Top Banner
Journal of Loss Prevention in the Process Industries 14 (2001) 283–306 www.elsevier.com/locate/jlp An assessment of the likelihood of occurrence, and the damage potential of domino effect (chain of accidents) in a typical cluster of industries Faisal I. Khan a , S.A. Abbasi b,* a Faculty of Engineering and Applied Sciences, The Memorial University of Newfoundland, St. John’s, Newfoundland, Canada A1B 3X5 b Centre for Pollution Control & Energy Technology, Pondicherry University, Kalapet, Pondicherry 605 014, India Abstract In the context of risk assessment and loss prevention in chemical process industries, the term domino effect denotes ‘chain of accidents’, or situations when a fire/explosion/missile/toxic load generated by an accident in one unit in an industry causes secondary and higher order accidents in other units. Most of the past risk assessment studies deal with accident in a single industry, more so in one of the units of an industry. But, often, accident in one unit can cause a secondary accident in a nearby unit, which in turn may trigger a tertiary accident, and so on. The probability of occurrence and adverse impacts of such ‘domino’ or ‘cascading’ effects are increasing due to increasing congestion in industrial complexes and increasing density of human population around such complexes. The multi-accident catastrophe which occurred in a refinery at Vishakhapatnam, India, on 14 September 1997, claiming 60 lives and causing loss of property worth over Rs 600 million, is the most recent example of the damage potential of domino effect [Lees F.P. Loss prevention in process industries, 2nd ed. Butterworths, 1-3, London; Khan, F.I., & Abbasi, S.A. (1999a). Major accidents in process industries and an analysis of their causes and consequences. Journal of Loss Prevention in Process Industries, 12, 361–378; Khan, F.I., & Abbasi, S.A. (1999b). The worst chemical industry accident of 1990’s–what happened and what might have been: A quantitative study. Process Safety Progress, 18, 135–145]. Recently, we have proposed a systematic methodology called ‘domino effect analysis’ (DEA). A computer automated tool DOMIFFECT has also been developed by us based on DEA [Khan, F.I., & Abbasi, S.A. (1998a). Models for domino effect analysis in chemical process industries. Process Safety Progress — AIChE, 17 (2), 107–113; Khan, F.I., & Abbasi, S.A. (1998b). DOMIFFECT (DOMIno eFFECT): a new software for domino effect analysis in chemical process industries. Environmental Modelling and Software, 13, 163–177.]. The methodology is based on deterministic models used in conjunction with probabilistic analysis. This paper illustrates the application of DEA and DOMIFFECT to an industrial complex. Out of 16 credible accident scenarios envisaged in four closely situated industries namely Madras Fertilisers Limited (MFL), SPIC–Heavy Chemical Division (SPIC–HCD). Manali Petrochemical Limited (MPL), and Tamil- nadu Petroproducts Limited (TPL), ten scenarios forecast domino effect. Further analysis reveals that accidents in the ammonia synthesis unit, secondary reformer, and urea reactor of MFL may cause domino effect. Similarly, accidents in the storage units of propylene oxide, ethylene oxide and mono propylene glycol at MPL, hydrogen storage units at SPIC–HCD, and the propylene oxide and fuel oil storage units of TPL are likely to cause a domino effect. The consequences of all these credible accidents have also been forecast. The paper makes a strong case for making DEA an integral part of all risk assessment initiatives. 2001 Elsevier Science Ltd. All rights reserved. Keywords: Domino effect; Chain of accidents; Industrial disaster; Risk assessment; Explosion; Fire 1. Introduction Accidents in chemical process industries involving explosions and/or fires can damage nearby reactors, * Corresponding author. Tel.: +91-413-655363; fax: +91- 41365227/65263. E-mail address: profF[email protected] (S.A. Abbasi). 0950-4230/01/$ - see front matter 2001 Elsevier Science Ltd. All rights reserved. PII:S0950-4230(00)00048-6 pipelines, or storage vessels. If the installations thus hit by the primary accident happen to contain hazardous chemicals, the damage my trigger fresh explosions/fires/toxic releases. This type of ‘chain of accidents’ may continue to propagate itself by taking into its fold other units within striking distance which, on being hit, may become the cause of fresh accidents thereby perpetuating the chain.
24

An assessment of the likelihood of occurrence, and the damage potential of domino effect (chain of accidents) in a typical cluster of industries

Mar 07, 2023

Download

Documents

Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Page 1: An assessment of the likelihood of occurrence, and the damage potential of domino effect (chain of accidents) in a typical cluster of industries

Journal of Loss Prevention in the Process Industries 14 (2001) 283–306www.elsevier.com/locate/jlp

An assessment of the likelihood of occurrence, and the damagepotential of domino effect (chain of accidents) in a typical cluster

of industries

Faisal I. Khan a, S.A. Abbasi b,*

a Faculty of Engineering and Applied Sciences, The Memorial University of Newfoundland, St. John’s, Newfoundland, Canada A1B 3X5b Centre for Pollution Control & Energy Technology, Pondicherry University, Kalapet, Pondicherry 605 014, India

Abstract

In the context of risk assessment and loss prevention in chemical process industries, the term domino effect denotes ‘chain ofaccidents’, or situations when a fire/explosion/missile/toxic load generated by an accident in one unit in an industry causes secondaryand higher order accidents in other units. Most of the past risk assessment studies deal with accident in a single industry, more soin one of the units of an industry. But, often, accident in one unit can cause a secondary accident in a nearby unit, which in turnmay trigger a tertiary accident, and so on. The probability of occurrence and adverse impacts of such ‘domino’ or ‘cascading’effects are increasing due to increasing congestion in industrial complexes and increasing density of human population around suchcomplexes. The multi-accident catastrophe which occurred in a refinery at Vishakhapatnam, India, on 14 September 1997, claiming60 lives and causing loss of property worth over Rs 600 million, is the most recent example of the damage potential of dominoeffect [Lees F.P. Loss prevention in process industries, 2nd ed. Butterworths, 1-3, London; Khan, F.I., & Abbasi, S.A. (1999a).Major accidents in process industries and an analysis of their causes and consequences. Journal of Loss Prevention in ProcessIndustries, 12, 361–378; Khan, F.I., & Abbasi, S.A. (1999b). The worst chemical industry accident of 1990’s–what happened andwhat might have been: A quantitative study. Process Safety Progress, 18, 135–145]. Recently, we have proposed a systematicmethodology called ‘domino effect analysis’ (DEA). A computer automated tool DOMIFFECT has also been developed by usbased on DEA [Khan, F.I., & Abbasi, S.A. (1998a). Models for domino effect analysis in chemical process industries. ProcessSafety Progress — AIChE, 17 (2), 107–113; Khan, F.I., & Abbasi, S.A. (1998b). DOMIFFECT (DOMIno eFFECT): a new softwarefor domino effect analysis in chemical process industries. Environmental Modelling and Software, 13, 163–177.]. The methodologyis based on deterministic models used in conjunction with probabilistic analysis. This paper illustrates the application of DEA andDOMIFFECT to an industrial complex. Out of 16 credible accident scenarios envisaged in four closely situated industries namelyMadras Fertilisers Limited (MFL), SPIC–Heavy Chemical Division (SPIC–HCD). Manali Petrochemical Limited (MPL), and Tamil-nadu Petroproducts Limited (TPL), ten scenarios forecast domino effect. Further analysis reveals that accidents in the ammoniasynthesis unit, secondary reformer, and urea reactor of MFL may cause domino effect. Similarly, accidents in the storage units ofpropylene oxide, ethylene oxide and mono propylene glycol at MPL, hydrogen storage units at SPIC–HCD, and the propyleneoxide and fuel oil storage units of TPL are likely to cause a domino effect. The consequences of all these credible accidents havealso been forecast. The paper makes a strong case for making DEA an integral part of all risk assessment initiatives. 2001Elsevier Science Ltd. All rights reserved.

Keywords: Domino effect; Chain of accidents; Industrial disaster; Risk assessment; Explosion; Fire

1. Introduction

Accidents in chemical process industries involvingexplosions and/or fires can damage nearby reactors,

* Corresponding author. Tel.: +91-413-655363; fax: +91-41365227/65263.

E-mail address: [email protected] (S.A. Abbasi).

0950-4230/01/$ - see front matter 2001 Elsevier Science Ltd. All rights reserved.PII: S0950- 42 30 (00)00 04 8- 6

pipelines, or storage vessels. If the installations thus hitby the primary accident happen to contain hazardouschemicals, the damage my trigger freshexplosions/fires/toxic releases. This type of ‘chain ofaccidents’ may continue to propagate itself by takinginto its fold other units within striking distance which,on being hit, may become the cause of fresh accidentsthereby perpetuating the chain.

Page 2: An assessment of the likelihood of occurrence, and the damage potential of domino effect (chain of accidents) in a typical cluster of industries

284 F.I. Khan, S.A. Abbasi / Journal of Loss Prevention in the Process Industries 14 (2001) 283–306

In the domain of chemical process industries, occur-rence of such a chain of accidents is called ‘dominoeffect’. The name is derived from the game of dominotoppling in which large numbers of small flat blocks,often made of wood, are so arranged in patterns thattoppling of one block over the next one starts a chain inwhich each toppling block makes the next one to toppleand so on. Upto several thousand dominos kept downthe line can be toppled by the initiating event of thetoppling of the first domino.

