Dead Space Thickness, and Minimum Ignition Energy in Premixed Methane-Air. Combustion ... The provisions of the Clean Air Act and its amend- mentst dictate ... This primarily inert exhaust gas lowers the peak .... closed-cycle operation was determined using a chemical equilibrium computer code (Gordon and. McBride ...
Rqrlnltld from
Combustion Scienct! and Tt!chnology, 1980, Vol. 23, pp. 1-15 0010-2202/80/2301-0001$04.50/0
© 1980 Gordon and Breach Science Publishers, Inc. Printed in Great Britain
The Effect of Exhaust Gas Recirculation and Turbulence on the Burning Velocity, Dead Space Thickness, and Minimum Ignition Energy in Premixed Methane-Air Combustion N. GAT The University of Cincinnati, Cincinnati, Ohio 45221 C. W. KAUFFMAN The University of Michigan, Ann Arbor. Michigan 48109
Q 1980 Gordon and Breach Science Publis hers, Inc.
Comb11stio11 Science and Technology , 1980, Vol. 23, pp. 1-15 00I0-2202/ 80/2301-000 I $04 50/0
Printed in Great Britain
The Effect of Exhaust Gas Recirculation and Turbulence on the Burning Velocity, Dead Space Thickness, and Minimum Ignition Energy in Premixed Methane-Air Combusti on 1
N. GA T The University of Cincinnati. Cincinnati. Ohio 4522 7
C. W KAUFFMAN The University of Michigan. Ann A rbor. M ichigan 48 709 (Received February 6 , 1979; in fi11al form January 9, 1980)
Abstract- The dilution o f a hom ogeneo us combustible mixture with products of combustion has proven to be a successful method of reducing the oxides of nitrogen produced by a combustion process. The Exhaust Gas Recircula tion (EGR) affec ts the flame velocity, ignition energy, and quenchi ng behavio r. An experimental program has been conducted to determine the effects o f EGR on these fund a mental aspects of a turbulent metha ne/air flame in a constant volume bomb. Data a re presented showin g that the decrease in fla me vel ocity caused by the EGR can to so me extent be compensated fo r by an increase in the turbulence intensity a t a price of increased ignition difficu lty. The effect of the EGR, as determined from bot h heat transfer and io nizat ion measurements, is to increase the s ingle-wall quenching dista nce.
INTRODUCTION The provisions of the Clean Air Act and its amendmentst dictate the control of automotive emissions. A method used for the control of oxides of nitrogen (NOx) emission is Exhaust G as Recirculation (EG R) where a fraction of the exhaust gas is recycled into the combustibles of the intake charge (Quadar, 197 l ; Deeter et al. , 1968; Newhall, 1967). This primarily inert exha ust gas lowers the peak flame temperature, a major factor in the formation of NOx. Theoretical and experimenta l evidence indicate that besides lowering the flame temperature, the dilution of the intake charge with the exhaust gas may have some negative effects such as lowering the burning velocity of the mixture, thickening the quench layer, and increasing the required ignition energy (Newhall, J 967 ; Morgan and Kane, 1953; Mellish and Linnett, 1953). Such problem s may lead to an undesirable increase in hydrocarbons and carbon monoxide emissions as well as to a drop in engine performance. Compensation for the first two problems is possible through the control of the cylinder turbulence. The level of turbulence may be limited, however,
tSee, for instance, The Annual Repor t o f the EPA to the Congress (1977). " Progress in the Preventio n a nd Co ntrol o f Air Po llution in 1976," d ocu ment nu m ber 95-75 .
by ignition difficulties (Balla! and Lefebvre, 1977; Swett, 1949). Extensive experimental work is needed before the above qualitative analysis can be cast into quantitative relationships which ma y be used by designers. Hence, the purpose of this research program was tq study the effects of exhaust gas recirculation and turbulence on the burning velocity, the single wall quenching dista nce, and the ignition energy . In addition, some effe: ts of hydrogen enrichment of the fuel were investigated . A constant volume bomb method was selected for the present stud y, since it a llows convenient control over most of the test p1rameters. In viewing the reported results, attention should be paid t ) the fact that the fuel used in the present stud y primarily consisted of methane which, un like gasoline, contains no carbon-carbon bonds.
