Document zJjRELO4wbO7aBJwa7qBoodz

" -:^3 cCO r r i r-( z.y GAS - \ c. W !R<. --\ 2 2- 35. 0561 n c'dfl S November 24, 1980 Docket 69-50 To: CGA Task Force, Uniform Fire Code Vinyl Chloride Committee Gentlemen: I have just received a copy of the 1977 report that you may not have seen entitled "Explosion Hazards Associated With Spills of Large Quantities of Hazardous Materials". It was sponsored by the Coast Guard and relates most closely to LNG/LPG cargos. Circulated for your information. Very truly yours. JCC:1s Enclosure J. C. Crawford o;\,0^$ O0'cA U.S. DEPARTMENT Of COMMERCE National Technical information Service omn0* AD - A 04 7 5 85 EXPLOSION HAZARDS ASSOCIATED WITH S P IL L S OF LARGE QUANTITIES OF HAZARDOUS MATERIALS PHASE II C. D. Lind, at al Department of Navy China Lake, California No ve mbe r 1977 Mod 00013?29 ADAO 47585 n. CG-D--35-77 EXPLOSION HAZARDS ASSOCIATED WITH SPILLS OF LARGE QUANTITIES OF HAZARDOUS MATERIALS PHASE II C. D, LIND J. C. WHITSON FINAL REPORT NOVEMBER 1977 KfltoouccDrr NATIONAL TECHNICAL INFORMATION SERVICE IL S- OCPAffTMCNT 0* CpMMtKI Document 1$ available to the public through the National Technical Information Service* Springfield, Virginia 22151 Freparad for DEPARTMENT OF TRANSPORTATION UNITED STATES COAST 6UARD Office of loaeerth and Development Weehinpton, D.C. 20590 MCD 000013730 NOTICE This document la disseminated under the sponsorship of the U. S. Department of Transportation in the interest of In formation exchange. The contents of this report reflect the views of the authors, who are responsible for the accuracy of the data presented. T*<Knt3l Report 0<jm*nrTion Pag* M*. mdl Acc*t<* N, 3. ft. CG-D-85-77 4. Tiff* *4 SmUiH* Explosion Hazards Associated with Spills of Large Quantities of Hazardous Materials Phase XI S. Rr1 0< November 1977 4- 0'lAltll*M C*4 7. Am***'*) 8. N*. C. D. Lind and J. C. Whitson . 0'9*n tif*M Ham* an* A*4r* Department of the Navy Naval Weapons Canter China Lake, California 93555 (0. W*rfc Umt M. (TRAIS) 3301.20.6 11. C**r*T Gr*T H*. DOT-CG-34095-A )3. 7yp* mi Rmmmri m*4 Pmum4 ** 12. Sp*.s**tnf A**Ar Nm* *4 Commandant (G-DSA-1/TP44) Final Report U. S. Coast Guard Headquarters Washington D. C. 20590 x. *****mf Af***r C*4m U. S. Coast Guard 12. SMppiMMivry N*** The tJ. S. Coast Guard Office of Research and Development's technical representative for the work perform herein vas LCDR M. W. TAYLOR. (G-DSA) 1*. Afcir*1 This report documents the results of Phase XI of a program aimed at quantifying the explosion hazards associated vlth spills of large quantities of hazardous material euch as liquefied patural gas and liquefied petroleum gas. The principal results of this phase of the work are: a quantitative empirical description of the burning behavior in fuel-air mixtures, an examination of flame acceleration processes, the observation that, in 17 large scale burn tests, no transition to detonation occurred, and that methane-air mixtures cannot be detonated with moderate size solid explosive boosters. 17. K*yW**d Explosion Hazard Spills Hazardous Materials 19. Clnif. (( --------*4*A Fern DOT F 1700.7 <-n a*1 o'-nOO *cs> It. Oiattibutl*** infmwt Document Is available to the Public through the National Technical Information Service, Springfield, Virginia 22161. 30. $wrty CU*4<. (W rM* pm^m) 21. N. mi 22. Pnmm TTnrl agsifl+4 ml ------------Re avttwf'taS 40 * 000013733 It s | o* I I ^ I *1 I1 on lit ot 01- O9k. - T*fM 0 01i. mw --*-- #. M<u R| ^ In pm *<1 t/ nmii t<i|3 5. Ir* tM (* M M N pi*l 9q! | *i*fnj --t 1*0 Ht4 9n9 r^ll n K 0 Nl II to* jnniBA iiiw *** jiy** IMI| WIH Ml wimnw it (* noil > ipvwO < i laMW t4 tsvN C* M 0 1 M I****4* M3 pw* Mn* MM *--* It >0 II III l,*00'*ll M"BM WIMtlll woM IW|M mM IWtttWIM tlMII vv PM* M| |P fp*l n WIMMn ii 11 9$9*mm M *00 MMMHtW H19M11 *1 i 1* *|*tM*l* ** t toioiMII |i|ip| aoi| >*|i*ji03 nnfitiKy 1 1 1 I'M 1 0 I-* M| m m V* M> |0|0l 1 m W * Z-- =-- - Z-- - -- m z~ m =-- * ** -- 5-- Zm ` " -- i- w m * r- :-- -- * t5H P- * 'Mt 'i*v '? |M) * *<mMHrin'vto im m 'ti mfi n ><w 'mhm <wnf i* H"I*.* WI|lMwM a. mill} lit OTHMiHf1* *0*11/1 womrt t. it) ivmvsuim * mN i Mi| i Mill i flf t win I" IWKIMIIW 1* M imiiin1* * coo f III (' I* n i * |W* M" --H0 MM* KM* hum Mil Wllll||ll wilim i*- |M M M 9 --m Ml o tMlO 0 tlWIM| Imtwpp i--M i" r MW |I|*I mniwii tniWt 0 IWWW|1^ m MW Miwnm bwimmiOT* mnioA * It m mu Ml IM|I pww* IMM UifCiMi) urn MM rt mm iwiM Ml IM| MU II MM imM Tlii n MM n c MM n. MM t* t" *" mrn r* M *19*11 r*M*i rti I* VW>M mo*I WA IMM l*OTt| tm*o*n *|i|og| 01 M0f*i*03 nnpnMi SHOmi N0ISH9ANO3 3HHH o s CONTENTS Introduction ................................... ,.................................................................................. 3 Phase I Results ................................................................................................................... 3 Hemisphere Bum Tests ....................................................................................................... 7 Experimental Equipment ..................................................................................... 7 Experimental Procedure ............................................................................................ 10 Results .......................................................................................... Augmentation of Flame Velocity by Turbulence ............................................................ 15 Hemisphere Detonation Tests ................................................................................ 16 Conclusions ............................................................................................................................. 19 Future Plans ......................................................................................................................... 20 Appendix A ......................................................................................................................... 23 Appendix B ........................................................................... 29 Appendix C ......................................................................................................................... 