ML20073H356

Provides Addl Info Re Natural Gas Collection Pipelines in Vicinity of Facility,Per 910429 Request

ML20073H356
Person / Time
Site:	Fort Saint Vrain
Issue date:	05/03/1991
From:	Crawford A PUBLIC SERVICE CO. OF COLORADO
To:	Weiss S NRC OFFICE OF INFORMATION RESOURCES MANAGEMENT (IRM)
References
P-91152, NUDOCS 9105070082
Download: ML20073H356 (54)

Text

{{#Wiki_filter:, J- ^ i; p 9 eueiic service-

=":: =.-

l P.O. Box 840 cene co so20m 2420 W. 26th Avenue, Suite 1000, Denver, Colorado 80211 A. Clegg Crawford Ncie ONations May 3, 1991 Fort St. Vrain Unit No. 1 P-91152 U. S. Nuclear Regulatory Commission ATTN: Document Control Desk Washington, D.C. 20555 Attention: Dr. Seymour H. Weiss, Directur Non-Power Reactor, Decommissioning and Environmental Project Directorate Docket No. 50-267

SUBJECT:

Natural Gas Collection Pipelines in the Vicinity of Fort St. Vrain

REFERENCES:

1) PSC Letter, Crawford to Weiss, dated March 27, 1991 (P-91111)

2) PSC Letter, Crawford to Weiss, dated April 23, 1991 (P-91139)

Dear Mr. Weiss:

In a phone conversation on April 29, 1991 between Mr. Richard Dudley, Jr. of your staff and PSC's Mr. Michael Holmes, the NRC requested additional information concerning the natural gas issue at Fort St. Vrain (FSV), primarily related to References 1 and 2. The purpose of this letter is to provide the NRC with the requested information. includes the NRC questions and PSC's response to each question. Attachment 2 is a document entitled " Natural Gas Explosion Concern at FSV." This document was prepared by a consultant (Mr. Stanley Martin) whose services PSC acquired upon the recommendation of the U.S. -Gas Research Institute (GRI) and the Federal Emergency Management Agency (FEMA), who identified Mr. Martin as a leading expert in the area of deflagrations and detonations of unconfined gas vapor clouds, and the structural effects upon target buildings of the deflagrations and detonations. PSC has found Mr. Martin's extens%e knowledge in this field to be enlightening and considers tnat the information in Attachment 2 will prove helpful in resolving the natural gas issue at FSV. Attachment 2 also includes a number of reference documents associated with Mr. Martin's evaluation. 9105070002 910503 PDR ADOCK 05000267 l l P PDR j m aas M I,

P-91152 Page 2 May 3, 1991 is Mr. Martin's resume, documenting his qualifications in this particular field of knowledge, g Should you have any questions concerning this submittal, please contact Mr. M. H. Holmes at (303) 480-6960. Very truly yours, 8'd g A. Clegg Crawford E_ Vice President L Li_ clear Operatione ACC/JRJ: bit Attachments cc: Regi nal Administrator, Region IV g Mr. J. B. Baird Senior Resident inspector Fort St. Vrain Mr. Roberc M. Quillin, Director Radiation Control Division Calorado Department of Health 4210 East lith Avenue Denver, Cn 80220 Mr. Steve Ruffin Priject Manager Irradiated Fuel Section Fuel Cycle Safety Branch Office of Nuc'. ear Material Safety and Safeguards

P-91152 Page 1 May 3, 1991 Responses to NRC Phone Questions As stated in PSC's letter dated April 23, 1991, NRC Question concerning this natural gas issue, an overpressure of 0.45 psi at the Reactor Building was calculated to result from postulated detonation of natural gas assumed to be released from rupture of the 4 inch line south of the FSV facility. By doubling this, the NRC arrives at an overpressure of 0.90 psi at the Reactor Building. The question is, what would happen to the blowoff panels, located above the fuel deck, in the event of such an overpressure condition at the Reactor Building? PSC Response - As discussed in Reference 2 of the cover letter, PSC takes exception to the NRC's doubling the calculated overpressure at the Reactor Building. This " doubling" only applies to actual detonations which produce shock waves with essentially instantaneous rise times to account for the effects of reflected pressure from a wall normal to the incident pressure wave. It is not applicable to generated from deflagration of an unconfined natural pressure waves gas vapor cloud. This is further discussed in Attachment 2, a document prepared for PSC by Mr. Stan Martin, an expert in vapor cloud detonations and deflagrations and associated ef fects on target structures, referred to PSC by the U.S. Gas Research Institute (GRI) and the Federal Emergency Management Agency (FEMA). In Attachment 2, Mr. Martin provides justification for his contention (and that of virtually all other experts contacted by PSC on this subject) that detonation of an unconfined natural gas vapor cloud will not occur, and discusses effects of deflagration of such a cloud, which is a credible occurrence. The blowoff panels respond to dynamic pressures, such as would be generated by winds, which give rise to drag forces. As noted in Attachment 2, winds generated by deflagration of an unconfined natural gas vapor cloud could reach speeds equal to the combustion flame propagation speed through the flammable portion of the cloud. Assuming highly turbulent conditions (much higher than would be generated by the assumed 1 meter per second wind speed used to reduce dispersion and maximize the amount of natural gas within a flammable concentration), a combustian flame propagation sp9ed of 30 meters per second would be possible. Thus, wind speeds at the portion of the natural gas cloud at the lower flammable limit (LFL) could reach 30 meters per second (67 miles per hour). Such a wind speed would have no impact on the Reactor Building, or its blowof f panels, designed to blow off at wind speeds greater than 202 mph. If a detonation were hypothesized to occur, which produced an overpressure shock wave at the Reactor Building of 1.0 psi, the blowoff panels would not be expected to break away from the wind girts which support them. This is demonstrated by the following assessment: The relationship of dynamic pressure (q) to peak side-on overpressure (P) in an ideal air blast is q= S/2 (P-squared) l (7 times the atmospheric press.) + P

P-91152 Page 2 May 3, 1991 This is the Rankine - Hugoniot equation for a shock wave, and can be found on page 97 of Glasstone and Dolan, 1977. There is a thumb rule associated with this equation, relating particle velocities (or wind speeds) to peak side-on overpressure in a shock wave, which is used for peak side-on overpressures less than about 10 psi: a 1 psi peak side-on overpressure equates to a 50 feet per second particle velocity, and for each additional psi of peak side-on overpressure, an additional 50 feet per second particle velocity is added. NRC Regulatory Guide 1.91, " Evaluation of Explosions Postulated to Occur January on Transportation Routes Neat Muclear Power Plant Sites" 1975 presents the following equation for relating dynamic pressures (q) to wind speeds (V): q = 0.002558 (V-squared) Where q is the dynamic pressure in pounds per square foot and V is the maximum wind velocity in miles per hour. Based on the Rankine Hugoniot equation and the above equation relating dynamic pressure to wind speed, the actual dynamic pressure can be calculated and the equivalent wind speed can be determined for a given peak side-on overpressure generated from a hypothetical detonation. For a 1.0 psi peak' side-on overpressure generated from a detonation, the corresponaing dynamic pressure would be 0.029 psi. This corresponds to the dynamic stagnation pressure developed by a wind speed of about 40 miles per hour. In order to achieve the dynamic pressure and drag forces associated with a 202 mph wind, and begin to break away blowoff panels, a peak side-on overpressure of approximately 5 psi would have to occur at the Reactor Building. For overpressure of 1.0 psi, or less, PSC considers that the Reactor Building structure would not he damaged. This is consistent with conclusions reached by the NRC in Regulatory Guide 1.91, Rev. 1 (For Comment), dated February 1978, which states "A method for establishing the distances referred to above can be based on a level of peak positive incident overpressure... below which no significant damage would be expected. It is the judgement of the NRC staff that, for the structures, systems, and components of concern, this level can be conservatively chcsen at 1 psi." For the maximum 0.45 psi overpressure which was calculated to occur at the Reactor Building in Reference 1, and for double this overpressure, PSC concludes that the Reactor Building blowoff panels would remain in place. NRC Ouestion - If it is assumed the Reactor Building blowoff panels above the fuel deck were torn off, what effect would this have on the capability to remove decay heat from the fuel in the PCRV and in the fuel storage wells? I

P-91152 Page 3 May 3, 1991 PSC Response - The method for decay heat removal of fuel in the core and fuel storage wells, relied upon in the FSV Defueling Safety Analysis Report (incorporated into Sections 3.11 and 14.14 of the Updated FSAR) and Section 4.3 of the FSV Fire Protection Program Plan (FPPP), is one loop of liner cooling. Two redundant liner cooling loops supply both the PCRV and the fuel storage wells. Water can be supplied through these liner cooling coils by System 46 pumps (2 pumps per loop) in the recirculate mode, or by a firewater pump in the once-through mode. For aither mode of cooling, the equipment relied upon is classified Safe Shutdown and protected so as to withstand the effects of the Maximum Tornado with wind speeds as high as 300 mph (FSAR Sections 1.4 and 14.1.2). FSAR Section 14.1.2 states that "At the 300 mph wind speed, the siding on the auxiliary and confinement buildings, above the refueling floor level, may be carried away, but the basic building structure will not collapse. The confinement and auxiliary buildings are not required for safe shutdown. However, equipment and systems essential to safe shutdown which are located above the refueling floor are protected from tornado missiles by redundancy of the components, with sufficient separation and missile shielding such that a single missile could not involve both components. Equipment and systems essential to safe shutdown which are located below the refueling floor level are protected from tornado missiles by special heavy steel siding on the building." The PCRV liner cooling water surge tanks and associated piping are located above the refueling floor (" fuel deck"). A concrete missile shield separates these two tanks. Even if both tanks were assumed to be ruptured, PCRV liner cooling and fuel storage well liner cooling could be readily established from either one of the two firewater pumps. Thus, loss of the blowoff panels would not jeopardize the capability to remove decay heat from the fuel, even in the event of 300 mph winds passing through the top of the Reactor Building. In a previous phone call, PSC stated that cooling NRC Ouestion could be secured to the fuel for about three weeks, without creating unsafe conditions. Could PSC document the basis for this statement? PSC Resoonse - Decay heat generation from the fuel is extremely low, due to the permanent reactor shutdown on August 18, 1991. FSV Technical Specification 4.0.4 defines the method used to compute the CALCULATE) BULK CORE TEMPERATURE (CBCT). The basis for this specifica tion explains that this calculation assumes all decay heat power gererated is retained in the active core, with no heat transfer to the reflector, PCRV internals or primary coolant. Specification 4.0.4 aces not permit the CBCT to exceed 760 degrees F, which corr'_sponds to the design maximum core inlet temperature. This limit assures there can be no damage to fuel or PCRV internal components, even in the absence of forced circulation of primary coolant helium.

