration valid for the fuels used in
these experiments is:
08
C o2 = 20.8 · (( a -15)/a) · , a~ 15 vol% Eq. (10.5)
where:
c0 2 = volume fraction of oxygen, %

a = air-to-fuel ratio
For an air-to-fuel ratio less than stoichiometric one may expect an oxygen concentra-
tion down to zero.
10.3 Smoke Flow and Dispersion

10.3.1 Introduction
Smoke from fires in naturally ventilated areas will escape through openings and
louvered walls. The hot gases are lighter than air, have buoyancy and will create a
smoke plume. The environment around the platform, characterised by wind direction
and velocity, will influence the smoke plume. Both the smoke plume and the wind
interact with the platform itself, creating areas with increased flow velocity and recir-
culating zones.
Scandpow er
Risk M an agement AS ~ SINTEF
I
Important questions to answer are:
Can smoke be transported downwards after leaving an opening from the fire
area, and be dragged below the platform?
Can smoke flow upstream the wind direction?
Will smoke enter the air intake for the living quarter ventilation system?
Will smoke penetrate into a pressurised living quarter?
How fast will smoke infiltrate a living quarter if ventilation systems are shut
down?
Which areas will be threatened by smoke in specified fire scenarios?
What is the major impact of smoke on people?
10.3.2 Basis for Prediction of Smoke Flow Patterns

In this context the flow patterns of smoke are supposed to be predicted by the com-
puter model KAMELEON 11, Ref. /10-6/, or other compatible computer codes.
KAMELEON II is a 3-dimensional general code for prediction of flow and dispersion of
gases in complex geometry, with wind influence and density differences.
The input to the flowfield modelling is a typical geometry of the actual object, here an
offshore North Sea integrated oil rig, wind direction and velocity, and smoke produced
by specified fires in various positions.
The scenario chosen is described in Table 10.1.
Tabl e 10 1 p arameters for Frre S cenanos srmu Iated bIV KAMELEON II

Fire location Wind
Direction relative to platform length axis Velocity at level 10 (m)
above sea
Process module Parallel, heading towards the flare boom 5 (m/s)
The rates of smoke flow represent the output of liquid pool fires. The pool inside a
module is 180 m2 , and it is located close to one end wall of the module. The offshore
platform module is 19 m wide, 47 m long and 27 m high. The smoke flows out of
openings with louvres at the end walls.
The leakage rate is chosen as sufficient to supply the pool at a rate proportional to the
evaporation rate. At a stationary situation the fuel burning rate was 21.8 kg/s. The fire
was then still well ventilated. The evaporation rate of fuel is determined by calculation
with the computer programme KAMELEON FIRE E-3D, Ref. /10-5/.
In the scenario the fire is located in a process module, between the drilling area and
the living quarter. The smoke leaves the module through the louvred walls in the
upper part.
Scandpower
(Gj)SINTEF Risk Management AS
NoJWegian Fire Research Laboratory
10.3.3 Results of Flow Field Modelling

Results from the calculation are shown in Figure 10.4 and Figure 10.5. The iso-
contour plots of temperatures, CO concentration and visibility represent time-averaged
values. Fluctuations will in practice occur. Smoke will then infiltrate other parts of the
platform than shown at the pictures, as gusts of dense and less dense smoke.
Figure 10.4 is a side view of the platform, taken at a position just outside the side wall.
This is indicated by Y-coordinate.
Figure 10.5 shows a view from above, taken at a level approximately at the middle of
the platform height.
The iso-curves represent a temperature, a carbon monoxide concentration and a view

range at the same time. The correlation between these properties of the smoke is
representative for scenarios as specified in this project, and should not be adapted
universally.
In parallel wind the calculated smoke plumes are almost identical at each side of the
fire module. The plumes have substantial buoyancy, and keep clear of both lower
parts of the platform and the lifeboat station at the lower part of the living quarter. The
smoke engulfing the upper part of the living quarter and the helicopter deck has a tem-
perature of less than 5 °c above ambient. CO concentration of the smoke engulffing
the platform is in the order of 75 ppm, which is of no importance as a threat to people.
Reduced visibility through smoke is the major threat to evacuating people.
More information of local temperatures is available in data files from the computations.
By means of a post-processor included in the computer programme KAMELEON II,
Ref. /10-6/, the user may easily look for these details on the screen of a personal com-
puter.
10.3.4 Scaling of Smoke Properties to other Fire Sizes

The smoke properties shown in the graphs represent a specific fire situation, with a
certain wind speed and direction. Accurate predictions of length of vision, temperature
and concentration of toxic gases in another fire scenario require a new calculation . It
is, however, possible to estimate the smoke properties for other fire sizes, provided
that the fires are not too different from the calculated fire situation. The fire which is
simulated in the process and drilling area is well ventilated. Even in well-ventilated
fires flames may occur at the outlets of the walls. A scaling procedure as demon-
strated here is suppositioning no flaming combustion from the reference point of the
scaling. The mutual relationship between the temperature and the concentration of
soot and carbon monoxide will remain constant throughout the flow field , provided that
the flow does not interact with cold objects.
Scandpower
I
r Handbook for Fire Calculations and Fire Risk Assessment in the Process Industry Page 10-9
E 0 --
T,miwr:11urc lll":11~
100 .oo
50.00
-
-
532.03
100 .oo
- -
0 30.00 50.00
N 0 CJ 10.00 - 30.00
0)
lilil
5.00
0.50
-- 10.00
5.00
0
lO CJ o.oo - a.so
0 Owtrlemi>. View fmJ CIH:Onl'mtr:atiora (vol%)
"
0
100
so
0.02
O.OJJ
O.IS
0.075
N 30 o.oso 0.0.&S
JO 0.15 0.015
0 +-.---,--j«Sl....-r-""'-,-"''--r-"",----r--+
80100120140160180200220240260
s
o.s ,...
0.28 0.007S
0.00075
r,..;1iJ>n: y (ml 64 • 50 X ( rn J
Figure 10.4: Isa-curves describing Temperature Field around an Offshore Oil Rig on
Fire. Smoke Dispersions calculated by KAMELEON II. (Side View)
--
Temper:iture sc=ile
100.00
so.co -- SGS.CO
-
iOO .00
CJ
30.00
10.00
-
-
50.00
30,00
s.oo
a.so
-- 10,00
0 s.oo
N CJ a.co - a.so
0
0
0
ro Overtnnp. View[m] CCH:oncentration (vol%]
0 100 0.02 0.1S
tl)
50 0.033 0.075
0
V 30 o.oso 0.045
10 O,JS 0.01S
0
N 5 0.23 0.0075
0 0.5 2.80 0.0007S
ao1001201401so1ao2002202402so
Posi1io11: Z 1ml 56.80 X (mJ
Figure 10.5: Isa-curves describing Temperature Field around an Offshore Oil Rig on
Fire. Smoke Dispersions calculated by KAMELEON II. (View from
above)
~SINTEF Scandpower
Risk Management AS
\
N
E 0
0
co
0
(\J -t-;---.---,"-"""'
0
--
Tempenuure scale
tZJ
D
30.00
10.00
5.00
-
-
-
a .so -
0.00 -
40.64
30.00
10.00
5.00
0 . 50
0
lO
0 Ovrrtemo. View(mJ Co-concentra lion [vol%}
'It 100 0.02 0.1S
0 so 0.033 0.o7S
(\/· 30 0.0SO 0.04S
10 0.1S 0.015
0
--t--r---,---,----<,,-LC--Y'--,_;...:=-,,:.:.:...:....:,.;;--'-+ s 0.28 0 .0075
0 20 40 60 80100120140 160180 0.5 2.80 0 .0007S
l'm,ition:
X Cm l 124 .00 Y Cm l
Figure 10.6: lso-curves describing Temperature Field around an Offshore Oil Rig on
Fire. Smoke Dispersions Calculations by KAMELEON II. (View to-
wards the Living Quarter)
The flow pattern of smoke in the flow field of wind will be determined by the mass flow
rate and the temperature of the smoke. The main pattern will, however, remain quite
similar, even if the temperature is increased. The rise of the smoke will be higher at
higher temperatures.
A correlation for another soot concentration or a carbon monoxide concentration will

then depend on the travel distance of the smoke plume. A first estimate will be to
make the soot concentration, the concentration of toxic gases and temperature directly
proportional to the source conditions. For a fire scenario with a fire in a process
module, as calculated by KAMELEON II, the temperature rise above ambient tempe-
rature is 483 °c at average at the ventilation opening. If the temperature is supposed
to be 600 °C alternatively, the excess temperatures in every position can be multiplied
proportionally by:
(600/483) = 1.24
Similar "scaling" exercise can be done with toxic gas concentration and with soot
concentration. The length of vision scales proportionally with soot concentration, so
this can be calculated by the equations in Chapter 10.4.1.
This type of "scaling" must be done carefully, considering all factors affecting fire
development and smoke flow. If for example a fire source in an enclosure is increa-
sed to ventilation control, the production rate of carbon monoxide increases dramati-
cally, and the ratio between excess temperature and CO concentration must be adjus-
ted.
10.3.5 Smoke Infiltration into a Building

A concept of infiltration of gas and small airborne particles into buildings is presented
in Ref. /10-11/. It is based on tests of onshore buildings, and presupposes that a
building tightness parameter is known. The equation is:
W/Wo =1 - [1 - e <·TJ•at>]/(11-~t) Eq. (10.6)
Scandpower
where:
W; is the dose to the species inside the building (units . h/m3)

Wo is the dose to the species outside the building (units - h/m 3)
11 is the ventilation rate (air changes/h)
t,.t is the time of exposure (h)
Natural ventilation rates for normal buildings onshore are measured to be 11 = 0.07 - 3
air changes/h, Ref. /10-11/. A ventilation rate of 0.5 air changes/hour is typical for a
building in a Norwegian coastal area, Ref. /10-12/. Supposing a similar ventilation rate
for a living quarter with the ventilation system shut down, and an exposure time of 1
hour, the ratio of dose to species inside and outside will be in the order of
W,/W0 = 0.20
This means that the dose of exposure inside the living quarter is 20 % of the dose out-
side.
This example calculation is a guidance to a method, and several critical parameters

are not verified. The ventilation rate, specified by number of air changes per hour, has
to be considered carefully. The tightness of the actual building can be estimated from
Ref. /10-12/, or may be based on more accurate ventilation calculations.
A typical living quarter with steel panels, sealed windows and air locks, will be more air
tight than onshore buildings. However, with the impact of a severe fire outside the
living quarter, a natural ventilation rate may be increased due to deformation of the
construction and doors.
The function of dose of gas exposure inside versus outside a building, depending on
natural ventilation rate, is shown in Figure 10. 7.
Dose of smoke inside a building

1.0 Function of natural ventilation rate
. 0.9 ..,..,.., ~ ... ,. . "

,. I" • •
... ...
••• ... . ••• ...
I " ' ... ,.
"
. . 0.8
. ."
..,• •"·
,xx xx•
..... ....... •••

,---
.
•• •
:B
'5 0.7
~ 0.6
.,,
.;
. .., X
X
X
.... ••"'
,• ,
C
•"'
.. "
X
,; 0.5 X
A
~
""0.4
X
•"" A IAA<> ~<>•
l •• ,...
]i 0.3 ~
12.
... X
"" . !,oOO IO~
~·
~ +++
0.2 ~o<> +++ +++ +++
0,1
A
...... <>9°
...... .. ++ +++ +++ +++ +++
~...: +++ +++

0.00
2 4 6 8 10 12
Time or &XJlO&Ul'8 (hours)
Ventilallonrate: +0.05 00.1 A0.5 •1 Y3
Figure 10. 7: Calculated Values for relative Dose of an Airborne Component inside a
Building, as a Function of natural Ventilation Rate
G)SINTEF Scandpower
Risk Management AS
NOf'Wegian Fire Research Laboratory
For a fixed, constant concentration of species outside the building, Ca, the dose inside
can be calculated by Eq. (10.7).
Wi =Co · {Llt -1 /11 ( 1 - [e<·l]·6 1>]} Eq. (10.7)
Resulting plots of the inside concentration for different ventilation rates is shown in
Figure 10.8.
Dose of smoke i ns i de a bu i l di ng
Concentrat i o n o ut srOQ : 15 0 0 pp
:50000
<100 00
,.,
C
e 30000
E
0.
a.
L.J
<D 20000
(0
10000
0
0 20 '10 tl O 80 10 0
Ti me or expos ure [ mrnutGGJ

a 0 .1 + 0.5 <> 1 A 3 V
Figure 10.8: Dose to Species inside a Building exposed to an outside Concentration

of 1,500 ppm. The Graphs represent different Ventilation Rates, see
below
A critical dose inside the living quarter can be found using the graphs in Figure 10.8.
A person staying outside the living quarter would experience a carbon monoxide expo-
sure of about 1,500 ppm in case of fire at the sea level. A dose value of 35,000 ppm .
min is critical for incapacitation. Details of this are given in Chapter 10.4. An expo-
sure of a CO concentration of 1,500 ppm would lead to human incapacitation after
(35,000/1 ,500) = 23 min. Inside a building, supposing a natural ventilation rate of 0.5
(air changes per hour), a dose critical for human incapacitation could be reached
within 1 hour 20 min.
This calculation has no experimental verification. Local variations of smoke infiltration

inside a living quarter will normally occur, due to external wind. The critical parameter
is to define a value for the natural ventilation rate for a typical living quarter, with wind
influence.
Scandpower
10.3.6 Aspects of Smoke Infiltration

The smoke hazard experienced inside a building depends on whether the ventilation
system works or not.
If the ventilation system is closed down, the infiltration of soot and toxic and irritant
gases will be determined by the natural ventilation rate, characteristic for the building.
This factor is the key to estimate smoke infiltration into a building.
The same offshore oil rig as used in the example fires in this chapter is chosen as a
reference. If one assumes a natural ventilation rate of 0.5 air changes/hour, which is a
normal infiltration rate for onshore buildings in a Norwegian coastal climate, a critical
dose of CO for incapacitation is reached within 1 hour 20 minutes, with a concentra-
tion of 1,500 ppm surrounding the living quarter. An actual ventilation rate may be
both lower or higher than this. Then the concentration of CO may be critical inside the
living quarter as well.
The soot concentration will probably be the critical factor inside the living quarter,
since reduced visibility occurs much earlier than critical CO dose.
10.4 Impact of Smoke

The hazard of smoke is characterised by three factors. The first threat is reduced visi-
bility due to soot. The second is that hot smoke can cause pain and injuries, and the
third is that a concentration of toxic and irritating components can lead to incapaci-
tation or death. The relative order of these factors can be found by comparison of
threshold values with actual exposure in a fire scenario. The threshold values are
treated in a special report, Ref. /10-8/. This is also treated in Chapter 8.2.
10.4.1 Reduced Visibility

When a soot concentration is determined by modelling of soot production rates and
dispersion, it can be converted to a range of vision. Soot particles block light passage,
and the range of vision is determined by the strength of the light source, the soot
concentration and to some extent the wavelength of the emitted light.
The visibility of an object is determined by the contrast between the object and its
background. Light emitting signs are for example two to four times easier to see than
light reflecting signs.
Scandpower
~
10 I ,A I
•
s
-
§. 2
>,
=
:0
•
·en 1
, owsed lamp in yeUow smoke7G
e F
5 VFocus&d lamp in black smoke19
0.5 Clllumlnafed s lgn21'
0 lllumineted sign2°
■ Oiffl•se nglft'<?
:;:;: Focused lamp in yellow smolcett' 0
it.Bacldighted sign-
0.2 Allluminated placard~
Ylllumlnated sJgnt
0.1,____,___.__._..............._ __,,~ .............,,....,,,,..................__....,___,,_.~......~

1
0.02 0.05 0.1 0.2 0.5 1 2 5

DJL, optical length of smoke (m·1)
1
Figure 10.9: Visibility of a Front illuminated Sign versus optical Density (per Meter
Smoke Thickness), Ref. /10-13/
A way to find visibility is to use the basic correlations for light obscuration described by
Beer- Lambert's law:
Eq. (10.8)
where:
I is the transmitted intensity of light, -

l0 is the emitted light intensity, -
1t). is the extinction coefficient of soot particles, m-1
Lis the path length of the light, m
According to Newman and Steciak, Ref. /10-14/, the soot volume fraction can be ex-
pressed by the equation
fv = 1t). • ')Jc Eq. (10.9)
where:
fv is volume fraction of soot, (-)

1t). is the extinction coefficient, (m- 1)
A is the wavelength of the transmitted light, (m)
c is a mean value for the ratio of extinction coefficient of particulate volume

fraction, found to be equal to 7.0 for various materials.
Scandpower ~
Risk Management AS I\Jv SINTEF
The attenuation of a light beam through smoke is also expressed by the so-called
optical density, D, (·).
The optical density per meter path length of light is defined:
D/L = -1/L (log,o(l/I 0 )) (m-1) Eq. (10.10)
A combination of Eq. (10.8), (10.9) and (10.10) gives an expression for the optical
density per meter path length:
D/L = fv ·Ch, · log, 0e Eq. (10.11)
For light seen by the human eye the range of wavelengths is between 0.4.10-s -
0.7-10-s m. A mean wavelength of 0.55-10-s mis used.
The following correlation can be used to calculate the optical density from volume
fraction of soot from hydrocarbon fires:
D/L = 5.5 • fv Eq. (10.12)
when fv is given in cm3/m3 .
The conversion from optical density to visibility can be done by an empirical corre-
lation, Ref. /10.13/, based on the visibility vs. optical density shown in Figure 10.9:
S = 1/(D/L) for a front illuminated sign Eq. (10.13)
S = 2.5/(D/L) for a light emitting sign Eq. (10.14)
This indicates a visibility of 1 m for an optical density of 1 m·1• Combination of Eq.

(10.12) and (10.13) gives a direct expression for the visibility of a front illuminated sign
and soot volume fraction:
S=0.18/fv Eq. (10.15)
when fv is given in cm3/m3 .
From calculations of fire development in an offshore module a soot mass fraction of

the smoke, fm, leaving the fire module was calculated. The corresponding temperature
was 500 - 600 °c. This means that visible flames are leaving the outlet from the fire
module. Supposing that the soot contents at the outlet is preserved in the smoke
plume, and only diluted with air, a soot volume fraction is calculated from the soot
mass fraction. In the calculations it is supposed ideal gas and negligible soot volume,
and a soot particle density of:
Ps = 1.1 (g/cm 3), Ref. /10-14/.
The smoke density is taken as for air at the actual temperature.
The soot volume fraction is then:
Eq. (10.16)
Scandpower
where:
Psmoke is the smoke density, taken as the density of air, (kg/m3 )

Ps is the density of soot, (g/cm 3)
Po is the density of air at reference temperature, (kg/m 3)
To is the reference temperature, (K)
Ts is the smoke temperature, (K)
Further on, a correlation between smoke temperature and soot mass fraction may be
derived from the calculated results of smoke flow and dispersion, performed by
KAMELEON II.
A correlation between actual smoke temperature and visibility valid for the visibility of
a front illuminated sign was derived for the actual situation:
S = 0.15 · (~T+T0 )/~T Eq. (10.17)
where:
S is the visibility of a sign with front illumination, (m)

~T is the smoke temperature rise, (K)
T0 is the reference temperature, taken as 273 K
Calculated view lengths based on this correlation are given in Table 10.2.
Corresponding concentrations of carbon monoxide are also derived from temperature,

and the resulting concentrations are given in Table 10.2.
The visibility vanishes quickly, and a temperature increase of 50 °C represents a visi-

bility of r:z 1 m. The corresponding CO concentration is then in the order of 750 ppm.
This is an indication of the importance of reduced visibility as the first occurring hazard
of the smoke.
The correlations for length of vision and concentration of CO are strictly linked to the
example scenario, and are not universal. If the concentration field and lengths of
vision are to be calculated for an other fire scenario, a representative estimate for the
smoke temperature, the soot particle concentration and the concentration of toxic
gases at the outlet of the fire compartment have to be found .
Table 10.2: Calculated Visibility of a Front illuminated Sign and CO Concentration

versus S moke over Temperature
Over-temperature, K 100 50 30 10 5 1 0.5
Visibility, m 0.6 1.0 1.5 4.2 8.3 41 .1 82.1
CO concentration, ppm 1,500 750 450 150 75 15 7.5
Scandpower ~
10.5 Smoke Control Systems

10.5.1 Introduction
Two different philosophies exist for control of smoke in fires, smoke ventilation and
pressurisation.
Smoke ventilation or thermal fire ventilation is based on removal of the produced

smoke, to prevent fire spread and to promote safe escape and manual fire-fighting.
The most frequent used systems are based on thermal buoyancy as the driving force,
with automatically opened hatches for smoke evacuation. Mechanically driven smoke
extraction systems are less frequently used, and normal ventilation systems are till
now scarcely used for smoke evacuation. The volume of smoke is too large for a nor-
mal ventilation system to deal with.
Pressurisation of escape routes and other important parts of a building is easier to

employ, if the normal ventilation system is used. The automatic control system may
be a prescribed procedure based on fire detection, or a manually run procedure con-
trolled by operators.
Extra ventilation systems to pressurise escape ways may also be used.
10.5.2 Pressure Differences

Normal pressure differences occurring in buildings are created by mechanical ventila-
tion systems, by thermal buoyant forces due to temperature differences and by exter-
nal wind influence. These forces decide the flow patterns inside a building, and domi-
nate the smoke flow pattern in an early stage of a fire. The flow and dispersion of
smoke from developed fires is dominated by the thermal buoyancy of smoke, which
may override some of the factors present in a building without a fire.
The pressure difference created by temperature difference between the inside of a

building and the surrounding atmosphere is called the chimney effect. This effect can
be calculated by an accelerating force, determined by the acceleration of gravity and
the difference of gas density. For normal conditions this buoyancy force gives pres-
sure differences in the order of 10 - 20 Pa.
«:, SINTEF Scandpower

Risk Management AS
•
Figure 10.10: Pressure Differences created by Temperature Differences between a
Building and the Ambience
The pressure difference created by a temperature difference may be calculated by:
~p = (po - Pi ) · g · H Eq. (10.18)
where:
~p the pressure difference between the inside and the outside of a building
Po ambient air density
Pi air density inside the building
g acceleration of gravity, m/s2
H height representing the pressure, m
The density of air or smoke is calculated by the ideal gas equation:
P = Pref (T retfl") E. (10.19)
where:
p actual density, kg/m 3

Pref density at reference temperature, kg/m 3
Tref reference temperature, K
T actual temperature, K
The density difference of hot inside air or smoke and ambient atmosphere may be
calculated by:
Eq. (10.20)
The pressure difference created by temperature differences may then be calculated

by:
~p = Pref {(Tref /To) - (Tref /Ti)} · g · H Eq. (10.21)
Scandpower
Typical pressure difference created by the chimney effect in fires in relatively tight
buildings may be in the order of 50 Pa.
In very tight enclosures, like submarines, concrete shells of atomic energy reactors or
concrete shafts of an offshore oilrig, the pressure due to a fire may be much higher
than this. The expansion of the enclosed gases due to increased temperature will be
added to the chimney effect.
This pressure increase may also be calculated, assuming ideal gas, constant volume
of the enclosure and negligible increase of total mass (mass burning rate is neglec-
ted):
Eq. (10.22)
where:
t,p pressure difference, Pa

T, initial room temperature, K
T2 actual room temperature, K
This type of pressure increase dominates totally over chimney effects, but will seldom
occur in normal fires, where leaky building construction will relieve the pressure.
As a comparison, the pressure differences making opening a door difficult is in the

order of 50-100 Pa, whilst the pressures created by thermal expansion may be in the
order of 1 bar (10 5 Pa)
10.5.3 Ventilation Systems

A normal ventilation system has a pressure drop from inside a supply channel to a
room in the order of 100 Pa. The same pressure drop is normal in extract channels.
A system operating at normal conditions provides an overpressure sufficiently high to
prevent smoke infiltration into the air supply channels. Pressure conditions in the ex-
tract system normally also protect against smoke-spreading into other parts of a
building.
The best protection against smoke-spreading through ventilation channels is to keep

the system operating in a fire situation. The only situation where the system should be
shut down is where the ventilation system has recirculation of extract air in the supply
system.
If a ventilation system is shut down in a fire situation, the channels may spread smoke.
To provide protection against this, and to provide a barrier against fire spreading
through partitions, fire dampers are installed in the ventilation channels. Such dam-
pers are normally automatically operated by thermal activators. Fire dampers should
withstand a fire for the same time period as the partition it is part of, and are subject to
fire classification tests.
Thermal fire ventilation is provided by hatches in the ceiling, normally automatically

operated by thermal activators. To keep a layer with relatively little smoke at floor
level, air supply hatches should be installed, operating simultaneously with the roof
hatches. Rules and regulations for installation of thermal fire ventilation exist, and
Scandpower
there are several computer codes which may be used to predict the position of and the
temperature of the smoke layer in buildings with a specified fire inside.
A computer programme specially developed to predict the position of a smoke layer

with specified smoke hatch area and building dimensions is available. This is named
RVENT, Ref. /10-15/, and has the option to specify various types of fire development.
10.6 Comparison with the Piper Alpha Accident

