Flow Apparatus for Determination of Spontaneous Ignition Delays JOSEPH L. JACKSON, RICHARD S. BROICAW, ROBERT C. WEASTI, AND MELVIN GERSTEIN National Advisory C o m m i t t e e f o r Aeronautics, Lewis Flight Propulsion Laboratory, Cleveland, Ohio
S
PONTANEOUS ignition processes manifest themselves in many reactions that are studied under the general classification of combustion. These processes may be deleterious, as a h e n they produce preignition in a reciprocating engine or create fires in the storing and handling of combustible materials, or they may be beneficial by serving as the ignition source in a Diesel engine or promoting smoother burning in a turbojet combustor. Some of the thermal theories of flame propagation suggest that a flame is propagated by virtue of continuous spontaneous ignition occurring on the leading edge of the flame front. I n any combustion process where self-ignition can occur, the time lapse before the flame appears is an important factor. As the time delay of ignition may be an inverse measure of the rate a t which the preflame reactions proceed, a study of the factors influencing the delays may provide information on the kinetics and mechanism that prevail in the ignition process. This paper describes an apparatus which may be used to study ignition lags of hydrocarbon vapors at elevated temperatures. Preliminary data on ignition delays of propane-air mixtures are also presented. Previous experiments to determine the effect of fuel-air ratio on ignition delay of hydrocarbons have given varied results. With the aim of a better understanding of the combustion procewes in reciprocating engines, several workers designed systems to determine ignition temperaturee and lags in rapid compression machines. Much of the work by this method was done a t high pressure, with special emphasis on conditions that produce cool flames and zones of nonignition. Jost (3) reported that the lag of the first region is relatively independent of fuel concentration for h e p t a n e - a i r mixtures, while more recent experiments, with iso-octane ( 4 ) , show minimum delays for stoichiometric compositions. In this type of apparatus, independent control of temperature and pressure is difficult, and the temperaturr and pressure a t which ignition occurs may be very different from those a t which the initiating reaction begins. A great deal of l\-ork has b e e n d o n e r5ith constantvolume static systems, where premixed gases are admitted into a heated vessel. Using 1
Case Institute of Technology,
Cleveland, Ohio.
.TEMPERATURE
this technique, llason and Wheeler ( 5 )found that ignition lags for methane-air mixtures increased with increasing methane concentration. This method has also been employed t o study the rates of slow oxidation of hydrocarbons. Maximum rates of oxidation, as measured by the rate of pressure rise, are usually observed a t compositions much richer than stoichiometric-for example, Newitt and Thornes ( 7 ) found the maximum rate of propane oxidation for a composition of about 55 volume % fuel inoxygen. Ignition delays have also been studied in flow systems which allow reactants to be preheated separately to the desired temperature. For very lean mixtures Mullins (6) observed no effect of fuel concentration on ignition lag when kerosine was injected from a small port into the center of a large duct of heated air. It seems likely that this method is not well suited to the study of concentration effects, as all mixtures from pure fuel to pure air may be present in the duct. Indeed, in the work of Olson (8), where an attempt a t rapid mixing was made, a definite shortening of ignition lag with increasing heptane concentration lvas noted. A program, the first part of which is reported in the present paper, was undertaken to clarify the effect of fuel concentration on ignition dclap in particular, and if possible to elucidate the chemistry of the ignition reactions. An apparatus has been designed and constructed which seemed better suited to the METAL TUBE particular investigation than HEATING ELEMENTS those previously reported. A flow system was chosen because it permits control of t e m p e r a t u r e, pressure, and PLE PROBE composition from the time of -EXHAUST mixture preparation until ignition occurs. The reactants were preheated separately and mixed a t the desired temperature in a time much shorter than the ignition lag. The initial results on propane-air mixtures presented here were made over a wide range of HEATING ELEMENTS fuel percentages at atmospheric pressure and at temperatures from 525" to 740" C. APPARATUS AND PROCEDURE
f R
Figure
1.