P-91152 Page 4 May 3, 1991 Based on current decay heat generation, the CBCT would not reach 760 degrees F if all cooling ceased for over 21 days. This is based on the assumption of adiabatic heatup of the fuel. If PSC were to perform an analysis which modeled heat transfer out of the fuel, it is quite likely that it could be shown that forced circulation and PCRV liner cooling could be secured indefinitely and would never need to be restored to prevent unacceptably high fuel temperatures. Due to the substantial heat sink designed in to the fuel storage wells (discussed in FSAR Section 9.1.2), fuel in these wells could easily withstand loss of all cooling for times beyond 21 days without 3 reaching the 750 degree F limit discussed in the basis for LC0 4.7.3, M " Fuel Storage Wells." NRC Question - In PSC's letter dated March 27, 1991, concerning this natural gas issue, a contradiction was noted at the bottom of page 2 of Attachment 2. PSC states the " shut-in valve" located at the wellhead automatically actuates to isolate the producer pipe at the wellhead, in the event of high or low casing pressures. But the next sentences state that pressure switches are set to open the shut-in valves at a pressure of approximately 350 psig and close the shut-in valves at a pressure of approximately 170 psig, for all ten wells that _ feed the gas collection system in the vicinity of FSV. Can PSC clear this up? There was a mistake in the first sentence the NRC PSC Response mentioned. The shut-in valves are automatically opened on high pressure and close on low pressure. They do not isolate the producer pipe at the wellhead on high pressure, as stated, only on low pressure. PSC apologizes for any misunderstanding caused by this mistake. Since the 6 inch manual isolation valve was closed, NRC Question which connects the 16 inch line to the 6 inch line, and flow between these two lines is through the 1.5 inch diameter line that bypasses this valve (as discussed in PSC's letter dated March 27,

1991, concerning this natural gas issue), has there been a significant pressure increase in the natural gas collection system in the vicinity of FSV?

PSC Response - Based upon data submitted to PSC by Panhandle Eastern, the owner of this natural gas collection pipeline system, pressures increased from an average pressure of about 130 psig to an average pressure of about 155 psig as the result of closing the 6 inch manual isolation valve and restricting flow to the 1.5 inch bypass line. NRC Question - What is the design pressure and test pressure of the 6 inch and 4 inch diameter piping in this collection system? PSC Response - Per discussions with personnel from Panhandle Eastern, all of the 4 inch and 6 inch piping in question has a design pressure of 720 psig and was hydrostatically tested to 1080 psig, one and one-half times the design pressure.

~. - ~ _ - - - -. - - P-911523 . Attachment 1-Page 5' May3. 1991' NRC Question - In PSC's letter dated March 27, 1991, concerning this natural gas issue, PSC stated-that the 6 inch manual isolation valve (discussedJ above) would be closed, except during maintenance 1or-surveillance activities.when Panhandle-Eastern would have the valve '" continuously manned _by an operator. who has been instructed to promptly close the valve in the event a pipeline rupture is observed or suspected." _ How will this operator know a pipeline rupture has occurred at either Break 1 or Break-2' locations? PSC Response - An operator stationed at the six inch manual isolation . l valve would be able tol hear and see the' effects of a large pipe rupture,_ such Las was postulated to occur at the Break 1 and Break 2-locations discussed in References I and 2 of the cover letter. The Break 1 ? location is approximately 3900 feet from the 6 inch manual isolation valveLand the Break 2 location is approximately 4800 feet from the 6 inch -manual isolation valve. Gas leaving the_ pipe at sonic _ velocity would. set up a shock wave that could be heard at much greater distances. The noise produced by gas escaping from a large rupture would-be comparable to that produced by a jet engine. The terrain between.the postulated rupture locations and the 6 inch manual-isolation valve -is :quite flat, without many trees or-structures which could interfere with and significantly attenuate the. sound. In addition-to'the sound generated from a large pipe rupture, the_ gas escaping at high pressure from a buried pipe would. create

a. large plume of dust which would be visible from the'6 inch manual _ isolation valve due-to the fact that Break 1 and Break-2 locations-are within the _-line of sight':of _a person stationed at the 6 inch manual isolation valve. Were,the natural-gas to. ignite, as the : result -of sparks at the rupture location, it would flare, which would;also be visible to a person stationed at the 6 inch manual isolation valve.

In. addition, Git is probable that the. noise' produced at the 6 inch 4 manual, isolation valve itself,ldue to :a sudden flow reversal-and-drastic. flow increase, would be sufficient to' alert the; operator-stationed at' the 6 inch manual isolation valve.

However, there is insufficient ; experience and-no1 means of~ testing this to arrive.at>

~ 'this conclusion withla high' degree-of. certainty. Based ~.upon -hearing a pipe rupture and/or.seeing the: effects of the

p'ipe rupture,-experienced per'sonnel at' Panhandle Eastern and Western-Gas' Supply Company 1(PSC's' natural gas subsidiary)Lare confident that an operator stationed at the 6

inch ' manual' isolation valve would rapidly '. become awareiof a ~ 1arge pipelin'e rupture near either,the

' Break;l or' Break 12 location.- 't r rn ~ > w 4-,e- + ww- -,2 -,,______________________.___._________.___._.__m. mm

,P-91152 l May 3, 1991 NATURAL GAS EXPLOSION CONCERN AT FSV Introductory Remarks By far, the preponderant majority of cases of acciden-tal or experimertal, unconfined vapor-cloud explosions have failed to produce significant airblast effects. Notable exceptions--the Flixborough and Port Hudson events are about the only examples--are often held up to those who would dismiss the prospect lightly, as a reminder that experience shows that the threat of structural damage and/or personal injury / death cannot be ruled out. These events were spe-cial, however, because the first involved a massive release of cyclohexane, while the second was propano, and may have involved detonation initiated within a strong walled build-ing. These two examples are therefore not directly appro-priate to the circumstances of current consideration, that is, a natural gas release on the open prairie, in the ab-sense of enclosures where strong shocks might be generated (PSC letters to S.H. Weiss, dated 27 March and 23 April 1991). Both experience and theory-based analyses says unconfined natural-gas clouds in air (lacking some peculiar, extremo, and very special conditions) are not a blant dam-age / injury threat. The Concent of TNT Eouivalence For years, attempts have been made to express the hazard of such uncondensed explosives as fuel-gas / air and fuel-gas / oxygen mixtures in terms of the energy release (or yield) equivalent to some standard condensed-phase explosive such as TNT. The concept has some merit, as long as the state of a uniform mixture of the reacting gases is well known, and the explosion-to-target distance is large com-pared to the size of the cloud (i.e., the target is suffi- -ciently remote that the extended source approximates a point source.) This is usually not the case, nor of practical interest, in accidental gaseous releases into the atmos-phere. The amount of gas that may participate in the com-bustion reaction, soon enough to support a pressure wave, is commonly much less than any calculation of the time-averaaed lower-flammability-limit envelope would suggest. Any esti-mate of yield is, therefore, very sensitive to variations in cloud geometry and-eddy fine-structure within the cloud, which in turn depends on wind-speed profiles, gustiness, terrain roughness, etc., that are hardly knowable. Fore-casts are quite beyond doing, with any confidence, at the present stage of development of the technology. Another very important source of variability in gas-J phase explosions is the kind of explosion: 1. deflagration; i

l 2. detonationt 3. thermal explosion. In the order given, the explosion energy yields increase from a small fraction of that potentially available (order of 0.01) to a large fraction (approaching 100%). For natural-gas / air, we can eliminate any prospect of a thermal explosion, because alkai.e hydrocarbons are stable in air at ambient tempera-tures; and detonations can occur only in very special cir-cumstances, such as in channels with high Reynolds' Numbers or in enclosures with strong walls. In pure methane / air, detonations can be ruled out. Higher hydrocarbons are more subject to detonation (more on this later), unsaturated hydrocarbons,.even more, and certain chemical additions to the hydrocarbon structure increase this tendency dramatical-ly. This explains, in part, the differences in experience-based energy yields of explosive vapors and gases. Refer to the table of yield factors (Table B.3) used with FEMA's ARCHIE Code. Alkane hydrocarbons are in the group having Y=V.03. This would be appropriate for deflagration of natural gas / air. The higher yields are for others, nitrated paraffins, olefins, and acetylenics. This is entirely consistent with the findings of Brasie and Simpson(1968). In the incidents they surveyed, including a wide range of hydrocarbon / air explosions, yield values were commonly less than 0.04 Deflagrations having yields as high as 0.1 require mixtures containing the more reactive components, listed above, or special circumstances. Analysts often use this more conservative value to cover any and all eventuali-ties, to be way over on the safe side. It would seem, however, inappropriate to do so here, unless it can be shown that detonations are both possible and probable. This issue is addressed in the following section. The Possibility of Detonation, There is consideracle, quality evidence that detona-tion cannot be directly initiated in a natural-gas / air mixture with anything less than heroic efforts (e.g., kilo-grams of high explosive or a strong shock emanating from a pipet see MIT-GRI LNG Workshop, 1982). Further, there is considerable doubt that, once initiated, a detonation could sustain itself in practical situations of nonuniform mix-ing. That may have happened at Port Hudson because of special circumstances, but alternative explanations for the damage have been offered. It must be remembered that the Port Hudson event was fueled by propane not natural gas. The only reason for not categorically denying the possibility of detonation at the FSV site is the presence of higher hydrocarbans in natural gases. (Refer to FSV gas composition in PSC letter, dated 22 February 1991.) Concen-trations greater than 15% (by volume) of ethane and/or pro-pane increase the sensitivity to initiation of automation. Results of all the tests conductec to the 1982 date of the MIT-GRI 4orkshop, to study the detonability of unconfined 1 vapor clouds, indicated the follnwing: 2

l e No detonations are possible in an unconfined stoichio-metric mixture of pure methane vapor and air, even under the initiating influence of 2 kg explosive Sustained detonations are possible in unconfined vapor o clouds containing stoichiometric fuel air mix.sres when the fuel vapor contains methane and propane in the molar ratios of 60/40, 70/30, and 85/15. Above 85% (molar) methane concentrations, no sustained detona-tions have been observed. e Propagating Jetonations (from a pipe) have been sus-tained in unconfined mixtures of methane / propane / air (stoichiometric) for methane volume fractions of less than 85% 1.i the fuel vapor. It is evident from these results that relatively small fractions of heavier hydrocarbons, such as ethane and pro-pane, in the tuel vapor increase the propensity for the detonation of unconfined vapor / air mixtures. This phenome-non is of great importance in evaluating the detonability consequences of natural-gas releases in the form of late-time boil-off from LNG, but usually does not apply to re-leases from pipelines, because gaseous releases are usually not rich in higher hydrocarbons. Tlie sensitivity of the natural gas at the FSV site to the initiation of detonation can be gauged by referring to attached Figure 7 from Bull and Martin (1977), which shows the mass of tetryl required to detonate binary methane / ethane mixtures. A 10-to-20% ethane mixture still needs a kilogram or more of high explo-sive to detonate. As a practical matter, as I understand the circum-stances of the possible release of natural gas at the FSV site, I conclude that the likelyhood of a detonation is essentially zero. Flame Generated Pressures In view of the large uncertainties in any attempt to establish an explosive-energy yield, an arguably better approach to setting an upper bound on airblast; pressures from unconfined gas-phase explosions is one based on the theoretical development of Kuhl, Kamel, and Oppenheim (1974 1 see attached figure.). A flame propagating through a pre-mixed gas of uniform composition and properties pushes unburned gas ahead of its flame front, causing the pressure to rise. In the free field, the flame propagates in three dimensions as a sphere. The compression is less than it would be if the flame were constrained by a nonyielding surface (e.g., the ground or a wall) to propagate with only two degrees of freedom, and still less than situations in which expansion is constrained to just one dimension. In any case, if the flame has a fixed speed--neither accelerat-ing nor deceleratirg--it will maintain a fixed pressure rise 3