This chapter has presented a systematic way to treat smoke hazard in fires, with the
focus on offshore platform fires. The fire example was in the process area, and the
threat from the smoke to people in escape routes and lifeboat stations outside the
platform has been treated. The report so far has described a calculation and judge-
ment procedure, combining a lot of different prediction tools. To evaluate the results
of this procedure, the only reasonable full-scale events to compare with, are acci-
dents. Some oil slick fires on sea surface, resulting from discharge of oil or gas con-
densate through the flare system, have occurred, but the only catastrophic event in
the North Sea is the fire and explosions leading to the loss of the Piper Alpha platform
in the British sector, Ref. /10-1/. In the accident, oil, gas and the whole platform were
burning, both on board and below the platform. The fuels were forming a fire more or
less like a massive fireball. Even though this situation is somewhat unlike the fires
simulated in this context, the smoke flow patterns may be compared .
The wind velocity during the Piper Alpha accident was low, in the order of 5 (mis), and
the wind direction was almost parallel with the platform length, heading towards the
flare boom. This is one of the worst situations one can expect regarding smoke flow
towards the living quarter.
Looking into the predictions shown by the graphs in Chapter 10.3, there are certain
points to compare with pictures from the Piper Alpha accident. The buoyancy of the
smoke was substantial in both simulation and in reality. There was little smoke flowing
downwards in the wake zone downstream Piper Alpha. The most significant observa-
tion from the accident is the massive fire itself, with flames from the sea level engulfing
the whole platform. The fire itself resulted from riser rupture, and involved all combus-
tible material on board the platform. This comprises both gas, liquid hydrocarbon and
solid material. Despite the massive fire, no indication of upwind smoke flow, except
where the smoke was hampered by the lower decks of the platform, was seen. The
same effect was seen in the calculated results.
One main conclusion is that one must expect upward smoke flow, and smoke tran-
sport in wind direction in fires. Local fires at higher levels will not affect lower parts of
the platform, except by heat radiation from flames. Fires at lower levels of the plat-
form will spread smoke in downwind positions, and may hamper evacuation from the
affected areas. In fires inside the platform body itself, there is a possibility of smoke
flow underneath the platform. This indicates that the position for air intake for venti-
lation underneath the platform is the best possible except for fires at sea level.
Fire on the sea surface beneath the platform results in total smoke engulfment in low
wind conditions. This gives the most severe smoke threat from smoke dispersion in
the open.
Scandpower
Risk Management AS {G)) SINTEF
/10-1/ J.P. Petrie: "Piper Alpha Technical Investigation - Interim Report", Department
of Energy, United Kingdom, September 1988.
/10-2/ J.P. Stensaas, K. Opstad, 0. Brandt, P.J. Hovde: "Modelling of Hydrocarbon

Fire Offshore. Studies of Pool Fires in a Room of Dimensions 1m x 1m x 2m",
SINTEF Report STF25 F86003, SINTEF NBL - Norwegian Fire Research
Laboratory, Trondheim, 1989.
/10-3/ K.O. Opstad, 0. Brandt, R. Wighus: "Experimental Modelling of Liquid Pool

Fires in an empty 1:4-scale Offshore Module", SINTEF Report STF25 F89001,
SINTEF NBL- Norwegian Fire Research Laboratory, Trondheim, 1989.
/10-4/ K.O. Opstad, 0. Brandt, R. Wighus: "Experimental Modelling of Liquid Pool

Fires influenced by Object in a 1:4-scale Offshore Module", SINTEF Report
STF25 F89015, SINTEF NBL - Norwegian Fire Research Laboratory, Trond-
heim, 1989.
/10-5/ J. Holen, B.F. Magnussen: "KAMELEON Fire E-3D - A Field Model for enclo-
sed Pool Fires", SINTEF Report STF15 F90010, SINTEF - Applied Thermo-
dynamics, Trondheim, 1990.
/10-6/ Professor B. Magnussen et al.: "Kameleon II: A Transient, 3-Dimensional

Computer Program for Fluid Flow, Heat- and Mass Transfer", developed at
NTH/SINTEF, Norwegian Institute of Technology (NTH), Division of Thermo-
dynamics, Trondheim.
/10-7/ "Scandinavian Star-ulykken, 7. april 1990", Norges offentlige utredninger,

NOU 1991: 1A og 18 ("The Scandinavian Star Disaster 7 April 1990", Norwe-
gian Official Ellucidations, NOU 1991: 1A and 18), Oslo, 1991.
/10-8/ J.P. Stensaas: "Human Tenability Limits with respect to Toxicity, Visibility and
Heat Stresses of Fire Effluents", SINTEF Report STF25 F88052, SINTEF NBL
- Norwegian Fire Research Laboratory, Trondheim, 1988.
/10-9/ R. Wighus: "Production of Soot and Toxic Gases in enclosed Hydrocarbon

Pool Fires", SINTEF Report STF25 A91006, SINTEF NBL - Norwegian Fire
Research Laboratory, Trondheim, 1991.
/10-10/ K. Opstad, R. Wighus, J. Holen, B. Hekkelstrand, J.P. Stensaas: "Modelling of

Hydrocarbon Fires Offshore - Final Report", SINTEF Report STF25 F89021,
SINTEF NBL- Norwegian Fire Research Laboratory, Trondheim, 1990.
/10-11/ F.P. Lees: "Loss Prevention in the Process Industries", Vol. I, Butterworths &
Co. (publishers) Ltd.), ISBN 0-408-10697-2, 1980.
/10-12/ Norwegian Standards Association: "Thermal Insulation. Calculation of the

Energy and Power Demand for Heating and Ventilation in Buildings", Norwe-
gian Standard NS 3031, 1987.
/10-13/ E.G. Butcher, A.C. Parnell: "Smoke Control in Fire Safety Design", E.&F.N.
Spoon Ltd., ISBN 0-419-11190-5, London, 1979.
Scandpower
(Gj)SINTEF Risk Management AS
/10-14/ J.F. Newman, J. Steciak: "Characterisation of Particulates from Diffusion

Flames", Combustion and Flame, (67), page 55-64, 1987.
/10-15/ R. Wighus, 0 . Meland: "Smoke venting of Liquid Pool Fires in a large Indus-
trial Hall", Proceedings from 6th International Symposium on Loss Prevention
and Safety Promotion in the Process Industries, Vol. 11, Oslo, 1989.
Scandpower
11. Fire Risk Assessment

11.1 General
Basically there will be a fire risk in any location where flammable materials are stored,
transported, processed or handled. The risk will vary depending on a great number of
factors:
Nature of the flammable material

Quantity of the flammable material
How the material is stored, transported, processed or handled.
Presence of oxygen
Presence of potential ignition sources
Probability of unintentional release of the material
Environmental conditions
Type of risk reducing measures that are implemented
What is at risk• people, investments, environment
Etc.
When evaluating the fire risk associated with a system or plant the problem must be
approach in a systematic manner in order to pinpoint factors which contribute to the
risk and where risk reducing measures can be implemented in a cost-effective way.
The topic of fire risk assessment is complex and a number of methods are available to
approach the problem. In the following a general and overall methodology is presen-
ted. Other, and more specific methods to approach a fire risk assessment are presen-
ted in the literature and some references are given at the end of this chapter.
11.2 Overall Approach used in a Fire Risk Analysis

The main steps of a general fire risk analysis are shown in Figure 11.1 and are briefly
explained in the following subsections.
The analysis should always start with a system familiarization period where the ana-
lyst systematically works through the available documentation of the system in order
to obtain a fair understanding of the system design and operational modes. In this
context is is important to focus not only on normal operation but also on start-up, shut-
down, inspection, maintenance, etc.
Based on the available information of the system and operational modes a systematic
hazard identification should be performed. Elements from failure mode and effect
analysis, HAZOP-techniques, checklists as well as normal engineering review can be
used to make the identification as complete as possible. Based on the hazard identifi-
cation all potential hazardous events are listed.
There will always be a question of how to define a hazard. A hazard could be defined
as a situation where the combustible fluid is in contact with oxygen. And the analysis
would then focus on
Scandpower
a) How can the combustible get in contact with oxygen?

b) How can the potential combustible mixture be ignited?
c) What is the potential consequences of the resulting fire?
The hazard should be defined based on the system in question, e.g. if the combustible
fluid is handled above its autoignition temperature, the hazard could be defined as a
process upset. As a matter of fact, the definition is not of great importance as long as
the methodology is consistent and covers the entire scenario from the normal situation
to the ultimate consequences. As a guide to the hazard identification the factors dis-
cussed in Chapter 4 of the handbook should be considered.
In a cause analysis the probability of occurrence for the identified hazardous

events is evaluated. The probabilities of occurrence can be quantified by means of
fault trees, statistical data, or models; whichever is found to be most suitable in the
different cases. In any analysis where an analytical approach is applied to estimate
the probability of occurrence of the hazardous events it is of importance to validate the
results, e.g. by comparing with experience from a larger population (e.g. leak records
from similar plants etc.).
Scandpower
Acceptance
Criteria
-
Plant
-
Operational
Design Modes
Qualitative
I
-
Hazard
Identification
I
Design or
Evaluation
of
Risk Factors
-
List of Accident
Procedural
Counter-
measures
I Initiating Events • -
I
I
I
I
-
Accident
Modelling
•
. -
I
I
I
I
-
Consequence
Assessment of
Probability of
Occurence
I Evaluations
I •
I ~
I
I Accident
I Scenario
I
I
I
.,,l •
I
-
Evaluation
Evaluation
of Risk
I
I
I
I
I
--
-
Acceptable
Risk?
No
I
I
I
..J... Yes
I • Risk Picture of Plant
I
I • Design Accidental Loads
L--D • Residual Risk Frequency
• Risk Reducing Measures
Figure 11.1: Methodology
~SINTEF Scandpower
Risk Management AS
D "Tl
iii'
'1<' u5· ~
C Detection Ignition Strong Fire Accident SEQ.PROB. CLASS Loss o/ Any Production No. o/ Fata l Acc. C0
s: CD
and Explosion Fighting Escalation Main Safety Delay above Fatalities outside Ace
CD
....
Q)
]g, ........ Isolation

Failure
In• to Other Fune1ion 1 Week Area (p) ....
.,, Q)
effective Equipment ~
3::, !'_) TE 1
.,, a. 2 3 4 5
::, "C
... 0
):, :IE -I
(/) ~ '< 1.00E+OO 1 0.00E+OO
"O O.OOE+OO O.OOE+OO O.OOE+OO
o·
Q) O.OOE+OO 2 O.OOE+OO O.OOE.OO
I 0.00E+OO O.OOE+OO
)
m I O.OOE+OO
O.OOE+OO 3 O.OOE+OO O.OOE+OO O.OOE+OO O.OOE+OO
<
-
CD
::, IQ,MC+M I O.OOE+OO 4 0.00E+OO O.OOE+OO O.OOE+OO O.OOE+OO
::i::
-I IO O/ll:+00 Ill
IO,MF+OO O.OOE+OO 5 O.OOE+OO O.OOE+OO O.OOE+OO O.OOE+OO ::s
Q,
CD CJ'
CD O.OOE+OO 6 O.OOE+OO o.ooE.oo O.OOE+OO O.OOE+OO 0
I
o'
.., 10.00E+OO
O.OOE+OO 7 O.OOE+OO O.OOE+OO O.OOE+OO
0
:ii"
1.00F+OO lo.nnc.oo O.OOE+OO
Q)
IO.Ml=+OO O.OOE• OO 8 O.OOE+OO O.OOE+OO O.OOE+OO O.OOE+OO
...0.,,
:!!
.., I
CD
10.00E+OO
O.OOE+OO 9 o.ooe..oo 0 .00E+OO O.OOE+OO O.OOE+OO
~-
::0 0
O.OOE+OO 10 O.OOE+OO Ill
cij" O.OOE+OO O.OOE+OO O.OOE+OO 0
"::,)> C
Q)
'<
O.OOE+OO
I
10.00E+OO
O.OOE+OO
O.OOE+OO
1t
12
O.OOE+OO
O.OOE+OO
O.OOE+OO
O.OOE+OO
O.OOE+OO
O.OOE+OO
O.OOE+OO
O.OOE+OO
-o·
iii'
::s
Ill
(/)
cij' LQ.mF+OO I O.OOE+OO 13 O.OOE+OO O.OOE+OO O.OOE+OO O.OOE+OO Ill
::s
O.OOE+OO
I Q MC+IVI
O.OOE+OO 14 O.OOE+OO O.OOE+OO O.OOE+OO O.OOE+OO .,,
Q,
I O.OOEtOO 15 O.OOE+OO O.OOE+OO O.OOE+OO O.OOE+OO

;'
IO OOF+OO
;o
0 OOE+OO O.OOE+OO 16 O.OOE+OO O.OOE+OO O.OOE+OO O.OOE+OO iii'
:ii"
lo .mi=+oo )>
O.OOE+OO 17 O.OOE+OO O.OOE+OO O.OOE+OO O.OOEtOO Ill
g:
Ill
O.OOE+OO O.OOE+OO O.OOE+OO Ill
O.OOE+OO
z0 3
l.
0,
::,
"Tl
Typical Event Tree for Analysis of Process Leaks .1212101\EXAMPLE.TRE 6-23-92
--
Cl>
::s
: ;·
::r
~· CD
"ti
.,,
;o
0
~~ 0
Cl>
g. (II Ill
Ill
r
0, z
- :i"
Q,
[ -I C
m
~
~ .,, Ill
~
The accidental effects in terms of accidental loads (heat flux, smoke etc.) should be
modelled for the different accidental events. The models presented in Chapter 5 on
fluid behaviour and the fire models presented in Chapter 6 and 7 should be applied as
found appropriate in these calculations.
The hazardous events can then be developed, e.g. by the event tree methodology
(see Figure 11.2), to reflect the possible escalation of the different events. (E.g.
whether a leak ignites or not, failure or function of the safety systems, exposure of
critical systems etc.). The effect of implemented risk reducing measures, active or
passive fire protection, depressurization systems etc, see Chapter 9, should be consi-
dered in the accident scenario evaluation (event tree) as well as in the consequence
assessment.
The consequences for personnel and/or exposed systems should then be evaluated
taking into account the accidental loads, the effects of safety systems, the assumed
distribution of personnel etc, see Chapter 8 and 10.
Based on the preceding steps the risk could be evaluated in terms of
risk to personnel
risk to the environment
risk to material and equipment
risk to production regularity
and compared with the established acceptance criteria.
One method to sort and evaluate the accidental events is to group the potential acci-
dents in two categories:
Design Accidental Events (DAE)

Events whose resulting physical effects are to be sustained by the design
Residual Accidental Events (RAE)
Events with intolerable consequences.
From the Design Accidental Events the loads that the system must be designed to
sustain in order to meet the acceptance criteria can be derived.
Based on the performed analysis recommendations can be given in order to meet the
safety acceptance criteria and/or to improve the safety level in general.
Based on a quantitative fire risk analysis the risk in terms of repair cost, production
outage, etc. could be expressed in monetary value.
The risk reducing effects of implementing additional safety systems (e.g. active/
passive fire protection, detection systems etc. etc.) can, hence, be calculated in mone-
tary value and compared with the investment and maintenance cost in a cost benefit
analysis. Such studies could be used as a supporting decision making tool to optimize
the benefit of investing in risk reducing measures. It should, however, in this context
be noted that such studies are encumbered with significant uncertainties and the
result should be treated with care.
Scandpower
Norwegian Fire Research LabOratory
Below is listed a number of publications which address the topic of fire risk analysis.
The list is by no means complete, but could serve as an introduction to more detailed
surveys in the field.
National Fire Protection Association (NFPA), Society of Fire Protection Engi-

neers (SFPE): "The SFPE Handbook of Fire Protection Engineering", 1988
Center for Chemical Process Safety of the American Institute of Chemical
Engineers: "Guidelines for Chemical Process Quantitative Risk Analysis", 1988
John Bond: "Sources of Ignition, Flammability Characteristics of Chemicals and
Products", 1991
F.P. Lees: "Loss Prevention in the Process Industries", 1980
"Risk Analysis of Six Potentially Hazardous Industrial Objects in the Rijmond
Area - A Pilot Study", 1981
A.E. Green: "High Risk Safety Technology", 1982
R. King, J. Magid: "Industrial Hazard and Safety Handbook", 1979
F.P. Lees, M.L. Any: "Safety Cases within the Control of Industrial Major Acci-
dent Hazards (CIMAH) Regulations 1984", 1989.
Scandp ower
Appendix A: Nomenclature Page A-1
APPENDIX A
Nomenclature
A = Area (m 2), the length of the downwind side of a rectangular pool (m)
A; = Area of the inner surface of the insulation material per unit length of the member
(m2/m)
AN, = "Section factor'' of steel section (m-1)
B = Length of the crosswind side of a rectangular pool (m)
Cd = Discharge coefficient (·)
Cp = Specific heat (kJ/kgK) 'l
D = Equivalent diameter (m)
D' = Downwind length of the pool including the flame drag (m)
Deq = Equilibrium diameter (the spill rate equals the burning rate) (m)
Ep = Surface emissive power (kW/m2)
E, = Modulus of elasticity for steel at room temperature (2.1-10·5 N/mm2)
E,r = Modulus of elasticity for steel at elevated temperature for cr = O (N/mm 2)
F = View factor (-)
FR = Fire resistance time (min.)
fR = Radiative fraction (-)
Fr = Froude number(-)
Fr' = Modified Froude number(-)
g = Acceleration of gravity (m/s2)
g' = Effective acceleration of gravity (m/s2)
H = Compartment height (m)
Ho = Ventilation opening height (m)
H, = Radial flame plume impingement under the ceiling (m)
h = Heat transfer coefficient (kW/m 2K) 2J
= he+ h,
he = Heat transfer coefficient by convection (W/m 2K)
h, = Heat transfer coefficient by radiation defined according to Eq. (8.32) (W/m 2K)
hg = Height of the hot gas layer above the floor (m)
hb = Beam vertical depth from ceiling (m)
h, = Height of the liquid fuel level in the dike forming a liquid pool (m)
~He = Heat of combustion (kJ/kg)
~Hv = Heat of vaporization of the liquid fuel (kJ/kgK)
k = Absorption/emission coefficient (m-1), or thermal conductivity (kW/mK)
ko = Constant applicable to the material in linear regression analysis of fire resistance test
data of insulated steel members
k1 = Constant applicable to the material in linear regression analysis of fire resistance test
k2 = Constant applicable to the material in linear regression analysis of fire resistance test
L = Flame length or characteristic dimension of the fire exposed target or the com-
partment length (m)
Lm = Mean beam length (m)
I = Length at room temperature (m)
1) In JlkgK when dealing with calculations of temperature response

2) In Whn 2K when dealing with calculations of temperature response
Scandpower
Page A-2 Appendix A: Nomenclature
Lil = Temperature induced expansion (m)

lb = Half the distance between two ceiling beams or half the width of a corridor (m)
M = Mach number of the jet flow(-)
Mt = Molecular weight of fuel (kg/kmol) or total mass of fuel stored (kg)
m = Number of carbon atoms in fuel molecule CmHnOp (-)
Lim = Total mass of smoke accumulated in the hot gas layer (kg)
ma = Ventilation rate either by natural or forced ventilation (kg/s)
mt = Mass burning rate and mass leakage rate of gaseous fuel under assumption of com-
mt
. =
plete combustion (kg/s)
Mass burning rate per unit pool surface area (kg/sm2)
n = Number of hydrogen atoms in fuel molecule CmHnOp or an general exponent(-)
p = Number of oxygen atoms in fuel molecule CmHnOp (-)
Pr = Prandtl number (-)
Q = Heat release rate (kW)
q" = Heat flux (heat flow density per unit exposed area) (kW/m2)
Re = Reynold number(-)
Ri = Richardson number (-)
r = Stoichiometric air to fuel ratio (-), radial distance lateral to jet fire flame plume axis
(m) or relative atmospheric humidity(%)
s = Equivalent diameter of a slot exit of length, "I", and width , "t", or the lift-off distance
of a jet fire (i.e. the distance from the exit hole where a jet fire can not burn due to
high gas flow velocity to allow stable combustion) (m)
T = Temperature (K)3l
LiT = Temperature increase above initial temperature T 0 (Kor 0 C)
t = Time (s)
Lit = Time step (s)
u = Velocity (mis)
V = Volume (m 3) or visibility (m) or causive variable in probit relations
VL = Volumetric content of liquid fuel stored (m3)
Vs = Volume of the member per unit length (m3/m)
Vt = Regression rate of fuel (mis)
VL = Volumetric liquid spill rate (m3/s)
Wo = Width of ventilation opening (m)
w = Width of trench fire (m)
X = Horizontal distance or coordinate (m)
z = Vertical coordinate or vertical distance above virtual source (m) = z' - Zo
z' = Vertical distance from plume source (m)
Zo = vertical distance of virtual source from plume source (positive z 0 is above plume
source) (m)
Greek letters:
(l = Tilt angle of jet fire flame (i.e. angle between hole axis and flame axis)(deg.), mole
reactants to products(-) or 'particulate optical density' (m2/kg)
0 = Wall thickness (m)
oi = Thickness of insulation (m)
E = Emissivity(-), strain (%0)
CJ' = Stefan-Bolzmann's constant= 56.7·10·12 (kW/m2 K4 )
TJ = Combustion efficiency(-)
O'yT = Effective yield strength at elevated temperature (N/mm2)
O'y20 = Nominal yield strength at room temperature (N/mm 2)
3
p = Density (kg/m )
3) In 'C when dealing with calculations of temperature response
Scandpower
NorwegianFire Research Laboratory
Appendix A: Nomenclature PageA-3
= Thermal conductivity of enclosing boundaries (kW/mK)4J

"'
\)
= Kinematic viscosity (m2/s)
µ = Dynamic viscosity (kg/ms)
'C = transmissivity (-)
X = mass of soot per kg fuel burnt (kg/kg)
~ = heat capacity ratio ((ep.1l;p;A;)/(2c;,.p.V.))
y = isentropic exponent of the fuel (-), or gas specific heat ratio (c,Jc.,)
Subscripts:
a = Air, average
ad = Adiabatic
b = Boiling
C = Convective, ceiling, combustion, centerline, concrete
er = Critical
e = Exit, equivalent
ef = Effective
eq = Equilibrium
f = Fuel, flame, floor
g = Gas
= Insulation
j = Jet
L = Liquid
I = Lower
m = Mean, maximum, number carbon atoms in the fuel molecule CmHnOp
n = Number hydrogen atoms in the fuel molecule CmHnOp
0 = Ambient, ventilation opening, initial
p = Pool, number oxygen atoms in the fuel molecule CmHnOp, proportional limit
R,r = Radiative
s = Source, solid, surface, steel, sonic
T = Elevated temperature
u = Upper
w = Wind, wall (enclosing boundaries offire compartment), sea water
y = Yield
20 = Ambient temperature (20 °C)
Superscripts:
* = Peak value
4) In WlmK when dealing with calculations of temperature response
Scandpower
Appendix B: Properties of Steel, Concrete and Insulating Materials at elevated Temperatures Page B-1
APPENDIX B
Properties of Steel, Concrete and Insulating
Materials at elevated Temperatures
,
o. a
o. 8
---r--- ~
0
~
~ c.?
I c.a " r,.._
~
...
•
,:;.
~
o. s
"· .. "" I',_

~
!i
0.3
o. 2 "' ['-_
o. 1
0
0 zoo ◄ 00 ..ao
" -----aoor--- 1000
ST"liiiliiiiL T ~C:Aill.TUR!i. T• CC:,
Figure B.1: EffectiveYyield Stress vs. nominal Yield Stress (cryrlcry2o) as a function of Steel
Temperature for different steel grades Fe310-Fe510. (Equations (8.17) - (8.18))
,
c.a
~
,___
c ••
Q
o.?
~
ti
\J
;:;
•
~
•
'Is
o.a
o. s
o ...
..
" "'- ~
~
o. 3
\
I 0.2
\
o. 1
D
0 200 400 "oo
Figure B.2: Effective Modulus of Elasticity vs. Modulus of Elasticity at normal Temperatures
(EsrlEs) as a Function of Steel Temperature for different Steel Grades Fe310-Fe510.
(Equations (8.19) - (8.20))
Scandpower
Page B-2 Appendix B: Properties of Steel, Concrete and Insulating Materials at elevated Temperatures
Steel properties acc. to EUROCODE 3- DRAFT prENV 1993-1-2, May 1993:
Reduction factor
1
Effective yield strength
ky,e = fy,e / fy
0,8
o,6 Slope of linear

elastic range
0,4 kE,8 = Ea,e / Ea
0,2
Proportional limit
kp,8 = ~,8 / fy
200 400 600 800 1000 1200

