SPE 69679 New Techniques for Risk and Hazard Management in the Petrochemical Industry

G.A. Chamberlain, Shell Global Solutions (UK)

Copyright 2001, Society of Petroleum Engineers Inc. This paper was prepared for presentation at the SPE Latin American and Caribbean Petroleum Engineering Conference held in Buenos Aires, Argentina, 2528 March 2001. This paper was selected for presentation by an SPE Program Committee following review of information contained in an abstract submitted by the author(s). Contents of the paper, as presented, have not been reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material, as presented, does not necessarily reflect any position of the Society of Petroleum Engineers, its officers, or members. Papers presented at SPE meetings are subject to publication review by Editorial Committees of the Society of Petroleum Engineers. Electronic reproduction, distribution, or storage of any part of this paper for commercial purposes without the written consent of the Society of Petroleum Engineers is prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300 words; illustrations may not be copied. The abstract must contain conspicuous acknowledgment of where and by whom the paper was presented. Write Librarian, SPE, P.O. Box 833836, Richardson, TX 75083-3836, U.S.A., fax 01-972-952-9435.

Abstract In today's regulatory and business environment in the petrochemical industry, the onus is placed increasingly on the plant operator to demonstrate that safety risks are tolerable and as low as reasonably practicable. The physical effects of accidents, such as explosion overpressure, heat radiation and toxic dispersion, have to be calculated to assess the threat to people and whether escalation and further threat are real possibilities. Accurate assessments based on realistic accident scenarios can avoid costly over-design in terms of protective measures and ensure that all hazards and effects are correctly identified. This paper reviews recent developments in the science, validation, quantification and application of these physical effects to improvements in plant safety. Some examples using the latest hazard assessment tools as applied to refineries, gas plants and offshore platforms will be discussed. Introduction The last quarter of a century there has seen significant progress in the science and understanding of the consequences of fires and explosions relevant to accidental releases of flammable fluids in oil, gas and petrochemical industrial plant. While some uncertainties remain, the application of the knowledge so far via a Hazards and Effects Management Process (HEMP) is normally sufficient to improve design and operation of plant in a significant cost efficient way. Recent advances in modelling techniques and validation by experiments performed over a range of scales have given much greater confidence to predictions of hazardous effects. Initiatives are already underway to address residual major areas of uncertainty such as the explosive behaviour of simulated real releases of flammable gas, dispersion in

congested areas, and the effectiveness of ventilation. After ignition, predictions of explosion loading on surroundings need to be able to account for the inhomogeneities and turbulence in these gas and aerosol clouds and the effect of high congestion by small scale equipment. A hydrocarbon fire can rapidly weaken engulfed surroundings. The global behaviour of confined and open fires is now reasonably well understood, but the conditions that control the severity of local hot spots in confined fires and smoke movement deserve further attention. In addition, the effectiveness of water based mitigation systems, fire relief and vessel blow-down strategies, and efficacy of passive fire protection should come under scrutiny. The response of plant and structure to such accidental loading, mitigation of accidental loading by protection systems, and the probability of such events are equally important. The cost effective solution to tolerable risk-taking must ultimately involve combination of all four aspects loading, response, mitigation and frequency, sometimes referred to as the Hazards and Effects Management Process, shown in Fig. 1. The rigorous application of such a process is key for enabling licence to operate by: indicating which risks and hazards are tolerable demonstrating that risks are as low as reasonably practicable (ALARP) facilitating compliance with local legislation and Company policy facilitating communication with stakeholders. It can also enable clever use of Capex by enabling: focused use of resources on higher risk contributors cost effective design and protection option selection by risk ranking. Safety engineers, risk analysts and designers will require user-friendly validated hazard assessment tools which link together scenarios from release, through ignition and explosion and fire, to response and escalation and individual risk in a clear way. This can only happen via a multidisciplinary approach to the overall problem and improved procedures for implementation of sound guidance. Initiatives along these lines are already under way. Overview of hazard consequences An accidental release of flammable hydrocarbon results in a variety of consequences, dependent on the type and initial state