in the unburned gas it pushes ahead. Unless the speed of the flame front is up around 40 m/s, the overpressure just ahead of the flame cannot exceed one atmosphere (14.7 psig). To shock up and (possibly) detonate requires much higher speeds. The required "run-up speed" for natural gas in air is not known, but certainly approaches--and may exceed--Mach 1 (Lee, 1977), that is, about 330 m/s, which is patently unattainable in the open. Typical burning speeds for meth-ane are 7 m/s, in quiescent conditions. Turbulence may increase that by a factor of four (to, say, 30 m/s; see Kanury, 1975, pp. 303,304). Thus, the upper bound of overpressure (for pure meth-ane), at the flame front, would be 0.4 atmospheres (about 6 psig). At the target distance (taken to be 930 ft), this falls to about 0.035 atmospheres, or O.5 psig. It must be noted that, in order to estimate the falloff of pressure with distance, it is again necessary to use an energy yield factor to compute a " scaled distance" or a nondimensional-ized " reduced distance." Here, we have used the ultracon-servative value of 0.1. A calculation using the more real-istic yield of 0.03 would, of course, result in smaller overpressures. In either case, estimates of dynamic (i.e., wind) pressure would be inconsequential. It is important to note that this air blast will not " shock up"--that is, the pressure wave will have a finite rise time, and it is, therefore, inappropriate to use a doubling rule for reflected pressure loading of walls. Burning speeds would need to exceed 230 m/s to develop any significant shocking of the pressure wave (Lee, 1977). The pressure rise time will be comparable to the time taken by the flame to transit the near-stoichiometric (and richer) portion of the cloud. If this were on the order of 100 ft, the rise time could be a second or two long. Structural Loadino and Response Actually, side-on overpressures seem to have little or nothing to do with structural response in this case. I am not familiar with the building or buildings that might be subjected to blast loading, but I suspect it is (or they are) more likely to resgond to drag loading than to diffrac-tion (or crush) loading If the rise time is long compared to diffraction times and clearing times (and for leaky structures, long compared to inside-to-outside pressure equalization times), then there would be little net force acting on a wall except for dynamic (i.e., drag) forces. An air-tight box would be subjected to crusning forces, but such structures are inherently strong--even within the For a fuller development of the mechanisms of airblast loading and responses of structures (and definitions of such terms as drao, diffraction, and.clearino times), refer to Glasstone and Dolan, 1977.) i 4

cloud, no pressures capable of crushing damage would be expected. The pressure-pulse duration (and the accompanying drag-phase flow) is undoubtedly long compared to the natural period (roughly 1/2 sec.) of the Turbine / Reactor Building Complex as described in PSC letter, dated 23 April. Pres-sure durations from uncondensed explosives are 1 to 2 orders of magnitude longer than ideal explosions (see Bodurtha,

1980, p.

108)) therefore, the dynamic pressure pulse can be compared to a wind gust. Accordingly, it is appropriate to consider the wind force of the dynamic pressure wave to be the principal mode for mechanical damage, and to compare its magnitude to the design value. In ideal blast waves with a true shock front, peak dynamic pressures can be appreciable in magnitude compared to peak (side-on) overpressures. For example, a4 psi peak (side-on) overpressure is accompanied by 0.4 psi (60 psf) peak dynamic pressure. That represents a particle velocity of 200 ft/s or 136 mph. A 6-psi overpressure is accompanied by 0.8 psi dynamic pressure, corresponding to a 260 ft/s particle velocity, representing a 177 mph wind gust. A useful rule-of-thumb The particle velocity associated with a 1-psi overpressure is about SO ft/s, and increase (or decreases) in linear proportion to the overpressure.

Thus, a 3OO-mph air blast (440-ft/s particle velocity) accompanies a peak overpressure of 8.0 psi.

Alkane / air deflagrative explosions, by contrast, do not generate shocks. Clearly, therefore, the peak particle velocity driven by the explosion cannot exceed the flame speed driving it. As a result, flame speeds in quiescent, premixed methane / air mixtures would be incapable of driving wind gusts above 16 to 18 mph at the edoe of the exolodino cloud! I have no evidence showing that this conclusion would not apply to the FSV gas as well. Allowing a factor of four for turbulence, the limit would still be a factor of 4 to S below the designed wind force for the Reactor Build-ing, and would rapidly dissipate with distance from the explosion. This is consistent with my previous statement that the independently computed dynamic pressure is inconse-quential. Conclusions These are my basic conclusions: 1. The likelyhood of a detonation is negligibly small, if not Zero. 2. The air blast from a deflagrative explosion would exhibit a weak pressure pulse having a long rise time compared to any characteristic building-response time. Thus, the response would be entirely like a response to natural wind forces. 5

4

3. Responses to side-on overpressures, if any, would be limited to window / door / light-panel breakage or deformation, but

))} (except for glass window breakage) this is expected only from overly conservative estimates of yield.

4. Dynamic pressures would be inconsequential relative to the

-3OO-mph design. S tar.l ey B. Martin 2 May 1991 -s m / References

Allan, D.S.,

and P. Athens, 1968: " Influence of Explosions on Design," CEP Technical Manual, Loss Prevention Vol. 2, AIChE, New York

Bodurtha, F.T.,

1980: " Industrial Explosion Prevention and Protection," McGraw-Hill, New York Brasie. W.C., and D.W. Simpson, 1968: AIChE, Loss Prevention, Vol. 2, pp.91-102

Brode, H.L.,

1955: " Numerical Solutions of Spherical Blast Waves," J. Applied Physics, Rh, 766 Glasstone, S., and P.J. Dolan, 1977: "The Effects of Nucle-ar Weapons," Third Edition, US DoD and ERDA, Gov't Printing Office

Kanury, A.M.,

1975: " Introduction to Combustion Phenomena," Gordon and Breach, New York

Kuhl, A.L.,

M.M. Kamel and A.K. Oppenheim, 1974: " Pressure waves Generated by Steady Flames," 14th (Int'l) Combustion Symposium, The Combustion Institute, Pittsburg

Lee, J.H.,

1977: " Initiation of Gaseous Detonation," Vol. 28, An. Review of Physical Chemistry

Lee, J.H.,

et al., 1977: " Blast Effects from Vapor Cloud I Explosions," paper presented at the 1977 AIChE Loss Proven-6

tion': Symposium, March 1977, Houston, TX .j L'i n d, C.D., 1975: "What Causes Unconfined Vapor Cloud j Explosions?" CEP Technical Manual, Loss Prevention Volume 9, American-Institute of Chemical Engineers, New York

Raj, P.K.

(ed), 1992: "MIT-GRI LNG Safety and Research Workshop, Volume III: LNG Fires - Combustion and Radiation,'" GRI 82/0019.3, Gas Research Institute, Chicago

Strehlow, R.A.,

and W.E. Baker, 1976: "The Characterization and Evaluation of Accidental Explosions," Prog. Energy Combustion Science Z, 27

Stull, D.R.,

1977: " Fundamentals of Fire and Explosion," AIChE Monograph Series, 73, No. 10, American Institute of Chemical Engineers, New York 2 4 a f 1 7 1 s v +v --u-s- ,m-- m -v--

~ r Z D E A L. BLAST V)/rW CALC OLAT/ DOS & & (MCS)co(vdth ySd*Mc4 d{"* E h*N U$ y e 0 Raves 9rou pdf bewee.c, ard fN& 9 bdh N r dec>t ascl)y Allmc N i Athens

• to predict the effects of vapor cloud explosions on structures.

i Brode's solutions can only be used for far field blast effects of vapor ~ i cloud explosions since a vapor cloud is not well represanted by a point source. However, estimated overpressures and durations at a distance j" of several cloud diameters are good enough for most purposes; the uncer-tainty of other factors in estimating explosive yields cau ing greater deviations, i a i Allen and Athens provided plots based on Brode's-solutions to determine j the pe(r/A-a)ppwrre et A p<xk. ere pressur s, the duration of the positive pres-akg.na dyna j J sure phase, and the na u e pressure and dynamic pressure impulses at any distance from the center of explosion. The plots are reproduced in Figures 2 through 4 and can be used to estimate blast wave characteris- ) , tics once the energy yield is known. !la 10 6 g 9,g cfif u a ,Frm c5 A

• h'*

y c2 5 ~ }/ o a .e 90 i @ $ 10_ \\ 6 P-TT-l.[o -l 5 g Z b.8-3 L @Y t a

a. 3 a = Po J

h3 \\ O O IO \\ I.b l ai =7 g l My It i kE _j 4P' u < Ps s 0.1 i i i t i t ti i r i t t ih i if tisi 4f [A ppe V j .01 0.1 'LO 2 y go REDUCED DISTANCE, A

% g 1 - (q cV)/3 c

3 FIGURE 2 /' 8-Peak overpressure and peak dynamic pressure vsrgdgeddistance, s E \\ p )

• Allan, D.S. and Athens, P., " Influence'of Explosions af' Design," CEP Technical Manual, Loss Prevention Vol. 2, AICh ( New York, 1968.

\\ l \\ ---o.ocI \\ ! \\ in Yo%f $& N ~ ( - udoa =' V) "S" l 's - r-n A,3 & ud y

_m.. f ca(&M<a, f (L. t?.abuJ Ohk , b, Citaa.n AFswe C2st>o g op 6#9 s

• /3 3'uro KerA f 9 cuy 83aSD xto'

t. ca.l W mieesei m w4~rh S

& 3 3. '15 e ro' kca ) .(.)S% -f4 ynalel - % + u s it pha f c/a/c,/4aks, (E=at) 8.'S2D to' M c,a?. eb-,k +da. *t%~f

• b(atf qw.

Cmeg f, jcades - -

3. 3 2 f a / o' 4 ced?

x 4.t 'l c (o 3 3/he~e Ee, = f. 3 9 '2. x ro ' J-8 = I A " l0I 3" (7 ex3c,t;) j ~ 'I et. = ( 3 9 2 "o Tiotssr W wQ = 5j,g m so 3 fase, 'i2 = 280 nr R = 2 rc w 2. e e -m ai6at e 4 A clacf '2 m m m g7g2,e,o g V, 2 = O.6 s f. 3 7 Kj[g'3 As~ +A -Q explacla) eW is a 4.asy& JK } &s new 6 9 k n. 41.e Q ag>ur k \\f = (t.1 )' % = c e s a v, ab -:= 2< es 2 ew b (R3+ V '_ (JM fLeoo t I>PFZ) a. m.c 2FC. 09 w

--E.4 3 Cf. C "

O S Q h s 9 k 'o V b, & e oQ<).b?g % Q Q S$~ak e-o.s e >t Q 'b h r=" coat ee sm -6 h 4 % s%A. {O)

. _. _. ~. .y 5 10 104 Detonation region - en .~I 103 O N m lii 102 10' f l %CH 100 50, 0 4 0 50 100 %C % 3 Figure 7. Detonability of stoicheiometric methane / ethane mixtures with air

s I I [3 4D 20-i i k l t\\e o P<* /'di w P!. AN E a 5 [.5 f t ~ cM f 8 LINE w 3, g '/ E. POINT g .~. -k-5 A#U SI'/k7a g INTERIOR n b'0 ) GAS / AIR CLOUD q S Pac"7tlck e (3-D ) s wried e U EXTERIOR I GAS / AIR CLOUD s 1 0-8 5 10 5'O N FLAME SPEED vt ( m s'l) A FUNCTION OF BURNING ! RATE FOR DEFLAGRATION