33 HCD 000013734 INTRODUCTION The nation's navigable waters provide an efficient means for transporting large quantities of bulk materials. Among bulk items_commonly shipped in large quantities are liquefied flammable gases such as ethylene, butadiene and petroleum products. Since the transport of such materials generally becomes more economical as the quantity shipped at one time is increased, there has been a trend to ever larger shipping containers. As a result of the recent advent of large tankers for shipping liquefied natural gas (LNG), there now exists the potential for accidental formation of extremely large vapor clouds. The need to know and understand the burning characteristics of such vapor clouds led to the investigation reported herein. Specifically, this study was aimed at determining the possibility and conditions under which a transition from burning to detonation of such a vapor cloud might occur. Phase l of this investigation was reported on previously (Ref. 1). The Phase I effort resulted in (1) a phenomenological description of an accidental spill, (2) an examination of the detonation properties of fuel-air mixtures, (3) a qualitative theory of nonideai explosions, and (4) a plan for Phase II of the program. This report provides a limited discussion of the Phase l results and details of the Phase il effort which endeavored to quantify the explosion hazards associated with large spills of hazardous materials such as LNG and liquefied petroleum gas (LPG). PHASE I RESULTS The description of an accidental spill was studied to determine the magnitude of the vapor cloud formed and the conditions that might affect the burning behavior. The primary fuel of interest was methane since it is anticipated that large quantities of LNG will be imported into the U.S. The detonation properties of methane were investigated both theoretically and experimentally to determine the properties for damage predictions as well as to devise instrumentation for later experiments to detect a detonation. The theory of deflagrations in fuel-air clouds was examined by Professor F. A. Williams of the University of California. San Diego. The conclusions reached by this study were J1. A damaging pressure wave could be produced by a deflagration, but damage would be restricted to approximately the area covered by the cloud. 2. The development of a detonation from a deflagration does not appear likely 3 MOD 000013735 Preceding page blank Ideal detonation properties of fuel-air mixtures have been calculated using a computer code. This code assumes chemical equilibrium in the detonation wave and calculates detonation pressure, temperature, and velocity as a function of initial composition. The calculations do not determine if the mixture will detonate; they do predict the properties of the detonation if it should occur. Some results of these detonation calculations are shown in Table 1. The detonation limits are generally narrower than the flammability limits and are dependent Fuel Ammonia Butadiene Ethylene Ethylene oxide Methane Propane Vinyl chloride TABLE 1. Calculated Detonation Properties. Composition.* 21.3 3.67 6J3 / 7.73/ 9.43 4.02 i:.:^ Temperature.* K 2320 3100 3100 29 SO 2730 2320 2320 Prci.su re,* ban 17.n 19.0 17.3 13.4 16.3 17.J 19.2 Velocity,* m/s 1330 134(1 1340 1830 1810 1800 1810 LI*L.e volume % ts.o 2.0 2.7 3.6 5.0 2.1 3.6 UFL,f volume % 23.0 12.0 36.0 100.0 15.0 9.5 33.0 9 Stoichiometric composition assuming H^O. CO* products, percent by volume. Calculated temperature, pressure and velocity *i the stoichiometric composition. c Experimentally determined lower and upper flammability limits from "l-bmmability Characteristics of Combustible Cases and Vapors.** by G. Zjbctakia. Bureau of Mines. Bulletin 627. 1965. on the strength of initiation. The detonation properties at stoichiometric composition are not widely different for the different fuels considered. However, differences are apparent when the curves for one of the properties, such as pressure versus concentration as shown in Figure 1, are examined. Note that fuels such as propane and methane have relatively sharp peaks; fuels such as ethylene oxide and vinyl chloride have rather broad peaks. This suggests that methane and propane might have narrow detonation limits, should they detonate, and ethylene oxide and vinyl chloride might detonate over a wider concentration range. The vertical lines on Figure 1 are the flammability limits. The detonation tube facility consisted of an 0.6-meter diameter steel tube with the following attachments: a gis introducing system, piezoelectric gauges for measuring pressure and velocity, and thermocouples for measuring temperature. The arrangement is shown in Figure 2. This facility was used tor two types of experiments: detonation and burning. For the detonation experiments the tube was 1.8 meters long, for the bum experiments 3.6 meters long. In the detonation experiments, an explosive booster was placed at one end of the tube, both ends were covered with polyethylene film, and a known amount of fuel was introduced. After mixing, the explosive booster was detonated and the results were sensed by transducers in the side of the tube. 4 36 DETONATION PRESSURE. BARS FIGURE 1. Calculated Detonation Pressure. (BOTH SNOB) FIGURE 2. Detonation Tube Facility. Results of these experiments are shown in Table 2. With methane, no reaction was observed using a small booster (5 grams tetryl); using a large booster (90 grams sheet explosive) gave the results shown in the table. Notice that the pressures and velocities were far below the calculated values. The detonation limits for methane appear to be 6-12.5%; although, with the low extent of reaction observed, it is difficult to determine the limits accurately. The results for methane are somewhat surprising since Kogarko (Ref. 2) obtained detonations with the calculated properties. The propane results were obtained with the 90-gram booster; with the ethylene oxide the 5-gram booster was sufficient. The purpose of these detonation experiments was not to obtain precise measurements but only to verify the calculated values and to demonstrate that a detonation could occur. TABLE 2. Measured Detonation Properties. Experimental Caleukted I'UCl Concentration, volume ? Pressure. bar Velocity. m/s Pressure. bar Velocity. m/s Methane (90 g booster) 3.0 6.0 7.0 8.0 9.0 10.0 11.0 12.0 3.4 3.6 6.3 3.4 7.0 7.87.1 6.6 950 930 1010 1030 1030 1050 930 910 11.2 12.6 14.0 15.2 16.0 16J 16.6 16.4 1490 1590 1675 1730 1780 1820 1830 1820 Prupune (90 g booster) 3.0 3.6 13.7 13.2 1710 1800 15.2 16.6 1675 1730 Kihyicnc oxide (3 g b<*o*tcr) 3.4 9.7 19.9 15.3 17.0 I9J 1770 1840 1810 13.3 19.6 19.9 1700 1880 1900 In the bum experiments, one end of the tube was closed with a steel plate and a spark was used to ignite the gas mixture at the closed end. A high voltage firing unit (3000 V, 10.5 J) was used as a spark source. Fine thermocouples (28-gauge) were used to sense the flame front for velocity calculation, and piezoelectric gauges were used to detect pressure. The results of these experiments are shown in Table 3. The pressures obtained were less than predicted by Kuhl, Kamel, and Oppenheim (Ref. 3), perhaps because of the shortness of the tube (3.6 meters). These experiments indicated the order of reactivity for the fuels as methane < propane < ethylene oxide. Also, the pressures generated were low, and the spark source, which would be used in later experiments, did not cause immediate detonation. 