A look at the history of accidents involving hazardouschemicals reveals that chain of accidents have beenrather common. For example at Vishakhapatnam, India(Khan & Abbasi, 1999a,b; Abbasi & Khan, 2000) a leakin one of the eight Horton spheres at a refinery storingcrude led to a fire/explosion which damaged a nearbysphere causing it to explode. The chain of accidentsspread to other storage vessels and soon there was fireand damage all around. Sixty people lost their lives andassets worth Rs 600 million were damaged. In 1989, avapour cloud explosion caused by a leak in the poly-ethylene plant of a Philips installation near Houston,Texas, USA, led to two other major explosions in isobut-ane storage vessel and polyethylene reactor. These inturn set off a series of seven more explosions. Similarinstances of domino effect have been recorded at Feyzin(France), Pernis (The Netherlands), San Juan (Mexico),Texas City, Pasadena, Henderson (USA), and elsewhere(Khan & Abbasi, 1998, 1999a). Whereas in the first fewdecades of the previous century, the industries handlinghazardous chemicals were fewer in number and used tobe located far away from populous neighbourhoods, inlater years the number of such industries has risen shar-ply—as a direct consequence of ever-increasing varietyof products and processes. Simultaneously, due to press-ure on land, the distances between such industries andhuman settlements have become lesser and lesser. Forexample in Manali Industrial Complex (Chennai). where18 major petroleum and petrochemical industries aresituated within a rectangular area of 10 km2, or at Pillai-yarkuppam-Kirumampakkam (Pondicherry), where asimilarly large number of hazardous industries arelocated, one can find closely populated villages begin-ning right after the industrial premises. Indeed such situ-ations are common throughout India and the rest of thedeveloping countries (Abbasi & Vineethan, 1997;Abbasi, Krishnakumari, & Khan, 1999).

This scenario, and the grim reminder of the catas-trophes of the type which occurred at Vishakhapatnam.calls for urgent attention towards making the study ofthe domino an essential component of all risk assess-ment initiatives.

2. Events that can initiate domino effect

Domino effect may be initiated by one or more ofthese causative events:

(a) Fire: pool fire, flash fire, fireball, and jet fire.(b) Explosion: confined vapour cloud explosions(CVCE), boiling liquid expanding vapour explosion(BLEVE), vented explosion, vapour cloud explosion,and dust explosion.(c) Toxic release: instantaneous or continuous releaseof toxic light-as-air-gases; lighter-than-air gases, andheavier-than-air gases; release of toxic liquids.

In order to find out the probability of occurrence of dom-ino effect, assess its damage potential, and forecast thesequence in which chain of reactions may occur, deter-ministic models have to be used in conjunction with pro-babilistic analysis. The former type of models are usefulin quantifying physical and chemical processes such assize of leak, chemical release mode and rate, dispersion

Fig. 1. Domino effect analysis procedure.

Page 3: An assessment of the likelihood of occurrence, and the damage potential of domino effect (chain of accidents) in a typical cluster of industries

285F.I. Khan, S.A. Abbasi / Journal of Loss Prevention in the Process Industries 14 (2001) 283–306

(including cloud formation), type of explosion and itsintensity, heat/missile/shock wave loads generated, etc.The latter type, which are based on careful recapitulationand analysis of past accidents, and site characteristics,are needed to work out the probabilities of equipmentfailure, direction of missiles, direction of fireball/jet fire,pattern of toxic plume dispersion etc (Latha, Gautam, &Raghavan, 1992; Lees, 1996; Khan & Abbasi, 1998).

While choosing the models and conducting analysisbased on them one has to be conscious of the following:

2.1. Choice of models

The choice of models, and their use, must be donewith care, keeping in view the following factors:

(a) the computed magnitudes of incident heat andshock wave effects would have a measure of uncer-tainty;(b) the stress/strain patterns in vessels affected by fireor fire–explosion combinations would be complex anddifficult to quantify with precision;(c) external and transient factors such as wind direc-tion can play a major role;(d) simultaneous occurrence of heat and mass trans-fer, made complex by transient and rapidly changingnature and magnitudes of the initiating events such as

Fig. 2. Manali (near Chennai, India) and its surroundings.

fire due to leakage in a storage vessel, can contributeto the uncertainty.

3. Procedure for DEA (domino effect analysis)

A detailed consequence assessment should be conduc-ted as the initiating step of DEA, for the primary event(missiles, heat load, overpressure) at the location of sec-ondary unit. For this, latest models for assessing theimpact generated at the source of the primary event,directional probabilities (in case of missiles and fire jets),damage radii, scenarios of vessel failure etc should beutilised (Khan & Abbasi, 1998a). The sequence of stepsfor conducting a typical domino effect analysis ispresented in Fig. 1. We have developed DOMIFFECT(DOMIno eFFECT) which is a computer-automated toolincorporating the DEA methodology (Khan & Abbasi,1998b). DOMIFFECT is a menu-driven, interactivesoftware with an in-built database. It is capable of thefollowing operations:

� assessing likely hazards: fires, explosions, toxicrelease, and combinations of these;

� study of interaction among different accidental events

Page 4: An assessment of the likelihood of occurrence, and the damage potential of domino effect (chain of accidents) in a typical cluster of industries

286 F.I. Khan, S.A. Abbasi / Journal of Loss Prevention in the Process Industries 14 (2001) 283–306

Fig. 3. Location of various industries in the Manali Industrial Complex (MIC).

(generation of domino or cascading accidentscenarios):

� estimation of domino effect probability;� estimation of domino effect consequences.

The software incorporates state-of-the-art models forstudying (a) the hazards of fire, explosion, toxic release,or combinations of these present in a chemical processindustry; (b) the damage potential of likely accidents,assessed on the basis of credible scenarios the tooldevelops; (c) the likelihood of second accident beingtriggered by the first, (d) the scenarios of the secondaryaccidents, their damage potential, and the probability oftheir causing a third accident (steps similar to a–cabove), and so on. A major feature of DOMIFFECT isthe module automatic analysis. This module uses aknowledge-base built into DOMIFFECT; estimates thedetailed consequences, and checks the probability ofoccurrence of secondary or higher order accidents at dif-ferent locations.

DOMIFFECT further works out the probability ofoccurrence of domino effect on the basis of the location,contents, orientation, and material characteristics of thelikely target units. The analysis employs, inter alia, ves-

sel failure and probit models. If the probability of occur-rence of a ‘hit’ is sufficiently high then an accident scen-ario for the secondary accident is developed followed byan estimation of its damage potential. This procedure isrepeated until the full chain of likely accidents is cover-ed.

DEA is performed at two levels. In the first level, ascreening of all the units of a chemical process industryis done in order to identify the units which may comeunder the spell of domino effect. For this purpose thres-hold values of different damaging effects reported inliterature are used. For example, it is reported that anoverpressure of 0.7 atm can destroy a unit by blast waveimpact, a heat load of 37 kW/m2 is sufficient to inducevessel failure, and a missile (sharp edged) having a velo-city higher than 75 m/s has sufficient potential to pen-etrate the target unit provided that it collides with theunit. If the estimated values of these parameters at thelocation of the target unit are higher than the thresholdvalues, then a detailed analysis, or the second level studyis performed.

At the second level, more detailed analysis is conduc-ted to verify the existence of domino effect, using thedamage potential of the primary event and the character-

Page 5: An assessment of the likelihood of occurrence, and the damage potential of domino effect (chain of accidents) in a typical cluster of industries

287F.I. Khan, S.A. Abbasi / Journal of Loss Prevention in the Process Industries 14 (2001) 283–306

Table 1The output of DOMIFFECT for the primary accident in ammonia synthesis unit of MFL and subsequent domino effects. DOMIFFECT F.I. Khanand S.A. Abbasi, Pondicherry-605 014, India

Parameters Values

Primary eventUnit: ammonia synthesis unitScenerio: BLEVE followed by fireball and dispersion of toxic gasExplosion: BLEVETotal energy released (kJ) :6.45e+07Peak over pressure (kPa) :2175.80Variation of over pressure in air (kPa/s) :1151.75Shock wave velocity (m/s) :1051.81Duration of shock wave (s) :24.5Missile characteristicsInitial velocity of fragment (m/s) :763.2Kinetic energy of fragment (kJ) :8.11e+05Penetration ability at 50 m from the location of primary accidentConcrete structure (m) :0.70Brick structure (m) :0.95Steel structure (m) :0.15FireballRadius of the fireball (m) :91.67Duration of the fireball (s) :37.46Energy released by fireball (kJ) :2.71e+07Radiation heat flux (kJ/m2) :9655.4Toxic release and dispersionBox instantaneous model: elevated sourceConcentration at distance of 200 (m) (kg/m3) :3.452E�04Heavy gas puff characteristicsGround level concentration of puff (kg/m3) :3.452E�04Ground level concentration on puff axis (kg/m3) :3.452E�03Cloud radius (m) :1.167E+02Maximum distance traveled by the cloud (m) :7.874E+02Maximum ground level concentration (kg/m3) :4.3Domino effect—I: Secondary accidentUnit: Primary reformer unitDistance of the vulnerable unit from the primary event (m) :55.0Domino effect probabilities:Due to heat load (%) :100Due to explosion (%) :100Due to missile (%) :45Domino effect impactsScenario: BLEVE followed by fire ballExplosion: BLEVETotal energy released (kJ) :12890.0Peak over pressure (kPa) :65.8Variation of over pressure in air (kPa/s) :0.67Shock velocity of air (m/s) :389.0Duration of shock wave (ms) :300No missile effectFire: FireballRadius of the fire ball (m) :88.1Duration of the fire ball (s) :36.4Energy released by fire ball (kJ) :1.2E+06Radiation heat flux (kJ/m2) :745.6Domino effect—II: Secondary accidentUnit: Catacarb reboiler unitDistance of the vulnerable unit from the primary event (m) :70.0Domino effect probabilities:Due to heat load (%) :60Due to explosion (%) :45Due to missile (%) :15