• -EXPLOSIONS IN A UNCONFINED VAPOUR CLOUD.

w -w 1-

WWlf, t[. A { (lf 75~) t 303 Flames in Premixed Gases Introduction to Combustion Phenomena 302 5 This equation shows that an increase in the ihme speed would result in a higher space heating rate. Rendering the flames turbulent, the flame speed may be increased far beyond the fundamental (i e., laminar) fbme speed. l*- l Added to this, the fbme surface area is greatly increased by turbulence as it -*r*-Larze scale f scate 0 muhiktes, folds and often breaks up the flame front. The result is a greatly l l 0 8 exaggerated space heating rate. The effect of turbulence on flame propagation was discovered quite o O e e 3 cccidentally. Experiments using tube methed indicated that the "funda-l l mental" fhme speed measured in tubes oflarger diameters is larger. This was y later realized to be a consequence of :urbulence which increases with an y 8, r O o increase in tube size. Note here must be made, however, a hat once turbulence l ~ is triggered on, what indeed we measure is no longer the " fundamental fbme a speed" whose definition requires the fhme to be laminar. While the funda-8[8o mental flame speed is a characteristic property of the reacting mixture alone. t the turbulent fbme speed (hereafter called burning speed, in order to dis- ' ) depends upon the dimensions of the tube as well. Transport of Jtingui heat and mass in bminar fhmes occurs by molecular diffusion phenomena. In turbulent flames, however, eddy mixing which is a function of geometry 0 4 8 12 16 20xt,000 contributes to this transport. Referring to Chapter 3, turbulence is characterized by the eddy size-Effects of Reynolds number on the flame speed (from G. Damkohler. Zerr. brge or small scale-and the mixing length. He eddy diffusivity e is a fleure s.24 measure of the mutualinteraction among the eddies in the same manner as the kinematic viscosity v is a measure of the molecular interactions. e is l i known to be approximately proportional to the Reynolds number in tube effect of these fine scale eddies is to enhance the intensity of transport proces-flow. (Refer to Chapter 7.) Transition from hminar to turbulence is expectedses within the combustion wave. Under these circumstances, transport of l in tube flow if Reynolds number (spud /p) is greater than 2,300. Thus, heat and species is proportional to the eddy diffusivity e rather than the flow in a tube becomes turbulent if the tube size or the flow rate is increased, molecular diffusivity D (or K/p,C). Equations 8.15 and 8.16 indicate that Figure 8.24 shows Damkohler's (Bunsen burner) measurements of flame the flame speed is directly proportional to,/K/p,C or,[5[. IIence it is speed at various Reynolds numbers. Ile found that the flame speed is (a)logical to expect the burning speed, of a small scale turbulent flame, to be independerit of Reynolds number when Re < 2,300,(b) proportional to the l square root of Reynolds number in the range 2,300 $ Re 5 6,000 andproportional to,[e'. nus (c) proportional to the Reynolds number if Re it 6,000. Obviously, only e :/2 { = [K/p,C f}"*[@e } * ?}) Sr e item (a) above obeys our definition of the fundamental flame speed; items ( f ' (b) and (c) are influenced by turbulence and hence the meastered ihme speeds depend on geometry and flow. Denoting the turbulent flame speed by theFrom Section 7.6, c/w ::: 0.01 Re for flow in a tuve. Therefore, symbol Sr. Damkohler explains his measurements as following. S' ::e 0.1 Re itz (g,gg) UO l (a) Small Scale Turbulence This equation indeed predicts the trend of Damkohler's small scale burning In the range 2,300 $ Re $ 6,000, turbulence is of fine scale; that is, the eddy speed measurements. uze and mixing length are much smaller than the flame front thickness. He l

364 Introduction to Combustion Pi:enomena Rames in Premixed Gases 30 l (b) Lzrge Scale Turbulence 8.8 Fisme Stabilization 9 When Re 2 6.000 the turbulent eddies are large, of dimensions comparable l with the tube diameter, much larger than the laminar fhme front thickness. In order to accomplish large thrusts in a turbojet or a ramjet engine, thi i nese eddies do not increase the difTaisitie+ as the small scale eddies do, supply vel city of the reactant mixture is desired to be extremely high but they distort the otherwise smooth " laminar" flame front as shown in it is not unusual for this velocity to be as high as ten times the maximun p ssible turbulent flame speed of a given mixture. Experience shows that the Figure 8.25. He influence of these folds in the flame front is to increase the i flamefront area per unit cross section of the tube. As a consequence, the flame is blown away when the supply velocity exceeds the flame speed.The apparent fbmc speed is increased without any change in the instantaneous maximum supply velocity with which fresh mixture may be brought to th-local flame structure itself. Damkohler estimated to show that the increase ihme front without blowing it away is known as blow-of telocity. Thi important limiting velocity depends upon a host of factors which includes the instantaneous suppi, now.eiocity prorne nature of the fuel and oxidant, their ratio, mixture temperature, combustier . chamber pressure, turbulence in the approach stream, burner geometry 2nstantaneous name propaganon erecuon burner wall roughness temp:rature, etc. / l instantaneous name front shape ^7 (a) Stability ofa Bunsen Flame X Figure 8.8 indicated the mechanism which determines the shape of a Bunsen x Y flame. Based upon this simple mechanism we can explain the phenomena of =[ blow-off and flash-back. Figure 8.26 shows flame speed and normal com. _= / 4, ponent of the supply velocity u,in four different situations. Only the regior. .~ h N ) ) S y .s _y r s,-s s Hgure 8.25 An exaggerated view of the turbulent 8ame front in surface area is proportional to the characteristic wrinkle (fold) size which rt= rim

• *b

C'*"* *8 is proportional to the magnitude of velocity fluctuation (i.e., of turbulence intensity). Since e is proportional to the product ofintensity and mixing length and since e/r :::: 0.01 Re, S .i r x area cc fluctuation oc e oc Re (8.19) } Y UO \\ This explanation describes Damkohler's brge scale burning speeds quite ' ' ~ ~ O satisfactorily. j l i It is possible to conceive that at very high Reynolds numbers the t urbulence rim rim may get so intense that the wrinkles and folds in the flame front ultimate!y "** h-b *' ' B1 '-*f8 end up breakeg the front. He resultant pieces" (islets or lumps) of I .~.- prorne or normal component or.* flame may now jump ahead of the mean ihme location into the fresh gas and conversely, lumps of the fresh gas may jump into the flame. His situa. prorne or riame spud tson as extremely complex and at present our understanding ofit is too sketchy. Drure 8.26 stability of a fiarre frons near a wan

l[* f{ y. 1 '1, I l -~ 4 Ex tent of damage: i i'i 4 i i 'i't' l 0'8 l r + ._ Probable 0.6 l destruction 'O ' *3 bpsig l-g.4 <C concrete l buildingt W 0.2 E i Senove l structural v> ~n ~ t w ! -~ 0.1 l'""i"I 'C'd e. i concrete y 7 (p<,i) I Eartn wwe 0.M f-0.06 O l-lE*

D b

0.04 Minor k ~ structural l-damega 0,02 - Glass break l J / at about 0.002atm j } / / t t i t i,i, overpressure 0.01 i 1 l-2 4 6 8 10 20 40 60 80 100 L f. FLAME PROPAGATION VEL'0 CITY-(m/s) lI L l l-Figure III-8: Maximum Overpressures Developed i -- at Various Flame Speeds in a Vapor l.- C10'ud i- ~'gt 2-48 i i ^

I will discuss only the first ratio, and cloud temperature are a few. three. Laboratory experiments at McGill University, LLNL, and elsewhere have shown that flame speeds can increase under the influence of partia confinement and turbulence. Figure III-15 shows the results of these These data show that flame speeds can increase dramatically, experiments. l but confinement seems to have much more influence than turbulen vapor burns we conducted at NWC, flame speed was one 3 In the 40 m variable we paid close attention to. Table III-10 shows the boundary conditions of each of the vapor burn tests. Times given are all measured from the beginning of the spill. Four techniques were used to measure (1) high speed optical cameras; (2) side-on infrared imagery; flame speed: All (3) overhead infrared imagery; and (4) in-situ vapor burn sensors. data has not been analyzed as yet and therefore the results I will give ~ are preliminary. In the Coyote 3, 5, 6, and 7 experiments, good flame velocity mea-surements in the downwind direction were obtained from the si 5.8, 10.0, 4,6, The wind speeds in these four experiments were red cameras. 12.6, 16.4, and 6.0 m/sec. The average downwind flame velocities were 11.9, and 18.9 m/sec respectively. If we subtract the wind velocity frem these numbers, we obtain flame velocities relative to the air of 6.8, 6.4, ( That is, the velocities were all around 7 m/sec, It must { 7.3, and 12.9 m/sec. except for the last experiment, when it was almost twice that. be noted that the field of view of the infrared-imager differed from I In' periment to experiment, so we should not yet draw general conclusions. the only flame pressure measurement we perfonned, the results from Coyote 6 showed pressures of about one millibar. The overhead IR imagery on Coyote 6 and Coyote 7 allows us to follow We see the the flame front in both the upwind and the downwind directions. C, data plotted in Figure III-16. Velocities appear to be very high near tf4 f 3-14

~ ~ ~ ' ~ -{ DET0t4AT10ft 1000 - BLAST CONFINED TUBE, OBSTACLES DAMAGE 100 CONF'INED CHANNEL, OBSTACLES OPEN CHANNEL, OBSTACLES FLAME 10 SPEED (M/S) uy OPEN CHANNEL 1 0-1 0 0.5 1 DISTANCE (M) Figure III-15: Effect of Continement and Obstacles on Vapor Fire Flame Speed

u.. .o I .. ~... J i l Table III-10: Sumary of the Lawrence Livennore National Laboratories' Vapor Burn Tests I NITION (0YOTE MATERIAL SPILL SPILL WIND G tp DATE DURATION SegED IGNITION IlME Spi'CH ) !EST (M / MIN) (MIN) (M/S) SOURCE (MIN) 3 (% NUMBER DATE 2 8/20/81 70 16 0.5 5.9 FLARE 1.2 Y 3 ' 9/02/81 79 14 1.1 5.8 FLARE 1.7 m 5 10/07/81 75 17 1.6 10.0 FLARE 2.0 6 10/27/81 82 17 1.4 4.8 FLARE 1.9 7 11/12/81' 99.5 14 1.9 6.0 JET 2.4 1 - w< .-<.._,,._.i.2.,2...._,_ 7,,_.,

a 6 0 4 0 6 0 - m' 2 l ~ ~

' ! ds 2

i1 1 1 = 6 7 E E 0 0 D e T T i 0 N O ' 0 O m Y D Y 1 1 i I T C E O D W O E h t C E N w i P E W P i t S n S O io D D P N D i no I W N i t I i W i gn I 0 0 fo ' 5 i 5 d n i m nw w o D d E n a C n d N iw p = U A i = T y t i S Il O oc - ' 0 l I e D V D-i ema E- = l E-F P' D, f = o S E n E o D-P, i t S D a N ir .W' D, N I a I V N W 0 O 6 - ' 5 I W i SP 1 U I .. ~ I I = re u g n a Lh 0 .~ 0 i 7 7 F 0 0 0 o O 6. 0 0 0 0 0 o 0. 0 - 4 4 3 2 61 4 3 2 / f 1 b' Qwq% t 394k gioww4W W2qak u "b F; i

l.