6 mcd 000013738 TABLE 3. Measured Flame Velocities and Pressures. < 3000 V, ;0.i 3, spark ignition sourer). t uet Concentration, volume, Cc Velocity. m/i Pressure, bar Methane Propane hthylcne oxide 3.0 9.0 94 10.0 4.0 3.0 6.0 7.0 7.5* 3.0 43 63 38 35 123 91 116 233 270 - no 0.014 0.019 0.019 0.Q26 0.034 0.033 0.061 0.110 0.120 0.076 HEMISPHERE BURN TESTS Detonation properties and eariy ignition behavior can be studied with small scale experiments. However, the evolution of combustion in a fuel-air cloud, including flame acceleration and transition to detonation if they occur, can only be studied with larger scale tests. It would be desirable to study the burning behavior in completely unconfined clouds; however, it is difficult to control the position and concentration of the mixture when unconfined. Therefore, plastic film hemispheres were used to contain the mixture. The calculated internal pressure necessary to buret the film is very low (1.4 x 10"3 bare) and its thinness and low density should produce weak pressure wave reflections, so the presence of the film was predicted not to affect the burning properties, and no effect was detected in the tests. The hemispherical shape maximizes the flame travel distance before a boundary is reached for a given gas volume and film cost. EXPERIMENTAL EQUIPMENT The experimental arrangement for the 5-meter-radius hemisphere tests is shown in Figure 3. The arrangement for the 10-meter-radius hemisphere tests was similar, except a larger fuel line was used. The air was introduced into the hemisphere by a centrifugal blower via part of the instrumentation channel. It was found that the blower inflates the 5-meter*radius hemisphere in about 1 hour. Additional fans were used during the first part of the jinflation of the 10-meteiwadius hemispheres to decrease inflation time. The inlet to the hemisphere was covered by a "flapper" valve which directed the incoming air tangentially into the hemisphere, producing turbulent mixing. When the blower was turned off, the "flapper" valve closed, preventing escape 7 oooo^39 DETAIL Of HEMISPHERE BASE of the gas mixture. The partially inflated 5- and lO-meter-radius hemispheres can be jeen in Figures 4 and 5. FIGURE 4. 5 m Radius Hemisphere. 8 ooooi3740 rtCt> FIGURE 5. !Q m Radius Hemisphere. The fuel to be investigated was introduced into the hemisphere through a 30-meter-long tube (two tubes for the 10-meter-rudius hemisphere) from storage tanks containing the proper amount of fuel to give the desired concentration. For the fuels stored as liquids (propane, butadiene, and ethylene oxide), a hot water heat exchanger was used to vaporize the fuel. The fuel-air mixture was ignited by a spark. The equipment used was an exploding bridgewirc detonator tiring unit which consisted of a 3-af capacitor charged to 3 kV (13.5 joules stored energy) which discharged through 27.6 meters of coaxial cable to a 1.65-mm gap. This unit was used because it was already available at the test site, and because it gave a sufficiently bright spark to be seen on the photographs, thus marking the time of ignition. Recent work by J. H. Lee (Ref. 4) indicates that detonations can be directly initiated in some gas mixtures by a spark. Because of this possibility, the characteristics of the spark used were accurately determined. Details of these measurements can be found in Appendix A. The results indicate that the energy dissipated in the spark up to the tlrst quarter cycle was 2.79 SIcm and the power level was 0.361 MW/cm. Accord nig to Lee a spark with these characteristics would be sufficient to initiate a detonation in an acetylene-oxygen mixture. Since the mixtures used in the hemisphere bum tests are much less sensitive than acetylene-oxygen, they would not be expected to detonate directly from the spark. Photographic coverage for thetests was provided by 1000-2000 frame-per-second and 400 frame-per-second cameras positioned at ground level. In addition, for tests 4 through 19, a 1000 frame-per-second camera was suspended 4Q meters overhead. Because of the difficulty of precisely aiming the overhead camera, a short focal length lens was used giving a wide lield of view. Tests 4 through 19 were 9 oO conducted at night. In order to provide a distance scale on the film for these tests, a flashlight was placed at each side of the hemisphere and pointed at the cameras. For the daylight tests, the film was exposed for ambient light. For the night tests, the lenses were opened two f-stops. A time mark generator placed time marks on the film for accurate frame rate determination. The pressure measuring equipment consisted of eight piezoelectric gauges placed flush with or near the ground, and in a line from the center of the hemisphere to its edge. For tests 1 to 10, odd-numbered gauges were calibrated for a maximum pressure of 20 ban, even-numbered gauges for a maximum of 2 ban. The gauges were calibrated in this manner to ensure a recording of pressure whether or not detonation occurred. The odd-numbered gauges would provide a record of detonation; low level pressures would be recorded by the even-numbered gauges if there were no detonation. Since pressures were not detected in tests 1 to 7 and only a very slight indication of pressure was recorded in tests 8 to 10, ail ^uges were subsequently calibrated for a maximum pressure of 0.8 bar for tests 11 to 18. It was found that electrical noise in the system limited the lowest pressure that could be recorded to 0.1 bar. The output of the charge amplifiers was recorded on a high speed instrumentation tape recorder. One channel of the tape also recorded time signals so a time base was available. A signal from the ignition source (spark gap) was recorded and, since the flash from