2
APPARATUS AND PROCEDURES
The investigatio n was carried o ut in the simple environment (when compared to that of an engine) of a co nstant volume bomb illustrated in Figu re J. The test bomb, (1 ), a straight cylinder 50-cm high a nd 40-cm in diameter is equipped with a pressure relief di a phragm (2) which breaks whenever the p ressure in the bomb exceeds a pproximately 2 atm, a water injection port (3), a nd a strip heater
N. GAT AND C. W. KA U FFMA N
2
r --------- - --~
I
I
:
I
I
"
/
I I
I I I
I
__-- --- ____
:
)
.._
I }
o - J\IR c -
co2
' - 'I-
.. - co' < - AR
' - Oz
' - Hz .l -
fj
(w
r
f ?.ti;)
FIGURE I
FIGURE 2
Experimen ta l apparatus.
Igni tion electrodes amingement.
(4). The quenching surface (5), loca ted 13.3 cm from the cy linder axis, is equipped with a set of ionization gauges (6) .. Power and information lines are inserted into the bomb via a sealed feedthro ugh connector (7). The bomb is evacuated by means of a vacuum pump (8). A pressure transducer (9) is mounted on the wall. Turbulence is generated by moving a grid ( I 0) th rough the volume of the bomb. The grid is p ulled by a rod (11), and followed by an attached stainless steel trailing rod ( 12) which replaces the conventiona l ignition electrodes. The trailing rod is made of two pieces con nected with a 5 mm lo ng Teflon
insert of the same diameter as the rod . This Teflon insert serves as a n electrica l discontinuity across which an electrical spark can jump. The a rra ngement can be seen in Figure 2. On the way down the tip of the trailing rod clears several switches which activate, in that order, the recordi ng system and the ignition circuit. A magnetic pickup is used to monitor the speed of the grid . The grid is driven by a hydraulic system ( 13- 18), wh ich is activated by a switch (n). A mercury ma nometer (a) is used to monito r the partial pressure of the gases while the bomb is being filled . The supply of various gases (b-h) and a simulated EG R storage cylinder (j) complete the setup. A lamp (m) is used as a heater to ensure convective mixing of the gases stored in the EGR cylinder. For mo re details, see Gat ( 1978). The apparatus made the actual recirculation of the burned gases impractical. Therefore, the theo retical compositi on of the ex haust gas under a closed-cycle operation was determined using a chemical equilibrium computer code (Gordon and McBride, 197 l ). Assuming that the exhaust gas freezes at the adiabatic flame temperature equilibrium composition, a mixture sim1,1lating this composition has been prepared and sto red in a storage cylinder. U nstable species and trace elements such as H, 0, OH and ot hers were separated iato their basic constituents and added to the mixtu re in the for m of a more stable species (e.g. H 2 , 0 2) . This procedure ensured an overall conservat ion of mass in the system. The si mula ted exh aust gas did not contain any water since it would condense at the room temperature. The appro priate quantity of water was injected directly into the bom b which was kept at a temperature of 340 °K to prevent the condensation of the exhaust water. The fuel in the study was a city-supplied natural gas consisting of 96.6 percent methane, and 0.4 percent higher hydrocarbons. The test procedure was as follows . F irst, the bomb and the feed lines were evac uated . Then the appropriate quantities of fuel , air, exhaust gas, and water were introduced into the bo mb. The air-tofuel ratio was kept constant through a ll the tests at (A / F)v = 10, corresponding to an eq ui valence ratio of 0.965. The amount of EGR was varied between 0 and 30 percent accord ing to the following definition weight x 100 % EGR = _ __ .:::__ of_ exhaust. _ _ ,........:cgas _ _ __ weight of [fuel + air + exhaust gas]
(1)
EFFECT OF EGR IN PREMlXED METHANE COMBUSTION
To ensure a thorough mixing of all the constituents in the bomb, the mixture was stirred by moving the grid several times through the vessel. And finally, it was left to heat up to 340°K and to allow the turbulence in the gas to die out. Prior to ignition, a capacitor bank was charged to the desired energy, then the grid was pulled down at a preset uniform velocity, and a spark was set off at the geometrical center of the vessel. The resulting nearly-spherical flame propagated normal to the quenching surface located well within the interior of the bomb. Wall quenching was documented employing six ionization gauges, ranging in length from 76 to 1626 micrometers, and a fast response surface thermocouple. The geometrical arrangement and the ionization gauges circuit diagram are shown in Figure 3 with the surface temperature measuring circuit in Figure 4. A typical data record, reproduced in Figure 5, obtained from a single test contains the following information.