G.A. CHAMBERLAIN

SPE 69679

of the hydrocarbon and whether or not ignition occurs. These consequences are conveniently displayed by means of an event tree such as shown in Fig. 2. By means of gross simplification this tree covers ALL INITIAL outcomes - fire, explosion or safe dispersion. Further analysis reveals that just three event trees, determined simply by reference to the state of the fluid before release, are required to cover all the different physical effects from the possible range of fire types and explosion scenarios. The three fluid pre-release conditions determining the "source term" for the event trees are: 1. Liquid at ambient pressure, e.g. diesel, gasoline or oil tank, 2. Liquid at pressure above ambient, e.g. separator contents, pumped crude oil, oil-well blow-out, 3. Gas or vapour above ambient pressure, e.g. gas pipeline, vapour space in separator, gas blow-out. Complete definition of the source term then requires calculation of the outflow rate, and the physical form and dimensions of the release. For example, a flammable liquid release from an atmospheric storage tank will always result firstly in the formation of a liquid pool, immediate ignition of which results in a transient fireball which burns back to a spray jet fire and burning pool. If ignition is delayed, the pool first spreads and partially evaporates. The vapour cloud disperses. If the surroundings are confined, such as for a release inside a building or offshore module, then a vented explosion could occur on ignition. A cloud dispersing in the open produces a flash fire with no blast. An ignited release in a congested but unconfined plant environment could lead to the generation of fast flames with associated blast. The combination of confinement and congestion, such as in a typical offshore module, can lead to significant blast. However, dispersion of the cloud by natural ventilation may lead to a dramatic reduction in the blast severity. A liquid release from pressurised storage initially leads to a spray jet. The spray would consist of both vapour and liquid phases, in the form of droplets, if the liquid is volatile e.g. condensate, or simply droplets for non-volatile releases e.g. stabilised crude oil. Mechanical and thermodynamic forces both play a role in causing the jet to break up in to droplets. Just a few bar is required for atomisation of the jet1 independent of any flashing processes that may be occurring. The effect this has on subsequent dispersion, degree of liquid rain-out and fire behaviour is critical to an assessment of the hazardous consequences, and hence the extent and type of protection system required. Jet and pool fires can occur on immediate ignition after an initial transient fireball, but liquid rain-out is unlikely if a released liquid, such as LPG, readily flashes to vapour in the atmosphere. The fireball rises into the air, resulting in a transient radiation pulse to the far field if there is significant initial upward momentum. On the other hand, a horizontal release can only create a ground level fireball. The fate of a cloud undergoing delayed ignition

depends on the degree of confinement and congestion in the surroundings, as before. Catastrophic loss of containment and ignition of pressurised liquid stored above its atmospheric vapour pressure, such as LPG, results in a type of fireball known as a BLEVE (boiling liquid expanding vapour explosion). The underlying physical processes are now well enough understood2 for model formulation and prediction of the transient fireball heat flux and explosion overpressure pulse. A pressurised gas release normally results in a dispersing gas jet which may contain aerosols if the gas cools sufficiently during release to allow condensation of some or all of the components. Entrainment of air normally is sufficient to reevaporate the aerosol droplets. Ignition creates the classic jet fire, typified by refinery flares. In extraordinary cases the release of hot gas containing heavy components could lead to condensation and agglomeration of droplets and a rain-out pool. In all cases however, as before, delayed ignition results in either a cloud fire or explosion dependent on the geometry and layout of the surroundings in which the release occurs. Methodologies, models, rules and guidance now exist with varying degrees of sophistication to allow engineers to quantify ALL of the above physical effects. In most cases the models, usually in the form of computer tools such as FRED3, have been validated over large ranges of experimental conditions. Fire Jet Fires. A jet fire is a turbulent diffusion flame resulting from the combustion of a fuel continuously released with some significant momentum in a particular direction. In principle any pressurised fuel has the potential to generate a jet fire, the essential features of which are shown in Fig. 3. Jet flame stability. Not all releases give rise to stable flames. A correlation4,5 derived from extensive experimental data6,7,8 , shows that for unobstructed releases of natural gas from orifices less than about 30mm diameter, certain drive pressures will not sustain a stable flame. Thus accidental damage to small bore high pressure fittings might reasonably be expected not to result in a stable flame in open environments. In highly congested industrial plants however, the likelihood of flame stabilisation by impact on adjacent surfaces is high. The increased burning velocity associated with higher hydrocarbon gases results in greater stability and smaller critical diameters. On the other hand, jet releases of liquid kerosene up to 2.5 kg/s are unable to sustain stable flames when released as a horizontal jet because of insufficient droplet evaporation9 . Similar self extinction was evident in the crude oil jet releases in Phase 2 of the Joint Industry Project on Blast and Fire Engineering for Topside Structures (BFETS)10,11 .