the spark was visible on the photographic coverage, both the electronic and photographic recording systems had the same zero time. In an attempt to determine the effect of perturbations on the flame front, three types of obstacles were placed in the hemisphere in test 6: (1) a steel grid 0.5 m X 1.5 m placed flat on the concrete pad to increase surface roughness; (2) a steel grid 0.3 m x 0.3 m (25-mm-square grid holes) placed vertical to the ground and facing the ignition point to increase turbulence of the flame as it passed through the grid; and (3) a structure built of 50 X ISO x 150-mm steel blocks which presented a 0.3 X 0.6-m reflecting surface to the flame, and included a 0.5-m long tunnel with a 150-mm-square cross section for the flame to travel through. The open tubes of test 11 were placed on the concrete test pad radiating from the ignition point. The fan was placed on the opposite side of the ignition point facing the tubes to ensure the mixture entered the tubes. The wire screen cylinder of test 12 was 500 mm in diameter and 240 mm high, made of 0.7-mm diameter wire with 5.5-mm spacing. EXPERIMENTAL PROCEDURE The general procedure for the tests was as follows: 1. Measure and connect fuel supply. 2. Check pressure measuring system. 3. Load film and check photographic system. 4. Attach plastic film hemisphere to frame on ground. 5. Start blower. 6. When hemisphere is about half full, turn on flashlights, open fuel valve, and clear area. 10 & ocP' 7. When hemisphere is full, turn off blower and wait 10 minutes for turbulence to subside. 8. Start firing sequence. This sequence automatically starts cameras and tape recorder, discharges capacitor (10 J, 3000 V) to spark to ignite gas, and stops tape recorder and cameras. The test sequence and conditions are shown in Table 4. Trn no. 3 4 5 6 7 3 9 to 11 12 13 14 15 16 17 IS 19 1-uei Propane 1 1 Methane Propane Methane Ethylene oxide 1 Propane I * Methane Gthylcne Acetylene Methane Butadiene Acetylene Methane TABLE 4. Test Sequence and Conditions. Hemisphere Burn Tests Concentration, volume ' Ambient temperature. V Remark! 4.0 5.0 5.0 3.0 10.0 5.0 10.0 30 Daylight test: llame barely vtsibie. 30 Concentration increased to enhance visibility. Mixture pre ignited by static discharge. No data. 30 Marne barely vwblc; all subsequent tests done at night. 30 30 20 Obstacles placed in hemisphere to increase turbulence. 20 10 Meter Radius Hemisphere 7.7 7.7 7.7 5.0 4.0 10.0 8 8 Instrumentation failure. No data. 8 27 Open lubes 2 mcicrs long and 30*. 100*, and 150-mm diameter, placed in hemisphere, l-'an used in tha and all subsequent tests to insure uniform mixture and that mixture was in lubes. 27 Wire screen cylinder placed over ignition source to increase turbulence. 15 S Meter Radius Hemaphcru 6.5 3-S 10.0 3.5 7.7 10JJ 15 3 Acetylene cylinders did not completely empty, giving low concentration. 3 Detonation test. 10 10 30 Detonation test. The films from the tests were assessed using frame-by-frame analysis on a projection comparator. The frame rate was determined from the timing marks, and the magnification by measuring the image of an object of known dimensions. Time measurement was accurate to better than 2 ni and distance measurement accuracy was 20 mm to 100 mm (depending on image sharpness). The pressure record on the instrumentation tape was assessed by playing the tape back through an oscilloscope and photographing the trace. The oscilloscope trace could then be compared to calibration curves for each pressure gauge to obtain pressure measurements. RESULTS Detailed observations from each hemisphere test are provided in Appendix B. Qualitatively, ail tests appeared similar. The photo record showed the ignition spark followed by about a 100-ms period when the flame was difficult to detect. The flame then appeared as a blue, expanding hemisphere. As the flame expanded, it became rough with a "pebbled" appearance. This structure increased in size to about 0.4-1.0 meter with a finer structure superimposed. The plastic film tore at the base between 300 and 400 ms (for the 5-meter tests) and the entire hemisphere rose vertically. Many tests differed somewhat from this generalized picture. The plastic film always tore at its base where it was attached to the wooden ring; however, in some tests, the film tore unsymmetricaily and lifted off, due to expansion during the combustion process, by tilting or rotating around the region where it remained attached. Depending on how strongly the film was held, the flame shape was distorted away from the held region. This was not a serious problem, but it did represent a deviation from the ideal symmetrical combustion. Another difference in some of the tests was multiple ignition of the charge by flame getting into the instrument channel and accelerating because of confinement. This flame traveled ahead of the main flame and then broke out of the channel, igniting the main combustion mixture at a number of points inside the hemisphere. In some tests this behavior caused a series of spherical flames of different sizes to appear. These eventually coalesced without noticeable accelerations. It should be emphasized that, even though this behavior destroyed symmetry and produced a variety of odd shaped flames, the very complex flame propagation patterns that occurred did not produce noticeable accelerations of the flame or any other important unusual effects. In fact, it was surprising that, even when there was very unusual flame propagation, the final flame volume always retained a relatively hemispherical shape; this simply indicates that expansion of the products behind the flame tends to push the surrounding gas uniformly in all directions so that roughly spherical symmetry is maintained in the final flame volume. In none of the 5-meter tests was a pressure wave recorded. This result agrees with the subjective impression of personnel at the test site that little sound was produced. This result is perhaps not surprising since Kuhl et al. (see Ref. 3) have predicted that a flame speed of 10 m/s should produce a pressure wave of 0.1 bar and the lower limit of detection of the equipment used was 0.1 bar. Pressures were recorded during ethylene oxide tests; however the signals were barely detectable and difficult to distinguish from the signal caused by the heating of the gauges by the flame. An example of the measurements made from burn tests photo records can be found in Figures 6 and 7. Complete