Dispersion, Fire, and Explosion Risks Quantification

in an Ammonia Plant: A Case Study

Ammonia is the second most highly produced chemical in the world with its primary use in the
fertilizer industry. It also shows as a promising energy vector for the renewable sector. Currently, the
global production of ammonia is done via synthesis of hydrogen and nitrogen based on the steam
reforming of natural gas. The safe manufacturing of ammonia is paramount given the hazardous
characteristics of the feed, intermediates, and final products of an ammonia plant should accidental
releases occur. This paper aims to quantify the risk of dispersion, fire, and explosion in an ammonia
plant. The quantification approach is applied to a case study demonstrating how the results of the
assessment can be used to plan risk-informed prevention and mitigation measures in the associated
facility.

Khairunnisa
Gexcon Indonesia

Fadhilatul Aulia
Gexcon Indonesia

Teguh Cahyono
Gexcon Indonesia

Introduction was later ignited by the flare stack resulting in the

flames being propagated throughout the facility

W idely known as a toxic material, am-

monia can cause faint irritation to hu-
mans at 30 ppmv (AEGL-1). It is con-
sidered as a disabling irritant if one is exposed at
[3,4]. There were seven fatalities and 32,000 peo-
ple were evacuated due to this accident. One ac-
cident highlighting ammonia’s toxicity was a liq-
uid ammonia leak at Weng’s cold storage
220 ppmv (AEGL-2) for 10 minutes and lethal Industrial Company which killed 15 people and
(AEGL-3) at 2,700 ppmv for the same duration injured 25. The accident could have been pre-
of exposure [1]. Ammonia also poses fire hazards vented if proper ventilation was provided and the
whose flammability limit at ambient temperature workers had known the evacuation routes. Initial
is 16.6 – 27.2 % volume [2]. investigation revealed that the accident might be
due to a valve failure [5].
In 1989, 7,000 tons of ammonia contained in a
cryogenic tank at a Lithuanian fertilizer plant was Currently, the global production of ammonia is
released due to over-pressurization of the tank. done via the Haber Bosch reaction using synthe-
Large quantities of ammonia evaporated which sis gas (hydrogen and nitrogen) that is made via
steam reforming of natural gas. Hydrogen is a map the failure scenario development and
highly reactive flammable gas with a wide flam- compute the final event likelihood.
mability range of 4 to 74 vol %. For natural gas, 3. Consequence modelling. The release and
the flammability falls between 5 – 15 vol%. A consequence were modelled using semi-in-
safe manufacturing of ammonia is thus para- tegral consequence modelling tool,
mount given the hazardous characteristics of the EFFECTS 12.1. The toxic dispersion extent,
feed, intermediates, and final products of ammo- heat radiation exposure, and blast overpres-
nia plants should accidental releases occur. sure radius were calculated.
4. Risk calculation and evaluation. Coupling
This paper aims to quantify the risk of dispersion, the frequency analysis and consequence
fire, and explosion in an ammonia plant. The re- modelling results, the risk of having critical
sults of the assessment are then used to plan risk- targets exposed to an intolerable level of
informed prevention and mitigation measures in toxic, heat, or blast overpressure was esti-
the associated facility. mated. A target is considered impaired if it
is within the coverage of an unallowable
Methodology level of a specific period in a defined fre-
quency. Where relevant, risk-reducing
The analysis was performed following the work- measures were proposed to bring the risk to
flow shown in Figure 1. a more acceptable level.

Study Area
The case study of quantifying the risk of disper-
sion, fire, and explosion was carried out on Am-
monia Plant Z in Indonesia. The processing units
consist of Feed Treatment system, Steam Re-
forming system, Shift Conversion system,
Methanation section, Ammonia Synthesis sec-
tion, Refrigeration section, Ammonia Recovery
Figure 1 Risk analysis workflow section, and Hydrogen Recovery system.

