show can cause an internal explosion in a tank. Further- more, by desi ning for an explosion there is no need to ar-

The main point is to have a formal method to ensure t at the design and construction of a project is in accordance with recognized engineering practice. Thus, in more attention to methods of identifying and ana yzing hazards, we should not overlook the important role of engi- neering codes of practice as a basis for risk analysis and control. CARLOS RENALLES, National Scientific Services: Do you make any use of the Dow classification index or loss prevention measurements? JOSCHEK: No, we don’t use it. WALT HOWARD, Consultant: I want to emphasize one statement that Charlie Solomon made and go just a step further. You were emphasizing the need for hazard iden- tification, where we fail far more than we do on quantita- tive evaluation of hazard. There is no magic method for identifying or becoming aware of s ecific hazards. We’ve

and some of these companies promote some very hi hly.

work nor will any magic method provide those. Any appli- cable method is suitable if properly used and if people work hard enough with their think tanks. We cannot stop with merely good engineering design, good construction and good inspection of that construction. After all, we do have to keep oil in the engine of our car. We can’t build a good engine, put oil in it once, and then let it run forever. The same is true for equipment. We do have to do many inspections and checks on plant equipment. Unfortu- nately we also make changes in equipment and processes. So often those changes have no consideration of the safety aspects involved, and this again is one way in which we fail greatly.

!i

Pg

gue in great 1 epth over the probability of it occurrin

heard of special techniques used E y various companies

We cannot use any magic method to take the place of a ard

Sometimes we fail in the decision-making process. Many decisions have to be faced by manufacturing man- agement at all levels. The early life of the plant is desi

many years. We have to face many decision situations in o erations. In doing that we normally evaluate the effects

productivity, labor relations and union relations. We tend to forget to evaluate the safety aspects of those decisions.

My experience has been that a great many of our major accidents are caused, in the final analysis, by our failures to evaluate safety in the decision-making process. I don’t mean to criticize the decision makers. We need to criticize ourselves for not providing the proper training of all of our people. We give them all s~r t s of fancy training courses. But, we forget to include some training regarding consid- eration of safety as ects in making decisions. Let’s concen- trate attention in Lose areas instead of concentrating our attention on some fancy mathematics related to hazard quantification. I still think hazard quantification has its

lace, but again I agree with Dr. Joschek that it has very Emited use, in very specific small areas in a given situation.

and construction, but the major life is the operation P or

o f! many potential decision alternates including profits,

H. I. Joschek studied organic chemistry at the Uni- versity of Heidelberg, Germany, post doctoral fel- lowships in Paris and at Illinois Institute of Tech- nology, Chicago. Since 1966 he has been at BASF Co., Ludwigshafen, Germany with successive as. signments in research and development, environ- mental problems, licensing, safety and loss pre- vention. He is presently safety adviser at the Ludwigshafen site of the company.

A Study of Flame Arrestors in Piping Systems Even officially approved flame arrestors must be used only under the exact conditions for which they were tested and approved.

G. L. Broschka, I. Ginsburgh, R. A. Mancini, and R. G. Will, Amoco Oil Co., Naperville, Ill.

The ignition of flammable mixtures of hydrocarbons and air or oxygen in pressure vessels and piping can produce disastrous results [ I ] . The best way to eliminate explosion hazards is to avoid mixing of air and hydrocarbons, when feasible, and to require rigorous control of processes that mix air and hydrocarbons. This ractice has become

regulations that require collecting and processing hydro- carbon vapors that were previously vented to the atmos- phere. Air-pollution controI facilities required at refineries, terminals, service stations, chemical lants, and production facilities have increased the risi that flammable vapor-air mixtures may be in vessels and pip- ing. The use of commercial flame arrestors has been sug- gested as a way to reduce the risk of flame propagation within these facilities.

difficult in recent years because o I ) air-pollution control

ISSN 02784513-83-6775-05-$2.00. -he American Institute of Chemical Engineers, 1983.

In the work reported in this paper, commercial flame ar- restors intended for use on tank vents were tested to deter- mine their effectiveness in piping systems containing flammable vapors. The performance of an arrestor was found to be dependent on the location of the arrestor in the system, the ignition location, and the gas velocity in the pipe. Flame propagated through arrestors when flam- mable gas flowing at velocities above 15 feedsecond (4.6 mhecond) was ignited 3 feet (0.9 m) upstream or 15 feet (4.6 m) downstream from the arrestor. In one case, the housing of a commercial flame arrestor ruptured when tested in a flowing system with ignition 30 feet (9.1 m) up- stream. Limited tests were also conducted using sections of pipe packed with Pall rings in place of commercial ar- restors.

