Structure of Reacting and Non-Reacting Swirling Air-Assisted Sprays

-A detailed characterization of a methanol spray produced by an air·assist atomizer with swirling atomizing air has been conducted. This study is the third of a series which examines the structure of sprays produced by a standardized atomizer which can be operated in three modes. pressure swirl, non·swirling air-assist. and swirling air-assist Measurements of drop size and three components of velocity, three components of the gas phase velocity. the concentration of hydrocarbons within the spray, and time resolved droplet measurements arc obtained at axial locations of 7.5, 15. 25. 35. 50, 75. and 100 mm. These measurements arc obtained for both reacting and non-reacting cases. In addition, the atomizing air flow in the absence of the spray is characterized. Primary observations from the present study arc that (I) the presence of the drops alters the structure of the gas phase turbulence. including the degree of isotropy. (2) the presence of reaction strongly impacts the axial and radial velocity components, while having little impact on the azimuthal component. (J) reaction reduces the mean diameter of the distributions at all locations. (4) a strong dependence of the axial and radial velocity upon drop size exists, whereas li11le dependency is observed for the azimuthal component. and (5) detailed citaminalion of the droplet arrival indicates clustering of drops. Finally. it is observed that the findings from the present study both agree and contradict with the results of others. This is indicative of the inherent complexity of reacJing sprays, and suggests that a more methc>dical approach to studying the impact of swirl and geometrical changes is required 10 understand the effects of atomizing air. swirl. vaporization. and phase interaction such as that undertitken in the present effort.


INTRODUCTION
Detailed studies of droplet behavior in spray flames are necessary to (I) develop understanding of the physical processes of evaporation, fuel-air mixing, and transport phenomena. and (2) provide additional data for further development and verification of numerical codes. Such information is required to continue the current efforts to ( 1) mitigate environmental impact and (2) enhance performance and efficiency of liquidfueled continuous combustion systems. Unfortunately, the spray combustion problem is confounded by the presence of many interdependencies. It is difficult to separate the effects of vaporization, combustion, cransport, and phase inceraction on the resulting drop size and velocity distribucions. Also, typical specification of an atomizer (diffraction based line of sight SMD, cone angle, and flow number) can lead to two atomizers which are identical based upon these parameters, buc are significantly different in the detailed structure of the spray. As a result, the present study was undertaken to examine the structure of sprays produced by a standardized atomizer which is capable of operating in a pressure swirl or air-assist mode.
This paper presents results from the third part of a series of experiments directed at better understanding the behavior of sprays, both reacting and non-reacting. The first cwo parts provide results from a non-reacting pressure atomized spray (McDonell and Samuelsen,199 I a), and reacting and non-reacting, non-swirling, air-assit sprays (McDonell,Adachi,and Samuelsen,199la,b).
The objectives of the current paper are lo provide (I) better understanding of sprays produced by swirling air-assist atomizers, (2) a detailed data base suitable for model verification and development, and (3) provide a detailed study against which other •Visiting Scientist. Horiba Ltd. 225 studies may be compared. In support of the third objective, Table I provides details from studies which provide spatially resolved measurements in reacting and nonreacting sprays. The information provided in these studies is used in an effort to summarize current understanding about spray flames and their relationship to the non-reacting case.

EXPERIMENT
The facility used in the present study is the same as described elsewhere (McDonell and Samuelsen, 1991 a;McDonell and Samuelsen, 199 I b). Briefly, the spray is injected downwards in the center of a square duct which is 457 mm on a side. Three degrees of freedom are provided to the atomizer via a traverse system. The optical diagnostics remain in place. Two-component phase Doppler interferometry (PDT) is used to measure the drop size and velocities along with the velocity of the gas phase, and infrared extinction/scattering (IRES) is used to measure the concentration of vapor within the spray. Both instruments, as used in the present study, are described in detail elsewhere (Bachalo and Houser, 1984;McDonell and Samuelsen, 1991a;Adachi, McDonell and Samuelsen, 1991).
The downfired arrangement is used to retain flow field characteristics which were used to help stabilize the reaction in the non-swirling cases previously examined (McDonell et al.,l99la,b).

RESULTS
The results are presented in two sections, gas phase behavior and droplet behavior.
In each case, reacting and non-reacting sprays are considered. Tn addition, the behavior of the single phase flow is considered within the gas phase results to provide a baseline against which the gas phase behavior in the presence of the spray can be compared. To orient the reader, Figure I presents photographs of the spray under non-reacting and reacting conditions.
The photographs reveal important changes in the overall structure of the spray which are caused by the presence of reaction. These changes include (I) a reduction in the axial extent of the spray due to evaporation and consumption of the drops and (2) a " U-shaped" void of drops which is associated with the location of the reaction zone. This picture is consistent with group combustion theory (Chiu and Liu, 1977), which indicates that, for a group combustion number, G ~ I, a sheath of reaction will surround a cloud of droplets. Note that this structure is similar to that observed previously in twin-fluid atomizer spray flames (e. g., Mao et al., 1986;McDonell et al., 199lb;McDonell and Samuelsen, 199lb). Figures 2-8 present a comparison of the gas phase mean and fluctuating velocities for (1) the single phase flow, (2) the gas in the presence of non-reacting spray, and (3) the gas in the presence of the reacting spray.