The steps are briefly described as follows: The Ammonia Plant Z is located in the vicinity
1. Hazards identification. The hazards were of Ammonia Plant X. A Formic Acid Plant is a
limited to flammable/toxic gas or liquid, i.e., neighboring plant on its east. A residential area is
jet fire, pool fire, explosion, and toxic gas present on the south side of the Ammonia Plant
dispersion. The hazard inventory sections Z. Flammable/toxic exposure in these areas was
were divided considering provided isolation evaluated in the study.
systems (e.g., phase interface, isolation
valves, normally closed valves, etc.). Later, Hazards Identification
credible failure scenarios for each isolated
section were developed. The potential hazards were identified taking into
2. Frequency analysis. The leak frequency for account the substance characteristics, operating
each isolated section was estimated based on condition, and process limitation. In Ammonia
the leak frequency per element in the section. Plant Z, there are three main hazardous materials,
The number of elements was calculated us- i.e., natural gas of 88% methane, hydrogen gas,
ing the parts count method. An event tree per and ammonia (liquid and gas).
isolatable section was then established to
There were ten isolated sections determined with Iso Iso Sec- Main P T
one modelled stream for each. For conservatism, # tion De- Compo- (kPa) (oC)
a stream with higher flammable/toxic compo- scription nent
nents and/or higher pressure was selected as the (vol. %)
representative. The summary of the isolatable Regenera- (74% hy-
sections is shown in Table 1. tion drogen)
Table 1 Definition of isolatable sections
Iso Iso Sec- Main P T
# tion De- Compo- (kPa) (oC) Frequency Analysis
scription nent
(vol. %) In the actual event, there is an infinite probability
1 Feed Prep- Natural 4,224 371 of a leak size. Selecting an appropriate repre-
aration gas (88% sentative hole size is thus one important step in
methane) carrying out a risk analysis. There are several ap-
2 Primary Process 3,205 23.7 proaches to determine the hole size, i.e., using the
Reformer gas (60% upper limit of a range, arithmetic mean hole di-
hydro- ameter, arithmetic mean hole area, or geometric
gen) mean diameter or area. The geometric mean ap-
3 Secondary Process 3,126 47.0 proach was used in the analysis since, in terms of
Reformer gas (73% fatalities, the historic probabilistic distribution
hydro- and modelled correlations of hole sizes tend to
gen) fall towards the lower end of the range except for
4 Methana- Synthe- 13,455 -17.8 the highest leak size range [6]. The hole sizes
tor sis gas used in the study are summarized in Table 2.
(65% hy-
drogen) Hole Size Representative Leak
5 Cold Prod- Ammo- 13,406 -17.8 Range Hole Size Category
uct nia liquid (mm)
(99%) 1–3 2 Very small
6 Warm Ammo- 1,862 30.0 3 – 10 5.5 Small
Product nia liquid 10 – 50 22.4 Medium
(99%) 50 – 150 86.6 Large
7 Ammonia Ammo- 1,820 65.0 > 150 As per the largest Rupture
Recovery nia gas piping connection
(99%) size with a maxi-
8 Hydrogen Hydro- 6,527 50.0 mum of 299 mm
Recovery gen Table 2 Representative hole sizes
(94%)
9 Ammonia Mixed 1,655 -17.1 The leak frequency per elements (e.g., process
Recovery – gas (12% equipment, instrumentation connection, valves,
side stream ammo- flanges, piping) was adopted from IOGP 434-01
nia, 49% [6] and HyRAM v4.0 [7] for units handling am-
hydro- monia and hydrogen, respectively. Note that the
gen) leak causes considered in the analysis were lim-
10 Hydrogen Synthe- 6,86 50.0 ited to the general causes such as corrosion, vi-
Recovery sis gas bration, mechanical failure, failure during con-
struction or human error during installation.
Process specific releases were not accounted in The yearly total leak frequency distribution is de-
the study. picted in Figure 2. It is revealed that accidental
leaks are most likely to occur in the very small
The leak frequency breakdown per hole size and hole size with a frequency of one leak within 19
element type are given in Table 3 and Table 4. months. As can be seen in Figure 2, flanges be-
come the main contributor of the leak sources.
Leak Category Leak Frequency Priority of integrity check or inspection for
(/year) flanges connection is recommended.
Very Small (1-3 mm) 6.207E-01
Small (3-10 mm) 1.610E-01
Medium (10-50 mm) 7.089E-02
Large (50-150 mm) 2.799E-02
Full Bore (>150 mm) 4.805E-03
Total 8.854E-01
Table 3 Total leak frequency (/year) per hole
size