BACKGROUND When a quiescent flammable gas-air mixture burns in a

tube, a portion of the heat of combustion is absorbed by the

January, 1983 5 Plant/Operotions Progress (Vol. 2, No. 1)

wall of the tube. If the tube is of sufficiently small diame- ter, the wall can absorb enough heat to prevent the propa- gation of flame, i.e., quench the flame. Experimental procedures have been developed for determining the quenching diameter, or distance, for quiescent flammable mixtures. The quenching distance for gasoline vapor-air mixtures, for instance, is of the order of one-millimeter [2]. This rinciple has been utilized in the design of commer-

consists of a screen pf fine mesh, a bundle of small tubes, or a sandwich of closely spaced lates.

Factory Mutual, or other recognized organizations are in- tended for service in the vents of oil storage tanks or oil tanker ships [3] . Underwriters’ standards, for instance, re- quire that listed flame arrestors be subjected to an ex lo- sion test [4] in which ignition of a quiescent gasoEne va or-air mixture occurs in an open vent pipe on the outlet

15 feet (4.6 m) in length. The length specified in the list- ing is often the largest tested and hence provides no infor- mation on the performance of the device with longer outlet pipes.

Clearly, those test conditions do not reproduce, nor are they intended to reproduce, the operating conditions for facilities where potentially flammable mixtures must be transferred at reasonable flow rates through long pipes. Furthermore, Underwriters’ caution that a listed arrestor may be ineffective if the outlet pipe is longer than specified in the listing [3] .

Since the effectiveness of an arrestor de ends on remov-

quantity ofthe burning mixture %rough the arrestor can be important considerations. If an ignition occurs in a very long pipe, the flame can accelerate to a very high velocity, and can produce a significant increase in pressure that can drive the flame and/or hot combustion products through the arrestor. The flame acceleration can be very rapid, and sonic or even supersonic flame velocities may be attained in a relatively short length of pipe [5], particularly if the gas is initially turbulent.

Literature on the performance of commercial-type flame arrestors in pipes is somewhat limited. A report on work performed in Great Britain indicates that an effective in- line arrestor has been developed for flammable mixtures of town gas and air [6]. However, except in specific cir- cumstances such as cited above, it is not generally consid- ered effective to install a flame arrestor in a long pipe to

revent flame propagation in case of ignition [a. There Lave been cases in which flame arrestors have failed to stop flame propagation in piping systems.

cial 1 ame arrestors where the flame-arresting element

Flame arrestors listed by Un B erwriters’ Laboratories,

si B e of the arrestor. The outlet pipe is most often limited to

ing heat from the combustion rocess, t \ e velocity and

OBJECTIVES OF EXPERIMENTAL WORK

The objectives of this study were as follows: Determine the effectiveness of commercially availa- ble flame arrestors at conditions similar to those that would be encountered in vapor-recovery system pip- ing. Determine the limitations of such flame arrestors, if they were found to be effective under a limited range of conditions. Determine if alternate piping arrangements, such as locating the arrestor in an enlarged section of pipe, would reduce any limitations found. Accumulate a data base for the development of any novel conceDts generated as a result of the studv.

It was not our int&t conduct a comprehensive studi of all available commercial arrestors or to perform precise measurements of flame propagation mechanisms.

Flame arrestor tests were performed in a 3-in (76-mm) pipe with flammable mixtures of butane and air..Station- ary and flowing mixtures, up to 20 feetlsecond (6.1

6 January, 1983

&second) line veloci , were ignited with a high-voltage

and downstream of the arrestors. Technical-grade butane, whose composition was con-

firmed by GC anal sis to be at least 97 mole percent bu-

100 psig (792 kPa) plant air s stem. Air and butane flows

the mixture composition and flowrate. The lines from the flowmeters terminated at a mixin tee connected to the

that contained several tees. The tees were used as elbows and contributed further turbulence to insure good mixing. The feed-preparation system was protected from the ef- fects of the ignitions in the flame tube by a s stem of solenoid-operated block valves, swing check va r ves, and blow-out plugs located on the turbulence generating tees. Gas compositions were checked before each test by sam- pling the mixture leaving the flame tube with a calibrated MSA (Mine Safety Appliances) Gascope Combustible Gas Indicator Model 53. Sandbags were placed around the ar- restors and other potentially vulnerable locations in the test equipment.