results are given in Appendix C. The horizontal radius of the flame is measured from the ignition point to the flame boundary at a distance of about 0.2 meter up from the concrete pad so that distortion due to the surface would not affect the measurement. The horizontal radius versus time curves were straight lines in all cases indicating a constant velocity. It is apparent (especially in Figure 6) that the curves do not pass through the origin. It is possible that some time is required to build up to constant flame velocity due, perhaps, to the 12 MOD 000013744 13 MCI) 000013745 VERTICAL RADIUS, m HORIZONTAL RADIUS, development of turbulence. If this time to build up to constant velocity is not reproducible, it would explain why the curves appear displaced. Indeed, if the time axes are shifted, many of the curves can be superimposed. The vertical radius is measured from the ignition point vertically to the top of the flame. The vertical position versus time curves are not straight lines but show that the velocity is accelerating. However, the velocity appears to be reaching a limiting value especially in the lO-meter-radius tests. Table 5 summarizes the results of the hemisphere tests. TABLE 5. Summary of Results of Hemisphere Tests. Tw* Size. no. meter* fuel Concentration, volume ` ' Horizontal velocil y. m/s 5 3 Methane 75 13 ii) f ! 5 Propane 12 10 33 65 43 1 1 10 10 10 lOhvlene oxide 8 10 '' 14 5 I'thyiene 15 5 Acetylene 18 5 '' ! 7 5 llutadicnc 10.1) io.n 10.0 4.0 4.0 3.0 5.0 5.0 5.0 7.7 7.7 6.5 3.5 7.7 3.5 5.8 4 V2 h 6.1 4 6.9 H.J 9.6 13.4 14.7 H.8 3.6 23.7 3.9 Vertical VCilH.1l V at 3 m. at 8 m. m/* m/s 7.3 7.3 6.5 6.3 7.K 7.4 9.5 10.2 9.9 15.2 16.0 17.3 4.6 35.4 55 8.9 ... 10.6 ... 12.6 22.5 22.4 ... ... - 9 Burning fuel in (h iiwnimcmatian channel distorted the shape >> itc tlame jnd no Horizontal velocity muld be obtained. ^ Test performed in daylight and [lame ba.w was insuilicicntly visible fur horizontal velocity to be obtained. It was expected that several flame acceleration processes would be observed during the tests. Boundary layer acceleration along the surface of the pad was expected and observed as a distortion of the hemispherical shape of the flame at the base. The acceleration was not pronounced, however, even in test 6 where a steel grid was placed on the pad to increase roughness. It is possible that sufficient turbulence was already present so that boundary layer-induced turbulence did not appreciably increase flame velocity. This turbulence, apparent over the entire flame hemisphere, did increase the velocity well over that expected from a laminar flame. A discussion of the cause of this general turbulence can be found in the next section. The general turbulence also masked the turbulence produced by the wire screen placed over the ignition point in test 12. Wagner (Ref. 5) has observed that a screen grid increases flame velocity greatly. 14 Mod ooooi3?46 The buoyancy of the hot combustion products is probably the cause of the difference between the horizontal and vertical flame velocities, A rough calculation based on Scorer (Ref. 6) yields approximately the velocity difference observed. The obstacles placed in test 6 and the open tubes in test 11 did not appreciably influence the flame; however when the flame reached the 2- by 6-inch (5* by 15-centimeter) wooden ring to which the film had been attached there was a noticeable acceleration. AUGMENTATION OF FLAME VELOCITY BY TURBULENCE When a volume of a fuel-air mixture is centrally ignited, the flame moves radially outward from the ignition point. One effect of the flame is to raise the temperature and, thus, for a constant pressure process, expand the gases within the flame envelope. This expansion causes the flame to have a higher velocity than it would have if no expansion occurred. Thus, the measured velocity is made up of an expansion component and the basic flame speed (S). This flame speed may be calculated by multiplying the observed flame velocity by the density ratio across the flame. The density ratio can be calculated from the temperatures and average molecular weights before and after burning. The results of such calculations for the gases tested are shown in Table 6. The flame speed so calculated is labeled S'. The literature value for the laminar flame speed is given as S. The ratio of calculated speed to laminar speed S'/S is labeled ii. It is proposed that this is is the augmentation of the flame due to turbulence and, as can be seen, varies from 1.6 to 2.1 for all of the fuels except butadiene. As noted in the description of the individual tests, Appendix B, the concentration of butadiene may have been low due to incomplete vaporization of the fuel. TABLE 6. Augmentation of Flame Velocity by Turbulence. 1 uel Density Horizontal Ha me Concentration. ratio (lame s'. S.J volume temp., *C across velocity. m/t m/s Hamc nt<s if Methane Kthyiene Acetylene Propane Butadiene Lthyienc oxide 10.0 6.5 7.7 AO3.3 7.7 I960 2100 2325 1980 2!(>0 2140 0.131 0.137 0.1 12 0.128 0.117 0.1 13 5.8 3.8 13.7 6.1 3.9 [4.0 0,76 1.21 2.65 0.78 0.46 1.58 0.37 0.75 1.56 0.43 0.60 1.0! 2.1 1.6 1.7 1.3 0.8 1.6 * Kd Harnett, H. C. jnd Hibbard. "Basie Considerations in the Combustion of Hydrocarbon 1-ueU with Air'*. K. National Advisory Commute tor Aeronautics, Report No. 1300. 1959. Pp- 256-*. 15 mod 000013V 4V The turbulence noticed in the flames could be caused by a mechanism proposed by Landau (Ref. 7). According to this mechanism, an initial incipient wrinkling of the flame front caused by concentration or temperature in homogeneities (or in the case of the present experiments by the ignition spark) is amplified because combustion products moving normal to the flame front converge behind any bulges and diverge behind depressions. Therefore, the pressure of the combustion products increases behind bulges and decreases behind the depressions. This pressure difference causes an increase in the wrinkling of the flame front, thus an increase of the burning surface area and an acceleration of the flame velocity. A description of this process is given by Shchelkin and Troshin (Ref. 8), They cite experimental evidence that this turbulence becomes significant when the Reynolds number becomes greater than 104. The Reynolds number (Rfl) is defined as follows: Re p SR v where p is the density, S the laminar flame velocity, R the radius of the flame and y the viscosity. For the fuels tested, the Reynolds number is