I
Engineering Approaches
1
H i g h - T e m p e r a t u r e Kinetics in
...
Turbulent Combustors
I
W. D. WEATHERFORD, Jr., J. P. CUELLAR, Jr., and ROBERT K. JOHNSTON Southwest Research Institute, San Antonio, Texas
Choice of fuel for aircraft gas turbine engines could depend on these relationships between fuel deposition and metal loss I N SELECTING FUELS for aeronautical gas turbine engines, the need for considering cleanliness of combustion and intensity of flame radiation is steadily increasing. As more severe operating conditions are encountered, problems resulting from carbon deposition diminish while difficulties associated with durability of combustor components become more important. Combustor durability is intimately related to smoking tendencies of fuel, and it is now generally believed that increased radiation accompanying increased smoke formation is a major contributor to decreased combustor life. An experimental study has been conducted to develop methods for evaluating combustion characteristics of hightemperature fuels. This effort included a study of deposition properties of various fuels. The resulting kinetic evidence, supported by combustor metalloss data from another laboratory (5: G), suggests that radiation heating is not solely responsible for combustor deterioration. This study indicates that weight loss may be directly influenced by reactions of fuel with metal. Lowpressure combustor deposition characteristics of hydrocarbon fuels are intimately related to higher-pressure combustor metal-loss processes. Proposed reaction mechanisms which lead to a satisfactory correlation of experimental data suggest that both deposition processes and metalloss processes proceed by way of chemisorption of fuel-derived intermediates on metal surfaces. Depending upon combustion conditions, adsorbate may be converted, with participation of existing deposits, into additional carbonaceous deposits, or adsorbate can undergo oxidation to produce metal-containing oxidation products. This article describes the proposed chemical mechanisms by which carbon deposition and metal loss occur and presents a kinetic analysis of experimental data in terms of proposed mechanisms.
398
Experimental Deposition studies were conducted with a small-scale combustor having a turbulent-vortex combustion zone. This microburner (below), developed by Phillips Petroleum Co. ( 4 ) , was constructed from drawings furnished by the developer. Extensive exploratory studies of operating characteristics of the microburner and its accessories led to numerous modifications in equipment, techniques, operating conditions, and procedures (3). Because of turbulence, the microburner may be considered somewhat similar to gas turbine combustors. Fuel and one fourth of the total air enter the burner through a removable I/pinch diameter deposit tube (Type 304 stainless steel) positioned axially in a 1l/ 4-inch diameter cylindrical combustion chamber. The remainder of the air enters the base of the combustion chamber tangentially a t near-sonic velocity, generating intense vortex actions. The fuel-rich vapor emerging from the upper end of the deposit tube is caused to flow downward along the outer surface of this tube by the down-flowing core of the vortex. Hence, the primary combustion is caused to occur near the deposit-collecting surface, and the combustion heat is partly utilized for vaporization of fuel spray within the deposit
tube. Extensive testing revealed that negligible deposits are formed on the inside surface of the deposit tube Tvith the fuels studied. The metal-loss data of Streets (5, 6) were obtained in a 2-inch diameter turbulent combustor (Type 304 stainless steel), using a ratio of combustion air to quench air of about 1 to 4. This combustor design includes features in common with full-scale aircraft gas turbine combustion systems, and experimental performance of the apparatus is considered typical of full-scale combustors. Mild-temperature operating conditions employed in this laboratory for the microburner deposition studies and extreme-temperature conditions employed by Streets for the 2-inch combustor metal-loss studies are summarized in Table I. Fuels, Ivirh the exception of toluene, were furnished by the Air Force for use in both of the described studies. Fuel compositions, determined in this laboratory, are presented in Table 11.
Discussion of Results Since duplicate tests demonstrated quantitative differences in deposition rates, an extensive study was made of experimental variables and of refine-
Table I.