When desired, flame velocities and pressures were de- termined from measurements made with piezoelectric pressure transducers. The output from the transducers went to an oscilloscope equipped with a camera. The os- cilloscope was triggered by an induction coil wrapped around the ignition wire. Ignition was accomplished with a spark plug at one of several locations along the flame tube. A pushbutton switch was used to fire the spark plug via a high-voltage transformer.

spark. Tests were con ar ucted with ignition both upstream

tane, was used in a K 1 tests. Air was supplied from a filtered

were introduced through Cali i rated flowmeters to control

test pipe by about 15 feet (4.6 m) o f p one-inch (2.5-cm) pipe

PROCEDURE

After the arrestor was placed at the desired location in the pipeline and the instrumentation was stabilzed, the flow of butane and air into the 3-inch (76-mm) i e was

locity. The mixture was maintained at a constant flow for a few minutes to purge the pipeline. Meanwhile, the mix- ture leaving the pipeline was checked with the Gascope. The mixture was then ignited and a photograph of the os- cilloscope screen was taken. If the flame propagated through the arrestor, a loud sharp blast and a flash could be heard and seen at the end of the pipe. The arrestor ele- ment was removed for inspection after the tests.

Prior to making stationary-mixture tests, the mixture was purged through the 3-inch (76-mm) pipe at 5 feethecond (1.5 mhecond) for several minutes. Before ignition, flow was stopped by manually closing a valve between the mixing tee and the 3-inch (76-mm) pipe. The mixture was allowed to “sit” for about 20 seconds before ignition.

Baseline data on the propagation of a flame in 3-inch (76-mm) pipe was first obtained without an arrestor in the line. Five flame-arrestor configurations were then tested. Upstream ignition was used in 2 through 6 (below).

adjusted to the desired mixture composition an 88 ow ve-

Two series connected 3-in. (76-mm) parallel-plate ar- restors in 3-in. (76-mm) pipe, ignition between arres- tors. 6-in. (152-mm) parallel plate arrestor in 3-in. pipe. 6-in. parallel-plate arrestor with 3 4 feet (1.07 m) of inlet 6-in. pipe and reducers in 3-in. pipe. 6-in. parallel plate arrestor with Pall-ring packed re- ducers in 3-in. pipe. An 8-foot (2.4-m) long section of 6-in. pipe filled with Pall rings. 3-in. Crimped Spiral Wound Arrestor in 3-in. pipe. 3-in. parallel-plate arrestor in 3-in. pipe-downstream ignition.

PlantlOperotions Progress (Vol. 2, No. 1)

RESULTS AND DISCUSSION Flame Propagation in 3-Inch (76-mm) Pipe

A. Flammable Limits. The flammable range of butane- air mixtures determined by the Bureau of Mines for hori- zontal flame propagation is 1.9 to 6.5 mole percent [S]. Flammable limits, however, depend upon several parame- ters, including the geometry of the equipment and the pro- cedure used. Most experimental work on flammable lim- its has involved stationary gas mixtures. Two publications on the flammability of flowing mixtures reported differ- ent effects of flow velocity on the limits [g, 101.

We conducted tests to determine the practical limits of flammability for butane-air mixtures flowing in a 3-inch (76-mm) pipe at velocities encountered in Amoco’s gaso- line vapor-recovery systems. A 50-foot (15-m) long section ofpipe was used for these experiments, Figure 1. The mix- ture was judged flammable if either the rubber stopper at the upstream end of the pipe blew out or flame was ob- served at the downstream pipe outlet. The ignition loca- tion was 15-feet (4.6-m) downstream from the loosely placed stopper. The data, Table 1, indicate that the flam- mable limits vary with the mixture flow velocity and that the variation in the upper limit is more pronounced than is the variation in the lower limit. In general, flow narrows the flammable range. The composition of the butane-air mixture resulting in the most violent combustion is also shown in Table 1. The velocities and pressures associated with flame propagation at the optimum composition are summarized in Table 2. The violence or severity of the combustion was judged subjectively, based on the inten- sity of the sound that issued from the pipe. The severity of combustion was very sensitive to the mixture composition, in that only a slight change in composition was sufficient to cause a large change in the noise from the combustion.

B . Flame Velocity and Pressure. Butane-air mixtures at the “most- violent” composition were ignited in the same 3-inch (76-mm) pipe, except that in some cases a pipe plug (“closed” pipe) was used in place of the rubber stopper (“open” pipe). Flame velocity and pressure measurements were recorded over a range of pipe lengths and fuel-flow velocities, Table 2. The data indicate the character that would be expected of the flame incident on the face of flame arrestors located in a 3-inch (76-mm) pipe. Flame velocity and pressure were affected by:

1) Gas flow velocity. 2) Location of ignition from the open end of the pipe. 3) Distance the flame travelled from the ignition

4) Presence.of a relief vent at the closed (upstream) end

The Of flame the px” id not propagate upstream when ignition was within 6 feet (1.8 m) from the open end of the pipe, and the mixture was flowing at a velocity of 5 feeusecond (1.5 m/second). However, at the same flow velocity flame propagated upstream when ignition occurred 8 feet (2.4 m) or more upstream of the open end. Table 3 lists the maxi- mum ignition distance from the open end for which the flame did not propagate upstream for a particular flow velocity.

source.