SPE 69679

NEW TECHNIQUES FOR RISK AND HAZARD MANAGEMENT IN THE PETROCHEMICAL INDUSTRY

Jet flame size and external heat flux. In typical plant environments free jet fires could be considered unlikely but, to date, most predictive models have been developed and validated for such fires, and it is worthwhile reviewing the progress made. Impacting and confined fire behaviour is the subject of ongoing research, discussed later. The main hazard from a free jet fire is thermal radiation to the surroundings. A simple expression for the radiation flux received by an object q is q=tVS, where t is the atmospheric transmissivity, V is the view factor of the flame surface and S is the flame surface emissive power. For most calculations the Shell jet flame model12,13 which represents the flame as a solid body - the frustum of a cone radiating uniformly over its surface adequately predicts the incident flux for safety engineering purposes, and is considerably more accurate than single point source representations (e.g. in API RP521) for objects close to the flame. The applicability of empirical models is however limited by the range of experimental data on which they are based. Computational fluid dynamics (CFD) models have been improved in recent years but require considerable computer power, computer time and expertise. Some results however are remarkably good14 . Additional output from the model includes such important information as convective and radiative components of heat flux within the flame and external radiation field, which have been validated by experiment for natural gas and propane gas flames. Jet fire internal heat flux. An accidental ignited release is likely to impact on adjacent objects. Thus the thermal loading is required to assess the risk of escalation by failure of pipework, or vessels, or structure. The object surface is subjected to convection from the high velocity hot combustion products and to radiation emitted by the flame. These components of total heat flux not only vary along and across the flame but also vary in time due to the turbulent nature of the jet and fluctuations in wind. Measurements15 in sonic natural gas jet flames up to 10 kg/s indicate maximum timeaveraged total heat fluxes to a cold surface of about 250 kW/m2 split roughly equally between convection and radiation. Hazard analyses often make use of these data extremes to predict the temperature rise of engulfed plant. Good agreement with experiment is found if a gas velocity of 40 m/s and a flame kinetic temperature of 1200 oC are used in the heat transfer calculation. In some treatments16 the convective heat transfer coefficient and flame radiative temperature can be derived experimentally and combined in physical models to predict the thermal load to fire engulfed objects. An alternative approach is to use the direct measurements obtained from large scale experiments taking care not to extrapolate too far from the experimental conditions. With partial funding from the European Community, Shell18 and British Gas17 carried out many tests with propane, butane and mixtures of natural gas and butane at flow rates up

to 28 kg/s for the single component fuels and 2.5 kg/s for the mixtures. In general, the flames were more radiative than natural gas and produced more smoke. They were also more buoyant and wind affected. For butane concentrations up to 40% the flame properties were similar to those of natural gas. With funding from the API, Davenport9 carried out similar tests with kerosene/natural gas jet fires. The fires were even more radiative than with butane mixtures, but pure kerosene releases could not sustain flames under the test conditions. Measured time-averaged heat fluxes to a cold surface reached about 300 kW/m2 at high levels (60-80% by mass) of either butane or kerosene. In Phase 2 of the BFETS10 the jet fire characteristics of unconfined releases of crude oil were measured. Releases of crude oil and natural gas consisted of separate fluid streams which combined several diameters downstream in the jet fire. The two nozzle arrangement was chosen to provide a convenient way of varying the gas-oil ratio (GOR) and is close to but not strictly representative of a release of live crude, such as could occur in a well blow-out. Some general conclusions can be drawn from these tests. The size of crude oil jet fires was similar to other hydrocarbon liquid and two phase jet fires. All the flames were highly radiative, with maximum time-averaged surface emissive powers from 203 to 409 kW/m2. Increasing the gas content in crude oil increased the emitted radiation. The fraction of heat radiated in the mixed releases reached about 0.3 consistent with other 2-phase jet flames. Increasing the gas content in crude oil increased the incident heat flux on an impinged object by increasing both the radiative and convective components. The radiative component was, however, always higher at the rear of the object. Spot maximum heat fluxes were around 350 kW/m2, similar to other mixed releases. Pool Fires. A pool fire is a turbulent diffusion fire burning above a pool of vaporising liquid fuel where the fuel vapour has very low initial momentum. The vaporisation of high boiling point fuels is controlled mainly by radiative feedback to the pool surface. The hazard effects are controlled largely by the size of the pool and the fuel type. These determine the mass burning rate and the size of the visible flame above the pool. Wind tilt drags the flame beyond a confining wall boundary and can bring the fire closer to downwind objects. Two distinct regions are created in the flame zone - a region of continuous burning and a more smoky region where flame appears intermittently. Empirical models19,20,21 usually represent the flame as a tilted cylinder which take into account these two regions by assigning different factors for smoke obscuration. The external radiation field is calculated utilising a view factor and flame surface emissive power as before. Relatively little information is available on the thermal loading from pool fires to engulfed objects. However, it is known that radiative heat transfer dominates (about 80%) and that flame temperatures reach about 1000oC. The internal heat flux of kerosene or JP-4 pool fires is around 150 kW/m2 but