Element Type Leak Frequency

(/year)
Steel pipe meter 6.908E-02
Flange 4.135E-01
Valve 3.280E-02
Instrument 8.815E-02
Process vessel 1.954E-02
Pump 1.400E-02
Compressor 4.640E-02
HE Shell & Tube (shell 2.624E-02
side)
HE Shell & Tube (tube 1.757E-01
side)
Total 8.854E-01
Table 4 Total leak frequency (/year) per ele- Figure 2 Total leak frequency distribution
ment type profile

The calculated total leak frequency in Ammonia An example of event tree used in the analysis is
Plant Z is 0.8854 leaks per year which is equal to illustrated in Figure 3. In the event of unignited
1 leak within a period of 13.5 months. This infor- ammonia accidental releases, it is for conserva-
mation can be used as an input for the owner of tism reason that the final outcome was defined as
the facility to determine the minimum period of a complete toxic gas dispersion since the flam-
inspection or maintenance required for the facil- mable hazards consequences have been accom-
ity. modated in the ignition branches.
 Figure 3 Example of event tree

The ignition probability for natural gas was di- Specific for hydrogen, the total probability was
rectly taken from the Scenario No. 8 of IOGP derived from HYRAM database [7], as denoted
434-06 [8]. Despite being flammable, ammonia in Table 6.
requires high ignition energy ( > 100 mJ or equiv-
alent to ~34x of methane ignition energy) hence Leak rate Ignition probability
it is considered as low reactivity gas [9, 10]. (kg/s) Immediate Delayed
Therefore, the ignition probability of ammonia < 0.125 0.008 0.004
was estimated from the correlation given in 0.125 – 6.25 0.053 0.027
IOGP 434-06 [8] with a multiplication factor of > 6.25 0.230 0.120
0.1. The scenarios adopted were Scenario No. 8 Table 6 Ignition probability for hydrogen
(for gas) and No. 10 (for liquid). The summary of
ignition probabilities used as a function of leak An event tree for every failure scenario from each
rate is indicated in Table 5. isolatable section was developed to calculate the
frequency of potential final event. Table 7 sum-
Leak Total ignition probability marizes the total frequency per type of final event
rate Natural Ammonia Ammonia in Ammonia Plant Z.
(kg/s) gas (gas) (liquid)
0.1 0.0011 0.0001 0.0001 Final event Total frequency Per-
0.2 0.0014 0.0001 0.0001 (/year) centage
0.5 0.0020 0.0002 0.0002 Fire (Jet/Pool) 1.757E-02 2.0%
1 0.0025 0.0003 0.0002 Vapour Cloud 3.336E-03 0.4%
2 0.0050 0.0005 0.0004 Explosion
5 0.0125 0.0012 0.0009 Flash Fire 5.004E-03 0.6%
10 0.0250 0.0025 0.0016 Toxic Gas Dis- 1.037E-01 11.7%
20 0.0500 0.0050 0.0027 persion
50 0.1250 0.0125 0.0050 Flammable 7.558E-01 85.4%
100 0.2500 0.0250 0.0050 Gas Dispersion
200 0.5000 0.0500 0.0050 Total 8.854E-01 100.0%
500 0.6500 0.0650 0.0050 Table 7 Total frequency of final event
1000 0.6500 0.0650 0.0050
Table 5 Total ignition probability for natural
gas and ammonia
Table 7 indicates that flammable gas dispersion
is the most dominant outcome in case of acci-
dental release in Ammonia Plant Z. Although
ammonia is the main product in the facility, it is
worth highlighting that hydrogen and natural gas
are present in most of the processing sections
leading to domination of flammable hazards. The
second most probable final event is toxic gas dis-
persion which may arise from the release of ei-