Figure 1. Pipe for studying flame propagation.

The average flame velocity between the ignition source and the location ofthe three pressure transducers was used to characterize the flames. These values are re orted in Table 2. The actual flame velocity at the end o P the pipe may be much higher than the average listed in Table 2, be- cause the flame is accelerating at high rates in the pipe. Since, for subsonic speeds, our instruments could have re- solved differences between the times of arrival for the first and third transducers, it is likely that detonation (su- personic) velocities were attained during at least some of the tests.

We measured transient pressures in the range of200-220 si (1370-1520 kPa) in some tests. This is a considerably

gigfer range than would be expected for a butane-air mix- ture assuming completely adiabatic conditions (flame pressure 6 to 10 times the initial pressure) and tends to confirm the existence of detonation conditions in at least some of our runs.

Flame Arrestor Tests

Eight flame-arrestor schemes were tested in a 3-inch (76-mm) pi eline. The flame velocity and pressure ap- proaching i e arrestor were controlled by varying the dis- tance between the i nition source and the arrestor, the fuel

pipe plug in the tee at the upstream end of the 3-inch (76-mm) pipe. Tests conducted with the pipe plug in place are termed “closed”-end tests, while those with the rubber stopper are termed “open”-end tests. Flame arrestors used for test series A, B, C , D, F, and H consisted of an alumi- num housing which contains a removable arrestor element of corrugated aluminum plates. The arrestor element for the 3-inch (76-mm) arrestor is 4 inches (0.1 m) deep and presents an 8-inch x 8-inch (0.2 m x 0.2 m) surface to an approaching flame. The crimped corrugated plates are spaced 0.049 inches (1.24 mm) apart and provide 660 passa es for the flame. Underwriters’ Laboratories listing

15 feet (4.6 m) of vent or downstream pipe. The 6-inch (150-mm) flame arrestor consists of an arrest-

ing element 6 inches (150 mm) in depth and presents an 11-inch x 11-inch (0.29 m x 0.29 m) surface to an ap- proaching flame. The crimped corrugated plates are spaced 0.037-inches (0.94 mm) apart and provide 2472

flow velocity, and t a e use of either a rubber-stopper or a

speci P es that this arrestor be installed with not more than

TABLE 1. FLAMMABLE LIMITS FOR BUTANE-AIR FLOWING IN A %INCH (76-~M) PIPE

Mixture Composition Producing “Most Violent Combustion,” Flammable Limits, vol.

Fuel Velocity, Wsec percent vol. percent

5 10 ia

2.9-5.7 2.74.3 2.8-51

4.1 3.1-3.2 3.3-3.7

SI Conversion: #sec X 0.3 = m l s e c

PlantlOperotions Progress (Vol. 2, No. 1) January, 1983 7

a l

TABLE 3. MAXIMUM IGNITION DISTANCE FROM THE OPEN END OF THE PIPE TO PREVENT UPSTREAM FLAME PROPAGATION IN A

$INCH (76-MM) PIPE

Maximum Distance For Which Fuel Velocity Upstream Flame Propagation

ftisec Did Not Occur, ft

5 10 20

SI Conversion:

Wsec x 0.3 = d s e c

28% 3 3 4

0 0 2s;

0 1 0 (Dw mu,

Ei w

6 8

12

passages to the flame (approximately 4 times as many as the 3-inch (76-mm) arrestor element). Underwriters’ Labo- ratories listing s ecifies that this arrestor also be installed

A. Arrestor Configuration A - Two 3-Inch (76-mm) Flame Arrestors in &Inch (76-mm) Piping. The test-pipe configuration for this series is shown in Figure 2. Butane- air mixtures at the optimum concentration were passed through the arrestors and ignited with a spark plug 3 feet (0.9 m) upstream of arrestor “A”. Both flame arrestors quenched flames when the mixture was flowing at a ve- locity of 10 feevsecond (3 dsecond). However, with amix- ture velocity of 20 feevsecond (6.1 mlsecond), the flame passed through arrestor “A”.

No significant damage was done to arrestor “A”. How- ever, an imprint of carbon was found on the incident face of the element. The imprint suggests that the flame may have impinged on a 3-inch (76-mm) diameter portion of the 8-inch (0.2-m) square area of the-arrestor element.