G.A. CHAMBERLAIN

SPE 69679

increases to about 250 kW/m2 for large LNG or LPG pool fires. Confined fires. The behaviour of jet and pool fires can be significantly modified when the fire is confined by walls and ceilings, such as the boundaries of offshore compartments. In addition, hot combustion products rapidly fill ceiling spaces and impose fire loadings on any objects located there. If the compartment openings are small or the release rate of fuel is high, the fire may not be able to entrain enough air for complete combustion inside the compartment. The fire is then said to be ventilation controlled. As well as the hazards already discussed, additional ones exist22 . These include external flaming from compartment openings, impaired visibility along escape routes, increased CO hazard, explosion hazard from unburnt fuel if the fire terminates due to lack of oxygen, and possible increased thermal loading to objects due to greater amounts of hot soot. Recent large scale studies23,24 have confirmed that fire global stoichiometry is a useful correlating parameter. Gaseous propane jet fires burning in slightly fuel rich conditions represent a worst case because they combine high heat release with enhanced soot production. The heat flux was about 50% radiative in the impingement zone of the propane jet totalling about 350 kW/m2 in near stoichiometric propane jets. Pool fires, on the other hand, are more severe in over-ventilated conditions. When ventilation controlled the pool fire burning rate is governed by the size of the ventilation opening and the evaporation rate tends to drop to match the lower burning rate. Thus the pool fire limits itself to the slightly under-ventilated regime. Persaud et al25 show that the effect of insulation on the compartment walls is to raise the average smoke temperature by 200-400oC. The results of Phase II of the BFETS project, carried out at SINTEF NBL, confirmed and extended these findings to condensate jet and pool fires and larger scales26,27 . For jet fires it was found that there was no significant dependency of scale. Gas temperatures were between 1100 and 1300oC. Maximum heat fluxes were 170 kW/m2 for fuel rich jet flames and 250 kW/m2 for fuel lean flames. Explosion Overpressure generation. The burning of a vapour cloud in the absence of confinement or congestion does not generate significant overpressure. In order to generate what may be regarded as an explosion, there must be either congestion, i.e. obstacles in the path of the flame, or confinement of the gas cloud. In a typical offshore module, for example, both of these will be significant. In a typical onshore plant only congestion by vessels, pipework or support structures will be present. In the absence of turbulence, the (laminar) burning velocity of a hydrocarbon flame is quite low, around 0.5 m/s. As the flame ball grows this can increase owing to instabilities, but only by a factor of two to three. However, the expansion pushes gas ahead of the flame. If the resulting flow passes obstacles (e.g.

vessels, pipework), turbulence will be generated. The flame speed in a turbulent flow is much greater than the laminar flame speed, principally because the turbulence wrinkles the flame, creating a much greater area for reaction. Therefore, when the flame reaches the turbulent region downstream of the obstacles, it burns faster. This faster burning in turn creates faster flow and so a higher level of turbulence downstream of the next group of obstacles. The more intense turbulence results in even faster burning, and so on. This process is known as the Shchelkin mechanism (Fig. 4). Thus, even where there is little confinement, high flame speeds may be generated by congestion. The high velocities associated with the rapid burning are sufficient to generate high pressure, i.e. an explosion. Understanding of the mechanism described above allows characterisation of the types of congested but unconfined plant likely to cause high overpressures even without confinement.. These are typically chemical or refinery plant with large amounts of pipework or other obstructions. Locations where obstacles block more than 40% of the path of a flame, or where there are closely repeated rows of obstacles, are particularly bad. On the other hand, lightly congested plant no longer needs to be treated as if it were as hazardous as the most congested regions. Furthermore, since the explosion is generated only by the congested region, the volume of the explosion source is related not to the whole flammable cloud, but only to the volume of the congested region. This means that the overpressures experienced at a distance are lower than they would be if derived from a larger dispersing cloud. An explosion in a module is confined by the walls, which restrict the escape of the extra volume of gas generated by combustion. Such burning in a completely enclosed volume will produce an overpressure of about 8 bars because of this volume generation. Even if there are vents in the walls (partial confinement), high pressures may still be generated because the gas cannot flow fast enough through the vents to relieve the pressure. Explosions in areas which are more than 60% enclosed by a combination of walls or other obstructions may be treated as confined, vented explosions. With such partial confinement, the potential explosion overpressure is increased if the area available for venting is small, if the obstructions to the flow provide large blockage, or if the path from a possible ignition location to the main venting involves passing many obstacles. Another feature of such incidents is the "external explosion" (Harrison et al28). As the flame burns through a vented enclosure, most of the gas is pushed out of the vent ahead of the flame. This is a consequence of the 8:1 volume expansion. Outside the vent it forms a mushroom-shaped, highly turbulent "starting jet". When the flame emerges through the vent, it burns rapidly through this gas, creating an external explosion. The external explosion may cause damage at some distance from its source, and also has influence on the pressure inside the enclosure, so that the maximum internal pressure is usually reached after the external explosion.