ther gas or liquefied ammonia in the plant. In
general, the frequency analysis results show that
risk management should put more focus on the
prevention or mitigation of gas dispersion.
Figure 4 Targets of interest
Consequence Modelling
Event Threshold Description
The wind statistics and environmental condition Jet/pool fire 4 kW/m2 Injury after 30s
applied in the consequence modelling are shown [11] exposure
in Table 8 and Table 9. 35 kW/m2 100% fatality
Flash fire / gas 100% LFL 100% fatality
Month N S W E NW SW NE SE
August (2022) 10.3% 17.4% 22.2% 2.5% 8.1% 19.3% 14.2% 2.3%
dispersion [11] (if ignited)
July (2022)
June (2022)
11.8% 12.1%
17.8% 12.1%
19.6%
8.5%
3.9% 18.7%
12.8% 17.7%
29.1%
9.2%
4.7%
8.3%
2.3%
10.8%
Toxic gas AEGL-2 Disabling. Irri-
May (2022) 13.8% 11.6% 16.3% 5.4% 20.5% 17.5% 8.6% 8.0% dispersion [1] (600 s) tation in eyes
April (2022) 14.1% 15.2% 7.2% 14.8% 7.1% 6.2% 14.4% 14.1%
March (2022) 11.2% 12.1% 3.9% 21.2% 5.7% 4.2% 13.5% 10.8% and throat;
February (2022) 9.5% 7.1% 6.5% 12.3% 6.0% 4.5% 11.3% 18.8%
January (2022) 5.7% 8.0% 10.5% 21.2% 6.4% 5.0% 21.2% 23.9% urge to cough
December (2021)
November (2021)
3.4%
1.7%
1.8%
0.0%
0.7%
3.9%
1.5%
4.4%
2.1%
4.2%
0.0%
1.2%
0.9%
2.3%
2.8%
5.6%
AEGL-3 Lethality
October (2021)
Total
0.6% 2.7%
15.8% 10.2%
0.7%
6.9%
0.0% 3.5%
9.2% 12.8%
3.9%
15.3%
0.7%
20.1%
0.5%
9.7%
(600 s)
Table 8 Wind statistics of the facility Explosion 0.5 barg 100% fatality
(human)
Parameter Value [12]
Ambient averaged pressure 100.6 kPa Explosion 0.035 barg Minor damage
Ambient averaged temperature 28.2 oC (building) [12] 0.83 barg Total damage
Relative humidity 73.9 % Table 10 Consequence evaluation criteria
Surface roughness 0,5
Wind speed and Pasquill Class 2F 5D Exposure Maximum
Table 9 Environmental condition Type Exposure
Radius
A list of critical targets where the exposure was (m)
reviewed is given in Figure 4. The acceptable Jet fire radiation: 4 kW/m2 782
level for each final event is provided in Table 10. Jet fire radiation: 35 kW/m2 381
Pool fire radiation: 4 kW/m2 56
The maximum exposure radius for each final Pool fire radiation: 35 kW/m2 << 1
event is indicated in Table 11. The toxic concen- Flash Fire: 100% LFL 199
tration contours as the worst consequence mod- Flammable gas dispersion: 199
elling results are depicted in Figure 5. 100% LFL
Toxic gas dispersion: AEGL-2 6,030
Toxic gas dispersion: AEGL-3 1,270
 Exposure Maximum Risk Estimation and Evaluation
Type Exposure
Radius The risk evaluation was carried out as impair-
(m) ment analysis if two criteria below are met:
Explosion (human): 0.5 barg 283 1. Target is exposed at a level above the allow-
Explosion (building): 0.035 656 able limit.
barg 2. Target is exposed at a total frequency as fol-
Explosion (building): 0.83 barg 268 lows:
Table 11 Maximum exposure radius per final • Building: higher than 1E-04 per year
event at specific threshold • Individual (onsite): higher than 1E-05
per year
• Individual (offsite): higher than 1E-06
per year

The risk evaluation results for each target of in-

terest are summarized in Table 12. Note that only
the worst threshold is accounted for the impair-
ment. Given the risk evaluation at all targets of
interest, Ammonia Plant Z gives the highest risk
in exposing individual to ammonia toxic gas dis-
persion.