MAN I F O L D S
Schematic Drawing of Spontaneous Ignition Apparatus
2547
For this research, an apparatus was desired which would preheat the gases, air, and combustible, separately to a predetermined temperature; mix the gases a t this temperature in a t h e very much shorter than the ignition lag, provide
INDUSTRIAL AND ENGINEERING CHEMISTRY
2548
a reaction volume with the predetermined temperature constant throughout the region where ignition can result; and permit complete control over the fuel-air ratio. These requirements dictated a flow system; because none of the apparatus reported in the literature met these specifications, a new setup was designed as shonn in the schematic sketch of Figure 1. Fuel was supplied from a cylinder through a pressure iegulator t o a critical flow orifice. The size of the orifice and upstream pressure determined the flovi rate. B three-n ay solenoid valve located just ahead of the test section normally bypassed into the exhaust line. When the valve mas energized, the fuel was diverted into the mixer. Air was supplied from the laboratory line through a piessure regulator and critical flow orifices to another three-way solenoid. This valve normally passed the air to the mixing chamber but exhausted into the room when energized.
V Y C O R TUBE, 5 0 mm. I . D .
SECTION
Figure 2.
VYCOR TUBE, 25 mm. I.D.
A-A
Mixing Chamber for Spontaneous Ignition Apparatus
4 drawing of the mixer is given in Figure 2. Fuel and air xere supplied to separate concentric chambers, where temperature equilibrium was established with the system. There were 16 holes drilled tangentially into the small inner cup, and connected alternately w-ith the two chambers. This mixing cup had a volume of about 1 ml., which was very small in comparison viith the volume of the ignition section of the apparatus. For all data in this paper, the residence time in the mixing cup \?-as less than 2% of the ignition delay. The temperature of the unit, as indicated by a thermocouple i n the mixer, was controlled by a heating element and variable voltage transformer. From the inner cup the gases passed through a 3 / ~ i n c hhole and expanded to the wall of the tube through a diffuser section fitted with 200-mesh stainless steel screens to remove the swirl. The ignition zone consisted of a 25- or 50-mm. Vycor tube 36 inches long (see Figure 1). The exhaust fitting a t the top of the unit contained a movable thermocouple probe t o check the temperature axially through the reaction volume. Constant temperature throughout the reaction volume was accomplished by w a p ping resistance heaters in six ascending sections on a metal tube surrounding the Vycor. (Each section was individually con-
Vol. 46,No. 12
trolled by a variable voltage transformer. ) The original design had viewing windows in t,he outer insulating jacket and 3/4-inch slits in the met,al tube which permitted visual observation to provide added information about phenomena occurring in thc reaction chamber. Metal-to-glass seals were formed with compression fittings utilizing asbestos and talc as a packing mcdium a t the mixer and exhaust couplings. Several runs were made with this design t o observe t,lie ignitiori and subsequent events. However, it was decided to forego t,his feature in favor of a positive seal bet,wcen the reaction vessel and the surroundings. This niodificat.ion entailed the replacement of the metal tube by one without t,he vieving slits. The seals a t t,he mixer and exhaust sections were then made with soft. copper gaskets. Figure 2 shows this arrangement as well as the nozzle and screen adapters connecting the mixing cup to the 25- or 50mm. Vycor tubes. It was now possible to change the tube sizes without removing the insulating jacket. The results obtained werc the same wit'h either arrangement, suggesting that no appreciable leakage through the metal-to-glass seal occurred. The bulk of the heat required was supplied by air from a 25kn.. electric heater. This air was injected upward from a ring a t the bottom of the chamber between the ignition tube and the insulating jacket (Figure 1) and exhausted by a collector ring a t the top. The heating elements on the metal tube were adjusted to remove any temperature gradients. The pressure in the heating chamber was kept equal to that in the ignition section (atniospheric). The unit required 2 to 3 hours to