(continued on next page)

Page 6: An assessment of the likelihood of occurrence, and the damage potential of domino effect (chain of accidents) in a typical cluster of industries

288 F.I. Khan, S.A. Abbasi / Journal of Loss Prevention in the Process Industries 14 (2001) 283–306

Table 1 (continued)

Parameters Values

Domino effect impactsScenario: continuous release followed by pool fireFire: Pool fireRadius of the pool fire (m) :47.0Burning area (m2) :91781.8Burning rate (kg/min) :574.8Heat flux (kJ/m2) :332Domino effect—III: Tertiary effectUnit: Ammonia stripper unitDistance of the vulnerable unit from the secondary event (m) :90.0Domino effect probabilities:due to heat load (%) :25due to explosion (%) :30due to missile (%) :11Domino effect impactsScenario: BLEVE followed by toxic load build upExplosion: BLEVEEnergy released during explosion (kJ) :2.4E+06Peak over pressure (kPa) :1151.4Variation of over pressure in air (kPa/s) :481.4Shock velocity of air (m/s) :795.6Duration of shock wave (ms) :41.2Missile characteristics:Initial velocity of a typical fragment (m/s) :235.6Kinetic energy associated with a typical fragment (kJ) :1.5E+05Penetration ability at 50 m from the location of tertiary accident:Concrete structure (m) :0.06Brick structure (m) :0.11Steel structure (m) :0.01Toxic release and dispersionBox instantaneous model: Elevated sourceConcentration at distance of 200 (m) (kg/m3) :1.755E�04Heavy gas puff characteristicsGround level concentration of puff (kg/m3) :1.522E�04Ground level concentration of puff axis (kg/m3) :1.522E�03Cloud radius (m) :1.897E+02Maximum distance traveled by the cloud (m) :6.504E+02Maximum ground level concentration (kg/m3) :4.1

istics of the secondary unit. The following characteristicsare considered:

� the shape and constituent material of the unit,� the chemicals involved and the operating conditions

under which they are used,� quantities and physical properties of the chemicals

involved,� location of the unit in terms of terrain characteristics

and distance from other units, and� meteorology, especially the prevailing wind direction.

4. Study of occurrance of domino effect in atypical cluster of industries

In India, and indeed most Third World countries, it isvery common to find several chemical process industries

situated close to one another. Large clusters of indus-tries, each sharing its compound wall with one or moreother industries, occur throughout India. Such, ‘indus-trial complexes’ occupy several square kilometres ofland area.

More often than not, densely populated villages andsuburban neighbourhoods are situated just outside theperiphery of industries complexes. Indeed neighbour-hoods spring up even between the industries if blocks ofland happen to be available (Abbasi, 1999; Abbasi &Vineethan, 1997).

Manali Industrial complex (MIC) 15 km from down-town Chennai (Madras) is one such cluster of industries.The largest of the industries in this cluster is a petroleumrefinery—Madras Refineries Limited (MRL)—whichhappens to be one of the largest such industries in India.The cluster has several large petrochemical industriesbased on feedstock generated by MRL (Fig. 2). It isabout 3.5 km west of the Bay of Bengal, and is sur-

Page 7: An assessment of the likelihood of occurrence, and the damage potential of domino effect (chain of accidents) in a typical cluster of industries

289F.I. Khan, S.A. Abbasi / Journal of Loss Prevention in the Process Industries 14 (2001) 283–306

Table 2Impacts of initiating events and the probabilities of causing domino effect in MFL

Parameters Scenarios and their likely impactsBLEVE with fire ball and CVCE followed by fireball in BLEVE followed by toxicdispersion of toxic gas in secondary reformer release in absorber unitammonia synthesis unit

Explosion: BLEVE CVCE BLEVETotal energy released, KJ 4.38e+08 4.05e+07 2.55e+05Peak overpressure developed, kPa 20454.8 11354.9 587.3Variation of overpressure in air, kPa/s 10475.7 8675.5 301.04Shock wave velocity, m/s 3963.8 1673.5 631.6Duration of shock wave, s 41.4 20.7 14.5Missile characteristics:Initial velocity of fragment, m/s 7253.2 1553.2 182.2Kinetic energy of fragment, kJ 1.31e+8 2.45e+07 1.12e+05Penetration ability at study pointConcrete structure, m 0.50 0.35 0.01Brick structure, m 0.68 0.52 0.00Steel structure, m 0.10 0.08 0.00Fire FIREBALL FIREBALL Not applicableRadius of fireball, m 185.4 75.5 Not applicableDuration of fireball, s 87 29.4 Not applicableEnergy released by fire ball, kJ 6.59e+09 1.54e+08 Not applicableRadius of pool fire, m Not applicable Not applicable Not applicableBurning area, m2 Not applicable Not applicable Not applicableBurning rate, kg/s Not applicable Not applicable Not applicableRadiation heat flux, kJ/m2 654132.5 6070.7 Not applicableDomino checkingLocation of the unit from primary event, m 75.0 75.0 75.0Domino effect due to heat loadTotal heat received, kJ 1.30e+08 7.45e+07 Not applicableHeat intensity, kJ/m2 9306.0 4114.78 Not applicableProbability of domino effect due to fire 1.0 1.0 Not applicableDomino effect due to over pressure:Explosion energy, kJ 4.38e+08 4.05e+07 2.55e+05Peak overpressure, kPa 10576.08 9754.8 81.4Probability of domino effect due to overpressure 1.0 1.0 0.4Domino effect due to missileExplosion energy, kJ 4.38e+08 4.05e+07 2.55e+05Missile velocity, m/s 5.20e+03 9.04e+02 1.5e+01Probability of domino effect due to missile after 1.0 1.0 0.14meeting the targetToxic release and dispersion

(continued on next page)

rounded by the villages Thiruvottiyur and Ennore in theeast, Amulavoil and Vaikkadu in the north, Mathur andMadhavaram in the west, and Chinnasekkadu, Periya-sekkadu, and Selavayal in the south (Fig. 3). The MICand its surroundings within 10 km radius of the centreof the complex have a population of 4 million people.The terrain is flat, at mean sea level of 4.5 m. The meanroughness factor of the area is 0.3. Besides MIC, the areawithin 10 km radius of the centre of MIC has populousneighbourhoods, agricultural lands, and barren lands(Fig. 3).

In this paper we present a detailed study conducted toanalyse the probability and consequence of dominoeffect occurrence in the four major industries of MIC:

Madras Fertilisers Limited (MFL), SPIC–Heavy Chemi-cal Division (SPIC–HCD), Manali Petrochemical Lim-ited (MPL), and Tamilnadu Petroproducts Limited(TPL). To estimate the probability of secondary acci-dent, a distance of at least 75 m has been assumed amongdifferent units. In reality the distance may often be less.

Initially we conducted a thorough risk assessmentstudy of each individual industry (Abbasi & Khan,1999). Based on this study we identified various accidentscenarios likely to initiate the domino effect in differentindustries. These scenarios have been further studied indetail to estimate the probability of occurrence and theconsequences of domino effect.

