^i/vr/Vi7^^||(^ I U^''". X.SVV- *^ NATIONAL ADVISORY COMMITTEE FOR AERONAUTICS TECHNICAL MEMORANDUM No. r> ON COMBUSTIGK IK A ITIPBULEKT FLOV/v By K. J.. . >ielkin Jourixal of Technical Physics (USSR) Vol. XIII, Hob. 9-10, 19^*3 aciP^ . IMIVFRSITY OF FLORIDA Washington ^'^'^TS' tc nFPARTMENT P.O.BOX 117011 ,o,njs^ GAINESVILLE, FL 32611 fUii -::^ X ' - ' / /uh), the mixing being begins to affect not only the heat transfer (as for small scales) but also the time of chemical reaction in the flame. Thus are presented the laws of flame propagation at a small scale of turbulence, and the properties that appear as the scale increases are showed. It should be pointed out that the case where the scale of turbulence is small by comparison with the width of the flame front is rarely met with in pure form. Under normal conditions for homogeneous gas mixtTuree the width of the flame front is of the order of 0.1 millimeter. Under real conditions (gas furnaces, engine combustion chamber) scales of order of magnitude less than 0.1 millimeter can only accompany larger scales and occupy only a certain extreme position in the scale spectrum. Howevsr, NACA ™ Wo. 1110 7 immediately on passing to slowly burning mixtures (e.g., lean mixtures) the width of the front may considerably exceed the above-mentioned order of magnitude^ and the probability of existence of the case discussed increases. Bext, consider the case where the scale of tui'bulence is large in comparison with the width of normal combustion zone. An analysis of this case in its general foiia on the basis of dimensionality considerations permitted Y. Zeldovich to conclude (unpublished work) that the speed of turbulent propagation of the flame does not depend on th$- scale of turbi>- lence and can be represuntsd by a formula of the type UT = Ui^ f (ufl/v ' ) where f (uj^/v') is a function of the ratio of the velocities to be determined. In the case of large-scale tarlRalence the flame surface is cui'ved. As the velocity of the fluctuating components increases the curvature increases^ and finally the flame front begins to break away. At strong turbulence the elementai'y volumes of gas^ both burning and fresh,, move chaotically with respect to one smother. In the flame there remain "islands^ " centers of unbui-ne-d mxture, broken off by the fluctu- ations into pai'ts and annihilated by the flame. If the effect of the curvatiire on the speed of flame propagation is neglected^ the normal flame speed may be considered as constant. The surface of combustion, however, increases and hence also the total rate of combustion. Before determining the speed of turbulent combustion over the flame surface^ consider the limiting case of a strong large-scale, turbulence when the combustion zone is filled with islands of imburned mixture. In figure 2 the sppce between the sections AA and BB may be con- sidered as the reaction zone of the turbulent combustion. In front of the plane AA is the fresh mixture, and behind the plane BB are the products of combustion. From A to B the mean concentration over the cross- sectional areas of the products of combustion Increases, and the concentration of the fresh gas decreases. The distance betvreen the sections may be considered as the width of the t^jrbulent f lomo front. As was mentioned previously, the speed of the flame, from dimensional considerations independent of the mechanism of the reaction is a magni- tude of the order of u~ JxT^ (3) In a turbulent flame the p.art of the temperature conductivity is taken by the coefficient of exchange Zv* and the part of the reaction time by the mixing time Z/v', and this gives for the velocity:. 8 ■ MCA OM Eo. 1110 Thus the conclusion is reached: for large turbulence when the velocity of fluctuation v' is large 'by corrparison with the normal velocity v^ (only in this case will the combustion front have a form like that in fig. 2), "the velocity of tiu'hulent flame propagation is proportional to the nean fluctuating velocity and does not depend on the chemical nature of the gas mixture. The reaction time in a turbulent flame may be computed .in a differ- ent mannei". Tlie combustion time of an elementary volume l'^ may be determined as T - (P) where £5% -^^ ^^^ volume rate of combustion, the product of the flame area by the normal velocity. It should be taken into account that an elementary volume with area of the order I- on entering the combustion front is broken up into parts by the fluctuations. The breaking up will