above 104 for radii over 0.1 to 0.5 meter, depending on the fuel. Thus, the turbulence described would be significant in these experiments. It is interesting also that the calculated radii are the radii when the flames first become visible. Based on small burner experiments, Bollinger and Williams (Ref. 9) have reported augmentation ratios (^) for Reynolds numbers equal to those in the hemisphere tests between 1.5 and 2.0. Thus, the augmentation observed is in agreement with past experiments. HEMISPHERE DETONATION TESTS Two attempts were made (tests 16 and 19) to detonate a stoichiometric (10% by volume) methane-air mixture in 5-mctci^radius hemispheres. The experimental arrangement was the same as for the hemisphere bum tests except soiid explosive boosters were used and the cameras wen: run at 8000 frames per second. In test 16 a 1.35-kg and in test 19 a 2.05-kg hemispherical Composition B booster was used. The photo record from both tests appeared similar. On initiation of the booster, an intense light appeared from the detonation products. These products expanded rapidly to about 3 meters, then a blue flame front became visible. The flame front expanded much more slowly than the initial detonation product expansion. Figure 8 is one 16 Ol3748 FIGURE 8. Attempted Detonation 10CJ- Methane, frame from the high speed photo record of test 19. The position of the flame front in tests 16 and 19 are shown in Figures 9 and 10. respectively. The flame reaches 3 meters in 3 ms, then expands at a velocity of 34 m/s. This is a higher velocity than previously observed, perhaps because it is moving into shock heated and accelerated mixture. Figure 10 also shows the position of the shock FIGURE 'J. Flame Position. Detonation Tt *16. 17 MCI) 00U01374y AP. iAHS wave as recorded by the pressure gauges, and the time the piastic Him tears. Figure 11 shows the predicted pressure from a 2.05-kg solid explosive (Ref. 11), actual measurements from a booster only, and the measurements from test 19. P*fOICTSO 2.06 kg COMP B 18 MOD 000013750 Bcc.minc I tic idiock wave moves jw;iy from the ll.iine, and the lljnie reaches j con.slunl velocity far Ionn than l he Jctmiatiun velocity CI >S 10 m/sj, Lind the shock wave decays to far loss Hum the dclonalion pressure (lb.3 bars). the evidence indicates j sustained detonation did not occur. The pressures measured lor the methane-air mixture were greater than either predicted or measured I'or the booster alone. This could have been caused by partial reaction of methane in the shock wave. 'I lie failure lo obtain a detonation is in disagreement with the work ol Kngarko (Kef. 2) but is consistent with that of Hull et al. I Kef. 10). !he latter researches determined the amount ol solid explosive required to detonate mixtures of iiiethaue ami oxygen diluted with nitrogen. An extrapolation of their data indicates (hat a charge of 22 -kg of solid explosive would he required lo detonate a mcthane-.iir mixture. Consider a hypothetical experiment with a 2.3-meter-diawcler vertical tube open at the lop. "Hie tube and a region above the tube contain a methunc*uir mixture. By some means a detonation Is produced in the tube and moves upward and out of the tube. Tile pressure in the wave as it enters the region above the tube is the detonation pressure (lb.2 bars). This is also the* calculated pressure of the shock wave from the 2.05-kg solid explosive charge when it has expanded to this diameter (2.3 meters). In the experiment, even though the pressure was higher due to some contribution of the methane, no sustained detonation was produced. This implies that a detonation in an enclosed space* (possible because of confinement) would have to exit the space through an opening larger than 2.3 meters to produce a sustained detonation in an u neon fined methane-air cloud. It was not determined how much larger the opening would have to be. if indeed an unconfined detonation is possible, but only that openings smaller than 2.3 meters do not produce a detonation. In the above paragraph it was stated that the shock wave at 2.3-m-diameter radius had the same characteristics as a detonation wave in methane-air. This is not strictly correct since the detonation wave has a shock front of twice the detonation pressure. This so called spike pressure is usually not considered in detonation effects because it is of short duration. However, if the "spike" pressure is effective in producing a sustained detonation the above argument is still valid but with a limiting diameter of 1.5 meters. CONCLUSIONS The unconfined burning characteristics have been investigated for premixed fuel-air mixtures using methane, propane, ethylene, butadiene, ethylene oxide and acetylene fuels. Although attempts were made to optimize conditions for transition from burning to detonation in unconfined mixtures, no transition occurred. Flame accelerations did occur; however, the flames appeared to reach a constant velocity after an initial acceleration, and accelerations which would be expected to lead to a transition to detonation in larger clouds were not observed. 19 MCD 000013751 A sustained detonation in unconfmed methane-air mixtures is not produced with solid explosive boosters of up to 2 kg. This implies that, if a detonation occurred in a confined space due to confinement, the exit of the detonation wave from the confinement would have to be through an opening over 2.3 meters in diameter, or the detonation would not be sustained. How much larger the opening would have to be was not determined. FUTURE PLANS Several areas of work originally planned for Phase II will be reported separately. These include the following experiments and studies: 1. Long Tube Experiments. It is known that ignition of a flammable mixture at the closed end of a long tube will lead to a detonation. These experiments will determine the effect of venting on the transition to detonation. 2. Ignition Source Survey. A survey will be made of possible ignition sources that could be present in the vicinity of an accidental LNG spill. The strengths of these sources will be compared, and the likelihood that these sources could produce direct initiation of detonation will be determined. 3. Large Spill Feasibility Study. A study will be made of the feasibility of conducting very large spill tests of LNG and the benefits of conducting such spills. 