I Gas phase velocities
At the nearest axial location, Z = 7.5 mm, an on-axis recirculation zone is observed for all three cases, as indicated by the radial profile of the mean axial velocity, U, shown in Figure 2. The presence of spray reduces the strength of this recirculation zone due to the axial momentum in the streamwise direction of the drops. The gas in the presence of the non-reacting spray shows a consistently higher value of U along the centerline which is again attributed to the phase interaction which tends to reduce the strength of the recirculation zone. Near the edge of the spray, at axial locations downstream of the atomizer, the presence of the non-reacting drops nari·ows the spread of the gas jet.
T he impact of reaction is seen starting at Z = 25 mm, where the heat release causes an acceleration of the gas, resulting in an increase in U when compared to the non-reacting case. Note that the expansion is greater away from the centerline, suggesting that greater heat release is occurring in those regions.
Radial profiles of the mean radial velocities, V, for the three cases are presented in Figure 3. The presence of the non-reacting drops reduces and shifts radially the location of the maximum radial velocity at Z = 7.5 mm. This is attributed to the " prompt" atomization action associated with this atomizer, where a strong jet of air blasts a sheet of liquid. The presence of the liquid sheet tends to reduce the spread of the gas jet, which is consistent with the behavior of the axial velocity.
The presence of reaction causes a significant increase in the radial velocity, which is again due to the heat release and subsequent expansion of the gas phase. By Z = JOO mm, the radial expansion has subsided, and the difference between the reacting and non-reacting cases is Jess. Figure 4 presents radial profiles of the mean azimuthal velocity, W, for the three cases. Unlike U and V, W is not strongly affected by the presence of the drops for either non-reacting or reacting conditions. The differences observed are within the experimental error of the measurement. An explanation is that the only source of swirl in the three cases is the atomizing air which has high momentum only near centerline where the presence of small drops has little impact. In the reacting case, symmetry dictates that the flow cannot expand in a preferred azimuthal direction, thus, reaction does not affect the mean azimuthal velocities of the gas phase. This is consistent with observations in an air-blast atomizer spray (McDonell and Samuelsen,199lb). The fluctuating axial, radial, and azimuthal gas phase velocities (u', v', and w' ) are presented in Figures 5-7. The behavior of u' ( Figure 5) suggests that the presence of the spray impacts the levels of fluctuations in the gas phase. This modulation has been observed previously in sprays (e.g ., McDonell and Samuelsen, 1991 b). In the present case, the level of u' are increased along the centerline at Z = 50, 75, and I 00 mm, which is attributed to the mixing of drops of different sizes and velocities. This effect will be clarified in the section on drop behavior. Away from the centerline at Z = 50, 75, and 100 mm, an increase in u' at the edge of the reaction zone is observed. The location of the peak in u' for the reacting case corresponds to the location of the maximum in dU/dr at Z = 50, 75, and I 00 mm. This is not so strongly observed in the single phase and non-reacting cases.
Radial profiles of v' are presented in Figure 6. Similar trends are observed for v' as for u ' . However, v' does not show as much correlation with the mean flow gradients. Note that u' and v' appear similar in magnitude. One exception to this is in the reaction zone, where u' becomes significantly larger that v' . This indicates that the presence of reaction in the present case tends to reduce the levels of isotropy found in the gas phase. One explanation for this is that, in the axial direction, buoyancy acts against the general flow direction, resulting in greater ins~ability of this velocity component.
Profiles of w', shown in Figure 7, shows that the presence of drops tends to damp the turbulence near the atomizer. Farther downstream, however, no appreciable difference between the three cases is observed. The levels of w' are similar to those of v' and u', with the exception of the levels of u' in the reaction zone, as mentioned TABl.f Bachalo. Rudolf. and Sankar ( 1990) I lard;1lupu\. Ta}lor. and \Vhlld:I\\ ( 1990) \fao. Wang 64. 4.8, 11.2, 19.5 mm /) i, v z = 10. 15. 35.   In all the profiles of the fluctuating velocities at Z = 7.5 mm, a secondary peak exists at radial locations, r, of -I 0 to -12 mm. This is attdbuted lo the presence of drops in high numbers with large variation in velocity and high slip velocities, which generates turbulence in these regions. To better quantify the impact of the spray and reaction upon the isotropy of the turbulence of the gas phase, Figure 8 presents ratios of the components. For the single phase and non reacting cases, u' and w' retain similar levels. The presence of reaction causes a consistent increase in u'/w', which is attributed to the downfired reaction and the expansion of gases axially. This same behavior has been observed in an air-blast spray (McD onell and Samuelsen,199lb).