In one test, at a flow velocity of 20 feevsecond (6.1 mlsecond), the ignition switch was actuated twice in a pe- riod of about four seconds; the first “ignition” failed to produce a ‘bang’ and it was not certain that the mixture had ignited, hence the igniter was activated a second time. The flame was observed at the downstream end of the pipe after the second ignition, but the flame also attached itself to the face of Arrestor “A” and continued to burn for about 10 seconds-at which time the flow was stopped. After the test, we examined the arrestor element and discovered that it was distorted and partially melted (Figure 3). No

with not more tK an 15 feet (4.6 m) of vent pipe.

Figure 2. Test pipe for configurations A & B.

8 January, 1983 Plont/OpQrotions Progress (Vol. 2, No. 1)

damage of any sort was observed on the element of arrestor “B”.

B. Arrestor Configuration B3-lnch (76-mm) Flame Ar- restor in 3-Inch (76-mm) Piping-Downstream Ignition. The pipe configuration was the same as configuration A. However, the flame-arrestor element from “A” was re- moved from the pipe and the ignition source was moved 3 feet (0.9 m) further upstream from the end of the pipe. A butane-air mixture, flowing at 20 feetlsecond (6.1 mhecond) was ignited 13 feet (4 m) downstream of flame arrestor “3”. A violent ignition was heard and the flame passed through and damaged the flame arrestor element, Figure 4. No flame velocity or pressure measurements were made, but subsequent tests showed that the arrestor plates begin to deform when subjected to 65 psi (450 kPa) static pressure.

C. Arrestor Configuration C-6-Inch (150-mm) Flame Arrestor in 3 Inch (76 mm) Piping. A 6-inch (150-mm) flame arrestor was placed in a 3-inch (76-mm) pi e (Figure

effective than the 3-inch (76-mm) arrestor, since it pro- vided more cooling surface and a larger housing diameter to reduce flame velocity at the face of the arrestor ele- ment. The flowing mixtures were ignited upstream from the flame arrestor, which quenched flame propagation initiated44 feet (13.4 m) (the maximum distance we tested) from the arrestor with fuel flowing at 5 feetlsecond (1.5

5). It was thought that the oversized arrestor mig hp t be more

Figure 4. Damaged arrestor element-configumtion B.

n

3”PiW i renor

3”rubb.r A nopper

Cornurntion C. blnchflmewnstorwith 6xr’reducen * Configuration D. 6-Inch fieme mnor with RIIringpecked6x3”reducen fl

bsii2‘i Configurnion E. Plpe .eFtion packed

with Rlirlnga

Configuretion F. Il-lnchflme~n8torwith 4 overmized plpeupmmn +7’+ k13

Configurdan 0. 3-inch arrestor m Figure 5. Flame arrestor test configuration.

PlanWOperations Progress (Vol. 2, No. 1)

m/second). At higher fuel-flow velocities, the arrestor’s performance was dependent upon the ignition location, Table 4.

D. Arrestor Configuration D-6-lnch (1 50-mm) Flame Arrestor with Reducers Packed with 518-Inch (16-mm) Pall Rings. A 6-inch (150-mm) flame arrestor was placed in the 3-inch (76-mm) pipe (Figure 5). Both 6 in. x 3 in. (150 mm x 76 mm) reducers were packed with %-inch (16-mm) alu- minum Pall rings to provide additional cooling surface area and to perhaps facilitate quenching flame propaga- tion. Ignitions 44 feet (13.4 m) upstream of the arrestor were quenched at mixture flow velocities of 0, 5, and 10 feetlsecond, Table 4. However, with a mixture flow veloc- ity of 20 feetlsecond (6.1 mhecond), the flame passed through the arrestor and damaged the arrestor element, Figure 6.

E . Arrestor Con guration E-6-lnch (150-mm) Pipe

of 6-inch (150-mm) pipe acked with %-inch (16-mm) alu- of the flame arrestor, was

placed in the 3-inch (76-mm) pipe (Figure 5). The Pall rings were held in an 8-foot (2.4-m) section of &inch (150-mm) diameter pipe with a screen. The packed pipe section consisted of:

Section Packed wit R %-Inch (16-mm) Pall rings. A section

minum Pall rings, in pace P

The Pall-ring arrestor quenched flame

2-yi feet (0.76 m) of 3-inch (76-mm) pipe 3-yi feet (1.07 m) of 6-inch (150-mm) pipe 6 in. x 3 in. (150 mm x 76 mm) reducers

propakation when ignition was 44 feet (13.4 m) upstream or ue mix- tures flowing at velocities of 0 ,5 , 10, and 20 feetlsecond (6.1 mlsecond), Table 5.