4. Methane-Propane Detonation Tests. A determination will be made as to what percentage of propane in methane is required to produce a detonable mixture. Phase HI of the Vapor Cloud Explosion Study will consist of a study of the burning characteristics of LNG spilled on water. A facility has been constructed with the capability for rapidly spilling up to 6 m3 of LNG onto a 50- by 50-meter pond. Measurements will be made of the LNG pool size, the heat transfer from the water to the LNG pool, the vapor composition, the flame size and shape, the radiation from the flame, and the spectral distribution of the radiation from the flame. There will be two series of tests. The first series will involve spilling the LNG and immediately igniting the vapor. The second series will consist of spilling the LNG, permitting it to vaporize, and igniting the vapor downwind. Six of the first series of tests have already been carried out; however data have not been fully analyzed. MCD 000013752 20 Rl-.I-TKINCIS 1. U.S. Coast (luard. "Explosion Hazards Associated with Spills ol Large Quantities of Hazardous Materials, Phase I," liy IX Liml. October 1974. (U.S. Coast Guard Report CG-D-30-75, NTIS Aj) A001242.) 2. S. M. Ko'jStrko. "Detonation oT Methane-Air Mixtures and the Detonation Limits of Hydrocarbon-Air Mixtures in a l.arge Diameter Pipe," Soviet Physics. 3. 1904. I95S. 3. A. L. Kuhl. M. M. Kamel, and A. K. Oppeuheim. "Pressure Waves Generated by Steady Flames," Transactions of the I4th Combustion Symposium, The Combustion Institute, Pittsburgh, Pa., 1973, p. 1201. 4. R. Knystantas and J. II. Lee. "Combustion and Flame,'' ^2- 221-228 (1976). 5. K. J. Dcrgc ami H. G. Wagner. "Acceleration ol Spherical Flames," Deuxiemc Symposium Huropccn sur la Combustion, Orleans. France (1975) pp. 253-258 6. R. S. Scurer. Natural Aerodynamics, Pergammi Press, N, Y. 1958, pp. 161-167. 7. L. D. Landau and E. M. Lilshits. "Fluid Mechanics." Addisan-Weslcy, Reading, MA. 1959. 8. R. !. Schclielkin and Ya. K. Trushin. "Gasdynamics of Combustion/" Translated by B. W. KuvshiuolT and L. HoUscliiag, Mono Book. Baltimore, MA. 1965. 9. L M. Bollinger ami D. T. Williams. "Effect of Reynolds Number in Turbulent-Flow Ranjse on Flame Speeds of Buasen Burner Flames," National Advisory Committee for Aeronautics, Report 932, 1949. 10. D. C. Bull. J. E. Elswortti, and C. P. Quinn. A Study of Spherical Detonation in Mixtures of Methane and Oxygen Diluted by Nitrogen, J. Phys. D: Appl, Phvs,. 9. 1976. 11. Wilfred E. Baker. "Explosions in Air." U of Texas Press. *1973. 000013753 MCD 21 Appendix A SPARK INITIATION SOURCE USED IN THE HEMISPHERE TESTS The source of ignition I'or the hemisphere tests consisted of a spark produced between the electrodes shown in Figure A-l by an exploding bridgewirc detonator firing unit. The essential features of the firing unit, manufactured by Reynolds Industries are shown in Figure A-2. The electrical characteristics of the discharge circuit are: voltage 3.08 KV. capacitance 3.05 yf. inductance A.40 yh, and resistance 1.10 Cl (with spark gap shorted). Most of the inductance and resistance are in the 27.6 m of coaxial cable. COAX CABLE FIGURE A-l. Spark Gap. Reynolds Industries. Inc.. P. 0. Box 1170. Marina del Rey, CA. 90291. MCD 000013754 23 Preceding page blank EGAG GP-90 SPARK GAP * In order to measure the characteristics of the spark a 500 MHz bandwidth coaxial current and voltage viewing network was inserted between the spark gap and the 27.6 m of coaxial cable. The current through the gap, the voltage across the gap and the output of a photodetector measuring the light output of the spark were recorded on oscilloscopes. Typical recordings are shown in Figures A-3, A-4 and A-5. to jA/div FIGURE A-3. Spark Current. 2d MCD 000013755 SO V/div 0.3 V/div I0i*/dv FIGURE Ao. Spark Light. A calculation of cite current through the gap (neglecting gap resistance) by the equations: 25 MOD 000013750 E exp -at sin u( j CJ R_ a 2L where: l w3 LC a* i - current (amps) t = time (sec) E = voltage (3.08 KV) L * inductance (6.4 x 10"6 henrys) R = resistance (1.1 2) C 3 capacitance (3.05 X 10"* farads) gave results in good agreement with the measured current. The gap resistance (r * e/i), t spark energy (E * le i At), and spark power (P * e*i) were calculated from the o current and voltage measurements. The results of these calculations are shown in Figures A-6, A-7 and A-8. 4 0246S TIMC, U* FIGURE A*6. Spark Resistance. 26 MOD ooo13?57 10 FICURE AS. Spark Power. 27 000013758 MCD Knystantas & Lee (Ref. 1) have proposed that a minimum critical power and 3 minimum critical energy must be reached during the first !/* cycle of a discharge for the direct spark initiation of a detonation. The only mixture for which these values were determined was a stoichiometric acetylene-oxygen mixture at a pressure of 100 torr. For this mixture the critical energy was found to be 0.1 J/cm and the critical power was 0.1 MW/cm. In these terms the values for the spark used in the hemisphere tests are 2.79 J/cm energy and 0.861 MW/cm power. These values are not strictly comparable since the electrode spacing of the spark gap used by Knystantas and Lee was comparatively large (37 mm) so that a cylindrical^ expanding shock wave was produced. In the hemisphere tests the electrode spacing was small (1.65 mm), and thus the shock wave would assume a spherically expanding shape and would not be as effective in producing direct initiation of detonation. Matsui and Lee (Ref. 2) have shown that the critical initiation energy (J/cm) increases as the electrode spacing decreases. The smallest spacing reported was 2.5 mm, and this spacing gave a critical energy of 1.5 J/cm for acetylene-oxygen mixtures. REFERENCES 1. R. Knystantas and J. H. Lee. "Combustion and Flame", 27, 221-228 (1976). 2. H. Matsui and J. H. Lee. "Combustion and Flame'*, 217-220 (1976). 28 0000^ 59 Appendix B DESCRIPTION OF INDIVIDUAL TESTS TEST l. 4'r PROPANE This was the first test and was a trial to check out instrumentation and photographic coverage. The burn was made in daylight. Because of ambient light the flame was barely visible, only the top of the flame hemisphere was visible and only a vertical velocity could be determined. There was some yellow incandescence at the base of the flame but because of the faint image it could not be determined if this incandescence affected the shape of the flame. TEST 2. 