SPE 69679

NEW TECHNIQUES FOR RISK AND HAZARD MANAGEMENT IN THE PETROCHEMICAL INDUSTRY

Explosion Modelling. Various methods which have been used for the prediction of overpressures associated with vapour cloud explosions have been discussed29 . Historically the first method, tried and used for many years, was the analogy with TNT explosions. However, it is important to emphasise that the TNT equivalence method is not appropriate for vapour cloud explosions since it does not correctly model the source and decay characteristics. The "congestion assessment method" 30,31 is one example of the new approaches currently being adopted for gas explosions in unconfined, congested regions. For confined, vented explosions essentially two types of models are used: 1. CFD models have the advantage that they can represent the detailed geometry of plant or offshore module. However, such models require long run times (tens of hours) on large computers and, as yet, they are not necessarily more accurate than simpler models. There are several CFD explosion models in existence; examples are EXSIM from Telemark, Norway, FLACS from Christian Michelsen Research, Norway (Van Wingerden et al32 ), and AutoReagas from TNO in the Netherlands. 2. The second category comprises models which idealise the geometry in order to simplify and speed up the calculations, but which are based on an understanding of the physics underlying the various processes involved in the explosion. Because such "phenomenological" models are based on physical understanding, they are better able to extrapolate outside of the range of existing data than earlier correlation-based models. A prime example is SCOPE (Shell Code for Overpressure Prediction in gas Explosions) by Puttock et al33,34 . CFD models of deflagration explosions have been used to assess potential hazard consequences for many offshore platform modules. CFD modelling enables the specification of the maximum overpressure for which blast walls are required to maintain their integrity and the potential size and magnitude of the blast wind flowing around vessel and pipework supports. The CFD model can predict the reduction in overpressure when a module is only partially filled with gas or the effect of different ignition locations. The model can also be used to study design modifications to assess whether they are likely to reduce potential explosion overpressures. An estimate of the statistical uncertainty in the predictions of major hazards is desirable in order to combat the risks effectively. Accordingly, the predictions of SCOPE were compared with the results from a full set of 308 experiments ranging from medium scale laboratory studies through to the recent full scale tests in the BFETS35 . It was found that overpressures in only 3.2% of cases would be under-predicted by more than a factor of two. For the full scale tests alone all the predictions are within a factor of 1.5. When the same exercise was repeated for the EXSIM model for blind predictions alone (8 tests), there was 95% confidence that the median values of peak overpressure are within a factor of 1.5. Furthermore, the distribution of overpressures, pulse