establish a desired temperature in the mixer and throughout the reaction volume, but once established, it did not drift v,-ith time. Ignition lags were measured by a photocell relay and electric tinicr or by a pressure recorder. When an ignition was indicated by both units, results rrere identical. If the light intensity mtio of flame to background v a s low (weak flames or high temperature operation) the photocell sometimes failed to indicate, whereas the pressure recorder always responded. With either syst,em, the timing cycle was started automatically on introduction of a coinbustible mixture illto the mixer. Only the pressure recorder could he used after modification of the apparatus. In general, lags were reproducible to about &5%. The customary procedure was to adjust the air flow through the burlier while bypassiug the fuel t o the exhaust. When constant temperature condit,ionewere reached, as indicated by the movable thermocouple probe, the fuel solenoid was energized, which diverted the flow t,o the mixer and started the electric timer or the recorder time-indicator. When t'he flame or pressure pulse appeared, the fuel valve was de-energized and the ignit,ion delay recorded. Three or four runs were made before t,he fuel-air ratio was changed. A conpt'ant,t'otal foil- of mixture JTas usually maintained for all determinations at, a given temperature, unless a change in flow rate was required to keep ignition within the central two thirds of t~hctube length (calculated from flow rate and ignition delay). The electrical connectioua were later changed to permit alternative procedures to be used. The first modification allowed the introduction of air int,o fuel rather than fuel into air. With either of these methods, a fuel-oxygen gradient exist,ed within the interfacial region. In order to eliminate this concentration gradient, the second modification allowed metered nitrogen to be sent through a solenoid into the mixing cavity; in this way all oxygen was swept,out of the system. A single switch cut off the nitrogen flow, diverted fuel and ail. into the mixer, and started t,he time cycle. The fuel-oxygen ratio was thus kept the same in t,he interface as in the mixture. RESULTS A N D DISCUSSIOh
EFFECTOF FLOWRATE. Several runs were made in which ignition lags were determined a t various flow rates. Results of a typical series of experiments are graphed in Figure 3, and it is
INDUSTRIAL AND ENGINEERING CHEMISTRY
December 1954
seen that the position for the minimum in the plot of lag os. flow rate shifts toward higher flow rates for richer mixtures. Investigations in which temperature and tube size were varied suggested that these parameters do not affect the volumetric flow rate giving minimum lag. As the flow rate effect is a function of volumetric rather than linear flow rate, it is perhaps associated with the mixing process. TUBE DIM, mm.
PROPANE,
7, VOI.
al
in
d
2
a
-I
z 1.0 2 .8 'I z .6 '3
2549
when the fuel lead method was employed, as there was no sharp appearance of flame or pressure pulse marking ignition-e.g., see Figure 4, propane lead, 20 and 30%. A t fuel concentrations above 25y0, no flames were observed, although a definite temperature rise a t the top of the tube indicated that some reaction was taking place. With less fuel in the final mixture, a luminous zone appeared shortly after air admission, which became more distinct as the interface passed up the tube (and the region near the mixer became leaner) until finally a well defined flame was formed. Presumably, the ignition reactions were started before an explosive mixture-that is, a mixture capable of reacting rapidly enough to give a sharp pressure pulse-entered the tube. I n a few cases the delay characteristic of the rich region was sufficiently long to permit an explosive mixture to enter the tube during the induction period. I n these experiments lags were similar t o those obtained for richer mixtures by the air lead method (see 10% propane lead experiment of Figure 4). It appears, then, that the shorter delays observed by the propane lead method are characteristic of the richer interfacial region rather than of the following fuel-air mixtures.
.4 PROPANE X V O l
I
I
Figure 3.
2
I
I
l
l
I
6 8 IO 20 FLOW RATE, I iters/min 4
I
RATE,
FLOW
IllRrSlmln
10
4
20
5
30
4
40
4
50
4
60
4
40
Effect of Flow Rate on Ignition Lag Temperature 6 4 9 O C.