i l ll i 'I f

M ignition source bhether it is a jet or a flare) and to fall off rapidly as the flame moves away from the source in either direction. While it is too early to draw very broad conclusions about flame s two things are clear. We saw much larger flame speedsthan did the Britt at Haplin Sands and our velocities were not at all constant throughout the cloud. Several questions important to detennining possible overpressure d from flame acceleration remain unanswered. How much confinement and tur. bulence is necessary to produce damaging flame speeds? Are there realist cloud sources and geometries that will result in fast flames? Nill a fast f',ame_ slow down after leaving the source region? How does flame velocity change with cloud size? il d. Thermal Damage O J Thennal damage from a fire can be estimated given the area engulfed [ in flame, the intensity and spectrum of the emitted thermal radiation, and the side-on area of the flame and its speed. Others have addressed issuei surrounding the emission of thermal radiation. I will address the shape of the fire. The fire's shape (i.e., the horizontal area engulfed in flame and ) width seen side-on) will be determined by the relative values of flame ( speed wind velocity, buouancy, and inertia. The balance among these f forces may change as a function of cloud size. The pre-ignition 5% concentration contour has usually been used as the horizontal fire area. 3 To see if this is true for clouds produced by 40 m spills on water, we measured the spatial concentration history of the clouds in the Coyote vag burns up to ignition. We will then compare the total burn area as measure < in the overhead infrared imagery to the total area within the pre-ignition 5% contour. The short video clips you see show representative data, it 3-18 1

fM ~(Q - Czyp) l,r O h e m .;;.,f *: ~ \\ g,/

• ,3,. g -

t is. 1 i 1 9 I f i r i, EXPLOSION ~ OF UNC0t&INED CLOUDS OF NATURAL GAS by i D. :C. Sull & J. A. Martin J. REED WELKER UNIVERSITY ENGINEERS, INC. 1215 WESTHEIMF.R DRIVE 1 NORMAN, OKLAHOMA 73069l .{ 1 2 t ' Paper to be presented at the American' Gas Association- -i Transmission Conference St. Louis May 1977 9 ] ; ). I-t j r, m...

SITMMARY This work deals principally with explosions and detonations in unconfined nat aral Gas clouds. It reviews the work of others and notes their conclusions that deflagratien to detonation transition is unlikely. The particular systems studied and reported concern direct initiation of hydrocarbon / oxygen / nitrogen mixtures and it is noted that initiation of spherical detonation of methane / air mixtures requires about 22 kilograms of Tetryl or'its equivalent. Other heavier hydrocarbons are more readily detonable and hence practical natural Cases may be expected to exhibit less stable characteristics than pure methane. Nevertheless initiation by the equivalent of about 1 kiloEram of Tetryl would be required for most natural cases and with this in mind it is extremely doubtful whether detonation within the confines of buildings represents a sufficiently ener6 etic source to initiate detonation within an unconfined gas cloud. Thus within the present state of our knowledge it seems that by all mechanisms detonation of unconfined natural Gas clouds is extremely unlikely. that detonation cannot occur.However, it is not possible to say at this present ju More attention is required to studies of initiation and propagation of detonation in practical unhomogenous Eas clouds. Similarly a better modelling is required of the blast potential of other than clouds of artificial geometric shape. F s m.. m.- -... -. - -. _.. ~.... _,

1. INTRODUCTION It is not the author's intent to present a detailed scientific account of the various circumstances in which natural gas may be made to explode, but rather to state the issues simply attempting to summerice current thinking and to point out areas where presently knowledge is inadequate. The term " explosion" is generally used for any gaseous expansion which gives rise to audible noise. Hence it may be made by, fer instance, LNG spilled on tc water, which in some cases with high ethane content gives rise to vapour explosions, or by uncor. trolled combustion of flammable gas / air mixture in confined surroundings. These two types of occurrence have been widely studied and it is not the intent to deal with them here, rather this paper will deal with explosive combustion of unconfined clouds of natural gas which might follow leakage from a damaged pipeline or accident involving a spill of LNG. We have studied the subject of explosions very much with LNG storage, shipment and handling in mind, since the Shell group of companies has for some time participated in LNG trading and wishes to develop it at a time when good supplies of clean gaseous energy are particularly required in many industrialised countries. The reasons for investigation into this subject are, firstly, a need to quantify the extent of hazard following a specified incident and gain some better impression as to the chance of it occurring and, secondly, by study of the mechanisms involved, hopefully to be able to reduce both the chance of an accident occurring and its magnitude should it occur. Two major incidents are uppermost in many peoples' minds and these are a major collision or grounding involving substantial damage to an LNG carrier and a major accident in,an LNG storage installation. One set of " ground rules" for size of disaster to be considered in the former was formulated by the US Coastguard. This requires consideration of one cargo hold lost instantaneously. For land storage the NFPA 59A regulation sets the maximum credible accident as the complete failure of the largest through wall connection to a storage vessel. In this case modern hi h bunded storage configuru' ions employing insulating 6 concretes and, if necessary, fire fighting foams can reduce vapour evolvements and disperse it in flammable concentrations within the property limits. Clearly it would be desirable to move the product to an undamaged storage vessel if this were available, but if this were not so it woujd be tempting to consider a suggestion put forward by Sliepeevich (U to deliberately fire the spill if this could be done with safety. The idea of firing a spill before a large hazardous vapour cloud has time to develop is even more tempting in the case of major damage to an LNG carrier, as here no means has yet been developed of limiting the speed of formation of the cloud. However, there are obvious problems which have yet to be overcome concerning the safe evacuation of the carrier crew and also the crew of any other vessel involved in the collision. Further it must be realised that the time available before mixing produces a dangerous cloud is very short, typically a few minutes, and thereaf ter this method could precipitate an accident. We have all read accounts in which the explosive potential of a i complete LNG cargo is compared to that of a large high explosive bcmb

2 or even nuclear warhead. Such statements are usually based on calculation of the gross thermal equivalent of the LNG which is then related to the heat release of some quantity of high explosive. Thia seems to be about as appropriate as grossing up the calorific content of all the natural gas used in the United States annually and on this basis arguing that natural gas usage should be prohibited since it is equivalent to 10,000 megatons of high explosive. Here the argument is clearly taken to extreme, but it does emphasise the point that the gas must be available in the right place at the right time, and mixed with the correct amount of air for an explosion to occur. As will be seen later these conditions do not apply to all the gas involved in a spillage. The next point concerns the speed at which the energy is released, and here we do not intend to dwell on the blast equivalent of gas clouds and TNT explosions, but the issue is largely the mode of burning - will it deflagrate or detonate? 2. DEFLAGRATION l The gas clouds evolved frrm LNG spills are far from homogenous (see Figure 1). There are mar.y models available for the calculation of the amount of gas in any cloud lying between stated concentrations in air. Shell has developed its own models which are applicable both for gaspoys releases as from a pipeline leakage (2,3) or from a pool source \\4/ and, as a model shows clearly, at any given time only a fraction of the gas is mixed in flammable proportions with air. Most mooels give a time average picture and opinions are very much varied as to the likely extent of the variation in peak-to-average gas concentrations at any given point. This is almost certainly a facter of spill size and of the distance of the point in question from the spill source. Figure 2 shows the fluctuations recorded on a highly sensitive instrument in a small gas spill. From the experimental work which we have done, it is clear that dependant largely on the size of the spill, in some circummtances the gravity spreading effects are dominant whereas in othere the atmospheric turbulence plays a more significant role. In calculations regarding major spills, we generally assume that the time average half-LEL contour defines the limit of flammable gas travel. Of course the upper flammable limit contour in no way represents a boundary-upon the gas which enters into deflagration. If it were to do so pool fires could not result. The convection currents set up once a flame is burning are sufficient to promote mixing of air with pockets of rich gas so enveloping virtually the whole over-rich portion of the cloud. Regarding flame speeds, in the small scale Ul0 spills which we and other experimenters have used involving quantities of a few cubic metres spilled on land or water, the gas has always burned quietly and undramatien11y despite the varied atmospheric conditions. Not infrequently in small spills the gas fails to light-back to the pool source, either because it-fails to ignite or local burn-out occurs. Although we do not plan to carry out tests at high wind speeds, it sometimes happens that wind develops during test and in theso --~.

i 3 circumstances we have noted that light-back to the source can occur at gust speeds of about 20 m.p.h. So clearly turkulent conditions prevail and it seems not unreasonable to predict that flame speeds of up to 10 metres per second are possible. In this connection it is interesting to note the results obtained by others with well mixed methane /pir systems in stoichiometric proportions. Lind and Strehlow(51 reported flame speeds of up to 5 8 metres per second horizontally and 7 3 metres per second vertically. We are aware that other workers are presently active, notably those at the Midland Research Station of the British Gas Corporation, and we understand in their comparatively large scale tests, flame speeds of less than 10 eotres per second were recorded. For evidence of higher flame speeds, it is necessary to turn to other workers using more reactive systems. Dorge and Wagner (6) examined acetylene : air combustion. Using a 0.2 metre sided cube, they noted flame speeds of up to 20 metres per second when the mixture was ignited with a weak spark.under conditions where high turbulence was generated by use of spherical nets. However the maximum peak over-pressure revealed was 0.2 atmospheres and flame acceleration was not sustained. Kuhl, Kamel and Oppenheim have given us a model relating flame speed to over-pressure (Figure 3). From this it is clear that pressures at the flame front of methane / air systems are likely to be less than 0.1 atmospheres, with scarcely detectable blast pressures. It is indeed unfortunate that almost without exception experimenters investigating spherical flames have used equipment which is much too insensitive to record the small over-pressures which are generated, and so have missed the opportunity of validating the prediction of the model. Munday (7) has fitted the damage caused within the cloud to an appropriate flame speed and matched the impulses to the TNT decay curves to align with damage in the far field. Using this or similar methods he and other workers have assigned flame speeds of up to 50 metres per second to violent explaions such as those at Flixborough - which of course involved a cyclohexane. A survey by Strehlow and Baker (8) of large scale natural gas escapes which was subsequently ignited has shown no evidence of high flame speeds. 3 DETONATION Thus so far in deliberate tests, lor in small scale accidents involving unconfined natural gas clouds, these have burned quietly. However, it is possible to postulate that in very large clouds more violent thermal stirring would occur and hence more vigorous combustion would follow - although the mode by which very high flame speeds would result is far from evident. The real issue is will detonation of a free cloud occur? Conceivably it might follow deflagration or be directly caused by a more energetic sources, such as an explosion. 31 Deflagration to Detonation Transition As has already been mentioned those workers who have investigated

4 the former mode have come up with very unexciting conclusions. Wagner (9) observed acceleration in propane / air mixtures, but this was not sustained when the turbulence was removed.