shape and peak durations at each pressure sensor, approximately 30 in each test, are also well predicted, giving confidence that the essential physics has been correctly modelled. Traditionally, the available science only allowed a "worst case" explosion to be calculated. The genuine worst case is an event of extremely low probability if not totally incredible. Clearly a more realistic approach is justified in order to demonstrate that the risks have been reduced to as low as reasonably practicable. Explosion exceedance analysis36 is a powerful means of displaying the risk of gas explosions. It is designed to take into account all possible accidental gas release scenarios and examine each for its explosion potential. A combination of gas dispersion calculations (to model the gas cloud sizes), and Monte Carlo technique (to sample the distributions of clouds, ignition locations, and explosion modelling uncertainty) is used. A typical analysis can take as many as 10,000 explosion simulations. This renders computer intensive techniques such computational fluid dynamics for dispersion and explosion simulation unsuitable. We have developed a "random walk" method for dispersion and a physical model for gas explosions (SCOPE) which can run in seconds and are fully validated by all large scale data. Uncertainties in the overall methodology are managed by making appropriate conservatisms in the methodology. The results are shown as curves for the frequency of exceeding a given overpressure or impulse against that overpressure or impulse. The design strength for blast protection or the risk of protection failure for an as-built construction are simply read from the curve. The methodology has already saved Shell Companies many 10s of millions of $s in cost effective safe design and has been used to demonstrate compliance with local safety criteria. Application to Plant Safety Many of the fire, BLEVE, and explosion consequences of accidental releases already described can be readily calculated in proprietary software packages such as FRED3. The validation and scientific basis of the models is given in external publications some of which are referenced here. FRED allows individual hazard scenarios to be assessed. If the results are combined with the frequency of event then a risk picture can be developed. This is illustrated in Fig. 5a. Here the total individual risk contours have been built up from contributions of cloud fires, jet fires, pool fires, BLEVES, toxic gas and explosions for a gas plant in the design stage. The risk to workers on the plant and neighbouring facility, and to the local society are clearly displayed as risk contours. Also the risk hot spots can be readily identified for further analysis using HEMP. Risk reduction measures were then evaluated and the resulting societal risk picture of Fig. 5b was developed. These safety measures included: location of the site to minimise impact on residential areas, traffic flows, and neighbouring industry.

G.A. CHAMBERLAIN

SPE 69679

separation of process trains to minimise risk of cross escalation. optimisation of fire and gas detection using maps of flammable gas clouds. appropriate emergency shutdown and blowdown to minimise releasable inventory in the event of flame impingement, and optimisation of emergency shutdown valve locations. reduction of inventories of flammable gas and liquids. reduction of ignition probability by using waste heat from gas turbines for the heat transfer fluids rather than using direct fired heaters. relocating the boundary fence and resident populations. In another study the explosion exceedance methodology was used to demonstrate ALARP status of a control building on a refinery. Assume the building has been designed to withstand 0.4 bar static blast pressure. A worst case analysis predicts that the building could be exposed to an overpressure of 0.73 bar. This seems to suggest that the building is unsafe. The exceedance analysis however calculated the risk of only 4 x 10-5 per year of exceeding 0.4 bar (Fig 6) and when allowance is made for the blast impulse the risk is reduced by nearly two orders of magnitude. This suggests that the risk associated with existing design is ALARP. Concluding remarks This paper cannot do justice to the rapid advances of recent years in the assessment of the combustion hazards posed by accidental releases of hydrocarbons. The interested reader is strongly urged to read the references cited herein. However many validated methods are now available to the engineer for cost effective safe design and research continues to address remaining areas of uncertainty, such as explosion prediction for example, in aerosols, effect of many small scale obstacles, effect of initial turbulence, ignition probability, gas clouds partially filling congested plant, linkage to response calculations, chemistry of compartment fires, general aspects of turbulent combustion for example, for mixtures of gases, to improve the formulations for sub grid obstacles in CFD models, water droplet/flame interactions for fire and explosion mitigation, integration of hazard consequence modelling with event probability in a consistent and demonstrable way by the use of computer tools. Acknowledgements The author is indebted to the members of the Hazard and Risk Management Consultancy at Shell Global Solutions UK whose continuing efforts make significant impacts to knowledge in this field.