Page 8: An assessment of the likelihood of occurrence, and the damage potential of domino effect (chain of accidents) in a typical cluster of industries

290 F.I. Khan, S.A. Abbasi / Journal of Loss Prevention in the Process Industries 14 (2001) 283–306

Table 2 (continued)

Parameters Scenarios and their likely impactsBLEVE with fire ball and CVCE followed by fireball in BLEVE followed by toxicdispersion of toxic gas in secondary reformer release in absorber unitammonia synthesis unit

Instantaneous/continuous Instantaneous Not applicable InstantaneousPuff/Plume characteristics: Puff characteristics Puff characteristicsPuff concentration at centre of cloud, kg/m3 3.452E�03 1.26e�02Concentration at cloud edge kg/m3 3.452E�04 1.42e�03Maximum ground level concentration, kg/m3 1.15E�01 2.45E�01Distance when maximum concentration occurs, m 787.4 1050.4

Scenarious and their likely impactsCVCE followed by fire ball in urea reactor BLEVE followed by toxic BLEVE followed by fireball in

release in pre-neutralizer of drier unitNPK plant

CVCE BLEVE BLEVE1.11e+07 2.35e+05 1.45e+0.41556.7 517.3 355.2575.7 271.4 144.0865.5 611.6 382.614.5 14.5 9.51405.4 155.1 65.91.51e+07 1.05e+05 4.1e+030.30 0.01 0.000.41 0.00 0.000.07 0.00 0.00191.6 Not applicable 67.30154 Not applicable 21.452.71e+10 Not applicable 4.50e+06Not applicable Not applicable Not applicableNot applicable Not applicable Not applicableNot applicable Not applicable Not applicable82325.4 Not applicable 4285.475.0 75.0 75.0

Not applicable5.45e+07 Not applicable 2.21e+05111215.4 Not applicable 609.51.0 Not applicable 0.751.11e+07 2.35e+05 1.45e+043296.5 434.5 53.51.0 0.85 0.251.11e+07 2.35e+05 1.45e+042185.25 6.55e+02 9.31.0 0.25 0.0

Not applicable Not applicableNot applicable

1.90e�042.13e�051.54e�02120.0

5. The industries likely to be involved

5.1. Madras Fertilisers Limited (MFL)

MFL is one of the leading fertiliser producers in Indiasituated at MIC. It produces raw ammonia, urea, andnitrogen–phosphorous–potassium (NPK) fertilisers as itsmain products.

At MFL, ammonia is produced through naptha–steam

reformation process, urea through total recycle process,and NPK through Dorr Oliver NPK granulation process.The main units and the corresponding sections at MFLare as follows:

Ammonia plantDe-sulphuriser sectionReformer sectionCO2 Absorption section

Page 9: An assessment of the likelihood of occurrence, and the damage potential of domino effect (chain of accidents) in a typical cluster of industries

291F.I. Khan, S.A. Abbasi / Journal of Loss Prevention in the Process Industries 14 (2001) 283–306

Fig. 4. A typical sequence of accidents that may be triggered by an accident in (A) ammonia synthesis unit, (b) secondary reformer, (c) ammoniaabsorber, (d) urea reactor, (e) pre-naturaliser, and (f) drier units.

Compression sectionAmmonia synthesis unitUrea plantCompressor sectionUrea reactorDe-compressor unitAmmonia absorber unitEvaporation and prilling unitNPK plantNeutralising sectionDrying sectionsScreens and crushing sectionCooling sectionCoating section

5.2. Southern Petrochemical Industries Corporation–Heavy Chemical Division (SPIC–HCD)

SPIC Heavy Chemicals Division, Manali, (formerlyKothari Industrial Corporation Limited) was com-missioned in 1979. The major raw materials used in

SPIC–HCD are sodium chloride, water, and ammonia.The various products are: (i) caustic soda lye—48%, (ii)caustic soda flakes, (iii) chlorine (gas and liquid), (iv)hydrogen (gas), (v) hydrochloric acid—33%, (vi)ammonium chloride, (vii) sodium hypochlorite.

The plant operations constitute the following sections:

1. Salt storage and handling2. Brine treatment3. Membrane cell house4. Chlorine treatment and drying5. Chlorine compression and liquefaction6. Hydrogen treatment7. Hydrochloric acid plant8. Ammonium chloride plant9. Caustic concentration plant10. Waste air de-chlorination and effluent treatment

plant11. Utilities12. Storage

Page 10: An assessment of the likelihood of occurrence, and the damage potential of domino effect (chain of accidents) in a typical cluster of industries

292 F.I. Khan, S.A. Abbasi / Journal of Loss Prevention in the Process Industries 14 (2001) 283–306

Fig. 5. A typical sequence of accidents that may be triggered by anaccident in (a) hydrogen treatment unit, (b) hydrogen storage, and (c)chlorine storage.

5.3. Manali Petrochemical Limited (MPL)

The main products of MPL (Fig. 3) are propyleneoxide, propylene glycol and polyol. The industry com-prises of three plants, each consisting of several sections.

5.3.1. Propylene oxide plantPropylene oxide is produced by saponification of the

propylene chlorohydrin with lime and then is recoveredby distillation. The process is carried out in four stages—chlorohydrin production, saponification, purification ofPO, and purification of DCP.

5.3.2. Propylene glycol plantPropylene glycol is produced by hydration of propy-

lene oxide. The plant comprises of: (i) reaction unit, (ii)glycol concentration unit, (iii) glycol fractionation unit.

5.3.3. Polyol plantThe polyol plant comprises of the following sections:

� Initiator preparation section� Polymerisation section� Treater� Water washing� Vacuum drying

5.4. Tamilnadu Petroproducts Limited (TPL)

TPL is engaged in manufacturing linear alkyl benzene(LAB). Recently, an epichlorohydrin plant has beenerected by the TPL management just opposite its existingLAB plant. The annual production of LAB is 80,000MT. The main raw materials for the production of LABare kerosene and benzene.

The process involves two major steps. The first stepis the extraction of N-paraffin of the desired range fromkerosene and the second is reaction of paraffin and ben-zene to form LAB. These two steps are carried out inthe following units:

I. Pre-fractionationII. HydrotreatorIII. MolexIV. PacolV. Detergent alkylation

6. Development of the accident scenarios

The following credible accident scenarios weredeveloped using DOMIFFECT, and the criteriadescribed elsewhere (Khan & Abbasi, 2000).

6.1. Madras Fertilisers Limited

Ammonia synthesis unitBLEVE followed by fireball and dispersion of toxicgasSecondary reformer unitCVCE followed by fireballUrea reactorCVCE followed by fireballPrimary absorberBLEVE followed by toxic releasePre-neutraliserBLEVE followed by toxic releaseDrier unitBLEVE followed by fireball

Page 11: An assessment of the likelihood of occurrence, and the damage potential of domino effect (chain of accidents) in a typical cluster of industries

293F.I. Khan, S.A. Abbasi / Journal of Loss Prevention in the Process Industries 14 (2001) 283–306

Table 3Impacts of initiating events and the probabilities of causing domino effect in SPIC-HCD

Parameters Scenarios and their likely impactsFireball in hydrogen CVCE followed by fireball BLEVE followed by toxictreatment unit in hydrogen storage vessel release in chlorine storage vessel

Explosion: CVCE BLEVETotal energy released, KJ 5.45e+08 3.33e+07Peak overpressure developed, kPa 22550.1 8764.3Variation of overpressure in air, kPa/s Not applicable 15475.5 4010.5Shock wave velocity, m/s 4255.8 1205.2Duration of shock wave, s 106 18.5Missile characteristics:Initial velocity of fragment, m/s Not applicable 8415.3 1065.4

Kinetic energy of fragment, kJ 2.46e+08 1.9e+07Penetration ability at study pointConcrete structure, m 0.55 0.31Brick structure, m Not applicable 0.78 0.41Steel structure, m 0.17 0.07Fire FIREBALL Not applicableRadius of fireball, m 75.6 215.4 Not applicableDuration of fireball, s 31.5 128.6 Not applicableEnergy released by fire ball, kJ 6.52e+07 9.54e+09 Not applicableRadius of pool fire, m Not applicable Not applicable Not applicableBurning area, m2 Not applicable Not applicable Not applicableBurning rate, kg/s Not applicable Not applicable Not applicableRadiation heat flux, kJ/m2 5441.6 755150.3 Not applicableDomino CheckingLocation of the unit from primary event, m 75.0 75.0 75.0Domino effect due to heat loadTotal heat received, kJ 1.14×106 3.30e+09 Not applicableHeat intensity, kJ/m2 1141.3 11235.1 Not applicableProbability of domino effect due to fire 1.0 1.0 Not applicableDomino effect due to over pressure:Explosion energy, kJ 5.45e+08 3.33e+07Peak overpressure, kPa Not applicable 15413.2 1994.4Probability of domino effect due to overpressure 1.0 1.0Domino effect due to missileExplosion energy, kJ 5.45e+08 3.33e+07Missile velocity, m/s Not applicable 7.14e+03 815.4Probability of domino effect due to missile after 1.0 1.0meeting the targetToxic release and dispersionInstantaneous/continuous InstantaneousPuff/Plume characteristics: Puff characteristicsPuff concentration at centre of cloud, kg/m3 Not applicable Not applicable 5.145E�02Concentration at cloud edge kg/m3 5.145E�03Maximum ground level concentration, kg/m3 1.54e�01Distance when maximum concentration occurs, m 875.4

6.2. SPIC heavy chemicals division

Hydrogen treatment unitFireballHydrogen storage unitCVCE followed by fireballAmmonia storage unitBLEVE followed by toxic release and dispersion

6.3. Manali Petrochemical Ltd

Propylene oxide storage unit:

CVCE followed by fireballEthylene oxide storage unitInstantaneous release causing BLEVE followed byfireballMono propylene glycol storage unitInstantaneous release leading to pool fireFuel oil storage unitPool fire

6.4. Tamilnadu Petroproducts Limited

Propylene storage unit

Page 12: An assessment of the likelihood of occurrence, and the damage potential of domino effect (chain of accidents) in a typical cluster of industries