Boni, Chapman and Cook (10) have reviewed the six distinct processes which Lind and Strehlow have suggested might result in a deflag-ration to detonation transition and conclude that the presence of a turbulent boundary layer or Rayleigh Taylor instability of the flame front, and buoyancy induced acceleration of large flames are very unlikely mechanisms:

neither do they think likely j flame acceleration due to other turbulent processes of a fluid i mechanical origin. This view is shared by Lee (11) who points out that the turbulent flame speed required for self initiation 1 of fuel / air mixtures is about 230 metres per second compared with laminar flame speeds for most hydrocarbon / air mixtures of the order of 0 5 metres per second. These workers then, conclude that weak initiation may not be i possible. Certainly in order to achieve it, an extremely folded flame front would be required and Lee doubts if this folded structure can be maintained long enough for detonation to occur. Of the 108 incidents reported by Strehlow and Baker, only one - that at port Hudson, involving propane, resulted in damage which was considered to be consistent with detonation having occurred. In view of the amount of evidence which has since been accumulated, it seems possible that such damage could be resulted from some lower flame speed followed by focusing of the blast effect. The workers from Science Applications Inc. have concluded that for deflagration to detonation transition to occur the only plausible mode would involve an explosion within some confined space which would progress to detonation within that space and which then might propagate into the cloud. 32 Direct Initiation So far in discussing deflagration to detonation transition, reference has been made to studies performed by a number of other workers. At Thornton Research Centre most attention has been directed to the alternative mode of detonation initiation - that of propagation from a shock wave established by use of an explosive charge. The first series of experiments has been reported elsewhere (12), but since new data will be presented in this paper, it is appropriate to briefly describe the experimental arrangements. Well stirred stoichiometric mixtures of methane and oxygen were used with various amounts of nitrogen as dilutant, and detonation-was initiated by electrically firing selected amounts of Tetryl explosive. The gaseous mixtures were contained in polythene bass, fabricated from 130 micron polythene sheet of specific weight 130 gm-2 and were of uninflated size 1.80m. x 1.80m. or 3 05m. x 1 52m. The gases themselves were of more than 99% purity, no hydrocarbons other than methane were detected and the principal impurity-in the methane was nitrogen.

i 5 The gases were simultaneously admitted to the test bags in pre-determined proportions using integrating gas meters, rotameters and a mixer / flame trap assembly. Thus on entering the bag the 6ases were saturated with water at ambient temperature. Figure 4 shows a laboratory mock-up of the polythene bag as it was suspended in the bomb chamber at ERDA, Waltham Abbey, (which was made available to us by the Ministry of Defence). A member of their staff assisted in some of the experiments. The progress of the detonation wave was monitored by use of e microwave Doppler unit operating at 10.687 GHz. This was coupled to a pyramidal brass launching and receiving horn operating to an aperture of 146 mm x 118 mm and again of 16 dB. The horn was fitted with a PTE plug and flexible wave guide coupling to give a measure of protection to the microwave Doppler module. An array of piezorlectric air blast gauges vere set on the same (horizontal) plane as the microwave Doppler unit. The Eauges were positioned at distances 0.6 m, 1.2 m and 1.6 m from the explosive initiator corresponding to distance of about 1.1, 1.8 and 1.2 m. from the centre of the bag. At the time the work was planned, the experiments of Kogarko(13) were known and so there was some hope that the full amount of nitroEen could be added to represent methane / air mixtures.

However, it soon transpired that more energetic sources were required than had been predicted by Kogarko.

Closer examination of Kogarko's experiment revealed that his measurements of detonation velocity were made from time of transit between two piezo Eauges set at radii so close to the initiator charge that the latter contributed 22% and 8% of the total wave energy. The experiment was therefore not suitable for determining whether or not the v+ve was self-sustained as a detonation; the fact that the de of transit velocity recorded was very much below C-J inuicates that in all probability it was not. This does serve to underline the importance of ensuring that the experimental path lenSth is sufficient that the wave observed is quite free from effects of the initiator For a spherical wave this obviously implies liberation source. of lar6e quantities of energy. As our experiments were limited by the amount of Tetryl and fuel which it was thought safe to use within the facilities, a technique of "end initiation" shown in Figure 4 was adopted. This allowed a much greater experimental path for any given gas volume and comparative tests with a centrally initiated detonation showed that the position of the initiator had no effect on the measured critical nitrogen content, or the pressure and velocity of the detonation waves. The experimental method was to use five charges of known weight and size and to determine the limiting nitrogen concentration for detonation to just be sustained. The results are given graphically in Figure 5 As has been intimated earlier, it was not possib1'e to experiment directly with methane / air systems due to the limitations of the experimental arrangements. Idmiting nitrogen : methane ratio at which detonation occurred with our largest explosive l char 6e w a about 5 5 : 1. The experimental velocity measurements I for all of the mixtures accorded well with the calculated C-J values.

.= 6 Within the experimental range the log of the initiator char 6e mass to limiting nitrogen concentration relationship is linear and there seems to be no reason from chemical kinetics or thermodynamic considerations why there should be a sudden departure from this. j Fi ure 5 also shows the results in which the bridge wire detonator E 1 was used alone as an initiating source. From the characteristics of the blast wave, it appears that this is equivalent to a charse of 0.25 grams of Tetryl. However, the rate of energy deposition is not necessarily the same as for a Tetryl charse. For these reasons the results of the detonator used alone have not been included in calculating the regression line shown on Figure 5 As the figure clearly shows, the projection of the straight line l relationship indicates that more than 20 kilograms of Tetryl would be required to initiate a spherical detonation in a stoichiometric mixture of methane / air. The work also points to the long path length (11 m.) which would be required in any experimental verification of the extrapolation. In more recent work the amounts of tetryl required for initiation of detonation in stoichiometric mixtures of other hydrocarbon fuels with air have been directly determined using the same experimental method,and these are shown in Figure 6. The extreme stability of methane in comparison with these other hydrocarbons is clearly demonstrated as might be expected from a compound with its particular atomic structure and relative inertness as exemplified by a high minimum auto-ignition temperature and hi h octane number. The 6 amount of Tetryl required to initiate detonation of methane is more than two orders of ma6nitude greater than that for the other saturated paraffin hydrocarbons and more than three orders of magnitude Breater than that for ethylene. Inevitably such a result mus't raise questions about the probable behaviour of natural Eas as opposed to pure methane. Natural gas is of course not a single substance which may be defined chemically, but commonly contains varying amounts of higher hydrocarbons than methane according to its origin and the treatment to which it has been subjected. It might be supposed that these smaller amounts of higher hydrocarbons would effectively " seed" the detonation and hence promote detonation in the whole gas. The Thornton team has therefore carried out a few supplementary experiments, wnich are still continuing, on simulated natural gas mixtures. In Figure 7 the figures for methane / ethane mixtures have been shown which would appear to confirm this supposition. It is intended to extend this work into mixtures containing hydro-carbons other than ethane, but from the evidence produced in Fi ure 6 there is little reason to suppose that, for instance, 6 propane will behave in manners significantly different to ethane and hence the data presented on Figure 7 may be considered to a first approximation as methane admixed with such hi her hydro-S carbons as are likely to occur in natural gas. At this stage,it'is appropriate to return to the question left unanswered at the beginning of this section; namely the possibility of detonations, originating in confined spaces, propagating into - -~

? free clouds of natural gas and hence initiating detonation therein. From the data presented in Figure 7 it is clear that an explosive source equivalent to at least 1 kilogram of Tetryl would be required to initiate such a detonation in most natural gases. At this present juncture it seems to be generally accepted that the critical conditions for the initiation of a detonation cannot be wholly defined by critical energy considerations alone nor even by some refinement which considers the application of power density to unit volume. With such reservations in mind and accepting the imperfection of our understanding, it appears that any explosion within a confined space, which might be relieved by the destruction of windows or even of the building itself, is most unlikely to provide the energetic source required. It is conceivable, however, that detonation in more confined surroundings which could generate higher pressures an, for example, in processing plant, could lead to the sufficiently energetic sources to trigger detonations in free gas clouds. In the previous section when considering deflagration, reference was made to the highly structured nature of unconfined gas clouds as exemplified in Figures 1 and 2. Certainly not all the gas lies within the flammable limits and the proportion which does is a functie: of time. When dealing with detonation, however, we are not concerned with gas within the flammable l'imits, but'"rather with gas within a narrower band. Only gas lying within these detonation limits will contribute to detonation since the time scale of events is insufficient to permit spectral changes. Again, due to the limits imposed by the test equipment, it has not been possible to' determine the amount of initiator required to promote direct detonation over a range of concentrations of methane in air. However, this has been done for ethane / air and the results are illustrated in Figure 8. There is no reason to suppose that the shape of -curve would be markedly different considering methane / air and for normal natural gases it may well be argued that the higher hydrocarbons effectively determine the sensitivity to detonation. From the ethane / air curve the heavy dependance upon stoichiometry will be seen and it will be noted that the amount of initiator required varies by almost two orders of magnitude according to stoichiometry. Thus considerable amounts of energy will be required to initiate detonation in gas mixtures lying at the extremes of the detonable limits. With these considerations in mind, there is clearly some uncertainty as to the amount of gas in any unconfined gas cloud which will contribute to detonation. In any one selected area only the amount of gas lying within the detonable limits will contribute. The question is, however, can such detonations propagate through discontinuities to other areas lying between these same detonation limits and cause sympathetic explosion. A subsidiary question which arises is therefore "If these discontinuities cause the wave to decay, to what level can it decay before it is no longer able to re-establish itself, when it once again encounters a detonable mixture? In this context

8 measurements were also made with critical initiation of ethylene / air which like those of Brossard and Edwards on acetylene / oxygen / nitrogen and propane / oxygen / nitrogen systems, respectively, indicate that there exists under conditions just critical for initiation, a region af ter the major decay of the very intense shock from the initiator where the average velocity of the wave is below the steady state Chapman-Jouguet value. These results are illustrated in Figure 9 where it may be seen that the reaction front velocity falls to a minimum of approximately one-third of the Chapman-Jouguet value before attaining a steady ctate velocity similar to that projected by the Chapman-Jouguet calculations. From this work it therefore appears that knowledge of the behaviour of a detonation under conditions marginal to its propagation coupled with better information concerning the instantaneous concentration of fuel in air within a gas cloud will define whether propagation of detonatioa through a spillage cloud is possible, and under what range of conditions the possibility could exist. It will also help to indicate the proportion of the fuel likely to be detonated and the nature and extent of the blast wave produced by the detonation. 4. DAMAGE p0TENTIAL The concept of a TNT equivalent of a gas explosion has in the past been useful as it gives some impression of equivalence. It is generally accepted, however, that this model must be used with considerable recerve, particularly when applied to the near field. In practice as has been seen from Figures 1 and 2, the unconfined gas clouds originating from LNG spills are inhomogeneous disc-like in appearance, vis they are thin compared to their diameter. Some models treat these as well mixed hemispherical clouds for mathematical convenience. In practice the energy liberated in the wave will parallel the inhomogeneities. Further, there is bound to be considerable interference from waves leaving the thin disc and, therefore, present analyeis of the blast effects within the cloud and in the f ar field are sadly very deficient. 5 REMEDIAL ACTION Although this work has been proceeding for a period of about two years there is clearly much still to be done and our understanding of the mechanisms involved is very imperfect. It has already been seen that large gas clouds are very lumpy and badly mixed and there is some question as to whether detonation can proceed throughout the whole mass gas lying within the detonation limits. This, therefore, suggests that if some mechanism can be found to break up the gas it may be possible to use such discontinuities to reduce the blast effect caused should a detonation ensue. From Figure 6 it is apparent that the tendency of gases to detonate is a strong function of their chemical reactivity. It is well known that the methyl halides act as combustion suppressants by inhibition of the chemical kinetic properties. -It was therefore hoped that they would be effective in suppressing detonation. The very limited amount of work which has been carried out to date in this area has proved disappointing. These methyl halides have acted in the same way as ' any

9 inert material such as nitrogen and therefore we must return to more fundamental studies and look elsewhere for solutions. It remains our hope however that some substance may be found which will prove an effective detonation suppressant and which may be safely and quickly spread on any major gas spilly;e.