Temperature [ °C]
Figure B.3: Reduction Factors for the Stress-strain Relationship of Steel at elevated Tempe-
ratures (e = 2 %). (Ref. /8.27/).
Effective Yield Stress vs. nominal Yield Stress (cryrfcry20), and Effective Modulus of
Elasticity vs. Modulus of Elasticity at normal Temperatures (EsrfEs) as a Function
of Steel Temperature for different for Steel Grades Fe310-Fe510
Scandpower
Steel properties:
20
15
1•
17 /
_/
16
15 /
14 /
13
/
~ 12
11 /
/
"
'-' V
10
,
~ 9
a
./
;v
./
! 7
/
8
i..---
5
./
4
,,/
3
2
V
,,/
1 __,.,.
0
0 200 ◄00 500 900 1000 1200
STEEL TS4PER~TUAE Ta C~
Figure 8.4: Thermal Elongation (iii/I) as a Function of Steel Temperature for different Steel
Grades Fe310-Fe510. (Equations (8.21) - (8.23))
5000 J/kg°C at 735°C

"1200
..,.,00
.,000
900
I
G j
~ 1100
~ /"' I'\. ..._
-
'-'
"'u
?00
_______. .....
-
BOO
j ~
...
---
500
~
"00
I 300
200
100
0
0 zoo ... 00 eoo aoo 1000 1Z00
STEEL TEMPERATURE T• CC,
Figure 8.5: Specific Heat Capasity (Cps) as a Function of Steel Temperature for different for
Steel Grades Fe310-Fe510. (Equations (8.25) - (8.28))
Scandpower
Steel properties:
so
,._____
,o
,.'d' --- r----. ,...____

V
➔O
;; ---- t-------- r------._

>
i8 30
2 20
.!!
~
'"
0
0 200 ➔ Oo 800 900 1000 '1200
STEEL TB.IPERATURE Te cc:i
Figure B.6: Thermal Conductivity (1's) as a Function of Steel Temperature for different for
Steel Grades Fe310-Fe510. (Equations (8.30) - (8.31))
Concrete properties:
1800
1S00
1400
1300
1200
G
ft
;'!.
1100
,coo
000 - --
I
u
aoo
700
i 800
ODD
~00
300
aoo
100
0
0 200 ➔ 00 aoo 900 1000 1200
Figure B.7: Specific Heat Capacity for Concrete (cpc) as a Function of Temperature.
(Equation (8.36))
Scandpower
Risk Management AS (G))SINTEF
2 ••
2, a
2,4
ft 2, 2
•::..... 2
-........_
-- -----
>- 1.e
,. -a.._
1, a
i8
1,4
1, 2
,..__ 11,....
--,~
-
r- ---,_
1
-
I 0,9
o."
0,4
0, 2
0
0 200 -400 ISOO BOO 1 000 1200
COncr•~o ~•,roer•~uro Cc:>

0 6f I TC:OOUO + ca I caf' c;,,,oug <> l..WA
Figure 8.8: Thermal Conductivity (Ac) as a Function of Temperature for different Types of
Concrete. (Equations (8.37) - (8.38))
Scandpower
Insulation materials: Fire ratings on structural steel according to hydrocarbon fire.
FIRE PROTECTION TYPE: lntumescents: CHARTEK Ill®

CHARTEKIV®
PITTCHAR XP®
Subliming materials: THERMO-LAG 440®
DENSITY: 1,200 - 1,400 kg/m 3
MINIMUM THICKNESS: 4mm
BASIS FOR TABLES: Results from fire tests + Constrado calculations (see
Equation (8.33) in Chapter 8.3.3)
Table B.1a: Necessary Insulation Thickness on Steel Members for certain lntumescent, and
subliming Fire Protection Materials to give different Fire Ratings when exposed
to Hy,drocarbon Fire.
. VaII'd ior I-sections
.
Section Insulation thickness to give ratings shown (mm)
factor {Ts,aitica1=450°, I-profiles)
(m-')
30 60 90 120 180 240
50 4-5 5-6 6-8 9-11 14-16 19-22
70 4-5 5-7 8-10 11-13 18-19 25-26
90 4-5 6-7 10-11 14 20-22 29
110 5 6-8 11-12 15-16 21-25
130 5 7-9 12 16-18 23-25
150 5 8-9 13 16-19 24-26
170 5 8-9 13-15 17-21 25-27
190 5-6 9-10 14-15 18-22 26-28
210 5-6 9-10 14-16 18-23 26-29
230 5-6 10 14-17 18-24 27-29
250 5-6 10 15-18 19-25 27-30
270 5-6 10-11 15-18
290 5-6 10-11 15-19
Scandpower
Risk Management AS (G))SINTEF
Table B.1b: Necessary Insulation Thickness on Steel Members for certain lntumescent, and
subliming Fire Protection Materials to give different Fire Ratings when exposed
t 0 HIVd rocarbon Fire. VaIl"d f or RHS-sections
.
factor <Ts,aitical=450°, RHS-profiles)
(m-1)
30 60 90 120 180 240
70 5 7 11 15 22
90 5 9 13 17 25
110 5-8 10-12 14-16 19 27
130 6-9 11-13 16-17 21
150 6-10 11-14 17-19 22-23
170 6-10 12-15 18-20 24
190 7-11 13-16 19-21 25-26
210 7-11 13-16 19-22 27
230 7-12 14-17 20-22 28
250 7-12 14-18 21-23
Note: lntumescent epoxy materials are "active" materials and may behave differently on different steel shapes. E.g.
required thicknesses for different ratings are normally different for I-sections (see Table B. ta) and RHS/CHC-
sections (see Table B.1b). I-sections require often the lowest thicknesses
Scandpower
Page 8-8 Appendix B: Properties of Steel, Concrete and Insulating Materials at elevated Temperatures
FIRE PROTECTION TYPE: Cementitious materials Dossalack HCF®

Mandolite 550®
DENSITY: 500-600 kg/m 3
MINIMUM THICKNESS: 10-12 mm
BASIS FOR TABLES: Results from fire tests + Constrado calculations (see Equa-
tion (8.33) in Chapter 8.3.3)
TableB.2: Necessary Insulation Thickness on Steel Members for certain Spray applied
Cementitious Fire Protection Materials to give different Fire Ratings when
exposed to H1y,d rocar bon F.ire
factor (Ts.Clilical:450°)
(m-1)
30 60 90 120 180 240
30 12 12 14-18 16-24 21 26
50 12 15 19-23 22-28 29 36
70 12-14 18-19 22-26 26-31 34 42
90 12-16 20-21 25-28 29-33 38 47
110 12-17 22-23 27-30 32-35 41 51
130 13-18 23-24 28-31 33-36 44 54
150 14-19 24-26 30-32 35-37 46 57
170 15-19 25-27 31-34 36-38 48 59
190 17-20 26-28 31-34 37-39 49 60
210 18-20 26-29 32-35 38-40 50 62
Scandpower
Risk Management AS (G))slNTEF
CEMENTITIOUS SPRAY APPLIED MATERIALS

Critical steel temperature 450 •c
70
240 min
60
E
.5.
r/1
r/1
Q)
50
C:
~
5 40
C:
0
;;:::,
ro
:i 30
r/1
.5
~
ro 20
r/1
r/1
Q)
u
Q)
z 10
0
0 40 80 120 160 200 240
Time of exposure (hours)
Ventilation rate: + 0.05, ◊ 0.1, A 0.5, ic 1, v 3
Figure B.9: Necessary Insulation Thickness on Steel Members for certain Spray applied
Cementitious Fire Protection Materials to give different Fire Ratings when
exposed to Hydrocarbon Fire, Time/Temperature Curve
Scandpower
FIRE PROTECTION TYPE: Ceramic Fibre blankets
DENSITY: 128 kg/m 3
MINIMUM THICKNESS: 10 mm
BASIS FOR TABLES: Results from fire tests + Constrado interpolations (see
Equation (8.33) in Chapter 8.3.3)
Table B.3: Necessary Insulation Thickness on Steel Members for certain Ceramic Fibre
Blankets to give different Fire Ratinas when exoosed t o Hydrocarbon Fire
factor (TS,aitical:450°)
(m-')
30 60 90 120 180 240
70 11 23 35 47 72 96
90 14 29 44 59 89 120
110 16 34 52 70 106 142
130 19 39 60 80 121
150 21 44 67 90 136
170 23 48 74 99 150
190 25 53 80 108
210 27 57 86 116
230 29 60 92 124
250 31 64 98 131
Note: The tabulated thicknesses have to be adoptet to the available thicknesses of the blankets. The 128 kglm3 blankets
are nonnal/y made in the following thicknesses: 6 mm, 10 mm, 13 mm, 19 mm, 25 mm, 38 mm and 50 mm
Scandpower
CERAMIC FIBRE BLANKETS

Critical steel iemperature 450 °C
150 n10 min
240miy
140 ef i('rno min V 150min
Vi f / / _,. ~120 min
130
./ ff' / / ~.,
120
/ '{ _)(' ✓ ~
110
100 J JI / /'f ..... i,.-- 90min
J!. , '{ ./ )I' l""-
-
,,_I. Y"""
E go ~
.§. 80 70 / ., ✓'(' /
~ 60
:c
../ 1 /
,,V-- ~ 'Y""
~ V
i.-- ........-
- 1---+""" 60 min
1- 50 8
I/ ~-
40
30
....... .....
~-- ~
- - 30 min
20 -
..,,_ ~ . - 0 -
10
0
40 80 120 160 200 240 280
Section Factor F/A (m-1)
Figure B.10: Necessary Insulation Thickness on Steel Members for certain Ceramic Fibre
Blankets to give different Fire Ratings when exposed to Hydrocarbon Fire
Scandpower
FIRE PROTECTION TYPE: Calsium Silicate Boards
DENSITY: 800-900 kg/m 3

BASIS FOR TABLES: Results from fire tests + Constrado calculations (See Equa-
tion (8.33) in Chapter 8.3.3)
Table B.4: Necessary Insulation Thickness on Steel Members for certain Calsium Silicate
Boards to give different Fire Ratings when exposed to Hydrocarbon Time/Tem-
perature C urve
factor (Ts,critlcat=450")
(m-')
30 60 90 120 180 240
30 7 10 13 17 23 30
50 9 14 19 24 34 44
70 12 18 24 30 42 54
90 13 20 27 34 48 62
110 15 22 30 38 53 69
130 16 24 33 41 57 74
150 17 26 34 43 61 78
170 18 27 36 45 64 82
190 18 28 38 47 66 86
210 19 29 39 49 69 88
Scandpower
CALS I UM SILICATE BOARDS

CRITICAL STEEL TB,IPER.I\TURE -4,ct:e
100
90
1--9
~
i
u 80
_.-r
Jr'
..
i
C
70
V
~
,....-
ti ~
.,:c 60 / ........ ~
~
/ ~
--
C
!!
., so
~ ,__ ~ --
;
•
Cl
-40 / x ~
./v/
!: _,Ar ~-
:,.
_[-..Y' ,_..,,
-- --
30
i / V
V
~
~ I-+-~
5 20
t
10
::-:v
a- ~
~
- - - - -
0
0 -40 BO 120 160 200 2-40
SECTION FACTOR .I.I/Va (m-1)
□ 30mrn + 60 1111 n O 90 min
c. 120 n11n X 180 1111n v 240 min
Figure 8.11: Necessary Insulation Thickness on Steel Members for certain Calsium Silicate
Boards to give different Fire Ratings when exposed to Hydrocarbon Fire
Note: The tabulated thicknesses have to be adoptet to the available thicknesses of the boards
Scandpower
FIRE PROTECTION TYPE: Magnesium oxychloride
DENSITY: 1,000 kg/m 3
BASIS FOR TABLES: Results from TEMPCALC® calculations + Constrado inter-

polations (See Equation (8.33) in Chapter 8.3.3).
Table B.5: Necessary Insulation Thickness on Steel Members for certain Magnesium Oxy-
chloride based Spray applied Fire Protection Materials to give different Fire
Ratings when exposed to Hydrocarbon Time/Temperature Curve. (Values not veri-
fied through Classification Tests - for Comparison onlv\
factor CTs.amca1=450°, I-profiles)
(m-')
30 60 90 120 180 240
30 10 10 13 16 22 27
50 10 13 16 20 28 35
70 10 14 19 23 31 40
90 11 16 20 25 34 43
110 12 16 21 26 36 45
130 12 17 22 27 37 47
150 12 18 23 28 38 49
170 13 18 23 29 39 50
190 13 18 24 29 40 51
210 13 19 24 30 40 51
Scandpower
Risk Management AS O)SINTEF
MA.GNESILM OXYCHLOAIDE F I AE PAOTECTION

OllTICAL STEEL TB.IPERATUAE 43Q::C
60
i 50
:----
- -
L.J
~
..
II J. . . .- ~
...
C
0
.,
t:
C
.,2
◄0
JO
✓
V
./
,,...... ..,...:--
~
- -
V 1---4- -a--
---
!!.
..
:,
5 ✓ ~ V
~
I-+- ..........
20 _..
:,..
✓
V v--
M
§ ......V
~ ~
i....-i---- - - - - -
10
!
0
0 ◄0 80 120 160 200 2◄ 0
SECTION FAC"TCR Al/Ve (rr>-1]
O 30 '"'" + 60 min <> 90 min

t. 120 min x 180 m1n v 2◄ 0 mln
Figure 8.12: Necessary Insulation Thickness on Steel Members for certain Magnesium Oxy-
chloride based Spray applied Fire Protection Materials to give different Fire
Ratings when exposed to Hydrocarbon Time/Temperature Curve. (Values not
verified through Classification Tests - for Comparison only)
Scandpower
Appendix C: Calculation of Section Factor of Protected Steel Profiles Page C-1
APPENDIX C
Calculation of Section Factor of Protected
Steel Profiles
(Ref. /8.28/).
r- - - - - - - , ; - - - - - - - - - - - - - - - ,'
S1cd s~c1iun r:ofilc prou:c1ion
L"mn,n! bc~r.,s, un,.-r,nl

C,.,:\lml'II ~ll~JC!sU
tfi.:m: ~nd c•mll11cd1
o! :r
'." 11, :B•::D•:(B-t) 9 .. :D•:(B-1) B•:d•fB-t) B•D•:tB- 1J!2
~.:a .. :0-:1 =3B•:0-:1 •:B .. :d-1 .,:B•D-1 '
S:r.,:1ur~ uu! 10Ucd 1ee1
~
uf±1
11, :8 • :o B•:D•(B- :)
=:B•:0-1
'
Anpes
I F B
;----; '
0
l r'I'
-ii-
11, :D•:D a.::o B-<:D•(B-1)
' =:B•:O-t
Ch,.nnch
,.....,
B
o[£
"· I :a•:o+:{S-1)
~..:e•:0-:1
:a•D<[B-1)
~..:B•D-:1
B•:D•:(B· tJ
•~B•:D-:1
11,.,11,,,. J,c:,.>m.
~:.;i;aic ,n n;:ar.;u!.r
,....,
B
0
TD
11::il""' ~ec:w:i,.
"• :a .. :::> B<O
©
;i,;;u!~r
r"j
i:;L~..,ple
0 nO
~) r,di!r r,ru:r;riun - 3 s1a!cd r,.p,vi~1r

:nir.g :B • :oJ • ;; ,:;::m
1,,:i,t,~c.,J.,-r.11 H,•~B-:D-:?1
Hp c ~,:,;,J.9•:~:o:..:- :,f.O "'-t>ll.;+41:.~-lh
a " ~c,; _,, r:::n: = Jl~l!smm ~ IJK~~m
1 ~ l.0t::'II. = 1:1:r::m = 1.:1:,:i H,1A = I r11,1f1 l-t1,t,.;" l~!J,m·'
Hr . .\"' 1.::1:'vo:,iv.-: ~ :~:-!m"'
(Gl)SINTEF Scandpower
Risk Management AS
Page C-2 Appendix C: Calculation of Section Factor of Protected Steel Profiles
S1eel stc1ion Box ,nd so lid pro1ec1ion
w m
L'n1 ..·a ~ b(~ms . univrrsil.l
colum n$ :ind JOists
(pllin 1nd c1s1cU21cd)
Por t1llly e xposed
:s + :o B• : O B•D 8
St:u..:tu::d l nd rof'1:J tees ..i .u Jcs
--·rrrlr
18 • :o B+ :o B+:O
:s . :o B• :o
u
C!1,nads
[D \\ t,'~ tu s1Jffi1
~B • ~D :o • D B• :o
lt-,llo "' sc,:aons. .' s,Jcs
[] Tr,1r
squJre or rectln gl.Jbr
H~ .I :f!_.,. !D B•lD
CJ
HoU\J""' s«laun s,
circ:.:br .\'vtr. The li! spl ..·e :reJud in bo ~!n£ l !.tct!or. improves the insubtion l nd
1 ,·,!uc of Hp · A. and 1he:efoce Hp, h:i;he: I ~.,n Ih,: ro, ? refile pro«;tio n
.... cu!.! be :ir.u :1aio us. Hc:ice I{~ u, LJke n 35, the Clrcumforencc or ihe 1ube
:rnl! nl.11 .;Q,
~o
c) Bu, . -' si.JcJ c.'\?,))Jtc
tt 0 = lB • :o = 40- -~ • ~ 1:A H0 = 8 • :O = :OJ.q • ~1 : .~