294 F.I. Khan, S.A. Abbasi / Journal of Loss Prevention in the Process Industries 14 (2001) 283–306

Table 4The detailed results of DOMIFFECT for accident in ammonia storage unit of SPIC-HCD and subsequent domino effects in other units. DOMIF-FECT F.I.Khan and S.A. Abbasi, Pondicherry-605 014, India

Parameters Values

Primary eventUnit: chlorine storage vesselScenerio: BLEVE followed by fireballExplosion: BLEVETotal energy released (kJ) :1.32e+07Peak over pressure (kPa) :1745.4Variation of over pressure in air (kPa/s) :814.1Shock wave velocity (m/s) :1245.2Duration of shock wave (s) :21.4Missile characteristicsInitial velocity of fragment (m/s) :445.3Kinetic energy of fragment (kJ) :1.66e+06Penetration ability at 50 m from the location of primary accidentConcrete structure (m) :0.50Brick structure (m) :0.68Steel structure (m) :0.10Toxic release and dispersionBox instantaneous model: Elevated sourceConcentration at distance of 200 (m) (kg/m3) :1.225E�04Heavy gas puff characteristicsGround level concentration of puff (kg/m3) :1.114E�04Ground level concentration on puff axis (kg/m3) :1.112E�03Cloud radius (m) :1.52E+02Maximum distance traveled by the cloud (m) :5.454E+02Maximum ground level concentration (kg/m3) :5.4Domino effect—I: Secondary accidentUnit: Primary reformer unitDistance of the vulnerable unit from the primary unit (m) :55.0Domino effect probabilities:Due to heat load (%) :100Due to explosion (%) :100due to missile (%) :100Domino effect impactsScenario: Instantaneous release and dispersion of chlorineBox instantaneous model: Elevated sourceConcentration at distance of 200 (m) (kg/m3) :3.452E�04Heavy gas puff characteristicsGround level concentration puff (kg/m3) :3.452E�04Ground level concentration puff axis (kg/m3) :3.452E�03Cloud radius (m) 1.167E+02Maximum distance travelled by the cloud (m) 7.874E+02Maximum ground level concentration (kg/m3) 4.3Domino effect—II: Secondary accidentUnit: Hydrogen storage unitDistance of the vulnerable unit from the primary unit (m) :65.0Domino effect probabilities:Due to heat load (%) :–Due to explosion (%) :60Due to missile (%) :30Domino effect impactsScenario: BELVE followed by fireballExplosion: BELVETotal energy released (kJ) :3.33e+07Peak over pressure (kPa) :904.7Variation of over pressure in air (kPa/s) :1010.5Shock velocity of air (m/s) :1705.2Duration of shock wave (ms) :21.5Missile characteristicsInitial velocity of fragment (m/s) :665.4Kinetic energy of fragment (kJ) :5.9e+06

(continued on next page)

Page 13: An assessment of the likelihood of occurrence, and the damage potential of domino effect (chain of accidents) in a typical cluster of industries

295F.I. Khan, S.A. Abbasi / Journal of Loss Prevention in the Process Industries 14 (2001) 283–306

Table 4 (continued)

Parameters Values

Penetration ability at 50 m from the location of secondary accidentConcrete structure (m) :0.67Brick structure (m) :0.80Steel structure (m) :0.15Fire: FireballRadius of the fireball (m) :97.3Duration of the fireball (s) :39.7Energy released by the fireball (kJ) :9.06E+07Radiation heat flux (kJ/m2) :11384.5Domino effect—III: Tertiary effectUnit: Ammonia storage unitDistance of the vulnerable unit from the secondary unit (m) :90.0Domino effect probabilities:Due to heat load (%) :–Due to explosion (%) :30Due to missile (%) :11Domino effect impactsScenario: BLEVE followed by toxic load buil-upExplosion: BLEVEEnergy released during explosion (kJ) :1.4E+06Peak over pressure (kPa) :515.4Variation of over pressure in air (kPa/s) :281.4Shock velocity of air (m/s) :495.6Duration of shock wave (ms) :24.2Missile characteristics:No missile effectToxic release and dispersionBox instantaneous model: Elevated sourceConcentration at distance of 200 (m) (kg/m3) :3.452E�04Heavy gas puff characteristicsGround level concentration of puff (kg/m3) :3.452E�04Ground level concentration of puff axis (kg/m3) :3.452E�03Cloud radius (m) :1.167E+02Maximum distance traveled by the cloud (m) :7.874E+02Maximum ground level concentration (kg/m3) :4.3

CVCE followed by fireballAllyal chloride storage unitPool fireFuel oil storage unitPool fire

7. Results and discussion

7.1. Domino effect at MFL

The domino effect analysis, involving forecasting ofthe primary accident as well as likely chain of accidentstriggered by the primary event, for a probable accidentin ammonia synthesis unit (accident scenario BLEVEfollowed by fireball and dispersion of toxic gasammonia) is presented in Table 1. It is evident that theintensity of shock waves, missiles, and overpressurewould be high enough over an area of �450 m to initiatesecondary and tertiary accidents in the units placed

within this area. This would cover most of the units ofammonia and urea plants and also the units of MPL andother neighbouring industries. As the primary absorberis placed 55 m away from the ammonia synthesis unit,it would be impacted by the primary accident. The mostcredible accident, which may result, is BLEVE followedby fireball. The intensity of shock waves and heat loadgenerated due to this accident would envelop an area ofover �200 m radius. The impact of secondary accidentwould trigger a tertiary accident in the ammonia stripperunit (primary and secondary) placed 90 m from theabsorber, generating a fireball (Tables 1 and 2). Thelikely chain of accidents for this scenario is presented inFig. 4.

A sudden release of hydrogen from the secondaryreformer, generating CVCE followed by fireball, maycause development of severe shock waves, and heat load(Table 2). The severity of shock waves for this scenariois similar to that observed for the accident scenario inammonia synthesis unit. The contours for 50% damagedue to shock wave span an area of �400 m radius,

Page 14: An assessment of the likelihood of occurrence, and the damage potential of domino effect (chain of accidents) in a typical cluster of industries

296 F.I. Khan, S.A. Abbasi / Journal of Loss Prevention in the Process Industries 14 (2001) 283–306

Table 5The detailed results of DOMIFFECT for accident in Propylene oxide storage unit of MPL and subsequent domino effects in other units. DOMIF-FECT F.I. Khan & S.A. Abbasi, Pondicherry-605 014, India

Parameters Values

PRIMARY EVENTUnit: Propylene oxide storage vesselScenario: CVCE followed by fireballExplosion: CVCETotal energy released (kJ) :5.45e+07Peak over pressure (kPa) :17743.4Variation of over pressure in air (kPa/s) :9424.3Shock wave velocity (m/s) :3155.5Duration of shock wave (ms) :34Missile characteristicsInitial velocity of fragment (m/s) :1894.2Kinetic energy of fragment (kJ) :4.54e+07Penetration ability at 50 m from the location of primary accidentConcrete structure (m) :0.35Brick structure (m) :0.52Steel structure (m) :0.08FireballRadius of the fireball (m) :155.5Duration of the fireball (s) :50.0Energy released by fireball (kJ) :1.06e+09Radiation heat flux (kJ/sq.m) :112841.5DOMINO EFFECT—I: Secondary accidentUnit: Ethylene oxide reactorDistance of the vulnerable unit from the primary unit (m) :55.0Domino effect probabilities:due to heat load (%) :100due to explosion (%) :100due to missile (%) :100Domino effect impactsScenario: BLEVE followed by fireballExplosion: BLEVETotal energy released (kJ) :3.45e+06Peak over pressure (kPa) :904.7Variation of over pressure in air (kPa/s) :1010.5Shock velocity of air (m/s) :1705.2Duration of shock wave (ms) :21Missile characteristicsInitial velocity of fragment (m/s) :665.4Kinetic energy of fragment (kJ) :5.9e+06Penetration ability at 50 m from the location of secondary accidentConcrete structure (m) :0.18Brick structure (m) :0.27Steel structure (m) :0.04Fire: FireballRadius of the fireball (m) :97.3Duration of the fireball (s) :39.7Energy released by fireball (kJ) :9.06E+07Radiation heat flux (kJ/sq.m) :11384.5DOMINO EFFECT—III: Tertiary effectUnit: Chlorine storage unitDistance of the vulnerable unit from the secondary unit (m) :90.0Domino effect probabilities:due to heat load (%) :100due to explosion (%) :100due to missile (%) :65

(continued on next page)

Page 15: An assessment of the likelihood of occurrence, and the damage potential of domino effect (chain of accidents) in a typical cluster of industries

297F.I. Khan, S.A. Abbasi / Journal of Loss Prevention in the Process Industries 14 (2001) 283–306

Table 5 (continued)

Parameters Values

Domino effect impactsScenario: BLEVE followed by toxic load build-upExplosion: BLEVEEnergy released during explosion (kJ) :2.4E+06Peak over pressure (kPa) :751.4Variation of over pressure in air (kPa/s) :481.4Shock velocity of air (m/s) :795.6Duration of shock wave (ms) :21Missile characteristics:Initial velocity of fragment (m/s) :235.6Kinetic energy associated with fragment (kJ) :1.5E+05Penetration at 50 m from the location of tertiary accidentConcrete structure (m) :0.06Brick structure (m) :0.11Steel structure (m) :0.01Toxic release and dispersionBox instantaneous model: Elevated sourceConcentration at distance of 200 (m) (kg/cu.m) :4.754E�03Heavy gas puff characteristicsGround level concentration of puff (kg/cu.m) :4.311E�03Ground level concentration on puff axis (kg/cu.m) :4.311E�02Cloud radius (m) :7.315E+01Maximum distance traveled by the cloud (m) :8.215E+01Maximum ground level concentration (kg/cu.m) :9.882E�01

whereas heat radiation effects would devastate an areaof �200 m radius. This scenario too has higher prob-ability of causing domino effect, The likely sequence ofaccidents is presented in Fig. 4. Primary reformer, shiftconvertor, and CO2 absorber units are likely to fail insecondary accidents, whereas neighbouring industriessuch as MPL, units of refinery, and units of MadrasPetro-Products Limited (MPPL) are likely to initiate ter-tiary and higher order accidents.