EFFECT OF TUBE SIZE. Under comparable conditions of volumetric flow rate, fuel conrentration, and temperature, lags observed in the 25-mm. tube were on the average very slightly higher than in the 50-mm. tube-e.g., see Figure 3. However, this difference was in general within the experimental error, and ' in many cases the order was reversed. I n obtaining data, an attempt was made to select flow rates such that the position of the ignition (as calculated from flow rate and ignition lag) was in the central two thirds of the tube. For the 50-mm. tube, this condition corresponded closely to the flow rates for minimum induction period; in the 25-mm. tube, somewhat lower flows were employed and lags mere correspondingly longer. EFFECTOF MODE OF INTRODUCTION OF REACTANTS. The apparatus was designed to permit introduction of fuel and air to the reaction tube in three different ways. Figure 4 shows typical pressure-time traces obtained by these methods; ignition is indicated by the first sharp rise in pressure. For some of the richer mixtures pressure pulses were very weak; so the recorder was used at maximum gain, as indicated in the figure. In the first procedure, the desired air flow through the ignition tube was first established. Next, propane was introduced, and the time required for appearance of a flame was determined. I n this method the interfacial region between the leading air and the following fuel-air mixture must contain mixtures of all fuel-air ratios leaner than the one under investigation. Should any leaner mixture have a shorter lag, a spurious result might be obtained. As richer mixtures had shorter lags by this method (see pressure-time traces of Figure 4), it was felt that the measurements were meaningful. However, to verify this point, a few experiments were performed with the order of air and fuel introduction reversed. When the propane precedes the air, the interfacial region contains mixtures of higher fuel-air ratio only. If, as suggested previously, richer compositions have shorter delays, lags determined by this procedure (fuel lead) might be expected to correspond t o the shortest lags measured in the alternative manner (air lead). I n most cases, it was impossible to make delay measurements
Figure 4.
Typical Pressure Traces for Ignition of Propane-Air Mixtures
In the nitrogen lead method a flow of nitrogen preceded both reactants into the tube; propane and air were introduced and the nitrogen was shut off simultaneously. Here fuel-oxidant ratio was presumably constant through the interfacial region, with nitrogen concentration as the only variable. In general, ignition delays observed with this technique were the same as lags obtained by the air lead method (compare nitrogen and air lead columns of Figure 4). At high flow rates, the nitrogen was not adequately preheated, so that some cooling of the system occurred, giving rise t o increased lags. However, the generally good agreement between results of the nitrogen and air lead methods indicates that measurements by the latter technique are characteristic of the mixture rather than the interface. On the other hand, the nature of the interfacial region has a marked effect on composition range over which ignition is observable. With an air lead, sharp pressure pulses and flames were noted a t all fuel-air ratios, although the violence of the explosion decreased gradually as the propane concentration increased. However, with a nitrogen lead the intensity of the pressure pulse diminished rapidly with increasing fuel-air ratio and, in general, no pulse was detectable when the propane concentration exceeded 40y0 by volume. If the fuel flow was continued after ignition, stationary flames were seated in the tube for mixtures
INDUSTRIAL AND ENGINEERING CHEMISTRY
2550
containing as much as 3070 fuel in air. For the rich region the weak pulses may have been caused by cool flames; however, the presence of cool flames was not established. (Either the temperature rise after ignition, or the composition of combustion products, nlight serve t o distinguish cool from normal flames.)
I
1
--'IiSBLE
c.4h'ES
PRESSURE
ZPRESSLRE
OBSERVED lU2 LEAG
PULSES
OBSERVEO
PULSES OBSERVED
lU,
4 LEPCI-?
__
(AIR LEA01
0
EFFECTO F
PROPANE
C O N C E S T R A T I O N .4ND
Vo!. 46, No. 12 TEMPERATURE.
Typical plot,s of ignition delay us. propane concent,ration are presented in Figure 5. The ignition delay decreases with increasing propane concentration, the effect being greatest a t low fuel percentages. This result is in marked contrast t o other flame properties of propane, such as quenching distance (1, 2) and minimum spark ignition energy (I), which exhibit minimum values near the stoichiometric composition. The concentration of propane appea,rs t o be the important factor in determining ignition delays a t constant temperature and pressure. The adiabatic flame temperature of the mixture (at its maximum near stoichiometric) plays a decisive role in determining ignition energies and quenching distances, but would be expected to have no effect on ignition lags. Increasing the temperature shortens the delays (see Figure 5 ) , as might be anticipated, eince in general the rates of chemical reactions increase with temperature. SUMMARY O F RESULTS
-- ----_
-3
5
Figure 5 .