continue as long as the flsme with normal velocity xx^i passes over a distance equal to the scale of turbulence I, this time is equal to Z/iN. During this time the total path traversed with the fluctuating velocity reaches the value L = l/n-^ v*. The ratio l/z shows how many times during the combustion of the volume Z'^ the latter is traversed by fluctuations. Each such traversal leads to the formation of a new fLnme area of the order to Z^. The mean area of combustion of an element Z"^ will be proportional to the magnitude Z^ vVujj (6) The averaging of this quantity with respect to the combustion time affects only the constant factor. The time required for burning the volume Z^ id found eaual to (7) ■Ug The same result is attained; namelj^, that the combustion tinie is propor- tional to the mixing time. NACA TM Wo. 1110 The case where v'» ujj has heen eonsidered. Tlie case of small tTjrhulence may he approached only geometrically. It is in this meinner that the problem is considered by Damkohler. Rightly seeing the reason for the increase in velocity of combustion in the increased area of the flame surface, he schematically represents the flame surface as consist- ing of conical surfaces with their bases at right angles to the directicn of flame propagation (fig. 3)» Taking the areas of the cones proportional to v' he arrives at the expression Urj] ~ v' The considerations of Damkolrler are not accurate and for small v' are not true. According to Damkohler, the absence of turbulence (v'.=0) leads to zero velocity of the flame. Actually a relation like (k) as shown above is obtained only for strong turbulence v'>ui^, and not for weak turbulence which is the case considered by Damlcohler. Consider this problem more in detail. The ratio of the speed of turbulent propagation of the flame to the normal speed will be equal to the ratio of the lateral area of the figure a (fig. 3) to its base. The height of the figure, which is assvmied as a cone, will be of the order Zv'/ujj. For an element of the flame front will be carried away by the fluctuation from the general flame front only during the time re- quired for the noi'mal of the flame to traverse the distance Z. This time is eaual to Z/ujj. The height of the cone will therefore be of the order of Zv'/uj,j. The lateral area is Slat - AZ2/1 + B(v'/uij)2 where A and B are nondimensional coefficients of the order of imity. The ratio of the velocities of propagation is }n ^ Slat ^ Ai/ 1 -H B(vVuii)^ (8) % ^ase For strong turbulence v' > u,^ (equation (8)) leads to the known relation Uij V • The absence of turbulence v' = should lead to the condition u^p = Uj^. This gives the value Ai = 1. For any tiirbulence, including small turbulence, the criterion shouii hold 10 HACA TM No. 1110 ■ (ld''.=- l-^B(^) (9) The value of the fluctuating component of the velocity is thus found to vary hypertoUcally vith the velocity of turhulsnt flame propagation (fig. h) . If v' is replaced by the product of the Karmaii munher by the mean flow velocity, the formula is obtained '-^-) - 1 + BK2 (10) in which all magnitudes are subject to direct measurement. It is of importar,Ge to note that in the finite expression (9) or (lO) the scale of turbulence dees not enter. Hence, the obtained rela- tion will be valid for any scale, provided the latter does not exceed the width of the normal combustion front. BOMa Pli^CTICAL REMARKS From formii.la (9) it follows that for lax^e ratios v*/ujj (or by formula (lO) Y^^j-x-^) unit y may bo neglected undei- the root sign. The conibuEtion velocity is then independent of the noriual flame velocity and therefore of the phyEicocheraical properties of the nixture, and is pro- portional to the fluctuating velocity. For a given value of the latter (large in comparison with \x.^) different fuels will burn at the same I'ate. If the I'atio v'/u (J?