The absorber unit is the most important component ofthe urea plant. It is also the unit most susceptible to acci-dents, as it handles toxic chemicals at extreme conditionsof temperature and pressure. The accident scenario forabsorber unit is developed as BLEVE followed by dis-persion of toxic gas. As the unit deals with large quan-tities of ammonia under high pressure, there is a highprobability of occurrence of BLEVE. The releasedchemical would disperse in the atmosphere, buildingtoxic load. The DEA results for this scenario aredepicted in Table 2. An intense over-pressure load alongwith a high shock wave velocity would be observed overan area of �150 m radius. The lethal toxic load wouldbe operative over a radial distance of more than �2500m. The probable chain of accidents due to this accidentin presented in Fig. 4. It is evident that both primary andsecondary reformer units are likely to cause secondaryaccidents. The compression unit, the ammonia synthesisunit, and the surge gas condenser may fail in tertiaryaccidents.

Table 2 presents the DOMIFFECT output for a prob-

able accident in urea reactor (confined vapour cloudexplosion followed by fireball). The shock wave gener-ated due to CVCE can cause injury as well as secondorder accidents by seriously damaging other units (Table2). These impacts would span an area of more than 350m radius. The vapour cloud generated by CVCE onignition may cause a fireball and hence severe heat radi-ation effect. Similarly the missiles generated by CVCEmay hit nearby targets and may lead to secondaryexplosions or release of chemicals. The heat radiationeffect with 50% probability of lethality would beobserved over an area of �350 m radius. The probabilityof domino effect occurrence due to this accident is sohigh that it is almost a certainty. The chain of accidentslikely to be initiated by this accident is presented in Fig.4. It is clear that storage unit of MPL and most of theunits of ammonia and urea synthesis plant are under seri-ous threats of secondary accident, while units of MPL,SPIC–HCD and MRL are likely to fail as tertiary acci-dent.

As per accident scenario in pre-neutraliser unit,BLEVE followed by dispersion of toxic gases wouldgive rise to shock waves, missile effects, and toxic load(Table 2). The intensity of shock waves, high enoughto demolish all objects coming in their way, would bepersistent in an area of more than 200 m radius, Theintensity of toxic emissions would be lethal over an areaof �800 m radius. The probability of occurrence of dom-ino effect due to this accident is estimated as unity. Asmay be seen from the likely chain of events depicted in

Page 16: An assessment of the likelihood of occurrence, and the damage potential of domino effect (chain of accidents) in a typical cluster of industries

298 F.I. Khan, S.A. Abbasi / Journal of Loss Prevention in the Process Industries 14 (2001) 283–306

Table 6Impacts of initiating events and the probabilities of causing domino effect in MPL

Parameters Scenarios and their likely impactsCVCE followed by BLEVE with fireball Pool fire in mono Pool fire in fuel oilfireball in propylene in ethylene oxide propylene glycol storage vesseloxide storage vessel storage vessel storage vessel

Explosion: CVCE BLEVETotal energy released, KJ 5.45e+07 2.89e+06Peak overpressure developed, kPa 17743.4 844.3Variation of overpressure in air, kPa/s 9424.3 515.0 Not applicable Not applicableShock wave velocity, m/s 3155.3 887.6Duration of shock wave, ms 34 21Missile Characteristics:Initial velocity of fragment, m/s 1894.2 196.9Kinetic energy of fragment, kJ 4.54e+07 1.41e+05 Not applicable Not applicablePenetration ability at study pointConcrete structure, m 0.35 0.06Brick structure, m 0.52 0.21 Not applicable Not applicableSteel structure, m 0.08 0.01Fire POOL FIRE POOL FIRERadius of fireball, m 155.5 85.7 5.0 5.0Duration of fireball, s 50 34.5 Not applicable Not applicableEnergy released by fireball, kJ 1.06e+09 1.24e+07 Not applicable Not applicableRadius of pool fire, m Not applicable Not applicable Not applicable Not applicableBurning area, sq.m Not applicable Not applicable 78.5 78.5Burning rate, kg/s Not applicable Not applicable 21.5 35.0Radiation heat flux, kJ/sq.m 112841.5 8787.2 89659.8 154446.57Domino checkingLocation of the unit from primary event, m 75.0 75.0 75.0 75.0Domino effect due to heat loadTotal heat received, kJ 7.69e+07 1.11e+06 1.83e+08 5.15e+08Heat intensity, kJ/sq.m 8914.5 3102.5 10404.5 21518.3Probability of domino effect due to fire 1.0 1.0 1.0 1.0Domino effect due to over pressure:Explosion energy, kJ 5.45e+07 2.89e+06Peak overpressure, kPa 1515.6 516.6 Not applicable Not applicableProbability of domino effect due to overpressure 1.0 1.0Domino effect due to missileExplosion energy, kJ 5.45e+07 2.89e+06Missile velocity, m/s 975.4 95.4 Not applicable Not applicableProbability of domino effect due to missile after 1.0 0.9meeting the targetToxic release and dispersionInstantaneous/continuousPuff/plume characteristics:Puff concentration at centre of cloud, kg/cu.m Not applicable Not applicable Not applicable Not applicableConcentration at cloud edge, kg/cu.mMaximum ground level concentration, kg/cu.mDistance where maximum concentration occurs, m

Fig. 4, drying unit, absorber unit, and units of urea plantare very likely to fail as secondary accident whileammonia plant, and units of MPL are likely to meet withtertiary accidents.

The accident scenario for drier unit is envisaged asBLEVE followed by fireball. The summarised output ofDOMIFFECT for this scenario is presented in Table 2.The intensity of explosion would be lower compared tothe previous cases. Yet intense shock waves, capable ofcausing secondary accidents, would effect an area of�100 m radius, whereas the lethal heat load (50% prob-

ability of causing damage) would be observed over anarea of �100 m. Even though this unit, too, is likely tocause secondary and high order accidents, the probabilityis comparatively lower than in previous accident scen-arios. The sequence of events likely to be triggered inthis accident scenario is presented in Fig. 4.

7.2. Domino effect at SPIC–HCD

An accident in hydrogen treatment unit (fireball)would generate heat load high enough to cause 50%

Page 17: An assessment of the likelihood of occurrence, and the damage potential of domino effect (chain of accidents) in a typical cluster of industries

299F.I. Khan, S.A. Abbasi / Journal of Loss Prevention in the Process Industries 14 (2001) 283–306

Fig. 6. A typical sequence of accidents that may be triggered by anaccident in (a) propylene oxide (PO) storage, (b) ethylene oxide (EO)storage, and (c) mono propylene glycol (MPG) storage, and (d) fueloil storage.

damage over an area of more than �100 m radius (Table3). As a result, units placed close to the autoclave arelikely to get affected and lead to secondary accidentsas illustrated in Fig. 5. The electrolyser, compressors,liquefaction unit and storage units are likely to causehigher order accidents but the probability of such acci-dents is low.

The results of domino effect analysis for a likely acci-

dent in hydrogen storage vessel (CVCE followed byfireball) is presented in Tables 3 and 4. It is evident thatCVCE would generate great over pressure and shockwave over an area of more than 500 m radius, simul-taneously, a fireball would give rise to severe heat loadwhich would cause burns (at 50% probability) over anarea of more than �400 m radius. The units placedwithin striking distance of these events would all fail,leading to fires and/or explosions. The probability ofdomino effect due to this accident scenario is estimatedas one. The likely sequence of accidents is presented inFig. 5. It may be seen that electrolysis section, chlorinestorage, and units of MPL are most likely to fail in sec-ondary accidents while several units of TPL and MFLmay fail in tertiary accidents.

Chlorine is a highly toxic and non-flammable chemi-cal stored under pressurised conditions at SPIC–HCD.The results of DEA study for a probable accident scen-ario in the chlorine unit—BLEVE followed by toxicrelease and dispersion—are presented in Table 3. It isevident that an instantaneous release of the gas (asBLEVE would) develop intense shock waves, the impactof which would be felt over a wide area (of more than�250 m radius). The damage impact of these shockwaves accompanied by missile effect would initiate sec-ondary accidents in the units placed near to it. Theimpact of lethal toxic load would be operative over anarea of more than �150 m radius.