57 PROPANE In an attempt to increase the visibility of the flame, the fuel concentration was increased to 57. (4% is stoichiometric). Before the mixture could be ignited with the spark ignition source it preignited. It is possible that cite propane flowing through the inlet cube generated a static charge, that arced to the metal of the instrumentation channel, igniting the mixture. To prevent this occurring the inlet tube was grounded to the instrumentation channel. It is also possible that the film hemisphere became electrically charged. A coating for the film was investigated to make if conductive, however, the coating decreased the transparency of the film and was not used. TEST 3. 57 PROPANE Test appeared similar to test one. The increased propane concentration did not cause increased visibility of the flame. Only vertical velocity could be obtained. AJ subsequent tests were done at night. TEST 4, 57 PROPANE This was the first night test. The blue flame was clearly visible. Yellow incandescence appeared at the base of the hemisphere. This incandescence appeared tc be of two types: a diffuse incandescence radiating from the ignition point outward ground level, and an inrense jetting from the instrumentation channel. The first type 29 000013760 MOD which did not cause distortion of the hemispherical blue flame or secondary ignition ahead of the flame, could have been caused by dust swept up by the expanding gases. The second type, which did cause ignition ahead of the flame and subsequent distortion of the hemispherical flame, couid have been caused by fuel burning in the instrumentation channel. It is difficult to determine the velocity of propagation in the channel because the time and location of ignition in the channel are not known. Another phenomenon was apparent on the film from the overhead camera. When the fiame reached the 5*m-radius wooden ring used to fasten the plastic hemisphere, there was an acceleration from 8.3 m/s to a higher velocity of 13.8 m/s. The higher velocity then remained constant to a radius of 6.5 m then slowly decreased. TEST 5, 10% METHANE This was the first test with methane. The flame did not appear as bright as with propane. As with test 4, there was yellow incandescence, which caused multiple ignition and distorted the fiame at later stages. TEST 6, 5% PROPANE As with tests 4 and 5, yellow incandescence caused secondary ignition and distortion of the fiame. Because of the distortion it was not possible to measure the increase of flame velocity as the flame passed over the obstacles. However, the obstacles did not produce a pronounced effect on the fiame. TEST 7, 10% METHANE This test appeared similar to test S. TEST 8, 7.7% ETHYLENE OXIDE This was the first of the tests in 10-m-radius hemispheres; all previous tests were 5-m-radius. There was much less incandescence present in this test, and it did not cause secondary ignition. The fiame started hemispherical in shape; then because of apparent boundary layer acceleration the hemisphere developed a conical base. The velocity of the flame at the surface was constant at 16.8 m/s as opposed to 14.7 m/s above the surface. During later stages the fiame became distorted because of the plastic hemisphere tearing at one edge. TEST 9. 7.7% ETHYLENE OXIDE This test appeared subjectively, to the personnel at the test site, to be similar to test 8. However, because instrumentation malfunctioned, no data were obtained. 30 MOD 000013761 TEST 10, 1.17o ETHYLENE OXIDE This test was similar to test S excepc the plastic hemisphere separated uniformly, and there was no distortion of the flame shape even at late times. The boundary layer acceleration caused a constant flame speed at the surface of 19.8 m:s as opposed to 13.4 m/s above the surface. TEST 1 1, 5% PROPANE Yellow incandescence was again apparent in this test with secondary ignition over the instrumentation channel. The tubes placed in the hemisphere had no apparent effect. TEST 12, 47c PROPANE This test appeared similar to te*t 11. although the yellow incandescence did not cause much distortion of the flame. The only apparent effect of the wire screen placed over the ignition source was to make the flame visible at a slightly earlier time. Decreasing the fuel concentration from 5 to 4% did not noticeably decrease the luminosity. TEST 13, 107c METHANE This test was a near perfect bum. There was almost no yellow incandescence. The flame was a hemisphere that evolved in later stages into a sphere on a conical pedestal. Because the flame velocity was so low. the buoyancy force caused the flame to rise and produce the shape of a sphere on a pedestal. The boundary layer velocity was only slightly higher than the velocity above the surface (5.4 m/s vs 5.2 m/s). TEST 14, 6-57r ETHYLENE Much incandescence and many secondary ignition points made this test difficult to assess. TEST 15. 3.57< ACETYLENE The acetylene supply tanks did not completely empty in this test, producing a low fuel concentration. The long bum time caused by the low flame velocity (3.6 m/s ) permitted Che buoyancy forces to raise the flame envelope off the surface. The complex flow pattern caused the measured horizontal velocity to accelerate and the vertical velocity to be constant: the reverse of that usually observed. 31 MCD 000013762 TEST 16. 10% METHANE This was a detonation test and is discussed in a separate section. TEST 17, 3.5% BUTADIENE This test appeared similar to test IS. The low velocity permitted the buoyancy forces to lift the flame envelope off the surface. There was some afterburning from the end of Che fuel line, so it is possible ail of the fuel was not in the hemisphere at the start of the test and the 3.5% concentration was not achieved. TEST 18, 7.7% ACETYLENE The flame in this test was much brighter than the other tests. Secondary ignition occurred from burning in the instrumentation channel. TEST 19. 10% METHANE This, like test 16. was a detonation test and is discussed in a separate section. 32 mcd 000013763 Appendix C FLAME, POSITION, BURN TESTS 33 MOD 000013764 3* MOD 000013765 HORIZONTAL RADIUS. n HORIZONTAL RADIUS. FIGURE C-4. Horizontal Flame Position 7.7% Ethylene Oxide. 35 MCD 000013766 FIGURE C-6. Vertical Flame Position 5% Propane. 36 MCD 000013767 VERTICAL RADIUS. 0 02 0.4 04 TIMS. StC FIGURE C-8. Vertical Flame Position 7.7% Ethylene Oxide. 37 04 MCI) 000013768 5 RADIUS. RADIUS, m 0 0.09 0.10 0.19 0.30 0-2S 040 049- 0v40 TIMC. SEC FIGURE C*9. Flame Position 6.5% Ethylene Test 14. 6r o oia .i... 0-4 i 04 o!s " llo TIMS, SCC FIGURE C*I0. Flame Position 3.5CJ Acetylene Test 15. 38 4 1.6 0013769 W * 39 MCD 000013770