ISO
a) The pressure history during the initial stages of combustion in the bomb (6P < 0.6 atm .). b) The quenching surface temperature history . c) Output signal from each o f the six ionization gauges located at various distances in front of the quenching surface. d) A signal from a magnetic pickup used to monitor the grid speed. e) A reference 500 Hz wave. The burning velocity was calculated from the pressure history according to the following formula (O'Donovan and Rallis, 1959) :
s=
1·a(Pt/P )1 1ru(dn/dl)
-
-
3 [1-(l-n) (Pi/P)ll r 11 ]213
(2)
where
dn
dP
(3)
dt
Voo
rb &AIJG£
l
-
The use of Eq . (3) is limited for as long as the fraction of the burnt gas in the bomb is small (Lewis and VonElbe, 1961 , pp. 367- 381 ). The flame propagation velocity, which is composed of the burning velocity and t he expansion velocity of the burnt gas, was calculated from the time derivative of the flam e radius , as given by
FUSE
I
3
= ra[l -(l -11) (PifP) 11r
11
]1 13.
(4)
L~IHS•
U . 5)1},_X...'°'. l . !~,
.ti;t ' ;:
lF12ll 11m
GAJ.J: COM! NG• [L[(lllJ)Jlllf U:." "11 I ,\IJlR[JSI] ' •J
FIGURE 3 Ionization gauges and electrical ci rcuit.
TttER111lCJXPL£""
----;:;?
~CE
/ rn Lc~~E SID[)
SURFACE
BURR- BRO\IN, 3626 I NSTRUMENTATION A~PLI FI ER. ••
MED THE R.~,
CO- AX MICROSECOllD SURFACE TEllPERATURE PROBE .
G. E.. METAL OXIOE VARISTOR, TRANSIEH VOL 1\GE SU PPRESSOli .
F IG URE 4 syste m.
Quenching surface tempe rat ure measurin g
O ' Donovan's model assumes, among other things, the existence of a spherical flame of a negligible thickness. However, a s the level of EGR or as the level of turbulence increase, flame thickness is expected to increase as well. To check the validity of the calculations, the flame radius, based upon the pressure history, wa s calculated (Eq. 4). The time at which the calculated flame radius is equal to 13.3 cm was co mpared with the ti me the ionization gauges were triggere.: (from the experimental traces in Figure 5). The difference between these two times never exceeded one percent validating the model's assumptions for all practical purposes. The measurements of the turbulence cha racteristics in the apparatus require equipment (e.g. LDV) which was unavaila ble at the time the study was performed . Therefore, turbulence characteristics were calculated using empirical co rrelations deri ved for grid-generated turbulence in wind tunnels (considered to be a nearly isotropic
4
N. G AT A N D C. W. K AUFFM AN
( a) 0% EGR, v''.:: 9.8 CM/ SEC.
#1 IONIZATION GAGE
MAGNETIC PICKU #2
]~
IGNITION TIME
,____, 0.02 SEC (b) 30% EGR, v'
~
10. 5 CM/ SEC
FIGURE 5 Typical da ta record [fo r a 0 % EG R (a) and for a 30 % EG R (b) test].
homogeneous turbulence) .. · Althou gh in thi s stud y the grid moved th rough a sta tion ary gas, whereas in a wind tunnel the gas fl ows past a grid , some an a logy exists betwee n the two cases, since turbulence- a function of the shea r stresses in the flu idd epends o n the relative velocity between the grid and the flu id . T he fo llowing turbulence decay laws were selected : T urb ulence 1951)
inte nsity
(Bain es
a nd
Peterson,
-v' =
u
( Ut ). -5/i
1. 12 -
d
.
(5)
Taylor's d issipation scale ~2= 7~ .
~
Integral scale of turbulence (Sato and Ya mamow, 1976)
lg
.
- = 0.0 114 R e;1i. 23 + 0.530. Ag
(7)
EFFECT OF EGR IN PREMIXED METHANE COMBUSTION
5
TABLE! Comparison between the range of turbulence characteristics in the present apparatus and in other systems according to Andrews el al. (1975) Parameter
v cm/sec
cm
Rei
Re 1,
Re ·r.