s h:O.: ,mn ... 0 .3 ;0 m • 616 _, "'"' = 0.6 16 :n
11 1,,'A = O.S: ,OOOoo-1 = 1:3.S:11- 1 H,1: A = 0 .o:c>O .OO~b-' = Q:_sm- 1
Scandpower
B Secuon f actor H/
p A '
t-;-+1
Table 1.1 Profile Box I
Universal beams
l±Jr
0
3 sides
•t_~O{//~/•/
4 sides
.,,;Ira
3 sides 4 sides
r·--·1
I
~I1
~~~
I l ·--·; ;·--· '' ''
''''
Designation
Depth Width Thickness
''
''
'' '':!'' ''' '''
'' ''
Arca
or of ''
'' :_~·~; C----~
Serial Mass per section section \\'cb !Flange or :r-=~~:
size mcm: D B t T sec1ion
mm mm mm I mm I cm' m·• m·• m·• m·•
mm kg
-1:0.s :1.s 36.6 49-1.4 60 70 ,5 55
91JxJJ9 38S 920.5
60
)-13 911.4 -WLS 19A 3'.!.0 437.J 70 80 50
91-1>:.305 :?89 926.6 3U7.S 19.6 32.0 368.8 75 so 60 65
253 916.5 JCJS.S 17.3 27.9 322.8 85 95 65 75
22-1 9!0.3 31>-i.l 15.9 23.9 285.2 95 l05 75 85
WI 903 303.4 15.2 20.2 :?56,J 105 115 80 95
S38X292 216 850.9 ::~m.s 16.1 26.8 :ss.i 85 95 70
,o
80
19•
176
s..:.0.1
SJJ.9
:?9;!.J
:?91.6 ,,
1.:.1 21. 7
18.8
:?Ji.I
22•1.1
100
110
115
125 90
90
IIIO
7b2X'!67 197 769 6 26S 15.6 25.4 :!50.7 90 IOO 70 85
,,,
173 762
753.9
266.7
265.3
1-U
12.9
21.6
Ji.S
2:0..i
!&.'ta
){15
IW
115
135
80
95
95
110
bS6x:?5J 170 69~.9 ~55.8 fJ.!i :?}.i :?16.S 95 I l(J 75 90
15:? blii.6 :!!-J.5 13.:! :?1.0 J93.S 110 l:0 85 95
lJO 6R3.5 :?53.7 l:?.4 19.0 178.6 115 DO 99
ti33
fll7.5
6(J9 6
311.5
307
JO-LS
18.6
14.1
11.9
31.J
:?3.6
JIJ.7
303.7
:?:!7.9
190.1
90
I IO
""
lflS
l:?5
70
NII
X{I
95
6HJx.::::?9 HO 617 :?JU.I 13.1 :!:?. I PS.3 I05 l:?O 8'1 95
l::!5 61 l.9 :?:?9 I 1.9 19.6 1.59.5 I 15 ]JI) 90 IO!i
1:s.2. I 1.:? 17.) J..1-1,4 130 1J5 ltlO 115
113 607.3
16(1 110 IJIJ
IOI 6112.2 :?:?7.6 10.6 1-l.X l.'.!.9.1 IJS
53Jx.'.!.IO I.'.!.:? 5-1-1.6 :?11.9 \:!.8 :?J.3 155.7 I IO 1:?0 85 95
1:u 95 110
.,
)119 539.5 :?W. 7 I l.6 IK8 I3S.5 135
536.7 :?JO.I 10.9 17.4 1:?9.7 130 i,5 !00 115
JOI
9' SJ.3.1 .W9.3 JO.:? 15.6 117,7 IJO 160 JIU l:'l
!(U.-1 155 175 l:?O ) J(t
5:?S.J :?nS.7 9.6 13.:?
457'xl91 9S -H,7.4 19:?.8 11.4 19.6 l'.25.2 l:?O 135 90 HIS
••
,_
S'
-163.6
4flll.:?
19:?
191.J
Jl}.(1
9.9
17.7
16.11
113.9
JIU.5
!JO
J.!()
145
16(1
100
105
115
l:?S
·"~
190.5 9.1 14.5 9..t.9S 155 175 Ill 135
J57x 152
"
67
J57.:?
.:5J.6
. u,5.l
Jljl}.9
153.5
8.5
10.7
l:?.7
IS.9
85.44
JI~-➔
liCJ
!JO
190
1Jl
130
IIJS
150
120
..161.3 15:?.7 9.9 17.11 9..:_99 1..:0 1~5 Ill 130
67 -157.:? 151.9 9.1 15.0 85.41 155 175 l:?5 I'S
6(J J5J.7 152.9 8.0 13.J i5.93 175 IIJS 1.:11 1611
52 J..i9.!-I 15?.J 1.6 W.9 66.J9 :wo 220 160 180
9.7 9-'.95 1-10 160 105 l:'l
4D6x 178 7' --11::.s li9.7 16.0
t..io
67 -109 . .J 17/-C . 8 8.8 1.:.J S5.J9 155 175 115
60 406.4 lii.8 ''i..8 11.H 76.01 liS 195 130 155
s, J0;!.6 17i.6 7.6 l0.9 6SA2 190
205
215
230
1'5
160
170
185
4U6x 1-10 J6 ..101.J l.t2.4 6.9 11.:: 56.96
39 391.J I ~1.8 6.:,:. 8.6 49.J.O :!•HJ ?iO 190 :?.:W
J56X 171 67 36' 173.2 9.1 1$.7 R!iA2 1.:0 160 )05 125
172, I s 13.0 72.18 165 19<1 1:'l 1,5
57 358.6
51 355.6 171.5 i.3 11.l 6-US 185 :?IU 135 165
45 352 171 6.9 9.' 56.96 .:?HI 2..io 155 185
356x 127 39 35:?.S 126 6.5 J(J.7 ◄ 9.JO ::?15 240 170 195
2-1."l.5 1~5.J: 5.9 8.5 41.SJ 250 2SO 195 2:'l
33
166.8 7,7 13.7 68.38 160 185 115 1,0
3U5x 165 54 310.9
307.1 165.7 6 .. i 1 J.S 58.90 JSS 210 130 160
'6
40 303.8 165.1 6.1 111.2 SI.SO 210 240 150 180
48 Jlll.4 1:?5.2 9.9 14.0 60.S3 )60 180 125 145
3115x 127
,2 306.6 12J.3 8 12.1 53.1S 11-0 205 !JO 160
37 303.S 1:?J.5 7,2 !ll.7 47.-17 WO 225 155 180
-II. 77 :?15 2J0 175 .WO
.
3115x 102 33 312.7 1112.4 6.6 IO.S
23 JOS.9 IOl.9 6.1 8.9 36.JO 2..:s 275 :mo 2:'l
ZS 304.S tol.6 5.S h.X Jl.39 285 )15 225 260
120 150
254xt.t6 ,3
37
:?59.6
256
147.3
IJ6.J
7.3
., 12.7
lll.9
8.6
55.10
47.45
40.00
170
195
230
195
2:'l
265
140
160
170
200
31 251.5 IJ6.1 6.1
254x I02 ::!S 260.4 J02.I 6-' Ill.II 36.19 2W 250 170 200
25 :?57 tcll.9 6.1 s., 32.17 245 ~'ill 190 225
:?2 :?S.J J[Jf.6 5.8 6.X 2H.J:? 275 315 215 250
211Jx 133 30 21>6.8 133.S 6.3 9.6 Js.nn 210 2--1$ 145 IJ!O
25 203 2 1.i3.J 5.r. 7.8 32.31 2.:n ::XS 165 2IO
270 175 210
:?OJx l02
r:-sx Hl2
23
19
ZOJ.2
li7.S
1!'-:!.t
IOl.6
Wl.6
XS 9
5.2
,.1
H
9.3
7.9
77
"
2J.2
:n.5
235
:?tiS
2711
305
.1111 ,. ,
190 230
~35
I
J~;!-.,:J.(C)
l ::~" 7t, "
D 1::, 7t, 2
"
7 t, 11'1:;. :!:'~ },::11 1"5 2JIJ
Scandpower
I•
B
.. Section factor Hp/ A
I
11 I
Table 1.2 ' ' Proiile I Bo~
"(It• :I,
Universal columns 3 sides 4 sides 3 sides 4 sides
Ocs ign>tion
Depth
oi
Wid:h
of
Dl ~(._
.
•,
Thickness
Arel
. 1//: "//////(///(//,.
-=- -''·---
''' '''
I •
,-------~
'---,I ,-
I
I I
I
--· I
:
'
I
I
I
'I
II
I
I
I'
I
I
,- -- ·- - -~
I ' ·-
' - -~ I '
I I
Sc:ial !\llss per sc:tion section Web Flange of
SC"Ctio n ·-------·
, - - _J
·-- - --· ' '
l,. _ _ _ _ _ _...
'l,! _ ___ _ _ J'
size mc:cre D B t T
mm kg mm mm mm mm I cm: I m·l
I m•I m·I I m·•
)56 x .:o6 634 474 ,7 4:4.l 47. 6 n.o' 808 . l 2S 30 IS 20
.lSI 455.7 41 8.S .i:.o 61.5 701.3 30 35 20 25
467 436.6 -'1 2.-' 35.9 ss.o 595.S JS -IQ 20 30
25 35
393
340
zs;
4 19.l
-!06.-'
393.7
407.0
-!03 .0
399.0
30.6
Z6.5
2:.6
49.2
42.9
)6.5
500.9
-'32.7
366.0
"°.isso 45
55
6;
30
30
JS
45
23; 331.0 395.0 18 .S 30.Z 299.3 65 75 ,I() so
356 " J6a 20: 37.J.7 Ji-U 16.S 27 .0 257.9 70 85 .is 60
Ii i 363.J 372. 1 1.i.s 21 .s 225.7 so 95 50 65
153 362.0 370.Z 12.6 20 .7 195 .z 90 110 ss 15
1:9 355 .6 36d. 3 10.7 17.5 164.9 10 5 130 65 90
1 360.4 .is 55 JO 40
305 ~ 305 25]
2.:0
365.J
352.6
3: 1.3
317.9
26.9
21 .0 Ii -"'. :
37.7 30 5.6 .lO 60 35 .is
so
19~
153
13;
33~.9
J:?i . .2
1:0.s
J 14. 5
314. l
310.6
JO~. 7
306.S
19.2
15.7
13.8
ll.9
! 31..l
25.0
21.7
18 . 7
252.3
201.Z
174.6
149.3
60
75
85
100
90
105
1:0
75 40
50
ss
60
65
70
85
11$
97 307.d 30-l.3 9.9 IH l Zl.3 1:0 145 15 100
25-1 x 25-1 167 23'>. 1 264.:S 19.Z j
1
Jl.7 ::?1:!A 60 15 40 :so
DZ 276 .J 26 1.0 IS .6 2.U 167 .7 15 90 50 65
107 266.7 253.3 IJ .O 1 20.5 lJ6.6 90 110 60 75
89 260-4 25;,9 10.S j 17.J 114.0 110 130 70 90
73 254.0 25-1.0 S.6 14.~ 92.9 130 160 80 110
1 80
WJ " 203 86 2~1.J 208.3 IJ.0 I 20.5 110. 1 95 11 0 60
71 21:S.9 2C6.2 10.J l 7 .3 91.1 110 IJ5 70 9:5
6<) 2C<J.6 205 .2 9 .J l-1.Z 75.3 130 160 80 110
s~ 206.1 ZOJ.9 8 .0 12.5 66.4 1m 180 95 rzs
-16 ~OJ .~ 203 .2 7.3 11.0 :Sd.3 165 200 10 :S l,I()
11: x 152 F 161.3 15-1.-1 8.1 11.5 4 i .J 160 190 100 135
30 157 .5 152.9 6.6 9. ~ 33.2 195 235 1:0 160
2J l:52.4 l!ZA 6. l 6 .3 29.3 2.is 300 155 205
Scandpower
I Section factor HP/ A '
rt ·r·
Table 1.3 Pre-file Bo,
Joists 3 sides
.,,,,,,,,,,,,,,!.,,'(,
•n;I~, iI]
4 sides 3 sides 4 sides
Dcsigna!ion
Serial
size
mm
Mass per
mc:rc
kg
Depth
of
section
mm
D
Wid:h
of
section
B
mm
Thkkness
Web
mm
I
Flanrc
T
mm
or
section
Arca
em•
.:
.'
''
·:: -~--..::
m·•
'I ''
.: ''
,I
:.:-:·~:::
m·•
''
'
I
'
'
I
'
I
'L: ____ '.JI
m·•
''
'
'
'
'
:
I
I
' ___ I
L_
m·•
J
I
'
::54 X 203 s1.es 2~4.0 203.2 10.2 19.9 104.4 9S l ll 70 90

254xll4 37.20 254.0 114.3 i.6 12.8 4i.4 16l 190 1)0 Ill
203x 152 52.09 203.2 152.4 S.9 16.5 66.4 11' 140 55 IOl
203x102 ::S.33 203.2 I0J.6 !.S 10.4 32.3 205 2ll 155 190
178 X l02 21.54 177.8 IOJ.6 l.l 9.0 27.4 :,5 260 165 205
37.20
J52x 127
l52x 89 17.09
152.◄
is:.4
127.0
BS.9 ,.,
10.4 13,2
8.3
4i.5
:1.a
130
245
Ill
285
90
180
120
220
152x 76 17.86 15::?.4 76.2 s.s 9.6 :~.8 215 245 165 WO
127 X 114 29.76 12i.0 JJ4.3 10.2 J 1.5 37.3 140 liS 100 130
J27x JJ4 26.79 12i.0 114.J i.4 11.4 34.J J~S 190 110 140
J27x 76 16.37 ]27.0 76.2 S.6 9.6 21.0 205 l<l ll5 195
127x 76 13.36 127.0 76.2 4.S i.6 17.0 265 310 195 240
ll4X 114 26.79 114.3 114.3 9.S 10,7 3.;.-4 1.;5 150 100 ill
10::!x 102 ::3.07 101.6 J0J.6 ,.s J0.3 29.-4 ISO 185 IOl 140
I02x 64 9.65 lOJ.6 63.5 4.1 6,6 12.3 295 345 215 110
J02x 44
89x 89 '·""
19.JS
101.6
SS.9
"'·'
8S.9
4.l
9.S
6.1
9.9
9.S
:?.;,9
320
ISS
365
190
260
IOl
305
14S
76x 76 14.67 76.2 80.0 S.9 8.4 19.J 175 ::20 120 165
i6x 76 11.65 76.2 i6.2 S.I 8.4 16.3 ios :!.!iO 140 lfil
Table 1.4
...,
,'B' Section factor Hp/ A
l __jr' Prolile Bo,
Channels
Ll:
3 sides 4 sides 3 sides 4 sides
0
I
rr.;__ .,
~---·T·---J re:'
1//(ffe/C, 1//,C@a,,!,~ ~ ,'////////,
T
i
Designation Depth
of
Width
of
Thid.ncss
Arca
:
[C'_:: ..JF'l''':L_f
_____ .,'
1~
''
.,
' L--,
''
'' '
''
'
·n···u· ''
Serial Mass per stction stc1ion
Web Flange of ~- :..
---~ ;_ __ J '
size mecre D B
mm I ,, I mm
t
mm I mm rnm
T sedon
cm 1 I rn•I m·• r.1•1 rn·I m·• m·• m·• m·•
.;3;:x l02 65.54 ,:JJ.8 101.6 12.2 16.8 83.49 135 95 ,s 145 Ill 75 75 BO
381 X l02 !-Sc.JO 381.0 101.6 10.4 16.J i0.19 145 !OS 8S 160 1:s BS Sl 140
305x 102 46.18 304.8 101.6 10.2 14.S lS.'3 145 110 8S 165 1:!0 85 BS 140
305 X 89 41.69 304.8 SS.9 10.2 )3. 7 53.J I l.!5 IIS 90 1-;5 130 90 90 150
254x 89 35.'i4 254.0 88.9 9.1 B.6 45.52 160 i::5 95 180 130 95 95 ISO
254x 76 :s.::9 ::!54.0 76.2 S.l 10.9 36.03 195 I'S 11S 215 160 Ill 115 185
::29x 89 32.76 :2S.6 86.9 8.6 13.3 41.;J 165 JlO 95 185 130 95 95 150
::!29x 76 26.06 228.6 ,6.2 7.6 I l.2 H.20 195 ISO IIS :20 160 Ill 115 185
20Jx 89 29.78 203.2 IS,9 8.1 1:.9 37.94 165 ;35 100 190 130 100 100 155
203 X 76 :3.82 203.2 i6.2 7.1 J l.2 30.34 195 155 l ll 220 160 I IS Ill 185
I78x 89 26.81 17'i.8 8S.9 7.6 11.3 34.]5 JiO 1'S IOS 195 130 !Ol IOS 155
nsx 76 20.64 li7.8 76.2 6.6 10.3 16.54 .205 170 I.ZS 135 165 125 l2S 190
J52x 89 :?3.84 152.4 86.9 i.l l 1.6 30.36 1,s 1!5 110 205 130 110 110 160
ISlx 76 J7.68 152.4 i6.2 6.4 9.0 22.i7 ::.20 155 J35 155 165 135 Ill 200
127x 64 14.90 127,0 63.l 6.4 9.2 18.98 215 185 D5 250 165 J35 135 200
J02x Sl J0.-42 I0J.6 50.8 6.1 i.6 13.28 245 210 155 285 190 ]55 155 230
l6x 38 6.70 'i6.2 3S.J 5.1 6.8 S.!3 :ss .'.!-!O 180 330 ns 180 180 170
i!J)SINTEF Scandpower
Norwegian Fire Research Laboratory Risk Management AS
......
I
Section factor HP/ A
Table 1.5 "
IL
P~oiile Box
Equal angles
rr~ r-· rr;

0 J sides 4 sid<s 3 sides 4 sides
Size Thickn.ss Mu.s

per
I D
Atta
or
I
I
l~r·
I
I
I
I
L,
'
I
I
I
Wfq'ffe«//////ft
'LI I
I I
I '
I t. . ----,
l!-----·
I I
I '
I'
~;
W///62'~//~.,,U
I
I
I
I
I
I
I
I
Ii.: _____ J
I
I
I
I
I
L:-----'
I
I
I
I
Dx D t melrr sec lion

mm mm I kg cm: m·I I m·• m •l
I m·• m·I
200 x 200 2-' 71.1 90.6 65 85 85 65 90
20 59.9 76.J 75 100 105 80 105
18 54.1 69. 1 85 110 115 85 11 5
16 48.5 61.8 95 125 125 95 1.30
150 x 150 IS -IO. l 51.0 85 110 115 90 115
15 JJ.S 4).0 100 135 135 105 PO
12 27.J 3-l.8 125 165 !iO 130 170
10 2J.O 29. J 150 200 200 I 55 205
120x 120 15 26.6 JJ.9 105 IJS 1-IO 105 1-IO
12 21.6 27.5 1~5 170 170 130 175
10 18.2 23 .! 150 200 200 155 205
8 IJ.7 18.1 135 250 250 190 2SS
IOO x !CO I5 21.9 27.9 105 IJ5 140 105 1-'S
12
8
"-~
,~•.?
22.7
15. S
130
155
170
250
I.,.,
cv
250
130
195
175
255
90 x 90 12 15.9 20. J IJO 170 175 135 175
10 13.J 17.1 ISO 200 205 ISS 210
8 10.9 13.9 190 250 250 195 260
6 8.30 10 .6 2.is JjO 330 255 3-IO
80,c so 10 11.9 IS . I 15l 205 205 160 210
s 9.63 IU 190 250 255 195 260
6 7.J-' 9.JS 250 JJ0 33 5 255 3-'0
70 x 70 10 10.J 13.l 155 205 210 160 21 S
8 8.36 10.6 190 250 255 195 260
6 6. JS 8. 13 250 JJO 335 255 JJO
60x 60 10 8.69 JI.I ISl 205 2!0 160 215
8 7.09 9 .03 190 2!0 i60 200 265
6 5.J? 6.91 HO l JO 335 260 J-iS
5 ,U7 5.82 JOO 395 -WO 305 410
50" so 8 5.8~ 7.4 1 195 255 260 200 no
6 ,U7 l.69 255 JJS JJO 260 JS0
s 3.77 4 .80 300 -WO -IOS ) 10 41 5
45 x 45 6 4.00 5.09 255 3)5 J~O 265 )50
s ) .JS 4 .30 )00 JOO .io5 )10 -'IS
4 2. 74 J.49 370 490 495 JS5 l!O
-IO )( 40 6 J .52 4.48 255 340 J4S 265 355
s 2.97 3.79 305 400 410 J I5 4:0
4 2.4~ J .08 )75 495 500 )90 SIS
25 -< 25 s t. 77 2.26 315 41 l 4]0 J]l 4"'5
us sis s-is
4
3 I.II
1.35
1.4!
390
sos
SU
630 I 6H
405
SJO 7 10
Scandpower
Appendix C: Calculation of Section Factor of Protected Steel Profiles
Page C-7
~
Table 1.6
'
-rr- Section faclor H•/A
·Tl
Profile Bo,
Unequal angles 3 sides 4 sides 3 sides 4 sides
Size
DsB
Des.ignation
Thickness
I
Mass
""
mem~
~
Am
of
section
=l
''
''
' 1., ___
1..-----
~ ~ ~
'L
~~.:.:.:.:: .
''
.,
''
r- r- ;r' L' ='L' r-· r-, ,r,
' '.
,,' 'I
. I
"''
~
: :
.. ___ ,'
''.. ____ J''
~ ~
''
1,.
' ___ .,I
'' i.; _____ ,
' ' ''L____''
' '
mm mm kg cm• m·• I m·• m·• m·• m·• m·• m·• m·• I m·• I m·•
200x ISO Jg 47.J 60.0 110 !JO 90 80 115 90 85 90 8l IIS
JS 39.6 50.5 135 135 JOl 95 135 !JO JOO 110 JOO 14()
ll 32.0 4-0.8 165 165 130 120 liO l3l 120 JJS 120 170
200x JOO JS 13.7 43.0
ll
JO
27.3
23.0
34.8
29.2
"'
165
195
135
165
195
115
'""
165
90
110
135
135
liO
200
115
]45
170
95
115
l3l
IIS
145
170
95
Ill
135
140
170
205
ISOx 90 JS 26.6 33.9 135 135 llO 95 140 ll5 95 Ill 9l 14-0
12 21.6 27.5 165 l 65 14() 115 170 14() 110 14() 120 m
10 18.2 23.2 200 200 165 140 205 no 170
JSOx 7S 15 24.8 31.6 135 135 I :5 120 '"" '"" 105
12
JO
20.2
17.0
25.7
21.6
165
200
165
200
140
170 "'
135
90
'""
170
:?OS
145
175
95
115
140
120
145
175
9l
Ill
14-0
140
175
210
J25x 75 12 n.s 22.7 165 165
10
8
15.0
12.2
19.J
15.5
200
245
200
245
165
20l
'"" "'
14()
170
170
205
:so
145
J70
210
120
145
J7S
145
170
210
120
l4l
liS
:?JO
260
175
IOOX 15 ll JS.4 19.7 170 170 135 125 175 1-lO 125 14() 115 180
JO B.O 16.6 200 200 160 145 205 165 ISO 165 ISO 210
8 10.6 13.S 250 250 200 180 2~5 205 185 205 185 260
IOOx 65 10 12.3 15.6 200 200 165 14-0 :?OS liO 145 170 1.:s 210
8 9.94 12.7 245 245 200 175 255 210 180 210 ISO 260
7 S.77 J l.2 260 280 230 200 290 235 205 23.S 20.S 295
80x 60 8 8.34 10.6 250 250 200 180 :':$5 210 190 210 190 26.S
7 7.36 9.38 285 28.S 22.S 205 290 235 215 ~JS 2JS 300
6 6.37 6.1J 330 330 26l 240 335 :no 250 :,o 250 345
7~x so 8 ;_39 9.41 250 250 205 180 260 210 185 :10 185 265
6 5.65 7.19 330 330 270 :?35 340 275 24() 275
65x 50
•
6
6.75
5.16
S.60
6.58
:so
33l
:so
335
205
265
185
145
260
34()
210
275
190
250
210
2":5
'""
190
250
345
:265
3l0
5 4.3.S 5.54 395 395 3J5 290 4()5 ;:s 295 3'5 295 .;1s
I
.:
~SINTEF Scandpower
Section factor H 1/A

Table 1.7 ~ P:ofile So,
Structural tees
LT
0
'fff
3 sid<1
Yif :
T :
·r l 'T:
from universal beams
V,L/~/t/..lU~ 'a'~/t,f(-X
I
I
.
I
I
. : I
I
I
I
I
I
I
I
I
I I
Serial \ I.us Wid:h De~th T hickncu Area : I '

I
I
I
I
I
I
I
I
I I
r 'I
I '
I
:
size per of of
section
of
se:tion
I
:,J r-
\..: _____-., '.
I
I
I
I
I
I I
IL,; _ _ _ _ ..,I I I
metre section CJ l. - - - - _, L-----'
8 D t
m·' I m·• I m·' m·• m·•
mm kg I mm mm mm I c:nl I I m· • I
I
55 55 70
.I 19x .151
,-,
1'14 .i:o .5
.!l.~ .5
.Jm).1
455.1
~!
19 j
5 2J7 :!
21:l. 7
55
60
ill
so
70
Sil 60 60 80
Sil 85 65 65 85
305><.!$7
,:,
145 307.:l
3115 5
46) 3
J~') !
l\l 6
I7 3
l:l.! .i
161.4
65
7:,
S5
95 95
105
15
85
7::.
85
95
105
11 : jll-1. ! -+55 .! 15.'l 1.1: 6 ll.l5
1:0
Jll).4 451.5 15 2 1:!!i.1 95 It., 115 95 95
2CJ.:!><-l!<J
ltll
II.~ 293.:l J:!).-' lt>. I ,~.j 3 so 9~ 100 80 6ll 100
er ::?lJ~.J J:ll l IJ. 7 l:!J.5 <;\I
11;0 ,.,
115 !Li
I: 5
90
100
90
IW
115
1:5
2r,7x)SI
~ll
'I')
2'11 6
!ht'\
.Ii i •.!
)~:l
31'I
I.I
15 6
I.! J
11: tl
I:; .1
l icJ 1
~II
911
,m
115
105
115
85
95
85
95
lll5
1:0
R7 21,i, 7
7.1 :1,5.3 371, 'I l:!'J 94 11.: ltl5 1.15 IJ5 110 11 0 us
l!ll Ill) '10 9ll 110
::!5.i xJJ _j :l5 :!55 s J.!n. 5 1.:.5 WX. 1 :<:'
7r, ~~J j j4} :I 1.; : \ll,,91 \)• 1:11 1:0 95 95 125
711
/ij
~~j
::-_;
7 .1~ 1 ,I
J.;9
I ~ .1
II 7
~9.Jl
; 9 SI
Ill:'
115
1;11
,~5 ,~,
I.'ll 105
l !5
l05
11 5
135
15tl
JU5x j05 11 11
1)1,
75
Jll.5
JtJi
3H4 S
.llli 5
JllS 7
JIJJ :<
,~
IS:,
I
II.II
151.S
11.,,9
\/;' ,Un
t,.I
SH
'J5
~ll
1(1:'
l ~,
105
l15
~u 6()
81)
95
6(1
80
95
85
11 0
I JI)
!:'I x JIJ5 iU 1.,o I ,u,u 1.1 I X'I l<l 11, 1:11 1:0 95 95 1:u
h.~ :::,, j()~ ~ II 'I 7'I 7i II':' l.'ll 135 Ill:' 105 135
145
5; :?:~ ~ JnJ 7 II ' 1: ~.:! II:' IJ• 1.15 115 11 5
11,,1 uo IJU 165
~ IU >< ::!h7
51
1,I
~.:! 7 . t,
:r J tJ
Jll l I
:1:..~
1111,
I .:! ,'\
6-J
7;
59
:it )
I''
95 '"''
1:11 l .. , 95 95 1:s
5:'
51
:!Ill 7
:1u I
::nlJ 7
::!h.' -l
I! I,
Ill '/
6') ~,,
1,.1/,.,1
1115
115 ,~,
L-.~ 135
1.i;
110
115
:s
1 lll
11 5
l -10
150
41>
41
:!IN .l
:u:--: 7
:no I,
::r-..: ::
111.:
')04J
:'S ~-~
s:: :::
I''
!>Cl ,-,
ll>ll ll>CI
l~ll
I
140
1:5
1~11
160
ISO
P>I x:~11 ,111 111:: :< :JJ. ~ I I. .! n: .nJ Ill:' IJ:' i;-; 105 !OS 135
::!.\l .:< Ill n 5n ~J~ II' l-'5 t-l 5 115 IJ j 150
-15 l'J"
JI 1111 ,; :;o I '> .'\IJ 5:. :n t: 5 INI ll>Cl ,:s 1:s 160
37 {'>41 5 ::::s n ~, 10 ..i-; J.') p-i r:·, 175 135 135 1,5
1:0 195
J-1 1:-t\J 9 ::!:f. :< S5 - '-
.J " ..... 151) l'/fl 1115 l:'IJ
15t)
,~:: ( ::::1, JI 15; 5 ::;: 5 IU "'! 5:. ::J I'' I .I.' l.15 1:0 1:0
]JIJ h y :-:,,, -1'7° ~l) t::5 IS.' 160 !jl) 130 161)
37 15: 7 ,~o
J4 15U 1:::0:.r, 9 . 111 .1:.10 PO (75 Ii 5 145 1.15
Jtl I 5::.<J :::i J s 3; <JI, 11-d 1'15 :uo 161) 11,0 2lMI
:1, 15:: •.1 ~~l ',J 7o 3; ::~ li<ll ~:o 2:5 IStl ISO 2::5
l ,X'< :03 _;; ,-,, ; :nn..i 9.7 .n .:1 I'" 11,il
,-5
ll>Cl l::!5 1~5 165
J4 17:l S :1~ 7 ."< s .i: 7J 1.15 175 I.Ill 140 l:iO
200
311 I ii.:< 10,;. ~ 7.:1, 3S.INJ 1;'11 l'I' 2lffl 155 155
27 ,-n, :nu 7 It :-~ :t 1n5 :?!5 :!:lJ 170 170 nu
1.1: J :01 : ()'} :!9. 4S 1~11 2)0 :?JIJ 185 ISS 235
140><:llJ ?J
111 I.ii :I l'lS.b Id 2l ii) :!15 '.!-o 210 2:u 2:0 275
}.j 173 1 I:<:! 9 ill J:. 11 I,, 11"1 165 1:; 125 165
171 x 178 145 195
~') 11:. 1 17'l.J s 3ti.ll'l I J5 I')I ) JI)() 145
171 5 177.S i .3 3: :?9 INI 2111 1.15 165 165 215
,2n
_,.. 171 171! t,\l 2S .11, INl 2~11 1.U ) IS5 IS5 2J5
171> .i 65 2~. 71l JlJtl :!JII 2-ll) IY5 195 :?JS
l~I>< 17!< 20 I ?It
17 115 -l 17.i. : 5.'J 211.IJI 1:u :?SO 2Sll 2~5 2~5 2:!5
ll>o S 155 .I 1. i 34 l\l 135 IS5 1~5 1411 J.!0 190
lft5 >< 15? 27
_,
r IM 7 153.5 r,, 7 29 -'5 l hil ?Ill ?l:i
2-!5
I fill
ISII
l oll
IW
:!15
2J5
211 11,5. 1 15 1.') 6.l 25.75 ISll ~-10
24 l:!5 ~ 155 : 8 .S•I )II JI 1411 l~tl !Sil 145 IJ5 1:15
1: ixl.i~ 210
21 1~-1; 1533 8 :?n 59 11-il :?05 105 160 1611
:?~5 2}I) ISH ISO 2.30
llCx 15.3
19
17
J.j
l~J 5
Ill? .i
ltll 'I
,~,,.,
151.IJ
15.l I
i .:!
1,1,
t,, I•
:!3 7.1
211.SS
IS 15
175
1115
~:5
2Jll
?75
245
2:-1)
21~)
2:!~
2lk)
2:?5
250
2.'!0
IOI 6 15: -' 5.!! 15 r,t ) 255 315 3:11 2Nl 260 3:5
! 13
,, 147 J l.:!'1 :-i 7.J :?7. 55 ,~; l'JS : tMI 151) 1511 2lkl
i.in x 1::-
l'J 1.11, J 1:s b.J
.,..., -,
~
,_ 11!5 2=s 2311 1; u 170 230
21l tkl l'J~ :!n5 170 2tkl ;I~) 270
In 1.11, I 1::u h. l 255
IO~x 1::7 IJ 10:1 ljU ~ t,, J IS.IN PJ~ z:-o ::55 2m 2fkl
1,, 1 1111),! !'.:t) ?~O ~ !'«) 2~5 225 2S5
13 1111.•) I~:< 5 J:o
IU I It I:1 5 :i IJ ;I 2"'5 315 .315 2~11 250
II
133,c 111: 15 1.U s 1113 .I l>.J l'I !XI 175 :!-'' ~ ..; 11111 !Sil
21()
2511
~')0
1.1 IJj J IOI It 5..'4 lit 15 :w; !:<5 2~) 1111
!~IJ 275 210 2 Ill 2:(11
i u:"( 111: 1: ltll ft 1111 I, 5.: I.I 5 :11.;
4 7 I: I :?:5 Jll;i Jl!I 2J fl 1Jt) 31 5
10: <!'<1 J 1(1 llll n /Ill •I
315 235 235 )~5
~N <ih II M~ IJ 7n. : .in 111: ~.\IJ JI 5
7h 1/M 1 7h 1 6' '
.j . !'C ,N 13' J:11 J:.- :m 2-'11 J.' ~
Scandpower ~ SINTEF
Risk M anagement AS Norwegian Fire Research laboratory
I
Table J.8 'I B
Section factor Hp/A 7
Structural tees
I Profile Bo, I
3 sides
f]r.
from universal columns
Serial Mass Width Dep1h Thid,ncss Area ·=r-'

1/t,//~~//¾: 'lc'&if/4"~
,Jl_,
_____
~T] 'T'
'. '
"$-'.
✓.:"'1-',,.c:1/4
''
;,~~b"...'a'/4:
'_J_'
''' ' 'T;
:' ' ;
' .'' ' ''
._ -..,
size P" of or or
" " ' i..-----.l ' i..: ____ ::,J t. _____ !
metre se:1ion se;:tion SC(lion

B D I
mm kg mm mm mm cmJ m·• m·I m·• m·• I m·• m·•
406)( J78 118 39~.o )90.5 JS.5 }49,9 50 50
J68x 178 IOI 374.4 187.3 16.8 129.0 55
"
85
"85 60
l0
60
80
Bl
89 3il.l 164.2 KS JJ2.9 65 95 95 65 65 )(l()
n 370.2 16!.0 JZ.6 9i.6 7l 110 110 75 7S Ill
65 368.3 Jii.8 10.7 82.5 85 130 130 90 90 130
305>' 152 79 310.6 163.6 15,7 100.6 60 90 95 65 65 9l
69 308.7 160.) JJ.8 Si.3 70 105 !OS 70 70 110
l9 306.8 157,2 11.9 74,9 so 120 JW BS 85 1'!5
1~3.9 61.6 95 l~O
49 304.8 9.9
S:4.S 65
'" J.:5 100 100
254 >': 127 66 261.0 IJS.2 ]$.6 90 95 65 65 95

258.3 133.4
54
45
37
155.9
ZJ4.0
130.2
12;,o
13.0
10.S
8.6
61U
Si,O
46.4
"
90
105
110
130
160
II l
Ill
160
"
90
110
75
90
110
I JS
us
IOl
103 X 102 <3 208.8 Ill.I 13.0 55.0 ,5 ,o
36
30
206.2
205.2
JOS.0
)04.8
10.3
9.3
O:!.S
3'7.9
90
105
110
135
)60
115
135
160
80
95
110
95
110
"'
140
165
26 203.9 103.1 8.0 33.2 120 180 JS0 125 J.~5 IH
03 203.2 JOJ.6 7.3 :Z.9.4 IH ,oo :ws )40 1•0 205
15:Z.x 76 19 1~4.4 S0.9 8.1 23.7 JJO )95 )95 IJS i:~s 200
15 152.9 -;~.7 19.J .:::.s :Z.40
)2 152.4 76.2
6.6
6.J 1'.9
160
:oo 300 30l
160
:OS ""
205
:.:10
310
Section factor Hp/A

Table 1.9 --'-' Bo,
Rolled tees ;. ~- ;,:, Profile I
·-• -,
"'..!..
3 sides .: sides I J sides 4 ~ides
'
Size Mau Width
"
Depth Thid.ne5s Arca
I _Tj _JL =T: ;T: ;T;
•,:::✓/c,'lP b'/2'N <!1 ·-'@/4'/4-'b'...X:::
. '
///4"(1/-'~"//t
r r
' '''
I B>D
P" of
me:rc section
B
or
se,~ion
D
'
of
sei::1ion i
I
,, ______ , '
u ''
'
''L-----.J''
r
'''
1,, _ _ _ _ _ J
'
I mm kg mm mm I mm I cn::1 I m·• I m•: I m·• m·I r.1•1

!
SJ X51 6.92 l0.8 9.5 17.5 :w :30 r;s :~o
I
50.8 S.82
4.76 S0.8 50.8 6.4 6.06 :50 .$25 135 :.so 335
I 44>'.44 4.JJ 44.4 4J.4 6.4
,.s
S.14
,.oo
:z.ss 3::!5 ,:..:o :~s :.:o
3.14 "·4 .:..;.4 :;s ~:o ~s :'H .:..:5 ___ ...,
-
Scandpower
iG»SINTEF Risk Management AS
!
Table 1.10 t;:--1 Table 1.10
Circular hollow
0,
concinued
sections Section Section
factor Hp/ A factor HP/ A
~a!
Profile o r Box Profile or Box
Design31ion
Outside
diameter
D
Thickness
I
Mass
per
metre
Are3
of
section
f' j , fQ ;
I
I
'- ✓
I
'
I '
'-- ---J
Designation
Outside
di~mctcr
D
Thickness
I
Mass
per
metre
Arca
of
section
{J
I
·- .
'·
~
,
'
'
I
I
'
'
1...----~
m •l I cm 1 I m ·l
mm mm kg cm' mm I mm kg
21.3 3.2 I.H 1.8! 370 ~.U.5 6.3 37.0 47. 1 165
8.0 46. 7 59.J 130
26.9 3.2 1.S7 2.38 355 100 57.3 73.7 IOS
33.7 2.6 l.99 2.5J J IS l::!.5 71.5 91.1 85
3.2 2.41 3.07 3-45 16.0 90. 2 115 65
4.0 2.93 3. 73 235 211.0 Ill 1-l l 55
4!.4 2.6 2.55 3.25 4 10
3.2 3 .09 3.9-4 3J0 273.0 6 .3 4 1. J 52.8 160