..-....7.--..-.-.-.-.-.-.=.-- . ~. I l

• s'it

.y j .k ..-11, f'[* 't. J.;i..? i e!

• f.

,~..,u c.. z. ~ ~- , J Y. ,If'd/..g.:, yhb^'lY),,,*" 5, y j_, j, kt b '. i, -. g(.. g ))! J. : i. \\. U:.'*hh &;. k ?, ? '- C \\ 'h. 1 l . 5 l Figure 1, Gadilla Jettisoning Trials: Portion of gas cloud

s 8 t-t 6 l i t 4 1 ~5 5 1 j h 4 1 E 5 I i l r 2 I I l I l I l t l O L. ._ A f d J i l e I J l' b / m 0 1 2 3 4 Time, minutes j Figure 2. f.iethans cancentration at single sample point in sma!I spillage cloud i

I' 1 1,0 5m 1 2 0.5 Flame y Front Eg O Shock Front 0 i t I i i I I i i I i 1 5 10 50 Steady flame speed (m/sec) Figure 3. Flame and shock overpressure as a function of burning speed for a spherical flame (Oppenheim)

l ( I l d- ~" A~ =@55 l M .,_.;-~~r..~. = m ~:~~-ih~ - ~ - - - - - - - - - - - ~, - - - - - - - - - - ?, t ~C Tt.J. " u;,,s., l ,.: n a.3 t. 3 ~.a. .c a ;-. f k% MEb ,h ^ ~~ [ h.,Y~ ._.s":'n?c;c M r.} y;): n.' Q 'l '$.l :z$ 5. $ $ % h -3',.h 0 c. l i h.y'% Y,- ',....'l ' ,? i y, ~~

e. -

v. cs N% y' ' '

j l l g, I tf,i s - %:7 \\ E,,,.... c. ; '{$3 ~.. j % -g ~ l .h $ft ) \\ e j hYr i u-6 g',' t i,. et C ONT $..; * ' ly,l l S l lA4.+ i -T:'- i i -L.. l l ) I i Figure 4. Test arrangement for direct initiation detonation trials i i i ) I i

100K s' / / / 10K / _ / / / / / / / V 1000 / O O xx Kogarko Detonation (TNT) O O++x x 100 / O O x / / = h O CDCDCD4+o'X 10 lE

• • U 9'*d*"

O 3 O +x 1 / ./ / / / '.hr d B '* #4 3 Oxx x f / l 7' 0.1 - / / / / 0.01 0 2 4 6 A 8 1

• (N /CH,)

R Figure 5. Detonability limit of CH, + 20 + xN, in the (x,M) plane 2 c' Cdetonation established (microwave and pressure); x detonation ck failed;+ pressure records indicate detonation, but microwave inter. 4p 7 ferograms do not,05% confidence limits to regression extrapolation

!lj,i,I il 1 s aoo ej _ x. na h t -mg;~ 'N e M m,ah v.

a. o

' t )g K ( d a er iuqer r e o n ta a e t i i e n u t n a B n a t i p u o e o B s v r i Oa e P n iso na lp h x t E E ene ly ht E eS. nego r e d n y e H ly tecA o.0o ~ n 1 _ 3CFs 9 n. E.<* a2O3E==s OqC2s>r 3 >,25 l l

lill,

. ~ _ -.. _. -. .i 105 I i l l' 104 l l Detonation region I o' _~ . C 103 o Nm i E l I 102 l l 101 i %C H, 100 - 50 0 i 0 50 100 %C,% Figure 7. Detonability of stoicheiometric rnethane/ ethane mixtures with air

103 7. Flammability limits I Detonation reglon en ~ C 102 o N= lE 101 i O5 1.0 1.5 2.0 Stoichelometric fraction in air Figure 8 Ethane / air detonabiltty

~ i i 4 i t (a) i 1 l I (b) 2000 1 1 i' i I (c) 1 r i I I System initiator I i (a) C H2 + 2.50, + SN, 105 J exploding wire j 2 h I I (b) C, H, + 50, + 0.4N 105 J exploding wire i E 1500 I-2 O (c) C H, + 30 + 12N ' 20 g Tetryl explosive 2 2 2 \\ g (Air) t l 1 i i I L 1000 - 1 i V i l I l r i I 500 O 0.5 1.0 l R, m r Figure 9. Reaction fror t velocitv/rarfine nints inr critical snie;=t;4a nr.4.- ~t 8a--+ -

.. - ~ .=y* ,,s REFERENCES 1. Velker, J.R...Wesson, H.R., and Sliepeevich, C.M.. " LNG Spills - To Burn or Not to Burn", AGA Distribution Conference Paper 69D23 (1969). 2. Ooms, G. "A New Method for the Calculation of the Plume Path of Gases emitted from a Stack", Atmospheric Environment 1972 ) Volume 6, pp. 899-909 3 Ooms, G. Mahieu, A.P., Zelis, F. "The Plume Path of Vent Gases heavier than Air", Proceedings 1st Int. Sumposium on Loss Prevention and Safety Promotion in the Process Industries. The Hague May 28/30 1974 4. te Riele, P. " Atmospheric Diapersion of Heavy Gases emitted at or near Ground Level", 2nd Int. Symposium on Loss Prevention and Safety Promotion in the Process Industries. Heidelberg September 1977 5 Lind, C.D. and Strehlow, B.A. Unconfined Vapour Cloud -Explosion Study", 4th International Symposium on Transport of Hazardous Cargoes by Sea and Inland Waterways. Jacksonville,- Florida, October 1975 6. Dorg, K.J. and Wagner, H.G. " Acceleration of Spherical Flam s" e 2nd European Symposium on Combustion, 253-238 (1975). 7 Mundy, G. " Unconfined Vapour Cloud Explosions: A Reappraisal of TNT Equivalents", I. Chem. E. Symposium Series No. 47 8. Strehlow, R. A. and Baker, W.E. "The Characterisation and Evaluation of Accidental Explosions",-Prog. Energy Combustion Science 2, 27 (1976). 9 Dorg, K.J., Pangritz, D. and Wagner,_ H.G. " Experiments on the Acceleration of Spherical Flames by Grids", Acta Astronautica 3 1067 (1976). 10. Boni, A. A., Chapman, M., Cook, J.L. and Schneyer, G.P. '"Cu Combustion Generated Turbulence and Transition to Detonation", AIAA 15th Aerospace Meeting, Los Angeles (1977). 11. Lee, J.H. " Initiation of Gaseous Detonation", Vol. 28 An. Review of Physical Chemistry 1977 12. Bull, D.C., Elsworth, J.E., Hooper, G. and Quinn, C.P. "A Study of Spherical Detenation in Mixtures of Methane and Oxygen Diluted by Nitrogen", J. Phys. D. Appl. Phys. Vol. 9 (1976). 13

Kogerko, S.M., Adoshkin, V.V. and Lyamin, A.G.

"An Investigation of Spherical Detonations of Gas Mixtures", Int. Chem. Eng. 6, 393 Y r

• a.

-i

)% 6/wW yg o,d JL 6 -/977 " gr y -\\ u N 'M asa e: Ast retNoutN4 tN Ala 4No sust a(E st asts T e uf N tC A am t n of st AM wau w wutNa The partde sdsdy 8"' M *"d is a sudden preswee discontinuity, t e. My WM

• M L*

3.50 S<miewhat related to the um-blast wave immediately ahne at I-ur a true nor ideau shd InwW. are denved duion of the surf ace are the effects of large userprnwres with long paisne. inun the Rankme Hugonma condinnms glen by objecn armt material paked up by the phase duration, the shock will penetrate based on the conscrsation of mass. en-ergy, and smwnentum at & dud fnw ,,M g + l_g f - blast wave Damage may be caused by wme datarne into the gnmal but blast These condituists. together with the W. h P, min es wch as roths, boulders. and waves whnh are weaker and of stuwier pebbin. as well as by smaller particles duratmo are attenuated nuwe rapidly. the equarne of state few air, pennd w that for air sush as sand and dust This partientate The mapw principal cress in the smi dertvarum cf the seguired relatawis an-5 matter carried ahmg by the bint wave well be nearly sertical anu about equalin voI*mg the sinnk wehony, the partale does net necessardy affect the overpres-magnitude to the air blast overpensure. for wind) selocity. the overPre"ure. the sures at the shak front. In dusty areas. Thne matters witi be treated in more dynamic prenure, and the densay ed the The denuty

p. of the air behind the however. the blast wave may pnk up detail m Chapter VI air behind the ideal shsk fret shoti front n related to the ambee:W ermugh du a to in rease the dynamic 3.52 Ih a high arr burst. the blast 3.54 The blag wave swoperties in g

prenure over the values corre emdmg to the userpreswee in an ideal brass oserprewurn are espected to be rela-the regmn of regular retienen are compke and depend on the MM ' T wase.1here may alw be an increase in fively wull at ground level, the effects wwnewha t angle of inmiesne of the wa*e *nh the v For a W M+W-IT the schicity of air parinles in the wave of ground sh sk induced by air blast due to precurwr actim Consequently.

• • tt then tw negligible flut if the oser-ground n! the o*cipreswre omt ct wrf ace busd.
• ben there as but 7,9p the eff ed on vrudures whnh are dam.

pree-ure at the wrface as large. rhe e a sseigle hemnphernal trnergedt wave.

• GQ aged mamly by dynarmc prenure udt may be damage to t,uried gruttures as stated in l ) 34, and in the Math lhe slynanm penwrc. 4 n Md y Howeser, even si the strudure n strong be cswrnemdmgly unreased. espe.

tegen below the erapk puna path im an oatly in regams where the precuruw n enough to withstand the effe(t of the air burst, she varmus blast wave (harac- ,.4 g ground skick, the sharp gott rnufting sin ing. terntics at the shock ' net are un>Psch fact the impact of the sNsk wave tan

• tha' " *d"*II) "* h **

related by the Rankine &gmad c4*8-y GROl'ND SHOT K IROM AIR fil AST inms it n for these omdatwn. m whn-h per uma solume of air imme cause infJay to occupanh and damage to hind the Ank front. thn quanney un hme equipment in areas mhere the air there n a sangle shenk front. that the nno the same dimenwms as prewure 3.51 Another aspect of the bl.nl blast prewure is high, certain pubhc follo*mg rewits are appinable d"'imn of the RankmeWgoamt equa-wase problem is the pomble effed of an utihtsew wth as sewer pipes and drams a r buru on undergn und sinneures as a made of relanvely r'Fkt maret als ar.J 3.55 Thesh=L " loot) U i'e '"""I"'"' "E*"* resuh of the transfer of wmc of the bless hsated at shallow depths, may be dam. pre sed by

• ase energy into the ground A rmnor aged by carth movement. but relatively

),y p ostdiation of the surface n espetsenced fleuble neal pipe will not normally be U% i

• y

~g p, 'I " h p + ty U lf and a gn.und sh ck is pnniuced The affected f or a surf ace burst in whnh where e, is the ambient speed of wund 4155 h urength of this shock at any poent is cratering trcurs, the usuatum is quire L-(ahead of the Ask inmil. P n the peak 2 7I. ' f determmed by ihe overprewure in the different. as mdi he seen m Chapter VI overpreware 4tehind the shsk tumil. P, n the ambient gwcwure law of