0.83
0.08
145.0
62.0
3.9
46.7
1.21
Bunsen burner
200
1.0
250
IC engine
100
2.0
400
Pulverized coal burner
200
30.0
850
Gas turbine combustor
350
2.0
550 (idle) 900 (max power)
>.y 15-1;4 (Re,..)-112.
(8)
(In subsequent equations, for convenience, the subscript g is dropped .) Gat (1978) provides the arguments for the selection of these equations. Due to the apparatus design, the various turbulence characteristics were limited. Table I summarizes the maxima of the turbulence parameters which were obtained during this test program . For comparison purposes the table gives some values of these parameters in practical combustion systems, according to Andrews et al. (1975).
3
'1
cm
Present system
Kolrnogoroff microscale 'f/ =
,\
cm
RESULTS AND DISCUSSION
Some general observations can be made in regards to the output traces (e.g . for 0 and for 30 percent EGR, in Figure 5). The ignition point is indicated on these traces and the time axis runs from that point to the left. First, the pressure rise during an explosion is related to the burning velocity and as the EGR level increases, the rate of pressure rise significantly drops (as a comparison between the appropriate traces in Figure 5 clearly shows). Next, as the flame travels toward the quenching surface (located 13.3 cm from the bomb center) it first triggers the circuit of the longest ionization gauge. Then, the rest of the gauges are triggered
in a sequence of decreasing length. When the flame is highly turbulent this triggering sequence is destroyed, however. The ionization curren t decreases with the decrease in the length of the ionization gauges (i.e., as the flame. a pproaches closer to the wall). As the level of turbu lence increases, so does the ionization current. Gauge number 4 was found defective, however, and the ionization current is always high since the insulating coating broke off this gauge. Quenching surface temperatu re shows a slow temperature rise starting at the ignition instant until the flame hits the wall (corresponding to the adiabatic co mpression of the gases in fro nt of the flame). At that instant a sha rp increase in surface temperature is noticed. Figure 5 reveals that this temperature rise is higher for mixtures without EGR. When the pressure inside the bomb exceeds about two atmospheres, a pressure relief d iaphragm breaks, resulting in an uncontrolled completion of the combustion in the bomb. This combustion reactivated the ionization gauges. 3.1
The Pressure Rise During the Initial Stages of the Combustion in a Constant Vo lume Bomb
The p ressure history, plotted on a log-log scale, ha s been inspected for a large number of tests. An analytical expression given below for the pressure rise as a function of time was obtained by curve fitting the experimental data. Two typica l curves
N. GAT AND C. W. KAUFFMAN
6
are shown in Figure 6. In general it is concluded that the relationship in such plots is nonlinear, and therefore an equation of the type (Bradley and Mitcheson, 1976; Perleeetal., 1974)
or explicitly,
C2 ( D.P = exp - C1 7 - ([Jn(t) I C4
C3]2
C4 2 ) 112
}.
.
(IOa)
This is an equation of a hyperbola which in the limit, as time becomes large, approaches the asymptote ln(D.P) io-1
(II)
/, /,
EGR
/,
Equation (I 0) was found to apply to turbulent as well as to laminar combustion in the bomb. The values of the parameters C1 through C4 have to be determined by a nonlinear regression analysis. As already mentioned, the burning velocity was calculated from Eqs. (2) and (3). The expression for the derivative dP/dt was obtained from Eq. (IO) after the determination of the coefficients C1 through C4 . Explicitly,
/,
/,
I.
~
h
~
30l
I
°'
I
~
10· 2
I
I
I
I
dP 10· 3
f3ln(t)
(9) ~
~
=
or
l_____l___L-L-.1._UL.L..U_
10· 2
JO-:
__.__.L._._,'-W._,__...,
dt
l.il
Cz t
ln(t) .- C3
[C4 2
+-(ln(t)
C3)2)112
!f':E. SEC
FIG URE 6
Pressure rise in a closed vessel.