7.3. Domino effect at MPL

Propylene oxide (PO) is one of the most flammablechemicals used at MPL; it being the main raw material,is stored in large quantities. CVCE followed by fireballis the most credible accident scenario envisaged in thePO storage unit. The damage characteristics of this scen-ario have been studied in detail for the possibility ofinitiating domino effect. The output of the study ispresented in Table 5. It is seen that CVCE shall developpowerful shock waves and the fireball shall cause intenseheat load. These impacts would reach an area of morethan �350 m radius; damaging other units in the areaand sparking a cascade of accidents. The probability ofthis domino effect is close to unity (Table 6) Thesequence of accidents likely to occur due to this accidentis shown in Fig. 6. Propylene storage, ethylene oxidestorage, and propylene oxide reactor shall fail in second-ary accidents and the polyol reactor, PG reactor and unitsof neighbouring hydrogen and chlorine plants of SPIC–HCD may meet with tertiary accidents (Table 5).

The domino effect studies for an accident in ethyleneoxide (EO) vessel are presented in Table 6. The likelyaccident (BLEVE followed by fireball) would causedevelopment of highly intense shock waves and lethalheat load. Individual and/or combined impact of theseeffects would lead to failure of other units (secondary

Page 18: An assessment of the likelihood of occurrence, and the damage potential of domino effect (chain of accidents) in a typical cluster of industries

300 F.I. Khan, S.A. Abbasi / Journal of Loss Prevention in the Process Industries 14 (2001) 283–306

Table 7Impacts of initiating events and the probabilities of causing domino effect in TPL

Parameters Scenarios and their likely impactsCVCE followed by fireball in Pool fire in allayl Pool fire in fuel oilpropylene storage vessel chloride storage vessel storage vessel

Explosion: CVCETotal energy released, KJ 3.25e+07Peak overpressure developed, kPa 8674.5Variation of overpressure in air, kPa/s 3955.1 Not applicable Not applicableShock wave velocity, m/s 995.3Duration of shock wave, s 18.5Missile characteristics:Initial velocity of fragment, m/s 995.4Kinetic energy of fragment, kJ 1.66e+07 Not applicable Not applicablePenetration ability at study pointConcrete structure, m 0.28Brick structure, m 0.40 Not applicable Not applicableSteel structure, m 0.07Fire Pool fire Pool fireRadius of fireball, m 175.4 5.0 5.0Duration of fireball, s 85 Not applicable Not applicableEnergy released by fireball, kJ 6.15e+09 Not applicable Not applicableRadius of pool fire, m Not applicable Not applicable Not applicableBurning area, m2 Not applicable 78.5 78.5Burning rate, kg/s Not applicable 15.0 34.0Radiation heat flux, kJ/m2 644315.2 6896.1.3 144181.9Domino checkingLocation of the unit from primary event, m 75.0 75.0 75.0Domino effect due to heat loadTotal heat received, kJ 1.15e+08 7.45e+07 5.05e+08Heat intensity, kJ/m2 9105.4 5004.8 20105.3Probability of domino effect due to fire 1.0 1.0 1.0Domino effect due to over pressure:Explosion energy, kJ 3.25e+07Peak overpressure, kPa 1876.4 Not applicable Not applicableProbability of domino effect due to overpressure 1.0Domino effect due to missileExplosion energy, kJ 3.25e+07Missile velocity, m/s 775.4 Not applicable Not applicableProbability of domino effect due to missile after meeting the 1.0targetToxic release and dispersionInstantaneous/continuousPuff/plume characteristics:Puff concentration at centre of cloud, kg/m3 Not applicable Not applicable Not applicableConcentration at cloud edge kg/m3

Maximum ground level concentration, kg/m3

Distance where maximum concentration occurs, m

accident). The results of DOMIFFECT reveal that unitsplaced 55–75 m away from EO storage unit are vulner-able to secondary accident. The probabilities of occur-rence of secondary accidents due to heat load, shockwaves, and missiles, are all very high. The units involv-ing storage of propylene oxide, propylene glycol frac-tionation, and PO purification shall meet with secondorder accidents. The impact of shock waves and heatload thus generated shall cause tertiary accidents in theunits of MFL, SPIC–HCD, and MPL. The adverseimpacts of these accidents may lead to higher order acci-dents (Fig. 6).

An accident in monopropylene glycol (MPG) storageunit of MPL—pool fire—would generate very high heatload, sufficient to cause 50% damage over an area ofmore than 250 m radius (Table 6). There may be boilingof chemicals (benzene, propane, ethylene, etc.) andconsequent release in instantaneous or continuousfashion. The escaping chemicals on meeting with anignition source would cause fire and/or explosiondepending upon the type of chemical and the mode ofrelease. Most of the units of MPL, and the purificationand electrolysis units of SPIC–HCD may be damaged,perpetuating a chain of accidents effecting units of TPL,

Page 19: An assessment of the likelihood of occurrence, and the damage potential of domino effect (chain of accidents) in a typical cluster of industries

301F.I. Khan, S.A. Abbasi / Journal of Loss Prevention in the Process Industries 14 (2001) 283–306

Fig. 7. A typical sequence of accidents that may be triggered by anaccident in (a) propylene storage, (b) allayl chloride storage, and (c)fuel oil storage.

storage units of SPIC–HCD and the eplichlorohydrinplant.

The results of DEA study for an accident in fuel oilstorage unit of Manali Petrochemicals Limited ispresented in Table 6. The results reveal that burning ofthe released chemical (fuel oil) as pool fire would causebuilding up of intense heat load over an area �250 mradius. The probability of occurrence of secondary andhigher order accidents due to this event is high enoughto be taken as a certainty. Most of the units of MPL,TPL and epichlorohydrin would be damaged in thehigher order accidents.

7.4. Domino effect at TPL

As with MPL, propylene oxide (PO) is the main rawmaterial at TPL, and is stored in large quantities. Themost credible accident scenario envisaged in this unit isCVCE followed by fireball. The damage characteristicsof this scenario have been studied in detail for the possi-

bility of initiating domino effect. The output of the studyis presented in Table 7. CVCE will develop intense overpressure (shock waves) whereas fireball will develophighly intense heat load. The impacts of these wouldpenetrate an area of more than �300 m radius. The otherunits placed closer to the PO unit will be involved inchain of accidents, of which probability is close to unity.The sequence of secondary and higher order accidentslikely to occur due to this accident, depicted in Fig. 7,reeveals that other storage units of TPL, epichlorophyd-rin plant, and units of SPIC–HCD will encounter second-ary accidents while chlorine storage of SPIC–HCD,polyol reactor, and other units of neighbouring MPL arelikely to meet with tertiary accidents.

The accident scenario of pool fire in allayl chloridestorage unit forecasts burning of allayl chloride leadingto high heat load over an area of more than �100 mradius (Table 7). The probability of occurrence of sec-ondary and higher order accidents due to this initiatingevent is high but comparatively lower than the dominoeffect probabilities in other units of MPL and TPL. Theunits that are likely to meet with secondary accidents arepropylene oxide storage, reactors, fuel oil storage, andthe storage units of epichlorohyrdin plant. There is lowpossibility of tertiary or higher order accidents.

The results of DEA study for an accident in fuel oilstorage unit of TPL is presented in Tables 7 and 8. Theresults reveal that burning of fuel oil as pool fire wouldcause building up of intense heat load. Heat flux suf-ficient to cause damage or boiling of liquid would beobserved over an area of more than �200 m radius. Sec-ondary and higher accidents due to pool fire in fuel oilstorage unit are almost certain to occur (Table 8). Theunits likely to face secondary accidents are allayl chlor-ide storage unit, reaction unit, chlorine storage, and alsounits of epichlorohydrin plant. The fractionation andother process units of MPL, SPIC–HCD, MPL andneighbouring industry IAL are likely to be drawn intotertiary and higher order accidents.

8. Risk estimation

To visualise the areas-of-impact of the primary(initiating) events, and the chains of accidents set off bythem, risk factors have been plotted over the MIC layout(Figs. 8–11). Due to the limitation of space we arepresenting discussions on the worst scenerio for each ofthe industries.