Operating Conditions
S I L I C A FLAME
Microburner Air temperature Air rate
A L x * L IN
The down-flowing fuel burns near the microburner deposit-tube surface
INDUSTRIAL AND ENGINEERING CHEMISTRY
Fuel-air wt.ratio Combustor pressure
550'
F.=
29.5-31.2 ft. /sec.b 0.07 Approx. atmospheric
Phillips 2-Inch Combustor (5) 800' F. 1.03 Ib./sec.
0.010 450 inches Hg abs.
Indicated within burner base. Based on 550' F. and 1 a t m . and area of l'ja inches I.D. swirl ring. a
1
ments in experimental procedures. This study revealed that these discrepancies m a y be t h e result of t h e i n h e r e n t n a t u r e of t h e deposition processes. When the average cumulative deposition rates for three representative fuels are plotted us. test duration (Figure l),cyclic patterns result, and the similarity of these patterns among various fuels is striking. When the logarithm of deposit quantity is plotted us. cumulative fuel consumption (Figure 2), average data for most fuels tested indicate linear relationships during the first deposition cycle. These relationships imply that the rate of deposit formation is directly proportional to the quantity of deposit present. Upon ignition of test fuels, preliminary, porous deposit layers are formed immediately. This intercept deposition results from precipitation of volatile or suspended matter from combustor gases during warmup of the deposit tube. Because of the cyclic nature of the deposition process (Figures 1 and 2), slight variations in experimental startu p procedures, operating conditions, and termination procedures could be responsible for much of the data scatter. The chance for this type of error increases as both the amount of deposit and the deposition rate increase. Hence, experimental slopes should be biased toward being low, as evidenced by the exaggerated decrease in slope of the over-all pattern of points a t higher fuel consumptions. In order to compensate a t least partly for data scatter, the arrays of points were correlated by drawing linear boundary envelopes in the lower fuel-consumption range. Typical deposition data (3) are shown in Figure 2 with linear boundary envelopes indicated, and average and maximum boundary-envelope slopes are correlated in Figure 3. Postulated Mechanisms. Initial slopes of averaged experimental data indicate that, during the initial cycle, deposition processes follow first-order kinetics. This deposition, which is less rapid than the initial intercept deposition, is assumed to proceed in several consecutive steps. Both existing deposits and deposit tube
surfaces are assumed to participate in this sequence of reactions. T h e proposed mechanisms may be summarized as follows. Certain fuel constituents or degradation products are chemisorbed on the deposit tube surface. These chemisorbed complexes may react with existing carbonaceous matter to form additional deposits, or they may be attacked by oxidant to yield metal-containing, volatile or loosely held products. Simultaneously, deposits are consumed by direct oxidation and mechanical carryoff. Detailed mechanisms and assumptions: CHEMISORPTION :
Assumption: Steady-state adsorbate concentration ( F M ) established by intercept deposition. CONVERSION OF ADSORBATE INTO DEPOSITS
s
Table II.
Boiling Range, O F.
Fuelsa DiendomethyleneDecalin 438-496 Toluene 231b Isopropyl bicyclohexyl 535-573 Aromatic solvent 276-521 Alkyl Decalins 418-469 Decalin 371-406 Nonaromatic oil 506-720 Diethylcyclohexane 340-388 Olefin-free. Handbook value.