(I 2)
(9) does not adequately describe the pressure rise in the bomb. On a log-log graph, Eq. (9) should give a straight line with a slope equal to {3. Perlee et al. (1974) used a bomb ofa relatively large volume and fcund that the slope of the pressure vs. time logarithmic plot changed from 3 to ~.3 as the pressure rise became larger than 0.5 atm . Such a change in the slope was not reported by other investigators who used a relatively small bomb. In light of the results from the present study it is believed that the combustion duration in a small bomb is too short to reveal the exact shape of the pressure history curve. The present results indicate that the slope of the logarithmic pressure vs. time curve increases continuously toward an asymptotic value. This phenomena is best de~c ribed mathematically by an equation of the form
3.2
The Burning Velocity
Laminar burning velocity was measured in the bomb in an identical way to the measurement of turbulent burning velocity except that the mixture was ignited only after sufficient time elapsed for the turbulence in the bomb to die out ("'IO min). The effect of EGR on the laminar burning velocity of a homogeneous natural gas/ air mixture at an initial temperature of 340° K and a pressure of 1.09 atm is given in Table Tl . According to Andrews and Bradley (1972) , the laminar burning velocity of a methane/air mixture can be predicted using the formula
SL = 10 -i 0 .00037I T 2 cm/sec.
(13)
For T = 340°K this formula gives SL = 52.9 cm /sec. Considering the use in the present study of natural gas rather than pure methane, the difference, of less than 2 percent, between the predicted and measured burning velocity (see Table 11, 0 percent EGR) is insignificant.
EFFECT OF EGR IN PREMIXED METHANE COMBUSTION
7
TABLE II The laminar burning veloc ities and other parameters in the turbulent burning velocity equation (see text) T; = 340°K , P, = 1.09 Atm., ¢ = 0.965 ---------------------~---
SL
Dpr
% EGR
cm/sec
k 111a
k2
0 5 10 15 20 25 30
53.7 41.4 32.7 25.2 19.2 13.3 9.3
3.7 1 2.99 2.60
0.0 ·,906 0.07224 0.07953
0.06906 0.06625 0 .06741
2.03 1-66 2.36
0. 10536 0.1 2473 0.25469
0.07481 0.07829 0. 14167
cm
k2/ Fx
0.043 0.055 0.070 0.091 0.119 0.172 0.247
______________________ ___ &.O
c
i!
Jenee has a stronger effect on a mixture of a lower burning velocity, a phenomena predicted by Eq. ( 17) derived and discussed by Skelkin (1943) for a wrinkled flame front :
ESR
5 • IU • 20
0
s.o
sc::
Qt
" 2S
o in
•.O
.............__
J, O
(17)
'5
-'"' ~~
~~ ~
2.0 l.0
0.0
'--~----'--_.____._
0.0
5. 0
10.0
iS.l'J
_
i{U
_..__ 25.0
.,q!Ult'it!.
_._____._ 30.~
:r
_ ~ •..ti
_
_.___~
!;~ ,.;
5'
I
l~ W~~ H Y. £ ~/SH
(v' I A)
FIGURE 7 Turbule nt burning velocity with exha us t gas recirculation.
The effect of turbulence on the burning velocity is shown by the data given in Figure 7. The lines in the figure are the best curve fit for an equati on of the form (Putnam, 1976) :
-ST -_ [ l +ki ( -v' SL
) "-]
l /a
( 14)
SL
It should be noted that as v ' becomes very small thi s equation gives (15) a nd as v' beco mes very la rge, a linea r relatio nship is obtained
ST -
SL
=
In order to determine the combustion mode in the present study, the Kovasznay ( 1956) parameter was calculated
k 9 v' "
(16)
Table II and Figure 7 show th a t the slope of the lines, k 2 , inc reases as the level of EGR increases. This implies that for v' of the order of SL, turbu-
r = -- . (SL/01)
(18)
The flame thickness was assumed to be equal to the preheating zone thickness which was calculated following Damkohler (1940)
ow =
4.6 i
.0
S.'
01 CG~
0
E
0
+ lD A:-0
x ·< ()}"':_
3.0
~ ;;] j,
U.O ' - - -' - -- ' - - - ' - - -' - -- ' - -- ' - - - ' - - - ' - - - - ' ' - -- '
... s
ij,11
s.:.
io .;,,
~.o
Sr/S1, versus Re.,,.
2".
2:;.a
30.o
TIJROOU:NCC I·nm l1Y'
"'" ·. YiFIGURE 11
t~ .o
FIGURE 12
JS.a
40.o
:.s.o
wm
Flame velocity corrected for EGR.