4.0 3.79 4.33 275 8.0 52.J 6o.6 IJU
10 .0 6-l.9 8!.6 105
43.J J .2 3.56 -I.SJ 335 1: .5 SIU IO! 85
4.0 4.37 S.5i 271) (;9
16.0 IOI 65
s.o 5.J.I 6.80 1:?5 2tJ.0 l:~5 159 S5
f,{l.J 3.2 4.5 1 5.14 .330 !5 0 153 195 45
4.0 S.55 7.07 270
5.0 6.Sc 8.6Q 2:0 3::3.9 6.3 49.3 61.9 160
76. 1 3. 2 5.13 7.33 315 s.o 6::.J 79.-l 130
IO.0 77.-l 91!.6 IU5
4.0 7. 11 9 .06 :?t,5 ,,,
S.'!.9
5.0
3.2
8. 77
6. 76
11.!
8.6!
!15
3:5
l'.:!.S
11\.0
::u.1J
91>.t)
1::1
150
·--
155
191
85
65
55
-l.0 8.J:I IO. 7tl 2f>ll ~5.0 IS-I 235 -'5
S.tl lll.3 13.: !IU
11-l.3 36 9.SJ 12.5 :?:-i5 355.fi :!.I) 6:i.fl 87.-l 13U
s.u 13.5 17.2 ! Ill ltlll S5. :: IU9 100
6.3 16 .S !1.J (71) I :.5 106 135 85
IJQ. 7 S.tl 16.6 ~ J.:! 1U5 Iii.I) IH 171 65
6.3 20. 7 26.~ 1115 ::n.tl l6ii !II 55
11.IJ 21>.0 3.).1 I~°' !5.tl 21µ 260 ~5
Ill.II 3:.o 40 .7 !Ill
J U6.J IO.O 97.l! 1:s 100
16/U s.o :ti. I 25. 7 :as L'.!.5 1::1 155 80
6 .3 15.! 37. I 165 16.ll 15~ 196 65
11.1) 3 1.b -l0.J !31) ::o.o 191 2-13 55
Ill.Cl JQ.O 49.7 1115 15 .tl 1JS 300 -l5
193. 7 SIi 23.3 2\1.li :os J:.11 !'15 376 35
6 .3 2Q. l 37. 1 165
ll.il 36.6 46.7 130 J57.ll lll.O l tll 1.10 105
Ill.ti -l5.3 S1.7 105 1::.5 137 17S 80
l!.5 55 .9 71.2 115 16.IJ 17-l 21~ 65
16.ll 70.1 89.J ill :?U.ll !16 2ij 50
219. 1 s.o 2f,A 33.6 1115 25.0
3! .U
266
335
339
.i:1
.j()
35
6 .3 33. 1 .i:. 1 Jt,5
8.0 -ll.6 53. I IJII 40.0 -lll 5!~ 25
lll.tl 51.li 65.7 IUS
1: .5 63. 7 81.1 85 SOl!.0 IO.O 1::3 156 100
16.0 80.l II)! 65 l'!.5 153 195 80
20.tl 9:!. 2 1!5 55 16.0 I'>-' 2J7 65
Scandpower
Risk Management AS {GJ)SINTEF
Norwegian Fire Research l aboratory
Appendix C: Calculation of Section Factor of Protected Steel Profiles PageC-11
7
I)
Table l.11 1---1 Table 1.11

Rectangular continued
hollow sections · o ·. -1 Section Section
factor H,/A
(square) :_ - - • factor H,/A
~--~.C....-~
3 sides 4 sides 3 sides 4 sides
□ ;□
·□·-.,□.
Dcsigna1ion D,si,sna1ion ~lass Area
Mass Arel '
!--· ' Thicl.:nt"U of :' :' .:_ __ --~'
___ ,: : '
'"
Sitt Thickness of Site
DxD t '"
metre scc1ion o,o I meire Section
m·• m·•
mm mm
2.0
••
l.12
,m'
1.J2 .;25
mm
!:!OX 120
mm
5.0
••
Hi.0
cm'
22.9 l!iS 210
2.S us t.i2 JSO 6.3 22.3 ::s.s 1::s 170
t,i,O 27.9 35.5 1(0 135
2.0
2.5
3.0
1.43
}.i4
2.(14
l.S2
2.:2
:.w
"°
?,-tO
:YO
1().0
12.s
;.-1.2
.;).6
43.5
53.0
RS
70
l!O
90
l()x30
3.2
1.S
3.0
2.15
2.14
2.51
2.74
'
i20
.. 275
330
,SO
JJO)( JJO 5.0
6.3
s.o
21. 1
26.3
32.9
26.9
33.5
-11.9
155
125
110
2JIJ
165
135
3.2 :.tis 3.'.\8 26S 10.0 .m..i 51.5 AA 110
:.92 3.72 325 12.5 --l9.S t,3,0 OS 90
2.5
3.0 3.45 -t40 :!75
s.o ::8.9 155 :JO
).2
,.o
3.66 J.t,6
'"'
:w ISIJ)( ISO 22.7
3b,0 125 165
.
4.-16 5.t>8 6.3 ::8.3
s.o .;5,1 135
.,
175 8.Cl 35.--l J(IU
5.-tO 6.t,;.8
.:i;u 55.S I Ill
1.5
3.0
,...
3.71
-U9
-t.72
5.bO
3W
~70
l0.0
12.5 53.4 oS.O
i-...s
"°ss
65
3.2
,.o 5.72
5.94
7.:8
•«
~i,5
16.0
"'
!~))(Hill 6.3 JJ.2 --l3.6 1::s !05
5.0 6.IJ7 S.88 170
KO .i:;.o 5-1.7 IJ(l
t>O><bO
6.3
3.0
6.49
s.~
JIJ.8
6.80
1'0
~(,5
w.o
12.5
53.ll
M.2
bi.S
83.0
'"'bS
,o IUS
,s
3.2 5.67 1.:2 ::-o 16.0 Xl.4 1!4 .<O 711
,.o 6.97 S,i,.f; :us
J(,5
5.0 b54 IU.9 .:t,;,6 1::s lti5
.
135
~oox::1x1 6.3 31\.2
6.3 10.5 13.3 ,.o Jfi.0 hl.1 HO IJ{I
70xiCI
8.0
3.0
1:.s
6.28
Hi.3
S.IIO
I 1(1
...., ,,
l(J.0
1::.s
59.3
no
15.5
93.0 .,"' HJ5
RS
J.6 7.◄ 6 950 H,.0 91.5 I 17 .<o 711
5.0 10.1 12.9 i65
6.3 1:.s 1~.9 130 ::~o>c::50 6.3 JIU 61.2 1::s Jt,S
l!O ,s
6.0
..,
15.3 19.5 1\.0 60 . 5 7i.l JJ(I
!US
SCIXf:O :to
3.6
s.o
6:;9 9.20
I0.9
:!bU
::::o
t{J.0
1:.s
16.0
75.0
92.b
l)7
95.S
118
1'9
"°
65
.<O
s.<
65
6.3
KO
Jl.7
1.:.J
li.S
U.9
lfiA
22.7
'"'HIS
1:-0
300x300 rn.o 90.7 ll6 so w.s
3.6 9.i2 12.4 ::::o 1::.s ]12 1.:3 OS
;o
R5
65
16.0 JJ2 181
5.0 13.3 16.9 J(,(J
6.3 16.4 20.9 130 :;,o)(JSO )(J.0 !06 136 75 HIS
]05
IO(IX IOO
8.0
,.o
5.0
20.4
1:.0
1-t.8
:S.9
15.3
18,9
195
160
12.5
16.0
132
167
168
::13
65
.10 "'
65
16.4 23.4 J)0 .!!)(J;w.mO 10.0 1:2 156 75 !US

6.3 60 ,;
8.0 2:!.9 29.J 105 12.s 152 193
16.0 ::.:s 50 65
.-
10.0 27.9 35.S OS
"'
Scandpower
Non.vegian Fire Research Laboratoiy
---,
Section factor Hp/A
Table 1.12 ~ 3 sides 4 side~
·-
Rectangular
hollow sections iQ-, ,~□" '(□
f□
I o
I
'
I .
o
. 'I
I
':
I
I
I
I
I
I
<
!I '
o '------ ___ ,
I • I
I
•
0
Dcsign3tion
M.us Area 'I .' ' '
't.; _____ ' :,J
Size Thickness per of 1..:-----·

D >< B t mc:re section
mm mm kg cm: I m·• I m·l m•l
50x'.!5 2. 5 2.i~ 3 J7 360 290 HO
.3.0 3.1:! -1. 10 305 1-l5 365
.'
J .• 3.-1 1 -1.}J :!9U 2}0 3.;5
50x3U 2. 5 2.9: 3.i:! .350 2\15 J 3f)
.3.0 3.-15 -1 . .:0 ~'15 250 365
.3.! J 66 -1.66 2Sll 235 JJ5
J .ll J .:r, 5.63 230 195 2SO
5 tl 5.JO 6.38 190 160 235
60x-lO Z.5 3.71 .i.n J.;() 295 -CS
JI) .; 39 5.6() 2S5 250 355
.3.2 -1.66 5.9-1 270 2}5 335
.;,I) 5.i :! 7.28 ~:u 19') 275
5.ll 6 97 s.ss !Sil 160 ~:!5
6. } S.J'! 10 .~ 1.-0 111) 185
80xJ1l 3.0 5.3~ 6 .SU 295 :!15 355
J.~ 5.67 7.:: 275 !~I) .330
-1.ll 6.•n g $.'! 2:; ISO 270
5.0 8 .5-1 I0.9 IS5 1-+5 2~0
6.3 IU.5 IJ . J t:'1) 1:0 ISO
8.1) 1: ~ lo J 1:5 ICd l-15
9Ux50 .30 6 :s 8t:ll 2~1) 2.:u 350
3.1> 7 ..:n 9 :'tl 2JI) 2G<l 295
5.1) Ill I 1:., ISU 1.;5 2lS
63 1: .5 15.':I IJ5 1:u 175
!Ul 15.3 1\1 5 ,:u 95 1-15
llklx50 3 I) 6.75 S.60 29') 135 .350
3. 2 7.1~ 9. 1-1 :!75 ~:o .331)
J 1) ll.:ll> I' .
' J ~:o I75 265
5.tl Ill.•) !J.'.I l~O IJ5 215
l>.J I. • 17.1 tJ5 115 175
J -
ll.ll ln.n 21.1 1:0 95 1.:0
ll~)><r,f) 3.tl
~ ,. 9.:o :!:-t:' 2-'ll 351)
' ·--
.\.n 1U<J It)_') 1~fl 2m ! 95
5.11 I 1.1 IJ.•I 175 150 21S
l>.J 1.:.: 111.J IJO 1:0 175
S I) 17 :l !2. 7 I 15 95 PO
i:ux64l 3.n 9 7! 1:. .: ~ .ti) 195 ! 91)
5 II D.3 16.<J ISO (JO 21 5
Id l1>A :?O ') P5 I 15 170
~-ti !OJ 15.<J I 15 95 IJll
_,., -
1:uxSil 511 IJ.A !11.9 ,;u 151) 2IO
6.J IS J ' us t:ll l iU
!1.0 2:.~ ?'). ( I Ill 95 135
10.tl :1 9 35 5 ')f) &) 115
1511:< 11)0 5.0 IS. i 23 ~ 165 1.15 !IU
l>.J ~.3 s :?<J 7 135 1:0 lill
~-0 2'J ! 37 I I HI 95 135
10 IJ J.U -l5.S 91) ,; llll
I! 5 J3 6 55 5 711 65 90
ll>Ox :lll 5.0 !S ll 1:.'> I i5 I.II) 2 l0
6.J 11 J :?M.5 f JI) !Ill 170
X.ll ! i .9 35. 5 I l~ 90 135
!11.U J.i. ! -'J.S QI) 75 I Ill
t:.5 -II t, 53.U 75 60 90
!IM)x ll~l 5.0 2~. 7 !~.9 t7j f.11) Zit)
l>.J !:O JI, I) ! JI) I ltl 11>5
II.ti 35.-l -15 I lit) 911 ! 35
11)_1) JJ t, 55 5 \)0 ill I IU
53.J 61!.0 75 60 91)
·~ j 6n J .15 7U
11>.ll !I.I 5 60
25ll>< l :ill 6.J JX. ! .i.~ 6 IJ:i 11 5 11>5
II.•) .IS.ti 61. 1 1115 91) D IJ
lllll 5'1.J 75 5 115 i5 10.i
1:.5 73.1) 9:11l 70 60 85
lid) '.II 5 11 7 55 -15 70
)1Ml><21~1 l>.J -IX I 61.! l.'O 115 ! 65
1111 1,1) 5 ii, ( IU5 90 IJO
Ill.II i So 95 5 !i5 15 1115
1: j 9_:_,. I I~ iCI 60 85
11, II ![ 7 IJ'I 55 .15 65
..U NI )( :nu lll ll ')0 .i l lh x, ;o lll5
1: .5 I",_ IJJ 71l 55 85
lttJ J 1-1: ,.~J 55 JS 65
-l~U <:5o 111 II U!n l Jo S5 7il lll5
t: 5 u: h '-"< 711 55 1!5
l h II 11,7 ~ 1,l .55 JS 65
Scandpower
Appendix C: Calculation of Section Factor of Protected Steel Profiles
Page C-13
I Table 1.13 Castellated Sections

Ca:-.1clla1c:d Uni\'cri.al Beam!> C.1,tc:ll:1l<'d t:nivcn-al Columnr. (";,,1dh1ll·d Joi,1:-. c~"11.•Jl;1h•J Ch.1nm:b
Scrial!-izc Mas~ Scrial!-i:tc M;m Seri.ii r.izc Ma~ Scriab,iie Ma~~

per
""'
metre mczrc
per
metre ''"
mcut·
Oricinal
mm
C.i.s1c:1l:i1cd
mm l,
Oricinal
mm
C:.Mclla1cd
mm ,, Oricinal
mm
Ca.,1clb.1cd
mm k,
Oricinal
mm
Ca~1c!la1cd
mm ,,
Y)J x J)9 1371 X JJ9 388 350 X .:!Jl'I 5J6 X ➔06 0~-1 :!5.: X 20) 3SJ x 203 1-iUiS JJ:? X 10:? h-:S x Ill:? ti55J
3J3 551 25.: X 114 381 X ]i-4 37.20 ),1\j X 1112 572 Y 1112 55.10
si1.: x :ms 1371 X 3115 289 J07
;;9:;
W3x 152 3U5 X 152 52.09 Jll5 X !02 J5S Y lll2 Jfl.11'
:53 2tl3 X 1112 3115 x 102 25.33 ,\{!$ X Jl.t.9
224
201
~.;o
2~7 l7SXH12 267 X ](J2 :!154 25J X S9
:5J X
"'76 J%x
JXI x
Ji,;J X
S9
S9
70
J5.7J
2lt:!9
152 127 :!::!:[,; X J27 37.20
,.
235 X
xJH X 292 ]257 X 2'J2 226 :2YX X9 .,2. 1t•
,,.
JJJ X
19' J.'i6 x ;:.tiH 5;\:l X Jf)S :'112
li7
J52x S'9
J52x 76
:2.sx X9
:2sx 76
lt.09
Ji.ft(! 2:?IJX
2UJ X X9
:lJJ X
Jll:>x X9
'"70 2fl.1Hl
2'J.7t-
153 127 X ll4 l'Jl >: 114 ::!9.76
ll-13 XX 267 211J
,.",..,
762 X 267 197 127 X 114 191 X Il4 26.79 X Jll5 X 7h 2.•.X:'
li3 l:''J 17.i-ix :!:hi 2h.$1
Min X 254 )1129 >: 254 '"
170
305 X JIJ,S 4:,/i X 31J.5 283
:'J!J
127x
127 X
76
76
191 X 76 16.37
J9J X 76 13.36 17X x
152 X X9
X
2117 X 7h
::::,: X S'J
'" 21UW
2:u:J
J'JS Jl4 X J14 171 X 114 :ti.79 ::;:,i;: X 76
l:>2
152
140
125
151'
137
IU2x rn:?
X9 X
••
153 x lll2 .. ;.u;
134 X S9 J<.1.J5
1::7
1112
X
X
X
"
'"
;,
I'll X
15J X ,,;,'"
1':.XJ\
1-1.911
JO-C
t,l(I X 3115, 'Jl5 x 305 23X
17'}
JJ9 2:"J >" :.c:..i JXI:,, 25-1
"'"'
Jt,7
7ox 76 l]J X 76 12.hS ih X 3~ 11-1 X t,_70
x .:!:29 132
l',]U 'Jl5 X :21/ JJ{I
1117
125
113
IOI
21,:~:,, ::oJ 3115 X 2113
'"73
5::,J X .:!:HJ ""
,.
l'ilKJx::;111 122 7]
1119
!01
lJ2
"'
52
l'i2
J;i7 X J'JI MihX IY)
152 >< l:'-2 :::i;: > 15:? ;7
9X
X9
:.11
S2
:J
74
•7
J57x 152 hS6 X 152 K'.?
.,"
JIii, X )7S t>ll9 X r:'X
'"
52
74
,."''
•7
Jf>f,X ]J{J
-"'
hi~ X 1411
;9
J5t,x 171 534 X 171 h7
;,
;,
,;
,.,.
35tox 127 5J◄ X 127 39
33
.•us x lh5 J5X X lt>5 ;,
Jll
Jll5x 127 -l~X 127
42
37
3115 X Ill.:? J5X X 102 33
25..J X 1-16 3)-;I X )Jt,

"
.:!5
43
J7
31
~5,J x Ill:? JXI x 1112
"
.25
22
-.
211.• x 133 )1/_'i, X J,33 J(J
.25 I
(1'.ll)SINTEF Scandpower
Appendix D: Supplementary Models on Release Calculations Page D-1
APPENDIX D
Supplementary Models on Release
Calculations
D.1 EFFECT OF PRESSURE LOSS ON LEAK RATES

The error introduced by using formula (4.2) for leakages from pipes may be checked
by estimating the linear gas flow velocity in the pipe as:
9 m RT 4
V= ---- Eq. (D.1)
M p 1t. D2
where
v = linear gas velocity in the pipe (m/s)

D = pipe diameter (m)
M = molecular weight of the gas (kg/kmol)
m9 = release rate (kg/s)
T = average temperature in pipe (K)
p = average pressure in pipe (N/m 2)
R = gas constant (8314 J/K · kmol)
The pressure drop due to friction in the pipe can then be calculated as (Flanning
Equation):
Eq. (D.2)
where
f = friction factor f =f (Re, EID) can be found from Figure D.1

Re = Reynold's number =pvD/µ
L = total length of pipe (m)
D = diameter of pipe (m)
p = average density of gas (kg/m 3)
µ = dynamic gas viscosity (kg/m-s)
£ = roughness of pipe (m)
v = linear gas velocity (mis)
Scandpower
I
\.
Page D-2 Appendix D: Supplementary Models on Release Calculations
r If L'.pf is (by this rough estimate) judged to be significant compared to the pressure P
as used in Equation (4.2) the error due to friction loss must be considered significant
and Equation (4.2) will overestimate the leak rate. A more accurate estimate of the
leak rate may in this situation be obtained by an iterative calculation of the leak rate
and associated friction loss.
0.035 \ I I I I I I I
0.030
0.025
I Dr,=4rr .2
IP fl:
i 0.020
- I
." '
0.015
\f\~ e
f 0.010 o
0.008 "~
OJ>06 .' ...
'
0.004
0.002
0.005 0.001
0.004
I
0.0004
0.003
0.0025
0.0020
~aq~
' . - 0.0001
0.00005
0.00001
r-~ 0.000001
0.0015
0.0010
102 2 4 a 10> 2 4 a 10' 2 4 a 10" 2 4 a 10" 2 4 a 10• 2 4 a 10"
rVD0
R&=-.i,-
Figure D.1: Friction Factor as Function Reynold's Number and Relative Roughness
of Pipe, Ref. /D-1 /
If the gas leak occurs in a pipe fitted to a process vessel, the friction loss in the pipe
should also be taken into account. The actual pressure loss in the pipe can then be
obtained by use of Equation (4.2), (D.1) and (D.2). (The calculation process is iter-
ative and may be solved either manually or by a simple computer model.)
Scandpower
D.2 GAS CONDENSATION

To check whether the pressure and temperature drop brings the gas into a two-phase
flow condition the pressure enthalpy diagram for the gas may be used, see Figure 0.2.
--• -rt - t J
: -: ! ,
! f. . ,I i/. I t/i
.
I
·r-r .
·~~ • lj- ! V I
.~
·-·
- I
·- . .... .<:Pl-
+,Y I ~
'""rr
,.. 1 - ;.i
I v
V
y
~
~ 1:- ~r,. '

. 'f ...,
4-+j·llf-.l- -tt-+-tt-t V t}t v ~
II ':T
'
1 V / l/ 1'71 . .'
~ ...., V ~ ,...
... I~
l
If. ~
I ~i ~ ~111 I:'.
. .
.
- -
~ .~ ~ ·I'
I,
p
'
Figure 0.2: Pressure Enthalpy Diagram for Methane, Ref. /0-2/
If the gas in question is e.g. methane at 75 bar and -50 °C the state of the gas can be
described as point A in Figure D.2 By isentropic expansion the gas will follow the
constant entropy line to point B corresponding to conditions at the location of the leak.
After expansion to atmospheric pressure the state of the gas can be described as
point E. From the phase envelope curve it can be seen that condensation will occur at
these points. The quantity of condensate and gas can in fact be calculated based on
the location of point B or E within the phase envelope.
Scandpower
The pressure-enthalpy diagram may also be used for calculation of the release rate of
the real gas (i.e. it takes into account the non-ideal behaviour of the gas).
From a total energy balance of the gas and by neglecting the friction loss and static
head of the gas it can be shown that the exit velocity of the gas is given by, Ref. /D-2/
Eq. (D.3)
where
h1 = specific enthalpy of gas in the vessel (J/kg)

h2 = specific enthalpy of gas at the point of release (J/kg)
The density of the gas at different conditions may also be found from the pressure -
enthalpy diagram as
p = 1/v kg/m 3
where
v= specific volume of the gas (m3/kg).
Example: As an example the release rate of methane gas at 100 bar and +50 •c
through a 2 mm 2 hole will be calculated by the two above described methods.
1) Eq. (4.2) gives with y = 1.3
.----o------,-1+1.3
2 1 13 100-105
m09 =2-10--4 -0.62 1.3(-- ) - · •
1+1.3 323-8314
16
⇒ mo9 = 2.0 kg/s
2) Using Eq. (D.3)
and the pressure enthalpy diagram for methane, Figure D.2, he initial condition of the
gas may be plotted as point C in Figure D.2. By calculating the pressure at the re-
lease point from p = l!I · P2
where
p = pressure at release point

P2 = pressure in containment
ljl =see Chapter4.3.1, Eq. (4.1)
and assuming isentropic flow, point D in Figure D.2 is obtained. The gas velocity is
then calculated as:
Scandpower
Risk Management AS G}SINTEF
V= .J2 (593 - 525) -10 3

= 368 mis
The density of the gas at release point can be seen from Figure D.2 as the inverse of
the specific volume:
1 1 3
p =- = - - =47.6 kg/m
V 0.021
By introducing the discharge coefficient, C0 , the leak rate is obtained as:
mo9 = Co · A · V · p
4
m 09 = 0.62 · 2-10 · 368 · 47.6
mo9 = 2.17 kg/s
This method may be considered more accurate as it takes into account the non-ideal
behaviour of the gas. However, the method is generally limited to single component
gases as pressure enthalpy diagrams for mixtures are not readily available.
Scandpower
D.3 LARGEST NON-PROPAGATING CRACK IN A

PIPE
According to fracture mechanics, cracks exceeding a certain critical size will be
unstable, i.e. they will propagate quickly into a full size rupture and relieve the internal
pressure. For leak size classification and consequence evaluation it is important to
know the critical crack size, or the largest, just non-propagating crack. Based on the
work presented in Ref. /D-3/ the critical crack size may be assessed as indicated
below.
For ductile propagation of through-wall cracks the following formula relate critical crack
length 2a 0 to CVP (Charpy V absorbed energy) (Ref. /D-3/)
Cvp 8- ;:;2-a 0 • A • In sec {"

= --=------''--- MF · crT}
- ---'---'- (J) Eq. (D.4)
,r - E 2 ;;:
The factor MF is the Folia's form factor
2 4}"2
{
MF= 1+1.255~-0.0135 ·
Rt
~c
R t2
Eq. (D.5)
where
cr = flow stress = s-cry (Pa)

cry = yield stress (Pa)
s = strain hardening factor
A = area of Charpy -V sample (m 2)
E = elastic modules (Pa)
R = pipe radius (m)
t = pipe wall thickness (m)
r1T = basic allowable stress = d-cry
d = design factor
2a 0 can be calculated from Eq. (D.4) for increasing values of CVP. As CVP and 2a 0
increase, the curve flattens out due to the lnsec - term getting very large:
Sc and power
2a
C
"a*C
-
- - - - - - ---==-=-----
"-----------------------~~CVP
Figure 0.3: Critical Crack Size as a Function of Charpy-V Absorbed Energy
2ac cannot be increased beyond the limiting value 2a~ , even by making CVP very
large. 2ac is no longer CVP-controlled, but stress-controlled.
The limiting value of 2ac can be found from inspection of Eq. (D.4):
1 1t MF · crr 7t
sec = -- ⇒ oo for - • - - - = -
cos 2 ; 2
that is for
Eq. (D.6)
{Pa)
Using
s = strain hardening factor
and
arlay = d = design factor
we get
or
2 4 }1/2
~= 1+1.255~-0.0135~
d { Rt R2 t2
Scandpower
This is a quadratic equation in a~ /Rt which can be solved
~~ =5.38·10- 3
+1/2-[1.16-10-4 +3.19·{(~J-1}]2 Eq. (D.7)
Except for d ➔ S (which is not realistic) this can be approximated by
2 2
5
a 1 3.19· { ( )
R~ =2 d -1}
=0.893· ✓(~J-1
or
Eg. (D.8)
By applying a strain hardening factors = 1.1 and a design factor d = 0.66 one obtains
2
a~ =2.18
../Rt
The critical crack length may then be expressed as
2a~ =2.18 ../Rt (m)
Assuming a length to width ratio of the leak area of 10: 1 one obtains
leak area A = 0.475 Rt (m2) Eq. (D.9)
If the pipe wall thickness is unknown, t may be estimated as:
t=P..B__(m)
cr y -d
Eq. (D.10)
where
p = internal pressure of the pipe (Pa)
Scandpower
References Appendix D
/D-1/ R.H. Perry: "Chemical Engineer's Handbook", 5th Edition, 1973, McGraw-Hill Kogakusha
Ltd.
/D-2/ A. Lydersen: "Kjemiteknikk", Tapir, 1972.
/D-3/ A. Almar-Ncess: "Fracture Criteria in Pipelines, a Survey, in Proceedings of the Intern.

Conference on Underwater Technology", Bergen, 1982.
Scandpower
) Appendix E: Hazard Checklist Page E-1
1:
I
!
i
I
I APPENDIX E
Hazard Checklist
Hazard checklists are one of the most useful tools for hazard identification. With this tool, hard-
won experience can be systematized and utilized.
Many different kinds of checklists exist, depending on the activity they are meant to serve. They
are applicable to management systems as well as to projects. The challenge is to use approp-
riate checklists both for different management systems and project stages. Experience shows
that unsuitable checklists have a tendency not to be used.
Examples of activities in which checklists may be used:
Organization and administration

Design
Commissioning
Protection and safety
Personnel protection outfit
Fire protection
Industrial hygiene
Supervisory participation, motivation and training
Ergonomical and psychosocial relations
Accident investigation, statistics and reporting procedures
Storage
Transportation
Etc.
A large number of checklists are given in the literature, some examples are quoted in this
appendix.
However, the main purpose for all checklists are to identify possible hazards and give a final
check that nothing has been neglected.
Scandpower
«jJ) SINTEF Risk Management AS
Page E-2 Appendix E: Hazard Checklist
CHECKLIST EXAMPLE A
"Creative checklist HAZOP examination procedures flowchart"
Scandpower
Risk Management AS (r,»SINTEF
Appendix E: Hazard Checklist Page E-3
Creative Checklist HAZOP Examination Procedures
FLOWCHART
Divide the plant or area to be studied into suitable sections and choose team
State the first section to be studied
Enter the chemicals and raw materials present in the section on the first part of
the Chemicals Checklists
Complete the first part of the Chemicals Checklist noting hazards or nuisances as
appropriate. (Minute questions as necessary)
Complete the second part of the Chemicals Checklist. Note figures and rate the
hazard/nuisance as appropriate. (Minute questions as necessary)
Enter plant (or area) section on Hardware Checklist
Complete Hardware Checklist ticking potential hazards or nuisances. (Each tick

should correspond to an appropriate question in the minutes)
Choose the next section to be studied
Monitor the progress of action being taken in response to the minutes.
Scandpower
Chemicals Checklist for "Checklist" Type of Hazard & Operability Study
List in the first column all raw materials, intermediate products and effluents. Tick in the approp-
riate column those materials which are potentially hazardous or require special attention.
Chemicals Hazard or Nuisance

0::: Fire Explo- Detona- Taxi- Corra- Radio- Smell Dust
w
LL sion tion city sion tion
w
0:::
List the appropriate cell of matrix, the relevant data.

Note where the relevant data are not available in order to initiate their collection.
Hazards for Chemicals refer nos

nuisances Types of data required
1 2 3 4 5 6 7 8
Flash point. Open cup °C
FIRE Potential to spontaneous
ignition auto-ignition °C
Maximum attainable pressure
EXPLOSION Explosive limits % in air
upper
lower
DETONATION Detonation department data
Exposure limits T.L.V. p.p.m
TOXITY
Ventilation required
Effect on personnel
CORROSION
Effect on materials of construction
RADIATION Maximum permissible dosage
Effect on personnel
SMELL
Effect on site or neighbourhood
Effect on personnel
DUST
Effect on site or neighbourhood
Scandpower
Hardware Checklist for "Checklist" Type of Hazard & Operability Study
List in the first column all sections of the plant including vents, flares, effluent disposal, etc.
Tick in the appropriate columns items requiring special consideration.
Within the works Outside the works
C
0
"' -
-
- .!:!
.gi
·-
ii
.!:!
-'lij
&l u
C
0
~ -
- ~ "iii
-
-i;:: C
W•-
GI Ill u ; 0
ti·- J!l "' :a
GI Ill u ; GI
Sections
of plant
.!!!
0
z
I!!
.!!
IL
Ill
:::,
C
"i(
~ ,~ u:I!! ~ iii
w
C
~ ~
GI - ;, z
.!!
0
I!!
"'
ii:
Ill
:::,
C
"i(
~ ,~
:::,
=
w
Scandpower
(G))SINTEF Risk Management AS
Norwegian Fire Research LaboratoJY
CHECKLIST EXAMPLE B
"Design Review Checklist"
Scandpower
Risk Management AS {GJ) SINTEF
Design Review Checklist