The rubber stopper in the tee (Figure 5) was replaced with a pipe plug. As in the “open” pipe configuration, all flame propagation was quenched (Table 5). In all, 70 tests (including du licates of those listed in Table 5) were per-

through the apparatus was observed. After about 30 tests, the Pall rings were removed for inspection, and only those rings in the 3-inch (76-mm) pipe section were damaged. The rest of the rings were coated with pi e scale and car- bon, but in other respects were in 00 condition. The

F . Arrestor Configuration F-6-Inch (150-mm) Flame Arrestor with a 6-Inch (150-mm) Pipe Section in 3-Inch (76-mm) Pipe. A 3-yi-foot (1.07-m) length of 6-inch (150-mm) diameter pipe was placed upstream of a 6-inch (150-mm) flame arrestor in a 3-inch (76-mm) pipeline

formed with J: e same Pall rings and no flame propagation

same observations were made after k J e final test.

Figure 6. Damaged arrestor element-configuration D.

January, 1983 9

Distance Between Ignition and Arrestor, ft Flame Arrestor Scheme

A A A B** C C C C C C C C C C D D D D F***

3 3 3

13 28 28 28 28 34 34 34 34 43 43 43 43 43 43 43

* Determined by direct observation of end of pipe. ** Downsbeam ignition location. ***This test was conducted with a "closed" end. SI Conversion:

Wsec X 0.3 = m/sec

TABLE 5. TESTS OF PACKED-BED ARRESTOR-CONFIGU~TION E-UPSTREAM IGNITION-ALL TESTS SUCCESSFUL*

Distance Between Ignition Flow Velocity Open or and Arrestor, ft fthec Closed End Pipe

28 28 28 28 34 34 34 34 43 43 43 28 28 28 28 34 34 34 34 43 43 43 43

0 5

10 20

0 5

10 20 0

10 20 0 5

10 20

0 5

10 20 0 5

10 20

open open open open open open open open open open open

closed closed closed closed closed closed closed closed closed closed closed closed

Fuel velocity, ft/sec Flame Pass Through

Arrestor?*

TABLE 4. PERFORMANCE OF PARALLEL PLATE ARRESTORS-OPEN-END PIPE

10 January, 1983 Plant/Operotions Progress (Vol. 2, No. 1)

10 20 20 20 0 5

10 20 0 5

10 20 0 5 0 5

10 20 20

No Yes Yes Yes No No No Yes No No Yes Yes No No No No No Yes Yes

forming a factory-assembled flame-arresting element. The flame-arresting element was bolted between two cast aluminum flanged ends to form the complete unit.

The flame arrestor was tested in a 50-ft (15-m) long sec- tion of 3-in. (76-mm) pipe (Figure 5). All tests were con- ducted with the flammable gas flowing through the ipe

were used to ignite the flammable gas at distances of $6, 9, 12, and 30 (9.2 m) feet upstream from the arrestor. Re- sults are summarized in Table 6.

The first test was conducted by igniting the gas-30 feet (9.2 m) upstream from the arrestor. This test produced two results: 1) the flame propagated through the arrestor; and 2) the shell of the flame-arresting element exploded. Fig- ure 8 shows a close-up view of the failed flame arrestor.

at a velocity of 17 feethecond (5.2 mhecond). Spark p P ugs

* Determined by direct observation of end of pipe. SI Conversion:

ft/sec X 0.3 = m/sec

(Figure 5). The sudden enlargement in the pipe was ex- pected to decelerate the flame. Butane-air mixtures flowing at a velocitv of 20 feethecond (6.1 dsecond were ignited 44 feet (13.4 m) upstream of the arrestor. The flame propagated through the arrestor and dama ed it.

G. Arrestor Configuration G-3 Inch C r i m p % Spiral Wound Arrestor in &Inch (76-mm) Pipe. A 3-in. (76-mm) flame arrestor was selected for testing. The flame- arresting element (Figure 7) consisted of a tube bank formed from two sheets of aluminum (one flat and one crimped) wound into a s iral and mounted in a fabricated aluminum shell. The tubPe bank was welded to the shell, Figure 7. Crimped spiral-wound arrestor element-Eonfiguration G.

TABLE 6. TESTS OF CRIMPED SPIRAL WOUND ARRESTOR-CONFIGURATION &UPSTREAM I G M T I O N ~ P E N -

END PIPE

Distance Between Ignition and Arrestor, Flow Velocity Flame Pass Through

ft ft/sec Arrestor?*

30 17 Yes (arrestor exploded) 3 17 no 6 17 no 9 17 no

12 17 Yes * Determined by direct observation of end of pipe SI Conversion:

ftkec x 0.3 = d s e c

This result illustrates the fallacy of the idea that lacing an

“can’t hurt.” A replacement arrestor was tested beginning with igni-

tion at 3 feet (0.9 m) upstream from the arrestor and subse- quently moving the ignition point to 6 feet (1.8 m), 9 feet (2.7 m), and finally to 12 feet (3.7 m) upstream. No evi- dence of flame propagation through the arrestor was ob- served with ignition at 3,6, or 9 feet, but flame propagated through the arrestor with ignition 12 feet upstream. Fol- lowing failure of the arrestor to quench the flame, it was disassembled for inspection. No evidence of physical damage to the flame arresting element was noted. The manufacturer of this arrestor supplies an essentially iden- tical unit with a stainless-steel housing. This might have prevented the rupture of the shell that we observed in the first test, but we have no reason to expect that its flame- quenching ability would be better.