• TICi1NICAl ASIYCTS OF Bl.AST WAVE PflENOMENA.

shwkL and y a the ratm of the spetehc between the peak dynamic pensure in mer and the peak overiwn" "ad "* If Y es l PROPt riles DE THE IDE AL fit ASI' chapeer, and the remaanmg sectums will heah of the medium. I e. aer The seriatums el **k W AVE be devided m4+nly to a wenideratuin of gaten as i4 whn-h as the salue me bierg prewure sehsety. pastkle bw Peak w md) ubn-vane of the quanntanve aspects 4 i blast ~ mnderate temperaturn. the equasnm for sty. and peak dynamic preswre wdh the 3.53 The characternian of the blast uave phermmena m air The bauc rela-the shsk vetooty becomes sea teset, e peak oserprewure 48 wase hase been dmussed m a quahta-twenhirn anmng the pmperties of a blass desned from the foregomg equarkes. 82 6p~ are simwn graphkally in hg -1 55 inc manne an the.arber path of thn wave hahmg a sharp front at whish rhere U"4 I

• 7 p*

Ihr=ama.as c.. s.4 %.empen % e-om.unt.., w.2 u, % i

P-91152 f. May 3, 1991 -lardn]? ffaf Mociata STANLEY B. MARTIN Consultant, Petrochemical Fires and Explosions SPECIALIZED PROFESSIONAL COMPETENCE Research and applications of research to engineering problems in a broad field of fire and explosion protection. Analysis of. vapor evolution and dispersion, both underwater and in the atmosphere; evaluation of growth and decay of explosion-limit envelopes with time in a variety of wind fields, detonation potential, consequences.of boiling-liquid-expanding-vapor explosions (BLEVEs), airblast overpressures and thermal radiation fields resulting from l both confined and unconfined explosions, damage-causing and life-threatening effects of explosions and subsequent fires; development of concepts for prevention and mitigation. EXPERIENCE Length: Over 30 years of relevant experisnce. Affiliations: Stan Martin & Associates, Proprietor and Principal Investigator (since 1982) Los Alamos Technical Associates, Senior Scientist (part time, since 1982) SRI International (formerly Stanford Research Institute), Director, Fire Research Depart-ment (19,69-1982). URS Research Co., Manager, Fire Research Group (1965-1969) U. S. Naval Radiological Defense Laboratory, Research Scientist (1950-1965) l Related Experience: Broad-ranging research and problem-solving applications in fire and explosion hazards (see general resume) Experience Specific to Petrochemical Fires and Explosions: While at SRI. Mr. Martin investigated the risks and damage potential of releases of LNG and LPG in maritime accidents (see: " Cost Effectiveness of Marine Fire Protection Prog-l rams," Final Report to U.S. Dept. of Commerce, Maritime. Administration, November, 1978). He managed and actively participated in several experimental programs in which fire-fighting effectiveness of various agents and techniques were evaluted in situations involving petroleum-fueled fires (see, for example: "Extinguishants for Aircraft Fire hbO $AkbN a Mt j, Ohls lj hf

l l l t [' STANLEY B. MARTIN Consultant, Petrochemical Fires and Explosions (cont.) Fighting: Auxiliary Fire Suppressants," International Seminar on Aircraft Rescae and Fire Fighting, Geneva, Switzerland, 13-17 September 1976). Since 1982, Mr. Martin has devoted a major part of his project activities to accidet:ts involving fires and/or explosions fueled with petroleum products. For Varian Associates (Palo Alto), Mr. Mar tin personally assessed the potential for damage / injury caused by accidental leakage of on-site storage of liquified propsne. He assisted Dames & Moore (San Francisco) in an analysis of the risks of damaging fires and explosions from the accidental release of various petroleum products in a Persian Gulf scenario. l For Mobil Oil Co.(Exploration Norway), he assisted Sierra l Consultants International (San Francisco) in a detailed l study of possible consequences of failure of the planned sea-bed, natural-gas pipeline from the Statfjord Field in i the North Sea, and more recently in a comprehensive i fire / explosion safety study of Canada's Hibernia Field development. Also for Mobil (Chemical Company, Phosphorous Division), he analytically recreated a runaway aerothermo-che mi c a l -tae chani c a l scenario that resulted in a damaging explosion at a trimethylphosphite plant. ( Mr. Martin assisted Environmental Research and Technology, I Inc., in the preparation of an environmental impact statement for a proposed shipping terminal in the Santa Barbara area, by forecasting airblast overpressure contours associated with a hypothetical supertanker explosion, and thermal radiation safe-standoff distances for large petro-leum fires. These studies have generated several novel software programs for reliable and convenient machine computation of the fire / explosion hazards accompanying petroleum and petrochemical accidents. Most of these studies have included recommending cost /ef fective fire protection concepts and state-of-the-art systems for fire suppression and damage limitation. March 1988 j

STAN MARTIN & A S SOCI A T LCS Consultants in Fire and Explosion Safety CALCULATION OF MAGNITUDE AND CONSEQUENCES OF LARGE VAPOR-PHASE EXPLOSIONS Codes -E o r Industrial Accident Risk and ConsecIuence Analyses by Stanley B. Martin of Stan Martin & Associates Presented to: Eleventh International Conference on Fire Safety January 13 to 17, 1986 l 860 Vista Drive Redwood

City, Calif ornia 94062 (415) 365-4969

STAN MARTIN & A SSOCI A TE'S Consultants in Fire and Explosion Safety NARRATIVE During the past two decades, or so, my colleagues and I have been engaged in experimental studies and large-scale field-test measurements of fire and explosion phenomena. Some of these have a direct bearing on industrial and transportation accidents; others, while less obviously relevant, have given results that are nonetheless helpful in anticipating the consequences of such accidents and suggesting the need for safety measures. More recently, Stan Martin & Associates has had frequent occasion to apply this technology to industrial safety planning, some of it aimed at the anticipation of life-and property-threatening consequences of unlikely (yet possible and even credible) large releases of hydrocarbon fluids. An analytical methodology has evolved, which I will describe and illustrate. While the applications addressed thus far have been mainly blowouts of we11 heads, high-pressure pipelines, and pre-refinery Process equipment, much of the analytical development is also applicable to a variety of accidents involving the release of flammable

gases, liquids, and/or two-phase fluids.

Computer codeJ have naturally accompanied these developments, particularly in those situations of repetitive, labor intensive calculations. This development is summarized in the attachment. To date, however, these individual coding efforts have not been pulled together into a general utility, user-friendly package, despite the prospects for its widespread use. The physical understanding that serves as a foundation for this analytical methodology seems to be pretty well in hand. There are, however, some still unresolved issues, which I have attempted to bring out in my talk. One of these is the onset of instability in pool fires as size increases and fuel supply limits the per-unit-area burning rate. Another is deflagration airblast from unconfined clouds of gases / vapors, and how the resulting overpressures depend on ambient air motion and the site of flame initiation. Full confidence in analytical forecasts must await a satisfactory resolution of such incompletely under-stood issues. Low-budget experimentation could lead to the resolution of the instability

question, especially if it were coordinated with analytical modeling.

The second issue may not yield significantly without a costly series of large-scale tests. 660 Vista Drive Redwood

City, California 94062 (415) 5S5-4969

-.. - = _. STAN MA2% TIN & ASSOCIAT.E'S Consultants in Fire and Explosion Safety S CODES DEVELOPED FOR INDUSTRIAL ACCIDENT RISK AND CONSEQUENCE ANALYSIS j The analytical methodology applied by Stan Martin & Assocs, to the evaluation. of fire and explosion hazards of accidental discharges o' flammable fluids from pressurized process equipment and transport lines and carriers makes use of several computer j codes of our own development. Examples follow: o mechanics of subsea releases of pressurized gases and/or miscible two-phase

fluids, including plume flow and separation at the sea surface.

This was written as a Key-stroke program for the - HP-410. Since then,.the equations have been rewritten for improved generality and convenience. i They have yet to be programmed in machine language, however. l e atmospheric dispersion of cryogenic fluid spills and high-1 rate. releases of neutral-density gases (IBM B ASIC A). I Version CONCALC2 treats dispersion from a point source at the atmospheric boundaryl CONCALC3 treats unbounded cases. . Wind variables are speed and fou' gustiness. categories. e explosion potential of unconfined gas clouds and explosive mixtures in weak-walled enclosures; estimates far-field overpressures and their decay with distance from e::plosion center. (IBM BASICA filename, EXPLOP. BAS) e radiant heat levels from a fire plume (as a plane, rectangu-lar source); calculates safe standoff distances, given endurable fluxes as input.- (IBM BASICA.

filename, RADHEAT. BAS) e.

rates of. gas discharge from pressurized pipelines and reservoirs; - flow' may be either sonic or friction limited. (Fortran 4 -- - filename,- PIPE.FOR ; IBM BASICA

filename, PIPELE A K.B A S).

These. ' codes have been - successf ully employed in several industrial accident-consequence ' a n aly s e s.' .In a recent

study, 660 Vista Drive Redwood

City, California 94062 (415)- 365-4969

four codes were linked to accomplish the following: 1. Estimate the steady mass flow and nozzle velocity of pressurized gas issuing from holes of specified size. 2. Interrelate momentum dissipation by air entrainment with composition and velocity change, along the length of the jet of gas issuing from the hole, from its supersonic origin to the region where its directed flow effectively melds into the wind field. This result is used to decide how much of a role jet mixing plays in the formation of explosible mixtures. 3. Describe the size, downwind extent, and explosion potential of the steady plume formed by processes of wind shear. 4. Estimate the airblast overpressure field that might reason-ably result from the deflagrative explosion of the described plume. The estimates of discharge mechanics are based on the well known equatior. for adiabatic, 1sentropic expansion of a gas (ideal) through an orifice with pressure drop sufficient to ensure at least sonic flow. The second analysis uses results of the theory of jets developed by A. H. Eanury and presented in his book: " Introduction to Combustion Phenomena," Gordon & Breach Science Publishers, New York, 1975, pp 217-267. The analysis of atmospheric dispersion 3s based on the bimodal gaussian model of Pasquil as subsequently applied by Burgess et al.(15t h Combustion Symposium, 1974). Overpressure estimates follow developments of Brode, Porzel, and others for situations of noncondensed explosives. The efficiency factor is judgemental, but based on guidance derived from reviews of accidents (e.g.,

Brasie, W.C.,

and

Simpson, D.W.,

AIChE. Loss Prevention, Vol. 2, 91-102 (1966)). More

recently, boiling-liquid, expanding-vapor explosions (BLEVEs) have been reviewed, and a theoretical basis for code algorithms was developed to permit estimating the airblast overpressure field around such accidents.

This has not as yet been translated to software. Modifications have been made to the radiant-heat model for hydrocarbon-pool fires to increase its versatility and realism. The flame column is now modeled as a cylindrical radiator that bends in response to the ambient wind. Thus the hazard added by an unfavorably directed wind may be evaluated. Target points may be selected at elevated locations. (IBM BASICA

filename, FLAME 2. BAS) Further modifications to make the code more " user f riendly" are in progress, 10 December 1965 860 Vista Drive Redwood

City, California 94062 (415) 365-4969

_ _ _ _ _ _ _ _ _ _ _ _ _ - _ _}}