///ui}) is of the order of 1 or less, the x-ate of coar bustion will depend on Utj. V.'ithin increasing fluctuating velocity the rate of combustion increases hyperbolically. For very small v' the effect of the turbulence may be neg3.ectod as a second- order magnitude. By comparing v'/% with unity, it should be remembered that the val;ie of B is equal to 1 only in. order of magnitude. Fuj^-thermore , it will evidently depend on the structure of the turbulence. For exeonple, in pip;es with roughness of various shapes determ-inlng the structure of the turbulence, different values of B may be expected. Examination of formula (9) permits drawing the conclusion that if the velocity of fluctuation is not very large in comparison with ujf the effect of the turbulence on the speed of propagation characterized by the relative increase in the combustion velocity (urp/ujj) will be greater the slower the combustion of the mixture of the same composition at rest. The lower the value of u^ the greater the given v' (gi-wen turbulence), the numerical value of the root in formula (9). MCATM No. 1110 11 This conclusion is confirmed "by the published tests on the effect of swirling and turbulence on the speed of comhustion (reference 3)- Swirliiig always moi^e greatly increases the speed of combustion of lean and rich mixtures than stochiometric mixtures. The conditions of the experiment do not permit a quantitative analysis of these data but qualitatively they correspond in any case to these results. To evalu- ate the possible effect of turbulence on the flame propagation, it is necessary first of all to know the normal speed of combustion Uj^ and, of course, the speed of the fluctuations or, if the turbulence is known (the Karman number), the mean flow velocity. In table 1 are given the normal flame speeds for certain fuel-air mixtures taken from the book of Jost (reference h) (the speeds correspond to the coiiiposition of the mixture of maximum combustion rate). As an exeoaple, there are also given the mean flow speeds for 5--percent turbulence (K - O.O5), the mean fluctuating speed of which is equal to the corresponding normal speed of the flajne. Table 1 Fuel (cm/sec) w (m/sec) Hydrogen 267 53. i^ Acetylene 131 26.2 Ethylene 63 12.6 Propylene i^3.5 8.7 Methane 37.0 7.4 n— pentane 35.0 7.0 n— hexane 32.0 e.k Benzol + O.5 per- cent Ha 38.5 7.7 Carbon monoxide +1.2 percent water 41.5 8.3 Table 1 gives an indication of the order of flame speed at 5- percent turbulence when there is to be expected a hyperbolic depend- ence of urp on ¥ or when this dependence may be considered as linear. Thus, for hydrogen-air mixtures of stochiometric composition a linear 12 KACA TM Ko. 1110 dependence (and independence of the speed of propa(i;atlon on the normal flame speed) may "be expected at f lov speeds exceeding approximately seven to eight times the speed giving this dependence in pentane or benzol - air mixtures. For hjdrogen these will ha speeds of the order of himdreds of meters per second; for benzol, methane, pentane, hexane, of the order of tens of meters per second. If the t-orbulence is in- creased tivo times, up to 10 percent, then to attain the same result half the flow speed would be required, and so forth. The practical independence of the speed of f Lsme propagation on the physicochemical properties of the fuel in the engine was also observed by Marvin (reference 5) • On figure 5 are given the curves of the flame path against the crank angle degrees obtained at constant engine speed. The slope of the curves gives the fl/me speed. In the center part of the combustion cBamber where the flame front is sufficiently developed the flame speed cksnges little over a large distance of the chamber, and for various fuels is approximate!^'- the saiae while the nomial combustion speeds of these fuels differ considerably/ from one another (table l) . Different speeds in the engine are observed onlj in the initial stage of the flame propagation. The results of Marvin cen readily be explained. The velocities of the fluctuations of che gas in the engine are large in comparison with the normal velocities so that the flame propagation is described by expression (U)^that is, the rate of combustion is proportional to the speed of the fluctuation, and does not depend on the normal flame speed. It is very probable that under the same conditions for hydrogen or acetylene the speeds would be different. It is possible that for these gases the same speed of the fluctuations would not be large by compari- son with ujf. These considerations ai'e in the nati^re of suppositions since the true value of the speed of the fluctuations under the test conditions of Marvin is not knoi/n. The absence of an accelerating effect of the tujrbulence in the initial phase of the combustion (well icnown from the engine literature) may from the author's point of view be explained by the fact that at the start of combustion when the dimensions of the flame are small in comparison with the scale of the turbulence the latter does not affect the rate of combustion. A center of combustion is displaced as a whole by the fluctuation, .and