H . Arrestor Configuration H-3-lnch (76-mm) Parallel- Plate Arrestor in 3-Inch (76-mm) Pipe-Downstream lgnitionapen-End Pipe. In complex piping systems, a flame front can enter a pipe from a branch connection at a point located between an arrestor and the outlet end of the

arrestor in an unsuitable location is acceptable i ecause it

Figure 8. Ruptured arrestor element-configumtion G.

#ant/Operationr Progress Wol. 2, No. 1)

ButandAir w-

1 ,;; r-----l

l----zo.+ k lC-1 4 I & - - - - - r

Figure 9. Test configuration Hdownstream ignition,

- pipe. Hence, propagation to an arrestor from the down- stream direction in a flowing system should, as the tests conducted in the flame propagation studies demonstrate, be considered a possibili . A 3-inch (76-mm) parallel plate arrestor identical to z at tested in Series A was in- stalled in a iping configuration described in Figure 9.

stream of the arrestor at a distance of one foot (0.3 m) from the flanges to provide an indication of whether the flame front reached and passed through the arrestor. Test results are summarized in Table 7. The flame generated by igniting a butane-air mixture 25 feet(7.6 m) downstream of the flame arrestor passed through and severely damaged the plates of the arrestor element when the mixture was flowing at 20 feethecond (6.1 misecond). Another arrestor was installed and tested with the igniter 10 feet (3 m) downstream. Flame produced by igniting a mixture flow- ing at 5 feetlsecond (1.5 dsecond) moved upstream but did not pass through the arrestor. However, the flame sta- bilized on the face of the arrestor. The fuel was shut off after about 15 seconds, and subsequent inspection re- vealed partial melting of the flame-arrestor element.

Thermocoup P es were placed both upstream and down-

CONCLUSIONS AND RECOMMENDATIONS

The acceleration of a flame propagating through a pipe and the pressure generated by the combustion of a flammable mixture in a pipe depend on several param- eters including:

a. Flow velocity of the flammable mixture. b. Location of the point of ignition with respect to

the point at which flame velocity and pressure are measured.

c. Provision of a means for pressure relief in the pipe.

Commercial flame arrestors of the type utilized on oil storage tanks and oil tanker ships may not prevent the propagation of flame in piping systems carrying flowing flammable mixtures. Flame arrestors are not an acceptable alternative to the avoidance of flammable mixtures in piping systems. The insertion of aluminum Pall rin s in an oversize section of pipe may be a more e if ective means of quenching flame propagation than commercial flame arrestors. Additional studies would be required to determine re- liable design parameters for a Pall-ring flame arrestor,

TABLE 7. TESTS OF PARALLEL PLATE AFUIESTOR-CONFIGURATION H-DOWNSTREAM

IGNITION~PEN-END PIPE

Distance Between Ignition and Arrestor, Flow Velocity Flame Pass Through

ft ft/sec Arrestor?*

10 5 no 25 20 Yes

* Determined by thermocouples located one foot from arrestor flanges. SI Conversion:

Wsec X 0.3 = mlsec

January, 1983 11

particularly with respect to scale-up to larger pipe diameters.

For our final comments, it is worth noting that one of the arrestors tested in this pro ram was listed or approved by seven organizations inclu%in two of the national testing laboratories.The devices wi f: 1 robably function effec- tively in applications for which %e listings were granted; i.e., when located sufficiently close to the outlet of a pipe or vent containing a quiescent mixture and when the igni- tion source is located at the outlet. However, our tests demonstrated that the user cannot place the arrestor any- where, merely because the device is listed or approved by a reco nized agency and is supplied with a flange on each

that for which it was listed or approved by a recognized or- ganization, then it must be tested at those conditions and in the exact size and mechanical form in which it will be used.

end. I f? an arrestor is to be used at conditions different from

LITERATURE CITED

1. “Hazards of Air,” American Oil Company, 5th Edition, Chicago (1964).

2. Lewis, B. and G. vonElbe, “Combustion, Flame and Explo- sions,” Academic Press, New York (1951).

3. Gas and Oil Equipment List, underwriters’ Laboratories (Oct., 1973).

4. “FIame Arrestors for Use on Vents of Storage Tanks for Petro- leum Oil and Gasoline,” Underwriters’ Laboratories, UL525, 3rd Edition (1973).