EFFECT OF EGR JN PREMIXED METHA N E COMBUSTION
3.3
9
The Dead-Space Thickness
A primary problem in investigating the dead-space thickness is that there is no formal definition of what the dead space is. Such a definition would depend on the geometry which in the present investigation was a flat-flame-front moving perpendicularly toward the wall (the spacing of tbe ionization gauges, see Figure 3, relative to the radius of the flame when it approaches the quenching surface is such that for all practical calculations a flat flame can be assumed). As the flame approaches closer to the wall it would quench at a certain distance from the wall. There are numerous optio ns for defining the dead-space thickness for su : h a situation . For instance, the distance between the wall and the luminous zone of the flame or the distance between the wall and the reaction zone (i.e., the inflection point in the temperature profile), could both serve as definitions. It is p::>ssible to use other definitions as well, however, not a ll definitions are amenable to a simple experimental measurement. Jn the present study two methods were employed to measure the dead-space thickness. One is based on the temperature gradient of the gas in the quench layer and the other based on the ionization phenomena occurring in a hydrocarbon flame. W ith the first approach, the dead-space thickness was estimated from the heat flux to the wall from the flame at the instant of quenching (maxi mum of the heat flux), via the relationship
ot*
The bo undary conditions are
Tw(O, t )
= j(t)
Tw(l , t )
=0
and the initial condition
Tw(x , 0)
q=
-K w
j x
Q
200
,\FLAME
I' QUENCHING I' I I
160 kw
120
Tf
80
(23)
'
I
40
(3)
T1-Tw 0. 0125
and K
-=
oq
h.
(24)
To find the heat flux, q, the following procedure was followed. The temporal heat transfer equation for the quenching surface was solved , using the experimental surface temperature history as a boundary cond ition.
(26)
-o
240
m
Tw
oTw
A typical plot of the measu red wall temperatu re and the calculated heat flux during the flame approach for a 20 percent EGR mixture is shown in Figure 13 .
·
i< = - -- - --
0
ox
q, 7
JK(T) dT
=
where f( t) is the experimentally determined wall temperature. The heat flux was found then by evaluating the following expression
(22) where
(25)
ox* 2
.025 T, SEC
.0375
T °F
68 66 64 62
FIGURE 13 T emperature his tory and calcula ted heat Au x at the quenching surface.
It is possible to distinguis h, in Figur~ 13, am o ng three r~ g ions. T he temperature rise in region one is due to the adia batic com p ression of t he gases in fro nt of the fla me. The fla me interacts with the wall
N. GAT AND C. W. KAUFFMAN
10
in region two, where the interaction time is approximately equal to the ratio between the flame thickness and the burning velocity. Finally, the slow temperature rise in region three is due to the contact with the hot reaction products, the postflame reactions and the isentropic compression of the burned gases . Quenching of the flame occurs at point Q when the heat flux is at its peak. The area under the curve in region two gives the total heat transfer to the wall during the flame-wall interaction. However, it was found that the total heat transfer is not a controlling parameter. The maximum heat flux, i/max at po int Q, was found to be nearly a constant for a particular mixture regardless of the turbulence in the bomb, a somewhat surprising result. The heat losses which a flame can sustain without quenching depend on the rate of heat release by the flame, which in turn depends on the burning velocity. Therefore, the above observation can be understood only if the flame approaches the wall at a constant burning velocity regardless of the turbulence . In other words, turbulence seems not to affect the flame in the close proximity of the wall. This occurs, perhaps, due to the presence of a laminar sublayer which results from the nature of the flow field in the vessel (i.e., a decaying, turbulence with a mean flow velocity equal to zero). This sublayer, however, must be much thinner than the dead space defined by the ionization signal. Actual quenching may occur in a laminar burning regime, dominated by molecular transport phenomena rather than turbulent diffusivity.
230 3~ 0
2200 2100
300
~
0
260
220
2iYJ(i 1900 l80ii 1700
0
· r ... J
ll _.)
h · IOJ [
00
l',i
r-1 X
J
_/
0 10
JS ~
FIGURE 14
-
70
2S
30
E.SR
Effect of EGR on heat transfer .
Figure 14 shows the change in the value of and the heat transfer coefficient. h, a s a function of EG R . The latter parameter .
i/max