References 1. Bowen, P.J., and Shirvill, L.C.: "Combustion hazards posed by the pressurised atomisation of high-flashpoint liquids", J.Loss Prev. Process Ind., (1994) 7, No. 3. 2. Shield, S.R.: "A model to predict the radiant heat transfer and blast hazards from LPG BLEVEs", AIChE Symposium Series, (1993) Vol. 89. 3. Shell FRED version 3.1, Fire, Release, Explosion, Dispersion, FREDHELP@opc.Shell.com. 4. Chamberlain, G.A.: "An Experimental study of water deluge on compartment fires", International Conference on Modelling and Mitigating the Consequences of Accidental Releases of Hazardous Materials, AIChE, New Orleans (1995). 5. Chamberlain, G.A.: Evaluating offshore fires and explosions, Proceedings of the International Conference on Changing Health and Safety Offshore - The agenda for the next 10 years, Aberdeen (July 22-24 1998) UK Health and Safety Executive. 6. Kalghatgi, G.T.: "Blow-out stability of gaseous jet diffusion flames. Part 1: Still air", Comb. Sci. and Tech., (1981) 26, 241. 7. Birch, A.D., Brown, D.R., Cook, D.K., and Hargrave, G.K.: "Flame stability in under expanded natural gas jets", Comb. Sci. and Tech., (1988) 58, 267. 8. McCaffrey, B.J., and Evans, D.D.: "Very large methane jet diffusion flames", 21st Symposium (International) on Combustion, The Combustion Institute (1986). 9. Davenport, N: "Large scale natural gas/kerosene mixed fuel jet fires. Final report to the American Petroleum Inst.", (1994) Shell Research report no. TNER.94.061. 10. Acton, M.R. and Evans, J.A.: Horizontal jet fires of oil and gas, report to the Blast and Fire Engineering Project for Topside Structures Phase II, Steel Construction Institute (1997). 11. Cracknell, R.F. and Chamberlain, G.A.: Interpretation of experimental data for unconfined crude oil jet fires, report to the Blast and Fire Engineering Project for Topside Structures Phase II, Steel Construction Institute (1997). 12. Chamberlain, G.A.: "Developments in design methods for predicting thermal radiation from flares", Chem.Eng.Res.Des. (1987) 65, 299. 13. Johnson, A.D., Brightwell, H.M., and Carsley, A.J.: "A model for predicting the thermal radiation hazard from large scale horizontally released natural gas jet fires", Trans. IChemE., (1994) 72, Part B. 14. Johnson, A.D., Ebbinghaus, A., Imanari, T., Lennon, S.P., and Marie, N.: "Large-scale free and impinging turbulent jet flames - numerical modelling and experiments", IChemE Symp. Series 141, Hazards XIII, UMIST, Manchester (1997). 15. Cowley, L.T., and Pritchard, M.J.: "Large scale natural gas and LPG jet fires and thermal impact on structures", Gastech 90, 14th International LNG/LPG Conference, Amsterdam (1990).

SPE 69679

NEW TECHNIQUES FOR RISK AND HAZARD MANAGEMENT IN THE PETROCHEMICAL INDUSTRY

16. Cracknell, R.F., Davenport, N., and Carsley, A.J.: "A model for heat flux on a cylindrical target due to the impingement of a large-scale natural gas flame", 2nd European Conf. on Major Hazards Onshore and Offshore, IChemE, (Oct 1995). 17. Sekulin, A.J. and Acton, M.R.: "Large scale experiments to study horizontal jet fires of mixtures of natural gas and butane", British Gas report to CEC, GRC R0367 (1995). 18. Davenport, N.: "Large scale natural gas/butane mixed fuel jet fires. Final report to the European Commission", (1994) Shell Research report no. TNER.94.030. 19. Considine, M.: "Thermal radiation hazard ranges from large hydrocarbon pool fires", Safety and Reliability Directorate Report SRD R297, UKAEA (1984). 20. Pritchard, M.J. and Binding, T.M.: "FIRE2: A new approach for predicting thermal radiation levels from hydrocarbon pool fires", IChemE Symposium Series No. 130 (1992). 21. Johnson, A.D.: A model for predicting thermal radiation hazards from large scale LNG pool fires, IChemE Symposium Series No. 130, (1992) 507-524. 22. Wighus, R., Meland, O., and Vembe, B.: "Smoke hazard in offshore platform fires", SINTEF Report STF25 A91007 (1991). 23. Chamberlain, G.A.: "An Experimental study of large scale compartment fires", Trans. IChemE, (1994) 72, Part B, 211. 24. Chamberlain, G.A.: "The hazards posed by pool fires in offshore platforms", Trans. IChemE., (1996) 74, Part B, 81. 25. Persaud, M.A., Chamberlain, G.A., and Cuinier, C.: "A model for predicting the hazards from large scale compartment jet fires", IChemE Symp. Series 141, Hazards XIII, UMIST, Manchester (1997). 26. Chamberlain, G.A., Persaud, M.A., Wighus, R., and Drangsholt, G.: Blast and Fire Engineering for Topside Structures. Test Programme F3, confined jet and pool fires. Final Report, Steel Construction Institute (1997). 27. Persaud, M.A., Wighus, R., and Chamberlain, G.A.: Blast and Fire Engineering for Topside Structures. Test Programme F3, confined jet and pool fires. Interpretation Report, Steel Construction Institute (1997).