Comments on additional Action

NA
sheet if necessary sheet
1. Statutory needs
- Identify relevant regulations

and codes
- Are there any exceptions to

statutory regulations?
2. Client requirements
- Identify design basis, design

changes and particular functional
requirements
- Are there any exceptions to

these?
3. Safety aspects
- Identify particular safety require-

ments, accidental loads and
requirements for safety document-
talion
- Check if design is in accordance

with these
DATE: REV.:
Scandpower
Design Review Checklist
Comments on additional Action

sheet if necessary
NA sheet
4. Weight Consideration
- Consider possible factors which can

increase/reduce the weight
5. Layout considerations
- Discuss layout aspects
6. Reliability
- Consider reliability
7. Maintainability and operability
- Consider maintainability and

operability
8. Constructability
- Consider constructability
9. Standardisation
- Consider standardisation
DATE: REV.:
Scandpower
Project
.
name& Material or equipment
number names
Date
Hazard types Reference number
Inventory
Main Sub
Detail
Explosive solids & dusts
Explosive liquids
FIRE
Explosive gases/vapours
&
Flammables
EXPLOSION
E Ignition sources
Detonation
N Exotherms
CHEMICAL Unstable
E Incompatible
High/low temp.
R
THERMAL Phase change
Decomposition
G
High/low pressure
y PHYSICAL Radiation
Electrical
Heights/depths
High speed
MECHANICAL
Relative movement and
vibration
Gases and vapours
I TOXICITY Liquid & solids
N (INCLUDING Animals & pathogens
T BIOLOGICAL) Dispersal
E Persistence
R
LOSS Corrosion
A
OF Erosion
C
CONTAINMENT Fracture
T
I Evacuation
0 Theft
SECURITY
N Blackmail
Sabotage
E Noises
N SENSES Smell
V Vision glars, fog
I Gases, vapours & sprays
R EFFLUENT Liquids, solids & dusts
0 Disposal
N
Flood, earthquake, dust,
M
wind, heat
E ELEMENTS
N
T
Scandpower
CHECKLIST EXAMPLE C
"Simplified Checklist for Hazard Analysis"
Ref. Du Pont Safety and Fire Protection Guidelines
Scandpower
Simplified Checklist for Hazard Analysis
1. Introduction
The following simplified checklist was derived from "what-if' questions and covers important
aspects of a production operation. The words or phrases in the list should serve to stimulate
questions concerning the subject.
2. Example
The phrase "Materials of Construction" should lead to such questions as:
"Have suitable materials been used in vessels, piping, instruments, instrument connec-
tions, agitators, dip tubes, valves, valve packing, vessel supports, flange bolts, expansion
joints, etc.?"
"Are corrosion tests needed or desirable?"
"Where plastic pipe linings or equipment are used, are the temperature and pressures low
enough or adequately controlled?"
Simplified process hazards analysis checklist "'

C
C
0
"'I!
·c
z
C
1j
"2
.c
I
.s ii
·a,
Categorv Subjects to be investigated 0
8. C
w
u
~
'" E
e8
ti)
~
:g;
0e Storage tanks Design, separation, inerting
u..JW Dikes Capacity, drainage
O~il;!! Emergency valves Remote control - hazardous mat'ls
Www
(!) ,_ ,_
~<Z Inspections Flash arrestors, relief devices
0 ::l!i -- Procedures Contamination prevention, analysis
,-3:~
ti)~ g Specifications Chemical, physical, quality, stability
e Limitations Temperature, time, quantity
~
0..
~ (!)
Pumps Relief, reverse rotation, identification
<ii!: Ducts Explosion relief, fire protection, support
ii: .J
we Conveyors, mills Stop devices, coasting, guards
,-z
Procedures Spills, leaks, decontamination
:iii~ Piping Railing, codes, cross-connections
,-:
z Procedures Start-up, normal, shutdown, emergency
Wet/)
::l!i z II! Conformance Job audits, shortcuts, suggestions
!!: <
::, ti) ::, Loss of Utility Elect. heating, coolant, air, inerts, agitation
awe
W i=W Vessels Design, materials, codes, access
ti) ::J u
0-0 Identification Vessels, piping, switches, valves
w u a:
8 if a.. Relief devices
Inspections, tests
Reactors, exchangers, glassware
Vessels, relief devices, corrosion
a:
0..
Scandpower
a,
Simplified process hazards analysis checklist II) C
C 'C
'H j
0
e8. z
C c GI
Q.
'61 .c GI
C
u 'ti E
Category Subjects to be investigated 0 w
GI
I- c8
Electrical Area classification, conformance, purging
Process Description, test authorizations
Operating ranges Temp.; press. , flow, ratios, concentrations
Densities, level, time, sequence
Ignition sources Peroxides, acetylides, friction, fouling,
Compressors, static elect., valves, heaters
Compability Heating media, lubricants, flushes, packing
Safety margins Cooling, contamination
Protection Barricades, personal, shower, escape aids
.Jz Ventilation General, local, air intakes, rate
WO
z-
z l- Exposures Other processes, public environment
o frl Utilities Isolation: air, water, inerts, steam
"' 1-
0:: 0
w 0::
Hazard manual Toxicity, flammability, reactivity,
a. a. Corrosion, symptoms, first aid
Environment Sampling , vapours, dusts, noise, radiation
Control
Ranges, redundancy, fail-safe
"'w
co
Calibration,
inspection
Frequency. adequacy
Z>
<W Adequacy, limits, fire fume
u, C
Alarms
Tests, bypass procedures
.J >- Interlocks
00 Adequacy, vent size, discharge, drain, support
0:: z Relief devices
1- W Dump, down, inhibit, dilute
Ze> Emergencies
0 0:: Block valves, fire-safe valves, purging
ow Process isolation
:::E Air quality, time lag, reset windup
w Instruments
Hang-fires, runaways
Hazards
.J
w< Ditches Flame traps, reactions, exposure, solids
I-
UJO"'
Vents Discharge, dispersion, radiation, mists
~~C Characteristics Sludges, residues, fouling materials
Sampling points Accessibility, ventilation, valving

Procedures Pluggage, purging
Samples Containers, storage, disposal
Analysis Procedures, records, feed-back
w
0
z
Solutions, equipment, procedures
~
w Decontamination Size, obstructions, access
1-
z Vessel entry, welding, lockout
ci
:::E
z
0 Fixed protection Sprinklers, deluge, monitors: adequacy
wti
~w
Extinguishers Type, location, training
LL I- Firewalls Adequacy, condition, doors, ducts
ia. Drainage Slope, drain rate
Scandpower
Risk Management AS {GJ) SI N TEF
CHECKLIST EXAMPLE D
"Comprehensive Checklist for Hazard Analysis"
Ref. Du Pont Safety and Fire Protection Guidelines
Scandpower
(tj}SINTEF Risk Management AS
Comprehensive Checklist for Hazard Analysis
The question listed should be used to stimulate the identification of potential hazards and should
not be simply answered by a "Yes" or "No". Some questions may not be applicable to the review
of a given production operation.
I. Process Checklist
NOTE: Consider the checklist in terms not only of steady-state operation but also start-up,
shutdown, and upsets of all conceivable types.
A. Materials
Have materials been defined as "hazardous" or "non-hazardous" (ingredients as well as

final and by-products)
What process materials are unstable or spontaneously ignitable?

* What evaluation has been made of impact sensitivity?
* Has an evaluation of possible uncontrolled reaction or decomposition been made?
What precautions are necessary to meet environmental requirements and health of per-
sonnel?
What data is available on amount and rate of heat evolution during decompositions of any
material in the process?
What precautions are necessary for flammable materials?
What flammable dust hazards exist?
What materials are highly toxic?
What has been done to assure that materials of construction are compatible with the
chemical process materials that are involved?
What maintenance control is necessary to assure replacement of proper materials, e.g. to

avoid excessive corrosion, to avoid producing hazardous compounds with reactants?
What changes have occurred in composition of raw material identification and quality?
What hazards can occur as a result of loss of gas for purging, blanketing, or inserting?
How certain is gas supply quality?
What precautions need to be considered relative to stability of all materials in storage?
What fire extinguishing agents are compatible with process material?
What fire emergency equipment and procedures are being provided?
Scandpower
B. Reactions
How are potentially hazardous reactions isolated?
What process variables could, or do, approach limiting conditions for hazard?
What unwanted hazardous reactions can be developed through unlikely flow or process
conditions or through contamination?
What combustible mixtures can occur within equipment?
What are process margins of safety for all reactants and intermediates? What are the con-
sequences of missing ingredients or wrong proportion of reactants?
What reaction-rate data are available on the normal, or abnormally possible reactions?
How thoroughly is chemistry of the process and any undesired reaction known?
What foreign materials can contaminate the process and create hazards?
What provisions is made for rapid disposal of reactants if required by plant emergency?
What provisions are made for handling impending runaways and for short-stopping an
existing runaway?
What hazardous reactions could develop as a result of mechanical equipment (pump,

agitator, etc.) failure?
What hazardous process conditions can result from gradual or sudden blockage in equip-
ment?
What raw materials or process materials can be adversely affected by extreme weather
conditions?
What process changes have been made since the previous process safety review?
Scandpower
l(j)SINTEF Risk Management AS
C. Equipment
In view of process changes since the last process safety review, how was adequate size of
equipment assured?
Are any venting systems manifolded an, if so, what hazards can result?
What procedure is there for assuring adequate liquid level in liquid seals?
What is the potential for external fire which may create hazardous internal process condi-
tions?
Is explosion suppression equipment needed to stop an explosion once started?
Where are flame arresters and detonation arresters needed?
In confined areas, how is open-fired equipment protected from spills?
What safety control is maintained over storage areas?
In the case of equipment made of glass or other fragile material, can a more durable mate-
rial be used? If not, is the fragile material adequately protected to minimize breakage?
What is the hazard resulting from breakage?
Are sight glasses provided only where positively needed? On pressure vessels are sight
glasses provided which have a capability to withstand the pressure?
What emergency valves and switches cannot be reached readily and safely?
When was pertinent equipment, especially process vessels, last checked for pressure
rating?
What hazards are introduced by failure of agitators?
What plugging of lines can occur and what are the hazards?
What provisions are needed for complete drainage of equipment for safety in mainten-
ance?
How was adequacy of ventilation determined?
What provisions have been made for dissipation of static electricity to avoid sparking?
What requirements are there for concrete bulkheads, or barricades to isolate highly sensi-
tive equipment and protect adjacent areas from disruption of operations?
What provisions have been made for relieving explosions in building or operating areas?
Do all pressure vessels conform to state or local requirements?
Are the vessels registered in compliance with state or local code requirements?
When were pressure vessels inspected visually, calibered, radiographed, hydrostatically

tested, etc.?
Has the history of use of all vessels been completely reviewed?
Scandpower
Risk Man agement AS ~ SINTEF
D. Instrumentation Control
What hazards will develop if all types of motive power used in instrumentation should fail
nearly simultaneously?
If all instruments fail simultaneously, is the collective operation still fail-safe?
What provision is made for process safety when an instrument, operating in process safety
as well as in process control, is taken out of service for maintenance: When such an
instrument goes through a dead time period for standardization or when, for some other
reason, the instrument reading is not available?
What has been done to minimize response time lag in instruments directly or indirectly
significant to process safety? Is every significant instrument or control device backed up
by an independent instrument or control operating in an entirely different manner? In criti-
cal processes, are these first two methods of control backed up by a third ultimate safety
shutdown?
Has the process safety function of instrumentation been considered integrally with the
process control function throughout plant design?
What are effects of extremes of atmospheric humidity and temperature on instrumenta-

tion?
What gauges, meters, recorders cannot be read easily? What modifications are being
made to cope with this problem?
Is the system completely free of sight glasses or direct reading liquid level gauges or other
devices which, if broken, could allow escape of materials from the system?
How has the area electrical classification been established and hardware and techniques
selected?
• What process details affect the classification, group and division?
• What "UL-approved" hardware is unavailable for this job? Does this require testing?
• Are any new techniques being applied on this job?
Is the electrical system simple in schematic and physical layout so that it can be operated
in a straight-forward manner? (This minimizes human error in switching for isolation and
load transfer)
What electrical equipment can be taken out of service for preventive maintenance without
interrupting production? How?
How is the electrical system instrumented so that equipment operation can be monitored?
Will this eliminate downtime due to equipment failures caused by unknown overloading?
What are the overload and short circuits protective devices?

• Are they located in circuits for optimum isolation of faults?
* What is the interruption capacity?
• How are they coordinated?
• What instructions are furnished for field testing during the life time of equipment?
Scandpower
(GJ)SINTEF Risk Management AS
What bonding and grounding is provided?

* Does it protect against static build-up?
* Does it provide lightning protection?
* Does it provide for personnel protection from power system faults?
Check lighting
* Adequacy for safe normal operation?
* Adequacy for normal running maintenance?
* Adequacy for escape lighting during power failure?
Is tankage grounding coordinated with cathodic protection?
What is being done to verify that instrument packages are properly installed? Grounded?
Properly designed for the environment?
What procedures have been established for testing and proving instrument functions?
What periodic testing to check performance and potential malfunction is scheduled?
E. Operations
When was the written operating procedure last reviewed and revised?
How are new operating personnel trained on initial operations and experienced operating
personnel kept up-to-date on plant operating procedures, especially for start-up, shutdown,
upsets and emergencies?
What process revisions have been made since the last process safety review?
What special clean-up requirements are there before start-up and how are these checked?
What emergency valves and switches cannot be reached readily? What procedures are
there to cope with these situations?
What safety precautions are needed when loading liquids into, or withdrawing them from
tanks? Has possibility of static electricity creation been adequately taken care of?
What process hazards are introduced by routine maintenance procedures?
What evaluations have been made of the hazards of severed materials during normal and
abnormal operation?
How dependable are supplies of inserting gas and how easily can supplies to individual
units be interrupted?
What safety margins have been narrowed by revisions of design or construction in efforts
to de-bottle-neck operations, reduce cost, increase capacity, or improve quality?
What provisions does the operating manual have for coverage of start-up, shut-down,
upsets and emergencies?
What economic evaluation has dictated the choice between a batch process and a con-
tinuous one?
Scandpower
F. Malfunctions
What hazards are created by the loss of each feed, and by simultaneous loss of two or
more feeds?
What hazards result from loss of each utility and from simultaneous loss of two or more
utilities?
What is the severest credible incident, i.e. the worst conceivable combination of reason-
able malfunctions, which can occur?
What is the potential for spills and what hazards would result from them?
G. Location and Plot Plan
Has equipment been adequately spaced and located to permit anticipated maintenance
during operation without danger to the process?
In the event of the foreseeable types of spills, what dangers will there be to the commu-
nity?
What hazards are there from materials dumped into sewers of neighbouring area?
What public liability risks from spray, fumes, mists, noise, etc. exist and how have they
been controlled or minimized?
II. Electric Checklist
A. Design
How complete does the electrical system parallel the process?

• What faults in one part of the plant will affect operation of other independent parts of
the plant?
* How are instruments for a plant protected from faults or other voltage disturbances?
Are interlocks and shutdown devices made fail-safe?

• What is the need for each interlock and shutdown used?
• Are interactions and complications minimized?
• Is continued use of protective devices insured?
* What requirements or standards were used for the hardware that has been selected?
* Is a "positive try" step provided for the Lock-Tag-Clear-Try procedure?
What is the probability of accessibility during mishaps of power disconnects starters, etc.?
Is communication provided to operate the facility safely? (Telephones, radios, signals,

alarms, etc.?)
Are spacing and clearances furnished for normal traffic maintenance, and for fire fighting?
Is there a schedule for checking operability of interlocks?
Where sequency controllers are used, is there an automatic check, together with alarms,
at key steps after the controller has called for a change, and is there a check together with
alarms at key steps before the next sequency change?
Scandpower
Ill. Piping and Machinery Checklist
A. Piping and Valves
Were piping systems analyzed for stresses and movements due to thermal expansion?
Are piping systems adequately supported and guided?
Are piping systems provided with freeze protection, particularly cold water lines, instrument
connections and lines in dead-end service such as piping at standby pumps?
Are provisions made for flushing out all piping during start-up and shut-down?
Are cast iron valves avoided in strain piping?
Are non-rising stem valves being avoided?
Are requirements of Engineering Standard S23G being met at interconnections where

possible cross-contamination can occur?
Are controllers and control valves readily accessible for maintenance?
Are bypass valves readily reached for operation? Are they so arranged that opening of
valves will not result in an unsafe condition?
Are any mechanical spray stem de-superheaters used? What are the consequences of
too much or too little cooling liquid flow?
Are all control valves reviewed for safe action in event of power or instrument air failure?
Are means provided for testing and maintaining primary elements of alarm and interlock
instrumentation without shutting down processes?
What provisions for draining and trapping stream piping are provided?
Scandpower
Risk M anagem ent AS (r,J) SINTEF
B. Pressure and Vacuum Relief
What provisions are there for removal, inspection and replacement of relief valves and
rupture discs and what scheduling procedure exists?
What need is there for emergency relief devices: breather vents relief valves, rupture
discs and liquid seals? What is the basis for sizing these?
Where rupture discs are used to prevent explosion damage, are they properly sized
relative to vessel capacity and design?
Where rupture discs have delivery lines to or from the discs, what has been done to assure
adequate line size relative to desired relieving dynamics? To prevent whipping f discharge
end of line?
Are discharges from vents, relief valves, rupture discs and flares located to avoid hazard
to equipment and personnel?
Is any equipment, operating under pressure or capable of having internal pressures deve-
loped by process malfunction, not protected by relief devices? Why not?
Is discharge piping of relief valves independently supported? Is piping as short as possible

and with minimum changes in direction?
Are drain connections provided in discharge piping of relief valves where condensate or
rain could collect?
Are relief valves provided on discharge side of positive displacement pumps between posi-
tive displacement compressor and block valves: between back-pressure turbines exhaust
flames and block valves?
Where rupture discs are in series with relief valves is the installed rupture disc next to the
vessel? Does the section of pipe between disc and relief valve have a pressure gauge or
sensor and pressure bleed-off line? Have any rupture discs been installed on discharge
side of relief valve?
What provisions are made for keeping piping to relief valves and vacuum breakers at
proper temperature to prevent accumulation of solids from interfering with action of safety