IO
15
20
25
X, PROPANE, 9, vol
35
40
I
I
45
50
\-ariation of Ignition Lag with Fuel Concentration
These experiments indicate t,hat the rate of the init,iating reaction, as measured by the ignition delay, is characteristic of the regions of highest fuel-air ratio, even though a homogeneous mixture of the same fuel concentration may be incapable of supporting a flame. When a flammable mixture exists near a zone of very high propane concentration (as in t,he case of air lead experiments), it, may be ignited by reactions initiated in t,hericher region. The phenomenon obtained in very rich mixtures appears t o be analogous to one observed by Prettre (9) in studying the oxidation of pentane in a static system at 230" t o 360" C . Three types of pressure-time curves were observed. One type was characterized by an induction period followed by explosion. In a second type, the induction period preceded an extremely rapid react,ion, which did not, however, lead to ignition. Preeumably, it was the lag for this type of reaction mhich was measured in the present investigation in the case of very rich propane mixtures with an air lead. The final type of curve observed by Prettre was one in which rate of pressure rise rvas slow through the ent,ire course of the reaction. Inasmuch as the air lead method gave results comparable t o those obtained with the nitrogen lead, it was employed in obtaining the data discussed in the follox-ing section, since sharper indications of ignition were observed by this technique and better temperature control could be maintained a t high flow rates (no cooling by the leading gas).
An ignition-delay apparatus has been designed and constructed mhich allows the reactant, gases to be preheated eeparately, niised rapidly, and caused to react in a tube at constant temperature and pressure. Preliniinary experiments wit,h propane-air mixtures at atinospheric pressure show that ignition lags vary somewhat with volumetric flow rate in this apparatus. Cnder comparable conditions of temperature, composition, and volumetric flow rate, lags observed in reaction tubes 25 and 50 mni. in diameter TTere the same wit,hin experimental error. The effect of the mode of introduction of reactants has been investigated, and the air lead method (air preceding the fuel-air mixture) has becn adopted. For the propane-air system, delays are shortened by increases in init,ial t,eniperature and by increases in propane concentration, even for very rich compositions. LITERATURE CITED
Blanc, XI.V., Guest, P. G., von Elbe, G., and Lewis, B., "Third Symposium on Combustion, Flame, and Explosion I'henomena," p. 383, Baltimore, Williams & Wilkins Co., 1949. Friedman. R., and Johnston, W. C., J . A p p l . Phys., 21, 791 (1950). Jost, W.,"Third Symposium on Combustion, Flame, and Explosion Phenomena," p. 424, Baltimore, Williams & Wilkins co.,1949. Jovellanos, J. V., Taylor, E. S.,Taylor, C. F., and Leary, W. A,, Natl. Advisory Comm. rleronaut., Tech. S o t e 2127 (1950). and Wheeler, R., J . Chem. Soc., 121, 2079 (1922). Mason, W., Nullins, B. P., F u e l , 32,234 (1953), Newitt, D. &I.,and Thornes, L. S., J . Chem. Soc., 1937, 1669. Olson, D.P.. "Spontaneous Ignition Time for Mixtures of Heptane Vapor in A4ir," Yale University doctoral disertatiori, 1961. Prettre, A I . , "Third Symposium on Combustion, Flame, mid Explosion Phenomena," p. 397, Baltimore, Williams & Wilkins co.,1949. ACCEPTED Septeniher 14. 1954. RECEIVED lor review M a y 28,1954. Taken in part iron1 a thesis submitted by Joseph L. Jackson t o Case Insiiiute oi Technology in partial fulfillment of the requirements for the degree oi master of science in chemistr>-.