:
FM
+D
M
+20
(2)
Assumption : Negligible rate under extreme conditions conducive to metal loss. Loss OF ADSORBATE AND ACTIVE SITES: F M f - 0 k3 , Adsorbate Loss and Metal Loss
(3)
Assumption : Negligible rate under mild conditions conducive to deposit formation. DEPOSIT BURNOFF: D
+ 0 A Deposit Loss
( 4)
DEPOSIT CARRYOFF : D
+ Turbulence 2Deposit Loss
(5)
Kinetic Analysis. The following kinetic relations are prescribed by the proposed mechanisms :
w
3 0z
80
I60
2 4 00
320
400
3
FUEL C O N S U M E D , G
Figure 1 . Different fuels yield similar deposition cycles
Net Deposition Rate :
+
(dD/dt) = ( d D / d t ) o (dD/dt)B -k (dD/dt)c (11)
The total concentration of active chemisorption sites may be expressed analytically as the sum of unoccupied sites and chemisorbed complexes:
(K) = (M)
+ (FM)
(12)
Under mild conditions conducive to deposition with negligible metal-loss, the rate of chemisorption should equal the rate of adsorbate conversion. Hence, Equations G and 7 yield : ( F W = (kl/kZ) ( F * ) ( M ) l ( D )
(13)
For extreme conditions conducive to metal loss with negligible deposition, Equations G and 8 give : (FM) =
(ki/k3)
(F*) ( M ) l ( O )
(14)
Combining Equations 12 and 13 with Equation 7 yields the following relation for deposition rate (under mild conditions) :
Chemisorption Rate : (dF*/dt) = ki(F*) ( M ) Deposit Formation Rate :
I
*
(1)
F*+M%FM
‘
(6)
(dD/dt)o = kz ( F M ) ( D )
(7)
Metal-Loss Rate : -(dM/dt) = k8(FM) ( 0 ) Deposit Burnoff Rate :
(8)
- ( d D / d t ) ~= kr(D) ( 0 ) Deposit Carryoff Rate:
(9)
-(dD/dt)c = k,(D)
(10)
Fuel Compositions
Lamp Sulfur, %
Aromatics,
0.000
... 0.006
5 I
.
0.012 0.003 0.006 0.040
5 75 2 2 1
0.000
1
Naphthalenes, yo 0
...
0.03 10.4
...
0.04 3.4
...
Figure 2. Boundary envelopes are used for correlating cyclic deposition data VOL. 53, NO. 5
M A Y 1961
399
. a W
A
32 -
B C D
in ~
9
E
i G 2 4 -
F
L E sz
G H
.
=[r
+ o w l16 i
DIENDOMETHYLENE DECALIN TOLUENE ISOPROPYLBICYCLOHEXYL A L K Y L DECALINS AQOMATIC SOLVENT DECALIN NONAROMATIC OIL DIETHYLCYCLOHEXANE
-
* W
2I= 0
5
5
-
L E G E N D H A V E R A G E ENVELOPE SLOPE
*.
/ /
/
H M A X I M U M E N V E L O P E SLOPE
//
0
Combining Equations 12 and 14 with Equation 8 yields the following relation for metal-loss rate-under extreme conditions :
For fixed conditions, elapsed time is proportional to fuel consumed, and the quantities (F*) and ( K ) may be considered as constants. Mild-condition, net deposition rates then may be represented by : ( d In D / d F ) =
kAK/[(D)
+ ( k ~ / k ~ ) l k s - kc -
(15a) which, for deposit-free surfaces plifies to : ( d In D / d F ) = kDK - ke - kc
sim(15b)
where the various k's are proportional to rate constants, and subscripts, A , B, C, and D, respectively, denote adsorption, burnoff, carryoff, and deposit formation. Similarly, for fixed conditions, extreme-condition metal-loss rates may be represented by:
where (0) signifies the total effective concentration of oxidant, including air and decomposed oxygen-containing or sulfur-containing compounds, and where subscript, M , denotes metal-loss processes. For turbulent diffusion flames, which are approximated in the microburner and in Streets' 2-inch combustor, primary combustion processes occur in fuelrich regions of the combustors, and therefore Equation 16a reduces to: -dM/dt
= k,wK ( 0 )
(16b)
The validity of metal-loss mechanisms assumed in the foregoing derivation (Equations 8 and 16, a, b) is partly confirmed by Street's additional 2-inch combustor studies (6) which used isoparaffinic heavy alkylate fuels. In these experiments, metal-loss rates were observed to be independent of time (following brief induction periods), in
400