device?
Scandpower
C. Machinery
Are adequate piping supports and flexibility provided to keep forces on machinery due to
thermal expansion of piping within acceptable limits?
What is done to ensure separation of critical and operating speeds?
Are check valves adequate and fast acting to prevent reverse flow and reverse rotation of
pumps, compressors and drivers?
Are adequate service factors on speed changing gears in shock services provided?
Are there full-flow filters in lube-oil systems serving aluminium bearings?
Are there provisions for draining and trapping stream turbine inlet and exhaust line?
Are there separate visible-flow drain lines from all steam turbine points?
Are driven machines capable of withstanding tripping speed of turbine drivers?
Are non-lubricated constructions of non-flammable synthetic lubricants used for air com-
pressors with discharge pressures of greater than 75 psig to guard against explosion?
What provisions are made for spare machines or (critical) spare parts for critical machi-
nes?
Are there provisions for operation or safe shutdown during power failures?
Are vibration switches on alarm or on interlock for cooling tower fans provided? Is sprin-
kler protection needed for the fan deck on induced draft combustion cooling towers?
IV. Fire Protection Checklist
If the building has enclosed walls and the construction or interior has combustibles, are
automatic sprinklers provided?
If the building has open walls and the construction or interior has combustibles, is water
spray protection adequate?
What hydrants serve the area or project?
What fixed or portable monitor nozzles (on hydrants or separate) are provided for cover-
age of manufacturing facilities or storage facilities in open area (not within open or closed
wall buildings)?
Have the underground fire mains been extended or looped to supply additional sprinkler
systems, hydrants and monitor nozzles? Dead ends should be avoided. What sectional
control valves have been provided?
Are small hose standpipes provided inside of buildings?
What type, size, location and number of fire extinguishers are needed?
What flammable liquid storage tank protection has been provided? Foam ? Dikes with
drain valves outside the dike?
Scandpower
Risk Management AS (GJ) SINTEF
Where have total flooding or local-application Halon 1301 systems been provided? If
Halon, is there any decision made about replacement?
Is load-bearing structural steel exposed to potential flammable liquid or gas fires fire-
proofed to a sufficient height above ground level to protect the steel?
Has adequate drainage been provided to carry spilled flammable liquids and water used
for fire fighting away from buildings, storage tanks, and process equipment?
What protection has been provided for dust hazards?
What is the capability of fire water supplies? What is the maximum fire water demand?
How long will supplies meet this maximum demand?
What is the estimated maximum probable loss?
What is the approximate "hold-up" of flammable liquids in the manufacturing equipment?

Are "hold-up" amounts kept to a minimum?
What attention has been given to protection of process equipment from external fire?
Are liquid inventory tanks near or under the ground instead of elevated?
Is the area pad or flooding designed to conduct spill liquid away from process equipment?
What facilities are provided for drainage?
How have major storage tanks or vessels been located to minimize hazard to process
equipment in the event of rupture or burning?
Are all structures made of non-combustible materials and fire walls, partitions or barricades
provided to separate important property values, high hazard operations and units important
for continuity of production?
Are operating units spaced to minimize potential damage from fires or explosions in adja-
cent units and to allow room for fire fighting activities?
Have suitable locations been designated for fire alarms?
Has key data been developed and adequate protection provided for high piled storage
areas?
Scandpower
Appendix F: Characteristic Mechanical and Physical Properties of GRP Materials Page F-1
APPENDIX F
Characteristic Mechanical and Physical
Properties of GRP Materials
Typical mechanical and physical properties of GRP materials.
Table F.1: Selected Short Term Mechanical Properties of various GRP Materials correspond-
ding Values for Selected Steel, Aluminium and Copper/Nickel alloys are included
for Comparison. Steel NS 12603 is comparable to the St. 52 Type of Steels. Steel
UNS 21254.IS a h.IQ h MOIV
I bd enum a IIoved Starn
. Iess
Materials Mat. Woven Fabric Unidirec Aluminium Steel CU/NI
(P) roving (P) (P/E) (P/E) AIMg2.5 NS12603 90/10
Fibre orientation Random Bidirec. Bidirec. Unidirec. (AA5052) UNS 31254
Fibre content wt% 25-40 45-60 50-70 50-90
Specif. gravity: (g/cm~) 1.4-1.6 1.6-1.8 1.6-1.9 1.6-2.2 2.68 7.85-8.0 8.94
Tensile strength: (MPa) 60-170 260-500 260-500 430-1730 170-300 450-650 300-380
Tensile modulus: (GPa) 6-12 16-30 16-30 22-62 70 210 130
Flex. strength: (MPa) 100-250 210-600 210-600 700-1860 65-220 1> 200-355 1> 100-160 1>
Flex. modulus: (GPa) 5-11 18-14 18-24 28-48 70 2> 210 2> 130 2>
Compr. strength: (MPa) 130-170 200-400 200-400 210-550 85 450 -
Compr. modulus: (GPa) 6-9 10-24 10-24 - - - -
In plane shear strength: (MPa) 60-90 - - 50-80
lnterlaminar shear strength: (MPa) 22-30 25-30 25-30 0.28-0.3
Poisson's ratio 0.1-0.3 0.1-0.4 0.1-0.4 0.2-0.4 0.33 0.342
In plane shear modulus: (GPa) 3 3.3-4 3.5-4 -
lzod impact strength: (J/cm) 10-15 15-30 15-30 60 20-35
Elongation at break: (%) 1.5-2 1.5-2 1.5-2 1.5-2 3-18 35-40
1) Yield strength data P =polyester, E =epoxy

2) Tensile modulus data
Table F.2: Selected Physical Properties of various GRP Materials. Corresponding Values for
Elece
t d SteeI, Al um,nium and C oooer/N.IC ke I aIIovs are meIud ed f or C omIJanson
Materials Mat. Woven Fabric Unidirec. Aluminium SteelNS CU/NI
(P) roving(P) (P/E) (P/E) AIMg2.5 12603 90/10
Fibre orientation Random Bidirec. · Bidirec. Unidirec. (AA5052) UNS
Fibre content wt % 25-40 45-60 50-70 50-90 31254
Specif. gravity: g/cm~ 1.4-1.6 1.5-1.8 1.6-1.9 1.6-2.2 2.68 7.9-8.0 8.94
Coeff. of thermal expansion:
1!°C-10-6 18-40 11-16 9-11 5-14 22-24 12-16.5 17
Coeff. of thermal conductivity:
W!m°C 0.2-0.3 0.2-0.3 0.3-0.35 0.3-0.35 120-190 16.5-36 40
Eletrical resistivity: ohm-m 1012_1014 1012_1014 1012_1014 1012_1014 49-10"9 24-10-8 19-10-8
Barcol hardness, 934-1 40-60 40-60 40-60 40-60 HB 47-77 HB<210 HV70
Max. service temp °C 150 1> 155 1> 150/ 175 l ) 150/175 1) 400 750 400
= =
HB Brinell Hardness, P Polyester, E Epoxy =
1) Dependent on type of resin and environment
Scandpower
(r,J}SINTEF Risk Management AS
Page F-2 Appendix F: Characteristic Mechanical and Physical Properties of GRP Materials
T a bl e F ..
3 S eece
I t d Mec hamca
. I an d PhIvsIca 1 erent Mt.
. I Prooe rfIesof th ree d'ff a rJX Materia
. Is
Property Cured polyester Cured epoxide Cured phenolic
Specific gravity: g/m3 1.15-1.30 1.10-1.40 1.20-1.25
Hardness (Rockwell M) 70-115 80-110 80-85 (Shore D)
Tensile strength: MPa 40-90 30-90 40-60
Tensile modulus: GPa 2-4.5 2-4
Compressive strength: MPa 90-250 100-175
Specific heat: J/g°C=W/sg°C 1.25 1.05
(cal./g°C) (0.3) (0.25)
Thermal conductivity: W/m°C 0.15 0.15 0.15
Coeff. of Jin.exp: 11°C 60-90-10"6 40-70-10-6 60-90-10"6
Shrinkage 0.004-0.008 0.001-0.004 0.004-0.010
Water absorption: % 0.15-0.60 0.08-0.15 13-15
Flexural strength: Pa 100-140 100-140 100
Flexural modulus: Pa 2-4.5 2-4 2-4
Heat Distortion: °C 50-150 75.175 115
Scandpower
Risk Management AS ~SINTEF/
-- - - -- - ..... .....
--
........
.,,. .,,. '
'·
..... ....
....
....... ...
,,
I
I
I
I
,'
I
I
1
270°+-----'f~~~
'II
I
\
\
\
\
''
.... ...
.,,.
.... .,,.
/
.,,. .,,.
/
- .... .,,. ,,
--- --- -
Figure F.1: Variation of Tensile Young's Modulus with the Angle of load, 0, relative to the
Pincipal Fibre Direction for Laminates with different Types of Reinforcement.
Unidirectional (UD) with 75 wt% Glass, Balanced Woven Roving (WR) with 50 wt%
Glass, and Chopped Strand Mat (CSM) with 25 wt% Glass
Scandpower
oo
700 [MN/m 2 J
,.,.,., ., .,- -- -----........ ....

' , 45°
I
/
/
', \
I \
I
I
I ' \
\
I I
I I
270°'-+---l---6~ 1--,!----4-90°
I
I I I
I I I
\ I I
\ I I
\ I I
\ / I
\ I
I
'
225° ', ....
✓
........ ...
700 [MN/m2]
180°
Figure F.2: Variation of Tensile Srength with the Agle of load, 0, relative to the Pincipal Fibre
Direction for Laminates with different Types of Reinforcement. Unidirectional (UD)
with 75 wt% Glass, Balanced Woven Roving (WR) with 50 wt% Glass and
Chopped Strane Mat (CSM) with 25 wt% Glass
Scandpower
Risk Management AS {G»SINTEFj
Norwegian Fire Research Laborato,y
Illustration of the Possibilities by using a Computer-Model for Fire Simulation
Example: GRP pipe insulated with 12 mm Kaowool (ceramic fibre)
This example shows a typical use of the model. A GRP pipe with inner diameter 50 mm (2") and
wall thickness 4.5 mm is exposed to a hydrocarbon fire. The pipe size and the simulated fire
situation are relevant for small diameter pipes in the deluge systems. During the first five minu-
tes of the fire the pipe is empty and then water starts to flow. The calculated time temperature
curve for the outside of GRP pipe during the first 10 minutes is shown in Figure F.3.
The temperature-profile through the pipe wall is of great interest because the structural be-
haviour of the pipe is temperature dependent. In Figure F.4 and Figure F.5 the temperature pro-
files after 5 and 10 minutes, respectively, are shown.
250
Outside of GRP
200
6
~
e 150
.a
I!
CD
a.
E 100
~
50
0 2 4 6 8 10
Time (min)
Figure F.3: The Graph shows the Temperature Development on the Outside of GRP Pipe
during the first 1O Minutes of the Simulated Hydrocarbon Fire. The Pipe is empty
for the first 5 Minutes and then the Temperature Drops when the Water is turned on
Scandpower
1000 - . - - - - - - - - - - - - - - - - - - .
Pipe wall Insulation Steel
cladding
800
~
~ 600
:I
~
~400
E
i!
200
25 30 35 40
Figure F.4: The Graph shows the Calculated Temperature Distribution through the 12 mm
Insulation and the 4.5 mm thick Pipe Wall after 5 Minutes of Testing, i.e., just
before the Water is turned on
1200 . . . - - - - - - - - - - - - - - - - - , - ,
Pipe wall Insulation Steel
1000 cladding
I
I
~ 800
I
I
I
~ I
:, 600 I
~
I
I
a, I
~ 400 I
I
i! 200
I
I
I
I
o .l=;:=:;:::;::::J.-r--,,.....,..-.-~..-.~-.-.. . . . .~L.1!
25 30 35 40
Figure F.5: The Graph shows the Calculated Temperature Distribution through the 12 mm
Insulation and the 4.5 mm thick GRP Pipe Wall after 10 Minutes Testing, i.e.
5 Minutes after the Water started to Stream through the Pipe
Another application of the model is to calculate the temperature development of the liquid inside
the pipe, tank or vessel. The mechanical strength as a function of temperature of a structure
can also be calculated and the optimal thickness of an insulating coating determined. The
model is being developed further to also include simulation of intumescent coatings.
The simulation program can be used to reduce the need for time consuming and costly fire tests.
Scandpower
Risk Management AS ~SINTEFI
Appendix G: Temperatures and Velocities of Buoyancy controlled Diffusion Flames Page G-1
APPENDIX G
Temperatures and Velocities of Buoyancy
controlled Diffusion Flames
The rising fire plume is as shown in Figure G.1 usually divided into the following three distinct
regions:
1. A lower region which is termed the continuous or persistent flame plume where there is a
persistent flame and an accelerating flow of burning gases
2. A middle region which is termed the intermittent flame plume. In this region there is inter-
mittent flaming and near constant flow velocity.
An upper region which is termed the buoyant gas plume. This region is characterized by de-
creasing temperature and velocity with height.
Buyant Plume
Intermittent Flame
Persistent Flame
Figure G.1: A schematic Diagram of a Fire Plume showing the three Regimes of the Plume
The temperature of the plume will vary considerably between each region and within each region
both along the centerline and in the lateral direction. However, it is the variation of the centerline
temperature with height above the fuel source that will definitely vary the most. Generally, the
temperature decreases with increasing heights above the pool surface. Empirical correlations
for the centerline temperature rise above the ambient temperature will be presented here only.
Scandpower
Page G-2 Appendix G: Temperatures and Velocities of Buoyancy controlled Diffusion Flames
1. Persistent Flame Plume Temperature Rise and Velocity:
Cox et al., Ref. /G-2/, present the following values for the persistent flame plume temperature
rise:
Q2/5 z
- - > 10.2 and 0.03 < 215 < 0.08: LlT = 980°C
o. Q
Eq. (G.1)
Q2/5 z
- - > 10.2 and 0.02 < 215 < 0.06 : ,1.T = 960°C
o. Q
For 0.02 < z/Q215 < 0.08:
u = 6.83. z 112 Eq. (G.2)
where
tiT = centerline temperature rise (K)

Q = rate of heat release due to the fire (kW)
z = distance from the virtual source (m)
De = equivalent diameter of the fuel source (m).
The rate of heat release is given by the following expression:
Eq. (G.3)
where
ri = combustion efficiency (-)

m1 = discharge rate of gas (kg/s)
tiH 0 = heat of combustion of the fuel (kJ/kg).
The location of the virtual source, i.e. the distance from the pool surface, is given by Ref. /G-1/:
Z0 =-1.02D 0 + 0.083Q 215

Eq. (G.4)
where
Zo = is the distance from the fuel source'> (m).
Since z in Eq. (G.1) is the vertical distance from the virtual source, then z is equal to z' - Zo where z
is the actual vertical distance above the fuel source at which the centerline temperature is to be
calculated.
'! Positive values of z0 correspond to locations above the fuel source and vice versa
Scandpower
Appendix G: Temperatures and Velocities of Buoyancy controlled Diffusion Flames Page G-3
2. Intermittent Flame Plume Centerline Temperature Rise and Velocity:
The following equations apply in this region, Ref. /G-1/:
For: Q215/De > 10.2 and 0.08< z/Q215 < 0.20:
Eq. (G.5)
u = 1.85 . Q 115 Eq. (G.6)
For: Q215/De < 10.2 and 0.06< z/Q215 < 0.12:
LiT = 20.5Q 213 z - 5 13

Eq. (G.7)
u = 1.65. Q 115 Eq. (G.8)
3. Buoyant Gas Plume Centerline Temperature Rise and Velocity:
The following expression applies in the buoyant gas plume rising above flames, Ref. /G-2/:
LiT = 22.3Q 21 3 z - 5 ' 3 Eq. (G.9)
U=1.04 · ( zQ ) 1/3
Eq. (G .10)
References Appendix G
/G-1/ Cox, G. and Chitty, R.: "Combustion and Flame 60", 1985.
/G-2/ Heskestad, G.: "Engineering Relations for Fire Plumes, Fire Safety" Journal 7, pp. 25-32,
1984.
Scandpower
{GJ) SINTEF Risk Management AS
Appendix H: Heat transfer from Buyancy controlled Diffusion Flames Page H-1
APPENDIX H
Heat Transfer from Buoyancy controlled
Diffusion Flames
The treatment of the heat transfer from fires in general differs principally in two ways depending
on the location of the exposed target in relation to the flame. Firstly, the target may be located in
the flame plume in such a way that it is fully engulfed by the flames or swept by the flame plume.
Secondly, the target may be located some distance from the flame. In the former case, the
target, e.g. an object, is exposed to both convective and radiative heat transfer. In the latter
case, the target is exposed to radiative heat transfer only, but can be cooled by convection).
1. Heat Transfer on Objects Engulfed or Swept by the Flame.
As already mentioned, the gross heat transfer to the object in this case, q 1 , comprises a
radiative component q; and a convective component q~:
q," = qc" + q," Eq. (H.1)
where q; and q~ are given by the following expressions:
q," = EO" (LI.T + T0 )4 Eq. (H.2)
Eq. (H.3)
where
LI.T = centerline temperature rise at the height z' above the fuel source at which the fire
exposed target is located (see Appendix G) (K)
To = initial steel temperature (normally equal to the ambient temperature) (K)
Ts = steel temperature (at time t = O: Ts= T0 ) (K)
E = emissivity of the flame at the actual location in the flame plume (-)
h = convective heat transfer coefficient (kW/m2 K)
a = Stefan-Bolzmann's constant = 56. 7-10· 12 kW/m 2 K4
Scandpower
Page H-2 Appendix H: Heat transfer from Buyancy controlled Diffusion Flames
Since the Eqs. (H .2) and (H.3) include the centerline temperature rise (which may be predicted
from the correlations presented in Appendix G), these equations apply strictly speaking to the
heat transfer in the middle of the flame plume at a specified height z' above the pool surface. If
the target is located a radial distance from the flame axis, this will lead to an overestimation of
the heat transfer. Further, in the case when the target does not engulf the target completely, the
predicted heat transfer by the Eqs. (H.1), (H .2 ) (H.3) will be the maximum heat transfer to the
object of which only a small area will receive heat energy. The location of this area of maximum
heat flux will differ along with the flame fluctuations.
The emissivity of the fire plume is given by:
E = 1- exp(-kLm) Eq. (H.4)
where
k = absorption-emission coefficient (also called the extinction coefficient) (m-1);::: 0.3-0.4

for hydrocarbon pool fires
Lm = is the mean beam length of the fire plume (m).
The mean beam length of the flame plume is given by the following expression:
Eq. (H.5)
where
= flame volume (m3 )

= area of flame envelope (m2) .
By assuming that the convective heat transfer occurs by turbulent forced convection, the
following expression may be used for predicting the average heat transfer coefficient:
Nu= hL = 0.037Pr 113 Re 0 ·8 Eq. (H.6)

k
where
Nu = Nusselt number based on the length L of the fire exposed surface(-)

h = convective heat transfer coefficient (kW/m2·K)
k = thermal conductivity of air (kW/mK)
L = characteristic dimension > of the fire exposed object (m)
1
Pr = cpµ/k, Prandtl number of air(-)

ReL = uUv, Reynolds' number based on L (-)
Cp = specific heat of air (kJ/kgK)
µ = dynamic viscosity of air (kg/ms)
V = kinematic viscosity of air (m2/s)
1
J The characteristic dimension of a target is its length for vertical targets (e.g. columns, beams or steel plates) or its
width for horizontal targets with the length direction perpendicular to flow velocity of the fire gases
Scandpower
The following simplified expression for the convective heat transfer coefficient h at the actual
height z' above the fuel source may be deduced from Eq. (H.6) 2> for Re> 3-10 5:
Eq. (H.7)
where
T9 = centerline flame temperature at height z' above fuel source (K) (from Eqs. (G.1),
(G.5), (G. 7) and (G.9)
u = centerline flame velocity at height z' above fuel source (m/s)
Eq. (H. 7) requires that the temperature and the velocity of the fire gases at a height z' above the
fuel source are known. Correlations for prediction of these two parameters as a function of the
height, z', above the fuel source are given in Appendix G.
For flow around a sphere with diameter d, the following expression may be used:
h = 1.5-10-4 d-1 +32.1-10-4 u 0·5 d-5 Eq. (H.8)
2. Heat Transfer onto Objects Outside the Fire Plume
In this case the object receives radiative heat transfer from the flame plume only. The radiative
heat transfer from the buoyant gas plume is usually far less than 20 kW/m2 which is the maxi-
mum emissive power of optical thick fire gases at a maximum temperature of 500 °c3>_ This
radiative contribution from the buoyant gas plume can be neglected when considering the heat
transfer onto a target some distance away from the fire.
Two different methods can be used when predicting the incident heat flux onto a target from a
flame plume, namely the 'point source model' and the 'solid flame model' where the former is far
more simple and less accurate than the other especially at locations close to the fire.
The Point Source Model
The 'point source model' is a method which simplifies the problem significantly by assuming that
the flame plume is represented by a point source of thermal energy. Further, it assumes that a
certain specified fraction of the released energy is released by radiation.
21 The conductivity k, kinematic viscosity v, and the Prandtl number of the fire gases can be assumed to be equal to the
values of dry air because fire gases comprise mostly of dry air. A regression analysis of tabulated values of k and v
according to the mathematical expression y = ax', yields the following expressions (with coefficients of regression
r2 =0.99):
k = 3.248 . 10-7 . rg- 77
6668
V = 1. 262 · 10-9 . Tg1·
The Prandtl number, Pr, varies insignificantly with actual temperatures of fire gases. A constant value of 0. 7 is pre-
supposed irrespective of the temperature. This value and the correlations above are inserted into E.q. (H. 6)
3
J Fire gases and products of temperature below 500 'C will not be luminous, and can therefore be considered as
smoke gases
Scandpower
Page H-4 Appendix H: Heat transfer from Buyancy controlled Diffusion Flames
However, the accuracy of the results may be insufficient, especially in the near field of a large
pool fire. The radiative heat flux to a target, q~ , may be expressed in the following way:
" QR
qr = -
2 Eq. (H.9)
4 7tX
Eq. (H.10)