its surface therefore does not become branched. The rate of combustion is determined only by the normal flame speed. In this way it is assumed that the tiu:'bulence in the engine is of relatively large scale. Certain results for comparison of the theorj'^ with experiment are given also by other investigations conducted on engines. There are, however, fuaidamental difficulties, in applying the theory previously presented, to combustion in the engine. Consider a few of these. MCA TM Wo. 1110 13 In the first place, with regard to the effect of the rotational speed on the combustion rate, it is not known whether the Karman number re- mains constant when the speed chsjiges. This can only be assumed on analogy with pipes and special determinations : of K, for engines are lacking.- In the second place, the true effect of the motion of the entire mass of the gas in the engine chamber on the combustion rate is not clear-. It may be supposed that owing to the small length and the strong turbulence -producing effect of the intake valve, the degree of turbulence of the flow diaring intake and then in the chamber will be very strong, and the pajrt pla^'ed by the regular motion of the gas is small. Flnal3-y, photographs give the flame speed with respect, to the walls of the chambex' and not with respect to the gas, which is of inter- est in the latter speed. In the latter case, however, the picture is not clear. It may, moreover, be supposed that rotational flows are pre- doiulnant in the engine. This is indicated, for example, by Marvin's photographs in which a curbing of the f.lame front is noted, explained by the circular motion of the gas in the cylinder head. These consideratioaa are confirmed also by the fact that generally over a considerable distance of the combustion chamber the photographs of the flame in the engine show an only slightly varying flame speed. In the case of the presence of a strong flow along the chamber 'from one end to the other) there should exist also a return flow. It would then be possible to expect large ir- regularities in the flame propagation. There is, therefore, some basis for supposing that the principa.1 cause for the increased rate of combus- tion in the engine is the turbulence and not a mass movement of the gas in the engine head. In this respect the opinion of engine specialists may be s^^scribed to; for example, Bouchard, F. Taylor, and E. Taylor, who write, "the rapid increase in the maximum combustion rate with ro- tational speed is due primarily to an increase in the small-scale in con- trast to the organized motion of the gas in the engine head but not in the sense of comparison with the width of the flame front of normal com- bustion, of higher gas velocities through the intak;e system." (See reference 6. ) The same authors investigated the dependence of the flame speed on the rotational speed. The curve of flame speed against rota-r tional speed is given in figure 6. The speed of the mixture on intake, and therefore the speed of the gas in the chamber, increases in propor- tion.',t.o the rotational speed. Hence, in figure 6 on the axis of ab- scissae' instead of the rotational speed there may "be laid off a magnitude proportional to the speed of the gas or under the assumptions made to the speed of the fluctuations. The shape of the curve corresponds strongly to the theoretical relation obtained (fig. h) . Unfortimately, for small rotational speeds the experimental curve is extra-polated (dotted) . Bouchard, F. Taylor, and E. Taylor drew this part of the curve on the basis of measurement of the time of combustion of 95 percent of the charge. The tests of these authors confirm (though indirectly with account taken of the assumptions made) the conclusions as to the effect of the speed of the flame propagation. An acciurate check of the theory requires, of course, special tests with parallel measurement of the speed of combustion, Karman nxanber and flow speed. ll; . ■ KACA TM Ho . 1110 COWCLUSIOKS 1. With strong tiirbulence as the scale of the turbulence in- creases to a value comparable with the width of front of normal com- "bustion, the rate of the combustion reaction begins to depend on the mixing. 2. For turbulence the scale of which exceeds the width of the front of normal combustion (I >\) , the speed of the flame propagation increases hyperbolically with the speed of the fluctuations. For large fluctuation speeds (v' > ujj) the dependence may be considered linear, and for small values (v'