The detailed calculation for the accident scenario inammonia synthesis unit of MFL reveals that contours ofvery high (severe) risk contour (risk factor�1*10�4) dueto combined effects of explosion and fire wouldencompass an area of �400 m (Fig. 8). This wouldimpact several vulnerable units of MFL (urea plant, NPKplant, storage farm) as well as other industries such as

Page 20: An assessment of the likelihood of occurrence, and the damage potential of domino effect (chain of accidents) in a typical cluster of industries

302 F.I. Khan, S.A. Abbasi / Journal of Loss Prevention in the Process Industries 14 (2001) 283–306

Table 8The output of DOMIFFECT for the primary accident in fuel oil storage unit of TPL and subsequent domino effects. DOMIFFECT F.I.Khan andS.A. Abbasi, Pondicherry-605 014, India

Parameters Values

Primary eventUnit: Fuel oil storage unitScenario: Pool firePool fireRadius of pool fire (m) :5.0Burning area (m2) :78.5Burning rate (kg/s) :34.0Radiation heat flux (kJ/m2) :144181.9Domino effect—I: Secondary accidentUnit: Allayl chloride storageDistance of the vulnerable unit from the primary event (m) :50.0Domino effect probabilities:due to heat load (%) :100due to explosion is (%) :0due to missile (%) :0Domino effect impactsScenario: Pool fireFire: Pool fireRadius of pool fire (m) :7.5Burning area (m2) :176.6Burning rate (kg/s) :1.75Radiation heat flux (kJ/m2) :234551.2Domino effect—II: Tertiary accidentUnit: fractionation unitDistance of study from the secondary accident unit (m) :95.0Domino effect probabilities:Due to heat load (%) :100Due to explosion (%) :0Due to missile (%) :0Domino effect impactsScenario: BLEVE followed by fireballExplosion: BLEVETotal energy released (kJ) :1.55e+05Peak over pressure (kPa) :302.0Variation of over pressure in air (kPa/s) :250.0Shock wave velocity (m/s) :485.5Duration of shock wave (s) :14.0Missile characteristicsInitial velocity of a typical fragment (m/s) :95.90Kinetic energy a typical of fragment (kJ) :41241.68Final velocity of a typical fragment (m/s) :65.90Penetration ability at 50 m from the location of tertiary accidentConcrete structure (m) :0.01Brick structure (m) :0.05Steel structure (m) :0.00FireballRadius of the fireball (m) :75.5Duration of the fireball (s) :25.5Energy released by fireball (kJ) :4.75e+06Radiation heat intensity (kJ/m2) :6305.45

MPL, MRL, and TPL. The likely secondary accident thatmay occur due to this event is presented in Fig. 8. Itwould endanger other units and thus may initiate tertiaryor higher order accidents. The individual risk contourswould, together, cover the entire industrial complex(Fig. 8).

SPIC–HCD is located in the midst of several other

industries and any mishap in the former would have far-reaching consequences. This is evident from the damagecalculations for an accident scenario in hydrogen storagevessel. The risk contours due to the combined effect ofheat and shock wave generated due to CVCE and fireballcover an area of �500 m radius (Fig. 8). This wouldencompass a number of industries: TPL, MPL, MRL,

Page 21: An assessment of the likelihood of occurrence, and the damage potential of domino effect (chain of accidents) in a typical cluster of industries

303F.I. Khan, S.A. Abbasi / Journal of Loss Prevention in the Process Industries 14 (2001) 283–306

Fig. 8. Risk contours and likely chain of accidents in MFL initiated by an accident in secondary reformer unit; curve A depicts risk contour forprimary accident; curves B–E represent risk contours for accidents triggered by A. The overall impact area is represented by F.

and MFL (Fig. 9). The secondary accidents in these unitswould initiate tertiary and higher order accidents. Theeventual area-of-impact would cover the entire Manaliindustrial complex.

Like SPIC–HCD, MPL too is situated in the midst ofseveral other industries. However, the extent of riskposed by this industry is less than by SPIC–HCD. Therisk contour for a probable accident in propylene oxidestorage vessel covers an area of �200 m radius (Fig. 10).This encompasses the entire plant as well as the processarea of other industries (MPL, SPIC, and MFL). Theoverall risk contour of the domino effect envelops anarea of more than 2000 m radius (Fig. 10) extendingover most of the complex.

Assessment of the likely consequences of an accidentin propylene storage vessel of TPL (accident scenarioCVCE followed by fireball) reveals that an area of �250m radius is under severe risk due to the combined effectsof fire and explosion (Fig. 11). This would encompassmany units of the parent industry and also the units ofSPIC, and MRL. The secondary accidents likely to occurin these units are depicted in Fig. 11. It is evident thatsecondary accidents will envelop other hazardous units

and thus may initiate tertiary accidents. The overall riskcontour considering domino effect is presented in Fig.11, indicating that most of the area of MIC would beunder severe threat if this accident occurs.

9. Summary and conclusions

1. The paper highlights the need to make domino effectstudies an integral parts of all risk assessment exer-cises. In such studies the probability of occurrence ofdomino effect should be assessed and the impacts, ifsuch chain of accident occurs, should be forecast. Theauthors are prompted by the history of several majorpast accidents including the most recent one—andone of the biggest ever—which occurred at theHPCL’s refinery at Vishakhapatnam, India, on 14September, 1997.

2. Five types of primary events many initiate dominoeffect: fire, overpressure/shock waves, missiles, toxicload, and a combination of one or more of these.

3. Studies conducted by the authors, and others, indicatethat it is not necessary that the unit of an industry

Page 22: An assessment of the likelihood of occurrence, and the damage potential of domino effect (chain of accidents) in a typical cluster of industries

304 F.I. Khan, S.A. Abbasi / Journal of Loss Prevention in the Process Industries 14 (2001) 283–306

Fig. 9. Risk contours and likely chain of accidents in SPIC–HID; curve A depicts risk contour for primary accidents, curves B–E represent riskcontours for accidents triggered by A. The overall impact area is represented by F.

which may cause the biggest stand-alone accident,will also be the one most likely to cause dominoeffect. It is because the probability of occurrence ofdomino effect depends not only upon the damagepotential of the primary accident but also on a numberof others factors such as the characteristics of theunits impacted by the primary accident and the prox-imity of other units to the one meeting with the initialaccident. Prevailing meteorology, especially winddirection, may also greatly influence the course of theinitiating event.

4. The paper describes an elaborate case study of dom-ino effect analysis conducted on four closely spacedpetrochemical industries in an industrial complex atManali, Chennai (Madras), India. Numerous credibleaccident scenarios have been developed in which anaccident in one of the units of the four industrieswould start a chain of accidents involving other unitsof the same industry, or the units of nearby industries.

5. The studies reveal the seriously hazardous nature ofthe prevailing situation as several forecasts indicatethat major accidents—some involving the entire clus-ter of industries—can occur.

6. The paper highlights the importance of conductingdomino effect analysis (DEA) besides indicating theapplicability of the DEA methodology developed bythe authors.

Acknowledgements

SAA thanks All India Council for Technical Edu-cation (AICTE), New Delhi, for instituting the Com-puter-Aided Environmental Management Unit whichfacilitated this study.

References

Abbasi, S. A. (1999). Environmental pollution and its control. Philad-elphia and Pondicherry: Cogent International, pp xii+349.

Abbasi, S.A., & Khan, F.I. (1999). E/A and sustainability studies inManali lndustrial complex: executive summary of final report.Centre for Pollution Control & Energy Technology, PondicherryUniversity, Pondicherry, pp 1–186.

Abbasi, S. A., & Khan, F. I. (2000). Computer-aided environmental

Page 23: An assessment of the likelihood of occurrence, and the damage potential of domino effect (chain of accidents) in a typical cluster of industries

305F.I. Khan, S.A. Abbasi / Journal of Loss Prevention in the Process Industries 14 (2001) 283–306

Fig. 10. Risk contours and likely chain of accidents in MPL; curve A depicts risk contour for primary accidents; curves B–D represent riskcontours triggered by A. The overall impact area is represented by E.

management. New Delhi: Discovery Publishing House, ppviii+424.

Abbasi, S. A., & Vineethan, S. (1997). Environmental impact of indus-tries on suburban environments. New Delhi: Discovery PublishingHouse, pp viii+145.

Abbasi, S. A., Krishnakumari, P., & Khan, F. I. (1999). HOT TOPICS:everyday environmental concerns. New Delhi and London: OxfordUniversity Press, pp xviii+208.

Khan, F. I., & Abbasi, S. A. (1998). Risk assessment in chemical pro-cess industries. New Delhi: Discovery Publishing House, ppxvi+364.

Khan, F. I., & Abbasi, S. A. (1998a). Models for domino effect analy-sis in chemical process industries. Process Safety Progress—AIChE, 17 (2), 107–113.

Khan, F. I., & Abbasi, S. A. (1998b). DOMIFFECT (DOMinoeFFECT): a new software for domino effect analysis in chemical

process industries. Environment Modelling and Software, 13,163–177.

Khan, F. I., & Abbasi, S. A. (1999a). Major accidents in process indus-tries and an analysis of their causes and consequences. Journal ofLoss Prevention in Process Industries, 12, 361–378.

Khan, F. I., & Abbasi, S. A. (1999b). The worst chemical industryaccident of 1990s—what happened and what might have been: Aquantitative study. Process Safety Progress, 18, 135–145.

Khan, F.I., & Abbasi, S.A. (2000). A criteria for generating credibleaccident scenarios Journal of Loss Prevention in Process Indus-tries, in press.

Latha, P., Gautam, G., & Raghavan, K. V. (1992). Strategies forquantification of thermally initiated cascade effects. Journal of LossPrevention Process Industries, 5 (1), 15–21.

Lees, F. P. (1996). Loss prevention in process industries, vol. I–III.London: Butterworth.

Page 24: An assessment of the likelihood of occurrence, and the damage potential of domino effect (chain of accidents) in a typical cluster of industries

306 F.I. Khan, S.A. Abbasi / Journal of Loss Prevention in the Process Industries 14 (2001) 283–306

Fig. 11. Risk contours and likely chain of accidents in TPLl curve; curve A depicts risk contour for primary accidents, curves B–E represent riskcontours triggered by A. The overall impact area is represented by F.