5. Henderson, E., “Combustion Gas Mixtures in Pipelines,” Proceedings of the Pacijc Coast Gas Association, 32,98-111 (1941).

6. Cubbage, P. A., “Flame Traps for Use with Town Gas/Air Mixtures,” The Gas Council, Research Communication GC63, London (1959).

7. h i s t e a d , G., “Safety in Petroleum Refining and Related Industries,” 2nd Edition, John G. Simmonds and Co., Inc., New York (1959).

8. Coward, H. F. and G. W. Jones, “Limits of Flammability of Gases and Vapors,” Bull. 503, Bureau of Mines, Washington, D.C. (1952).

9. Starkman, E. S., L. P. Hexby, and A. G. Catteneo, “A Study of Free Flames in Turbulent Streams,” 4th Symposium (Intern.) on Combustion. The Williams & Wilkins Co., Baltimore, (1953), p . 670-673.

10. Cresciteii. S.. F. Naoolitano. G. Russo. and L. Tranchino. “FlammabiliG Limit: on Flowing Gases,” presented at the 3rd International Symposium on Combustion Processes, Kazimierz, Poland, Sept. 24-27, 1973.

DISCUSSION

AL CORONA, Mobil Research: We found a disadvanta e in one type of parallel late arrestor. If you pulled Je placed the drawer carelessly or pushed it in too fast, the momentum would pack the plates up at one end and leave uneven spaces between the plates. This can result in a flame propagation through the plates. MANCINI: Yes. There can be man problems with these

TREVOR KLETZ: It seems to me that this connection of vapors from tank truck filling and so on is almost the clas- sical example of solving one problem by buying a worse problem. To prevent relatively small amounts of vapor be-

drawer to inspect the p P ates or clean them and then re-

devices, if they are not installed an B maintained properly.

ing discharged to atmosphere, we have developed ex - sive systems which roduce an explosion hazard, angew”e

also the classical example of what ha pens when well-

don’t fulfy understand and try to solve one relatively small environmental problem and land with the more serious ex- plosion problem. MANCINI: I think that is correct. To date I do not know of an incident in which an ignition in a vapor recover system

try to solve this prob P em by a basic research program. It is

meaning legislators interfere in the i elds where they

resulted in flame pro agation from location to r ocation. However, I do know o P cases where ignitions, or near igni-

terconnected system. Ifan ignition occurs, tf e absence ofa

tions, occurred at the vapor rocessing e ui ment itself. Hence, this possibility must E e considerel a z n g with the possibility of ignition at a particular loadin spot in an in-

flammable mixture in the At facilities where gasoline is flammable some of the time, from gasoline loading are less volatile products.

Gregory L. Broschka is currently a Specialist in In- dustry Supply Analysis with the Standard Oil Company (Indiana). He holds an M.S. degree in Chemical Engineering from Northwestern Uni- versity and has been employed by Standard, or its subs id iq companies, for 8 years. While he was with the Ammo Oil Company Research Depart- ment, he conducted research in flammability and flame propagation, synthetic fuels, and leak de- tection technology.

Irwin Ginsburgh is a Senior Research Associate with the Amoco Oil Company Research and De- velo ment Department with whom he has been empfoyed for 31 years. He holds a PhD in Physics from Rutgers University and has conducted exten- sive research in gas phase detonations and static electricity. Current research interests include un- confined vapor cloud explosions, vapor recovery, and advanced energy sources. He holds 42 patents, has been awarded four IR 100 awards, and has pub- lished numerous articles and one book.

Robert A. Mancini is a Research Supervisor and Process Safety Specialist with the Amoco Oil Com- lay Research and Development Department. He

01 s a PhD in Chemical Engineering from North- western University and has conducted research in reaction kinetics, process safety, environmental conservation, and probability analysis related to process safe and equipment reliability. He is a member of 3[, Static Electricity and Fire Safety Engineering Subcommittees of the American Pe- troleum Institute Committee on Safety and Fire Protection and is a member ofthe NFPA Explosion Protection Systems Technical Committee.

Robert G.WiII is Director of the Engineering and Environmental Researrh Division of Ammo Oil Company Research and Development Depart- ment and has been employed by Amoco Oil or its subsidiary companies for 29 years. He is a mechan- ical engineering graduate from Purdue University and has conducted research in oil spill control, pro- cess safety, and burner design. He holds five pa- tents, has been awarded an IR 100 award, and is a Registered Professional Engineer.

12 January, 1983 Plant/Operations Progress (Vol. 2, No. 1)