agreement with Equation 16. In addition, metal-loss rates were higher when sulfur compounds were added to fuel. Within experimental repeatability of the method, increased metal-loss rates caused by addition of sulfur compounds were proportional to the quantities of sulfur added, irrespective of the particular sulfur compounds used, which agrees with Equation 16. A rational explanation of observed cyclic deposition phenomena can now be made on the basis of the proposed mechanisms. During the initial deposition cycle, Equation 15a reduces to first-order kinetics. However, as the quantity of deposits increases, the deposition rate tends toward zero order and decreases, thereby upsetting the thermal balance and causing the deposit burnoff rate to change. Thermal lags, deposit carryoff, and the changed order of the deposition process prevents the establishment of new equilibrium rates; hence, the observed cycling results. Additional information supporting the postulated mechanisms has been reported by Anderson ( 7 , 2). His experimental studies also indicate that metal surfaces participate actively in gas-phase deposition reactions. and he concludes that, in the presence of oxygen, such participation could enhance reactions of oxygen with adsorbate and metal ( 2 ) . Correlation of Metal Loss with Deposition Kinetics. In order to relate mildcondition deposition to extreme-condition metal loss, it is only necessary to assume that, under fixed conditions, the number of clean-surface adsorption sites is identical for all fuels under consideration. On this basis, Equations 15b and 16b may be combined to give: -(dM/dt)'
Z(ko,vI/kD) X
(O)'[(dln D / d t )
kB
4- kcl
(17)
where the superscript, degree, denotes extreme-condition operation; the lack of a superscript reflects mild conditions; Z is a proportionality constant; and (0)"is constant for fixed conditions.
INDUSTRIAL AND ENGINEERING CHEMISTRY
If ko.Tf, k D is to be considered constant, it is necessary to assume further that the influence of fuel properties on the adsorbate conversion rate constant, k D , is the same as on the adsorbate burnoff rate constant, ko,. Figure 3 demonstrates that the logarithmic deposition rate-average slope of boundary envelopes-correlates well with 2-inch combustor metal-loss data reported by Streets (5). When the slope bias is reduced by considering only the steeper of the two boundary envelopes for each fuel, as illustrated in Figure 3, the metal-loss cxrelation becomes approximately linear. This degree of correlation supports the hypothesis that metal-loss processes are intimately related to deposition processes. This contrasts with the poor degree of correlation between metal loss and metal temperature or radiant-heat flux observed by Streets (5). The fact that the maximum-slope correlation is approximately linear may be fortuitous. However, either case provides added qualitative agreement with kinetic predictions and does not conform to the concept that only heat-flux-induced direct oxidation contributes to loss of metal.
Nomenclature D F F* FiM
= carbonaceous deposits = cumulative fuel weight
k
= reaction rate constant = total active adsorption sites = unoccupied active adsorption site = oxidant = time proportianality constant
K
IM
o t
= fuel-derived intermediate = chemisorbed intermediate
z =
Subscripts
A
B C D
M
adsorption burnoff of deposits = carryoff of deposits = deposit formation = metal loss = =
literature Cited (1) Anderson, R. C., Dent, J. H., AF O f . of Sci. Res. Tech. N o t e 58-50, University of Texas, December 1957. (2) Anderson, R. C., University of Texas, private communication, Sept. 30, 1959. (3) Johnston, R. K., Shamblin, J. E., Lt'eatherford, W. D., Jr., Schneider, K. H., Cuellar, J. P., Jr., WADC Tech. Rept. 59-776 ( I ) , Southwest Res. Inst. Rept. RS 302 (October 1959). (4) Streets, W. L., Phillips Petr. Co. Res. Diu. Rept. 7793-57R (May 1957). I (5) Streets, W. L., Phillips Petr. Co. Res. Diu. Refit. 2307-59R (March 1959). ( 6 ) Streets, M7. L., Phillips Petr. Co. Res. Diu. Refit. 2572-60R (February 1960). RECEIVED for review October 6, 1960 ACCEPTED January 31, 1961 Division of Petroleum Chemistry, 138th Meeting, ACS, New York, N. Y.,September 1960. Work supported by Propulsion Laboratory, Wright Air Development Division, under Contract 4 F 33(616)-5702.