Impairment (/year)
Target

JF PF EXP (1) EXP (2) FF/FD TD

Figure 5 The worst ammonia toxic concentra-
1 6E-05 no No no no 3E-02
tion contours 2 2E-04 no 1E-04 3E-04 2E-04 1E-01
3 no no No no no 9E-03
4 1E-04 no 1E-04 2E-04 1E-04 5E-02
The consequence modelling results demonstrate 5 6E-05 no 6E-05 no no 4E-02
that the worst exposure is given by the accidental 6 no no No no no 2E-02
release of ammonia toxic gas resulting in approx- Legend: JF: Jet Fire; PF: Pool Fire; EXP: Explosion; FF: Flash Fire; FD:
Flammable Dispersion; TD: Toxic Dispersion
imately 6 km of toxic gas dispersion. The con-
centration was measured as AEGL-2 meaning Note:
1. For individual: 0.5 barg
that individual exposed to 220 ppmv of ammonia 2. For building: 0.83 barg
gas for 600 s will suffer disabling irritation. The Table 12 Risk evaluation summary
other most concerning result is the AEGL-3 cov-
erage reaching 1,3 km away from the leak source Risk-informed preventive and mitigation
which may lead to fatality when one is exposed measures were proposed as follows:
to the 2,700 ppmv of ammonia for 600 s. Alt- 1. Control engineering
hough the frequency analysis indicates that leak • Relocation of HVAC air intake due to
frequency of the ammonia processing units is not potential gas ingress in case of flam-
the highest, preventive and mitigative measures mable and toxic gas releases at the
of an ammonia leak is of importance due to the plant
significant toxic consequences. • Review the necessity of having addi-
tional isolation sections to limit the
volume of accidental release. An inte-
gration to interlock system is recom-
mended.
 • Review of fire/blast rating existing de- 2. Harris, G. P. F, MacDermott P. E., Flamma-
sign on the impacted building bility and Explosibility of Ammonia,
IChemE Symposium Series No. 49.
2. Control administrative 3. ARIA No. 717, Rupture of a cryogenic am-
• Perform routine inspection and monia tank March 20th 1989 Jonova Lithua-
maintenance on equipment, piping, nia (USSR), March 2007.
and connection accounting for the leak 4. Pattabathula V., Nayak R., Timbres D. H.,
frequency profile. The Safety and Integrity of Ammonia Stor-
• Revisit the Emergency Response Plan age Tanks, Indian Fertilizer Scene Annual-
and relocate the potentially impaired 2014.
assembly point to a safer location. 5. Palak Jain and Dhruv Patel, Research Re-
port, What has the Industry Experience been
Conclusion with Ammonia Manufacturing Plants? What
is Their Track Record? What Root Cause
Quantification of flammable/toxic gas disper- Have Typically Led to Them?, 3rd December
sion, fire, and explosion at an ammonia plant was 2019.
conducted. As ammonia is produced based on the 6. International Association of Oil & Gas Pro-
steam reforming of natural gas via the synthesis ducers, Risk Assessment Data Directory,
of hydrogen and nitrogen, the analysis was also Process Release Frequencies, Report 434-
done considering the hazards of all chemicals, 01, September 2019.
not only arising from ammonia. Based on the fre- 7. Sandia National Laboratories, Sandia Report
quency analysis, it was found that flammable gas Hydrogen Plus Other Alternative Fuels Risk
dispersion could be the most likely final event to Assessment Models (HyRAM+) Version 4.1
occur in case of an accidental release. This might Technical Reference Manual, April 2022.
be due to the significant presence of hydrogen 8. International Association of Oil & Gas Pro-
handling units in the facility. Nevertheless, the ducers, Risk Assessment Data Directory, Ig-
consequence modelling results show a different nition Probabilities, Report 434-06, Septem-
trend. Ammonia toxic gas dispersion will give ber 2019.
the worst consequences, tentatively reaching 6 9. VROM, Methods for the Calculation of
km away from the leak source, despite being a Physical Effects – Yellow Book.
lesser event frequency. The finding serves as an 10. HSE UK, Failure Rate and Event Data for
emphasizing point where the coupling of fre- use within Risk Assessments (06/11/17).
quency and consequence will provide more ben- 11. International Association of Oil & Gas Pro-
efits in understanding the facility risk profile. To ducers, Risk Assessment Data Directory,
conclude, risk evaluation on the defined targets Vulnerability of Humans, Report 434-14,
of interest shows that all are likely to suffer from March 2010.
toxic exposure. As an effort to lower the risk to a 12. VROM – Methods for the determination of
more acceptable level, additional preventive and possible damage, CPR16E.
mitigative measures were proposed.

References
1. National Research Council (US) Committee
on Acute Exposure Guideline Levels, Acute
Exposure Guideline Levels for Selected Air-
borne Chemicals (National Academies Press
(US); 2008), Vol. 6, Appendix 2.