where
QR = energy released by radiation (kW)

X = distance from the flame centre (m)
fR = fraction of the heat released as radiation 4> (-)
m1 = burning rate (kg/s)
LlHc = heat of combustion of the fuel (kJ/kg)
The 'Point source model' has been used successfully for flames which have a large flame height
to diameter ratio (i.e. jet fires and diffusive flare fires from comparatively small openings com-
pared to the flame height) except very close to the fire. In the near field the point source model
will overestimate the incident heat flux, which is a great disadvantage when predicting safe dis-
tances for process equipment and human beings. Further, this method is unable to include the
effects of wind (e.g. flame tilt and flame drag) satisfactorily.
The Solid Flame Model
The 'solid flame model' is the most usual method used and which yields the most accurate
results (irrespective of fire scenario) both in the near and far field of any fire. This model con-
siders the flame as a body which emits thermal radiation. The shape or geometry of this body
may be idealized as a cylinder or a cone for all fires except the fireball scenario which may be
idealized as a sphere.
The incident radiative heat flux onto a target, q, is given by:
Eq. (H.11)
where
Ep = average surface emissive power of the flame (kW/m 2)

F = view factor between the flame and the target(-) (see Appendix I)
"C = atmospheric transmissivity (-).
Average Surface Emissive Power
The average surface emissive power is given by the following equation if the effective radiative
temperature, Tt, is known:
4
> fR varies between different types of fires (e.g. jet fires vs. pool fires), size of the fire, fuel type and exit velocity of the
ignited high pressure gas release. Typically fR varies between 0.05 - 0.35 for pool fires. The lower value is typical for
very large scale pool fires (D > 30 m) where the burning fuel consists of heavy hydrocarbon liquid fuel (e.g. stabilized
crude oil). The upper value apply to medium scale pool fires (i.e. D = 1 - 2 m). For jet fires the radiative fraction fR
varies between 0. 15 - 0. 25 mostly dependent of fuel type
Scandpower
Eq. (H.12)
where:
E = emissivity of flame (given by Eq. (H.4)) (-)

cr = Stefan Bolzmann's constant= 56.7-10·12 kW/m 2 K4
However, it is equally difficult to find a representative value for T, as for EP directly. Thus, it is
advisable to find a representative value for EP directly instead of using the equation above
because an error in the estimate of T, will result in a significantly larger error in calculated EP
according to Eq. (H.12) because T, is raised to the fourth power.
Atmospheric Transmissivity
Brzustowski et al. Ref. /H-1/ give the following expression for the atmospheric transmissivity ~:
16 16
~=0.79• [ -100]" · [30.5]"
--
r X
Eq. (H.13)
where
r = relative humidity of the atmosphere(%)

x = distance from flame surface to exposed target (m)
References Appendix H
/H-1/ Brzustowski, T.A. and Sommer, E.G.: "Predicting Radiant Heating from Flares, API
Proceedings of Refining", 1973.
Scandpower
Appendix I - Geometric View Factor Page 1-1
APPENDIX I
Geometric View Factor
The geometric view factor, F, in Eq. (H.11) is the fraction of the total radiant intensity that leaves
the flame surface (i.e. fraction of average surface emissive power Ep) which arrives directly on
an infinitesimal surface element of the fire exposed surface when assuming a transmissivity of
unity. The geometric view factor between the visible flame surface (termed A 1) and a receiving
element on the fire exposed surface (termed dA2) is given by the following general equation:
Eq. (1.1)
where r is the distance between the flame and the receiving element, 01 and 02 are, respect-
tively, the angles formed by the perpendiculars to the flame surface and the receiving element
with the line connecting the elements. The integration is carried out over the entire surface of
the flame.
From the definition of the view factor above, it is obvious that for locations sufficiently close to
the flame surface, the view factor will attain a value approximately equal to unity because the
target "sees" only the flame. That is, the target receives a radiant intensity equal to the surface
emissive power Ep. On the other hand, for targets sufficiently far from the flame plume, the view
factor may attain a value practically equal to zero. That is, the incident heat flux received by the
target will be, according to Eq. (H.11), also zero.
1. Ground level view factors
In the"solid flame model" the geometry of the flame is usually approximated by a cylinder1>. In
the calm wind situation there will be a vertical cylinder as shown in Figure 1.1. The integral
above can then be reduced to the following expressions for the view factor to a vertical receiving
element, Fv, and horizontal receiving element, Fh, on the ground in the downwind direction,
respectively:
11
According to Moorhouse, Ref. 11.21, a cylindrical flame representation will overestimate the heat flux by 10 %
because the flame will have a shape of something in between that of a cylinder and a cone. However, accor-
ding to Mudan, Ref. II. 11, there is some evidence indicating that a cylindrical flame representation seems to
predict the radiative heat flux to an external target most accurately
Scandpower
Page 1-2 Appendix I • Geometric View Factor
I
I
I
H
!I
I
I
RADIATION
/ RECEIV ING
I
I
ELEME NT
I
-x~
(al RIGHT CIRCULAR SOURCE
RADIATION RECEIVING
ELEMENT
(3
f+------x----....,o-11
Figure 1.1: Coordinate System for a right Circular (a) and an inclined Cylinder (b); A = H/R,
b= X/R (from Mudan et al. Ref. /1-1/)
For vertical cylinder and vertical receiving element:
a a 2 +b 2 + 1
1tFv () I 2 +(b+1)2a
=-b ba r/2 I 2 +(b-1)2j'2
2 112 12
1
x tan-1[{a + (b + )22 } (~) ' ] Eq. (1.2)
a 2 + (b -1) b +1
1/2
1t -1 a atan-' b-1
+- an r, )112 b ( b+1 )
B \b2 -1
Scandpower
NoJWegian Fire Research Laboratory
For vertical cylinder and horizontal receiving element:
F: t -1 (b + 112 a2 + b2 -1
1)
7t h = an b -1 - [a2 + (b + 1)2 r2 [a2 + (b -1)2 r2
Eq. (1.3)
2 112
2
1
x tan-1{[a + (b + )2 ] (~)}
a 2 + (b -1) b+ 1
In case of wind, the cylinder may be tilted and the angle of tilt is termed 9. The vertical view
factor, Fv, and the horizontal view factor, Fh, for a target in the downwind direction are given by
the following expressions:
For a tilted cylinder and vertical receiving element:
1I2
acose t -1 (b-1)
1tF = ~---~ an -- acose
+ ~---~
v (b - a sin 0) b+1 (b - a sin 0)
2
a + (b + 1)2 - 2b(1 + a sine)
2 2 2 2 2 2
x [a + (b + 1) sin 0 f1 [a + (b -1) - 2a(b -1)sin 0r
a + ~\b + 1)2 2 - 2a\b

~ + 1) sine )1/2 b -1 1/2]
2
t -1 cos e
[( 2 2 2
x an a +(b-1) -2a(b-1)sin0 (b+1) + [1+{b -1)cos 2 ef1
2
x { an
t _1[
2
ab -(b -1)sin e
2
{b 2 -1)11 {1 + {b 2 -1)cos 2 e )11
2
l
2
+ tan_1[ (b -1)sin 0 ]
2 2 2 2 2
Eq. (1.4)
{b -1)11 {1 + (b -1)cos e)11
Scandpower
Page 1-4 Appendix I - Geometric View Factor
For a tilted cylinder and horizontal receiving element:
e
1trh
., t _,
= an (b + 1)
--
b -1
112
2
1+\b -1)cos e
•
---,~,-----,--~r=-
+ -,,-[
sin
2 12
t -1[ 2
ab-(b -1)sin e ]
12
x { an (b 2 -1)1 (1 + (b 2 -1)cos 2 e)
2
+ tan-'[ (b -1)sin e ]
2 12 2 2 2
(b -1)' (1+(b -1)cos 0)1'
2
a 2 +(b +1) -2(b +1+ ab sine)
2 12 12
a 2 +(b+1) -2a(b+1)sine a2 +(b-1}2-2a(b-1)sin0
2 2 2
2
x tan-'[-'(b + 1)" {a + (b + 1) -2a(b + 1)sin e}" ]
2
b-1 a2 +(b-1) -2a(b-1)sin0 Eq. (1.5)
In Eq. (1.4) and (1.5) the angle of tilt is measured from the vertical. It is positive for downwind
locations and it is negative for upwind locations.
The vectorial sum of the vertical and the horizontal view factor is termed as the maximum view
factor and is given by the following equation:
Eq. (1.6)
From the Eq. (1.4) and (1.5) it appears that the view factor between a fire and an element outside
the fire depends on the shape of the flame (i.e. its diameter D = 2R and the height H), the dis-
tance between the flame and the receiving element, X, and the flame tilt, e. The Figure 1.2 and
Figure 1.3 show the maximum view factor for an object located at the ground from a right circular
cylinder and tilted cylinder, respectively.
Scandpower
ll\
I~
~~
\ ~, \ \
\\
I\
HEIGHT-TO-RADIUS RA110
0.5
1.0
I/ V 3.0
\ ll / V 6.0
/y )('.V
l/
y / '
\ ''
\ "
l\ \
\\
1
\ ~ \\' \
\ ' \
' \ \ \
0.002
1 2 4 6
\
8 10
\ \\ 20 40 60 so
NONDIMENSIONAL DISTANCE FROM FLAME AXIS (-)
Figure 1.2: Maximum View Factors for an Object located on the Ground from a right Circular
Cylinder (from Mudan et al. Ref. /1-1/)
~ 0.0s1----1----........-t"r."-~rr----,i----1
:i;
~
0 ·0 1,'--------.___ _ _s_ _......,.,o,,..............~20.,,......_
2
NONDIMENSIONAL DISTANCE FROM FLAME AXIS (-)
Figure 1.3: Maximum View Factors for an Object located on the Ground from a tilted Circular
Cylinder (Tilt Angle 8, from Mudan et al. , Ref. /1-1/)
Scandpower
{GJ) SINTEF Risk Management AS
Page 1-6 Appendix I - Geometric View Factor
2. View factors for target not located at the same level as the fuel source
In Figure 1.4 the target locations T1 and T2 are not located at a different level than the fuel
source. In this case the fuel source is a pool fire represented by a cylinder. The fuel source is
the liquid pool represented by the horizontal line a-b in this one-dimensional sketch. The
cylinder is represented by the tilted rectangle a-b-c-d. The target T1 is located in the downwind
direction at a higher level than the pool, whilst target T2 is located at a level below the pool.
A. Procedure for calculating view factors for targets of type T1:
In this case the target T1 is located at a height z1 above the fuel source. The horizontal line
through the T1 is dividing the tilted cylinder into the two cylinders a-b-e-f and f-e-c-d. The view
factor between the target and the tilted cylinder is equal to the sum of the view factors of cylinder
a-b-e-f and f-e-c-d. If the cylinder a-b-e-f is mirrored around the horizontal line, target T1 will be
an upwind target with respect to this cylinder f-e-b'-a'. The target T1 will have the same view
factor with respect to cylinders a-b-e-f as cylinder f-e-b'-a'. It has to be pointed out that the view
factor for cylinder a-b-e'-f is found by using a negative tilt angle in the Eq. (1.4) and (1.5). T1 is a
ground level target for both the cylinder f-e-c-d and f-e-b'-a' and can, thus, be found by the ex-
pressions given in this appendix.
T1
z,
a
,,,
,, z,
,/ ,
,' ,' TI
'------- --------~------------------,~-
g h
Figure 1.4: A Sketch of the two principal different Locations T1 and T2 of a Target not located
at the same Level as the Pool
Scandpower
The total view factor, F,, of an elevated target is given by the following expression:
F1 = Fa'-b'-e-f (a1, b1, · 0) + Ft-e-c-d (a2, b2, 0) Eq. (1.7)
where
a1 = 2z11/(Dcose)
b1 = =
b2 2(x - z 1tge)/D
a2 = 2(Lcose - z1)/(D cose)
B. Procedure for calculating view factors for targets of type T2:
In this case the target T2 is located at a height z2 below the fuel source. If the tilted cylinder a-b-
c-d is extended down to the horizontal line through T2, the cylinder g-h-c-d is achieved. The
view factor between the target T2 and the fire represented by the cylinder a-b-c-d, is equal to the
view factor between T2 and the cylinder g-h-c-d minus the view factor between the T2 and the
cylinder g-h-b-a. The target T2 is a ground level target for both these cylinders and can, thus, be
predicted from the formulas presented in this appendix.
Hence, the view factor F2 between cylinder a-b-c-d and the target T2 is given by the following
expression:
F2 = Fg-h-c-d (a3, b3, 0)-Fg-h-b-a (a4, b4, 0)

Eq. (1.8)
where
a3 = 2(Lcose + z2)/(D case)

b3 = =
b4 2(x + Z2tge)/D
a4 = 2z2'(Dcose)
References Appendix I
/1-1/ Mudan, K.S. and Croce, P.A.: "Fire Hazard Calculations for Large Open Hydrocarbon
Fires", The SFPE Handbook of Fire Protection Engineering, ISBN: 0-87765-353-4, Sept.
1988.
/1-2/ Moorhouse, J.: "Scaling Criteria for Pool Fires Derived from Large Scale Experiments",
Institution of Chemical Engineers North West Branch Symposium on the Assessment of
Major Hazards, Manchester, April 1982.
Scandpower
Appendix J: Smoke Production from Buyancy Controlled Diffusion Flames Page J-1
APPENDIXJ
Smoke Production from Buoyancy
Controlled Diffusion Flames
The rate of smoke production from buoyancy controlled diffusion flames in the open may be
predicted by the fire plume correlations of Heskestad, Ref. /J-1/. They predict the mass flow rate
of both the flame plume and in the buoyant gas plume as a function of height above the fuel
source. A basic assumption is that the mass flow at a particular elevation in the fire plume is
nearly completely attributable to air entrained at lower elevations. Furthermore the mass flow
contributed by the fuel is insignificant in comparison.
The limiting elevation, z 1, is the elevation in the plume which corresponds closely to the mean
flame length (specifically it is defined as the elevation in the plume where the temperature is
500 °C above the ambient temperature). For normal atmospheric conditions, the limiting eleva-
tion, z 1, can be expressed by:
215
Z1 =Z 0 + 0.166Q 0
Eq. (J.1)
where
z0 = the location of the virtual origin which is given by Eq. (G.4) (m)
Q0 = is the convective rate of heat release by the fire (kW)
For z ~ z 1, respectively z ,,; z 1 and normal atmospheric conditions, the mass flow rate in the
plume mg (kg/s), is given by the following equations:
For z ~ z 1:
ITlg =0.071Q 0 113

(z - z 0 )
513 3
[1 + 0.026Q/' (z - z 0 t 513
] Eq. (J.2)
For z,,; z1:
•
mg = 0.o054 Qcz Eq. (J.3)
215
0.166Q 0 +Zo
Heskestad, Ref. /J-1/, points out that Eq. (J.3) is strictly speaking limited to pool fires or
horizontal surface fires. Furthermore, that the flames must be predominantly turbulent which is
achieved when the base diameter of the flame is greater than 0.3 to 0.5 m. Eq. (J.2) is,
however, not limited with respect to the fire source.
The convective rate of heat release, Q 0 by the fire, which is included in the equation for the
smoke production, mg, (i.e. in Eq. (J.1), (J.2), (J.3)), may be predicted by the following ex-
pression:
Eq. J.4)
Scandpower
PageJ-2 Appendix J: Smoke Production from Buyancy Controlled Diffusion Flames
where Q is the total rate of heat release (kW) and fR is the radiative heat output fraction which is
given by the following expression by Moorhouse et al. Ref. /J-2/:
Eq. (J. 5)
where
EP,, = the average flame surface emissive power (kW/m 2)

m1 = the fuel evaporation or burning rate per unit pool surface area (kg/sm 2)
D.Hc = heat of combustion of the burning fuel (kJ/kg)
UD = flame length to diameter ratio(-)
The average surface emissive power, Ep, may be given by the following equation of Mudan, Ref.
/J-3 I primarily applying to pool fires:
EP = e-0.120Epl + (1- e-0120 F\- pu

Eq. (J.6)
where
D = equivalent diameter of the fire (m)

Ep1 = emissive power of the lower flame (kW/m 2) -= 150 (kW/m 2)
Epu = emissive power of the upper flame (kW/m2 ) -= 20 (kW/m2 )
(For diffusive gas fires and pool fires that do not generate much soot (e.g. LNG-
fires): Ep1= Epu)
Visibility in the Rising Gas Plume
The obscuration ability of a gas flow may be measured by the attenuation of a light beam. The
measure of this attenuation is the optical density per unit beam length, OD (m-1), which is de-
fined in the following way:
Eq. (J.7)
where
10 = the intensity of the light source (W/m 2)

I = the light intensity of the light beam after passing through the smoke over a
distance L (m) (W/m2)
Evans et al. Ref. /J-4/ give the following relation between the optical density, OD, and the tem-
perature rise above the ambient, D.T, in the rising gas plume at an elevation z (m):
Eq. (J.8)
Scandpower
Risk M an agement AS l'GJ) SINTEF
Appendix J: Smoke Production from Buyancy Controlled Diffusion Flames PageJ-3
where
Oc = convective fraction of the total rate of heat release (= (1 • fR) · Q where fR is given
by Eq. (J. 5)) (kW)
Cp = Specific heat of air at the temperature of the gas T9 (kJ/kgK). It is given by:
125
Cp"' 0.48T/·
Po = ambient density (kg/m 3 )
X = mass of soot produced per kg fuel burnt(-)
Figure 7.8 in Chapter 7 shows a plot of visibility as a function of OD. As shown a visibility of 2 m
corresponds to an OD of 0.5 m·1 and a visibility of 1 m corresponds to an OD of 1 m·1• It seems
as if the visibility is inversely proportional to OD. Consequently, the following expression may be
established for the visibility, V (m), as a function of the optical density perm:
V=0□- 1 Eq. (J.9)
Concentrations of Combustion Products
According to Heskestad, Ref. /J-5/, the ratio of the volumetric concentration of a gas species at a
point in the buoyant gas plume to the temperature rise at the same point, i.e. C;//1T, is expected
to be constant everywhere in the nonreacting plume. If the mass generation rate of species, m;,
is given, the concentration/temperature rise ratio can be calculated from:
C; = Mcpm;
LiT M; QC Eq. (J.10)
where
Mo= the molecular weight of air (kg/kmole)

M; = the molecular weight of species i (kg/kmole)
Hence, by predicting the right side of Eq. (J.10), the volumetric concentrations in the plume at a
point in the buoyant gas plume can be obtained by predicting the temperature rise at the same
point by the equations shown in Appendix G. However, a problem might be the estimation of the
generation rate of a species. Tewarson, Ref. /J-6/ lists the yields of numerous common liquid
and gaseous fuels with respect to CO2, CO and unburned fuel. When knowing the burning rate,
mr, the mass generation rate of these species can easily be predicted.
Scandpower
PageJ-4 Appendix J: Smoke Production from Buyancy Controlled Diffusion Flames
References Appendix J
/J-1/ Heskestad, G.: "Engineering Relations for Fire Plumes", Fire Safety Journal 7, pp. 25 -
32, 1984.
/J-2/ Moorhouse, J. and Pritchard, M.J.: "Thermal Radiation Hazards from Large Pool Fires
and Fireballs - A Literature Review", Institution of Chemical Engineers North West
Branch Symposium on the Assessment of Major Hazards, Manchester, April 1982.
/J-3/ Mudan, K.S. and Croce, P.A.: "Fire Hazard Calculations for Large Open Hydrocarbon
Fires, The SFPE Handbook of Fire Protection Engineering", ISBN: 0-87765-353-4, Sept.
1988.
/J-4/ Evans, D.D. and Stroup, D.W.: "Methods for Calculate the Response Time of Heat and
Smoke Detectors Installed below Unobstructed Ceilings", National Bureau of Standards,
NSBIR 86-3167, 1985.
/J-5/ Heskestad, G.: "Pressure Profiles Generated by Fire Plumes Impacting on Horizontal
Ceilings", Report OF0E1.RU, Factory Mutual Research Corp., Norwood, MA, 1980.
/J-6/ Tewarson, A: "Heat and Chemical Compaounds in Fires", The SFPE Handbook of Fire
Protection Engineering, ISBN 0-87765-353-4, September 1988.
Scandpower

Handbook For Fire Calculations and Fire Risk Assessment in The Process Industry. 2003.

Uploaded by

Document Information

Original Title

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

Handbook For Fire Calculations and Fire Risk Assessment in The Process Industry. 2003.

Uploaded by

Copyright:

Available Formats

Scandpower

Handbook for Fire Calculations and Fire Risk

Date 17 September 2003

2. PURPOSE OF THE HANDBOOK 2-1

3. HOW TO USE THE HANDBOOK 3-1

4. RELEASE OF COMBUSTIBLE FLUID 4-1

5. BEHAVIOUR OF RELEASED FLUID 5-1

5.4 Liquid Release 5-13

6. GAS FIRE 6-1

7. LIQUID FIRES 7-1

7.2.7 Duration of Fire 7-11

8. RESPONSE TO FIRE 8-1

9. FIRE PROTECTION 9-1

10. SMOKE 10-1

11. FIRE RISK ASSESSMENT 11-1

The project has been sponsored by:

Kjeller, 16 September 2003

2. Purpose of the Handbook

reduction of repair costs

3. How to use the Handbook

3.2 Flow Charts

Chart 1: Starting Chart and Underwater Release

The charts are self-explanatory, but some comments should be made:

Chart 1: Starting Chart and Underwater Release

All assessments start at the top of this chart.

Question 1. 1: Flammable Material?

Question 1.2: Release?

The phenomenon of a fireball or BLEVE (boiling liquid expanding vapour explosion) is

Question 1.3: Under Water?

Question 1.4 and 1.5: Fluid?

Question 1.6: Vapour Pressure?

Question 1.7: Vapour Pressure?

Chart 2: Liquid Releases

Question 2. 1: Vapour Pressure?

This question decides if 2-phase release calculations have to be performed in order to

Question 2.2: Onto?

Question 2.3: Shape?

Question 2.4 to 2.6: Release Mode?

Instantaneous release of a given amount (kg) of liquid

Most situations can be approximated by one of these two simplifying assumptions.

Chart 3: Gas Releases

Question 3. 1: Release Mode?

The criterion for selecting an instantaneous or a continuous release model depends

Question 3.2 and 3.3: Condensation ?

Form a spray (together with the gas)

If the formation of a liquid is believed to be substantial a condensation calculation has

Question 3.4: P2 > Pent?

p1 = outside (ambient) pressure

lj/ = -2-)'~' "'0.55 for natural gas

Question 3.5: As a Jet?

Question 3.6: Steady State?

Chart 4: Ignition and Fire

Question 4.1: Fluid?

See Question 1.4 and 1.5, Chart 1.

Question 4.2: Onto?

See Question 2.2, Chart 2.

Question 4.3: Open or enclosed?

Question 4.4: Shape?

Question 4.5: Release Mode?

Question 4. 6: Release Speed?

Gas Liq1..~d Gas

Liquid spread Bubble plume

Liquid Pool Models