28. Harrison, A.J. and Eyre, J.A.: "External explosions as a result of explosion venting", Comb. Sci. and Tech., (1987) 52. 29. Cates, A.T.: "Fuel gas explosion guidelines", Presented at 1991 Intl. Conf. on Fire and Explosion Hazards, Moretonin-Marsh, Inst. Energy. 30. Puttock, J.S.: "Fuel gas explosion guidelines - The Congestion Assessment Method", 2nd European Conf. on Major Hazards Onshore and Offshore, IChemE, Manchester (1995). 31. Puttock, J.S.: Improvements in guidelines for prediction of vapour cloud explosions, Intl. Conference and Workshop on Modelling and Consequences of Accidental Releases of Hazardous Materials, San Francisco, (Oct, 1999). 32. Van Wingerden, K., Storvik, I., Arntzen, B., Teigland, R., Bakke, J.R., Sand, I.O., and Sorheim, H.R.: "FLACS-93 A new explosion simulator", Presented at 1993 2nd International Conference and Exhibition on Offshore Structural Design Against Extreme loads, ERA Report 930843. 33. Puttock, J.S., Cresswell, T.M., Marks, P.R., Samuels, B., and Prothero, A.: "Explosion assessment in confined vented geometries. SOLVEX large-scale explosion tests and SCOPE model development", Shell Research report to UKHSE, OTO 96 004 (1995). 34. Puttock, J.S., Yardley, M.R., and Cresswell, T.M.: Prediction of vapour cloud explosions using the SCOPE model, International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions, Schaumburg, Illinois (Sept 1998). 35. Johnson, D.M., Shale, G.A., Lowesmith, B.J. and Campbell, D.: Blast and Fire Engineering for Topside Structures, Phase 2: Final Report on the explosion test programme, Steel Construction Institute (1997). 36. Puttock, J.S.: The use of a combination of explosion models in explicit overpressure exceedance calculations for an offshore platform, International Conference on Major Hazards Offshore - Practical Safety Implications, ERA Technology, London, 27-28 (Nov 2000).

G.A. CHAMBERLAIN

SPE 69679

A ct i vi t y under assessm ent I dent i f y pot ent i alhazardous event s

Feed Back

A nal yse Ri sk f rom H azardous event s

l i kel i hood per event /of exposure /consequences

D et erm i neO veral Ri sk t l o Peopl e( A LA RP) , A sset s( C ost ) and Envi ronm ent ( I m pact ) Ri sk Reduct i oni on sel ect i on / - opt change deci si ons/em ergency response
F ig . 1 -H a za rd a n d E ffect M a n a g em en t P ro cess (H E M P )

Release

Yes

Ignites?

No

Dispersing cloud

Yes

Ignites?

No

More obstacles Greater confinement Flame acceleration

Jet fire

Pool fire

Cloud fire

Fast flame

Internal explosion

Safe dispersion

Fig. 2-Event tree showing all consequences of flammable release.

SPE 69679

NEW TECHNIQUES FOR RISK AND HAZARD MANAGEMENT IN THE PETROCHEMICAL INDUSTRY

Buoyancy dominates flow Cold soot escapes End of combustion Thermal plume and smoke Ambient flow Radiative heat transfer

Flame lift-off point

Flame Impingement Source Jet expands via series of shock waves Air entrainment without combustion Turbulent combustion Large eddies entrain air Combustion occurs rapidly in thin sheets flamelets Flame initially blue (no soot) Amount of soot increases along flame

Boundary layer Convective heat transfer Radiative heat transfer

Fig. 3- Schematic of an impinging jet fire

Flame

Turbulence

Volume Production

Flow Blast

Fig. 4- The Shchelkin mechanism of flame acceleration.

10

G.A. CHAMBERLAIN

SPE 69679

Fig. 5a-Risk contours generated for all the flammable and toxic risks from a gas plant. The plant is located centrally with the import gas pipeline entering from the left. Areas of population are shown cross-hatched.

0.01 0.001 0.0001 Frequency per year 1E-05 1E-06 1E-07 1E-08 1E-09 1E-10 1 10 100 Number of fatalities 1000 10000 Current design Original design

Fig. 5b-F(n) curve (societal risk) for the current and original design illustrating the drive towards ALARP status by adopting the measures discussed in the text.

SPE 69679

NEW TECHNIQUES FOR RISK AND HAZARD MANAGEMENT IN THE PETROCHEMICAL INDUSTRY

11

2.50E-02 2.00E-02 Free field Reflected

Frequency /yr
1.50E-02 1.00E-02

5.00E-03

0.00E+00 0 100 200 300 400

Overpressure, mbar
Fig. 6-Explosion exceedance plot for free-field and reflected overpressure at a distillation control building on a refinery. The source of potential overpressures arise from explosions of gas clouds in the vicinity of crude distillers and their treating unit and pump-house sections of plant.