75mm
Comparing u' and v' shows that, near the centerline, similar levels occur for the single phase case and for the gas in the presence of the non-reacting spray. At radial localions of 35 to 55 mm, a sharp drop in this ratio is o bserved for these two cases.
Finally, m lhe ouler region o f the How, u' is significantly higher that v' which is attributed to lhe nature of the turbulence in the co-flowing stream. The presence of reaction tends to raise the value of u' /v' a t a ll locations relative to the non-reacting and single phase cases. In a ll cases. the turbulence is not isotropic. mean hydrocarbon concentration for the non-reacting and reacting cases overlaid upon velocity vectors for the gas phase velocity formed from the values of U and V.

1.2 Methanol gas concentration
In addition, temperature contours are provided for the reacting case as measured using a thermocouple. In the region of the reacting case where appreciable numbers of drops are present, the thermocouple measurements a re likely affected by impaction of drops, and must be considered qualitative. The origin in Figures 9 and 10 corresponds to the centerline at the exit plane of the atomizer, and is offset from the scaled injector schematic for clarity. Figure 9 shows the results for the non-reacting case. fmmediately downstream of the atomizer, along the centerline, the peak concentrations are observed. These concentrations correspond to an approximately saturated condition for the methanol vapor (McDonell a nd Samuelsen, 1991a). This occurs despite the stro ng dilution

235
associated with the atomizing air. The strong mixing which occurs downstream of the atomizer due to the presence of swirl in the atomizing air stream enables the vapor concentration to reach saturation levels measured. With increased axial distance from the atomizer, the dilution associated with entrained air combined with a reduction in saturation levels due to evaporative cooling of the air reduces the concentration of methanol vapor. The role of entrainment at the edge of the jet is shown by the vectors, which show significant flow of surrounding a ir towards the jet. Figure 10 shows the same results for the reacting case. In this case, the heat release occurring in the downfired geometry results in recirculation of products in the o uter regions of the Row. This heated flow mixes with incoming fresh air from above somewhere between 25 and 50 mm below the nozzle exit plane. Some of the recirculated products form the local "peninsula" of hydrocarbons which forms at the edge

1.3 Flux of vapor
The vapor concentration measurements c:an be combined with the gas phase velocity measurements to provide the Aux of vapor at each point. Jn the non-reacting case, this can be done with high accuracy since the conversion from mole fraction to mass fraction can be made without significant assumptions about the mixture molecular weight in the case of methanol. In the reacting case, enhanced vapor production is offset by consumption , so the vapor ftow rate is ambiguous in this case. As a result, these results are not presented here. Figure 11 presents these combined results for the non-reacting case. Figure 11 a shows the evolution of the nidial profiles of the vapor mass flux. The recirculation zone downstream of the atomizer causes the negative flux at the centerline al I o"'~o : Measurement of the flux of vapor has proven to be less susceptible to errors than is the measurement of the flux of liquid via POI (e. g., McDonell and Samuelsen, 1991 b), especially in complex flows such as the present one. The errors shown are based upon use of U from orthogonal radial traverses, and indicates the degree of symmetry in the axial velocity field . Measurement of the vapor concentration for different atomizer rotational positions indicated variations of less than I 0% at any given radial location.

Droplet Behm•ior
Measurements of the droplets provide the size distributions at each location, along with coincident measurement of the velocity of each droplet. Data are acquired in the radial direction until the point where the measured liquid flux is I% of the maximum flux along the profile. The results for the droplets are presented in terms of distribution means and size-velocity rela tionships. Farther downstream. greater differences are observed. especially near the edge of the spray. \\ith both means being lower for the reacting ca-;e. This implies that reaction tends to reduce the siLe of all drops significantly for this case. Unfortunately. in this case. it is difficult to separate the effects of evaporation and transport of the drops. as might be done in the case without swirl (McDonell er al .. 1991b).
However. this same type of consistent decrease in the distribution means has been observed previously in other types of sprays operated under reacting and non-reacting conditions (e.g .• McDo ncll and Samuelsen, 1991 b: McDonell e1 al .. 1991 b). In principle. ll is possible for the mean si/e to increase due to preferential vaporitation of small drops (<'-K . . Chin e1 al .. 1984). but in the studies considered in T able I. thts behavior has only been observed experimentally by Bachalo £'1 al .. 1990. Other researchers have found both increases and decreases in the distribution means for different regions of the spray (e.g .. Mao e1 al Hardalupus. Taylor. and Whitelaw, 1990;McDonell, Wood, and Samuelsen. 1986). Ambiguity arises in a few of the above studies due Lo  : Bachalo et al .. 1990: Hardalupus et al .. 1990) which further couples evaporation and droplet transport in determining the local distribution means. Further complexity is added as a result of strong size-velocity correlations, which makes it difficult 10 establish a result for the entire spray because of the variation in time scales associated wilh each size class.

Drop si::e l'elocity relationship
To better undersland the transport of different size classes. Figure I 3    and Samuelsen (1991 b) provide a similar explanation for this observation . Hardalupus et al. (1990) also provide similar reasoning for the observed differences, but also reason that slow moving drops vaporize within the recirculation zone, thus not contributing low velocities that they would provide in the non-reacting case. Again, the differences in geometry make generalization difficult. An example of this is shown in Figure I 3a which presents radial profiles of the mean axial velocity as a function of drop size, U(D), for the two cases. In the non-reacting case, by Z = 75 mm, the relative velocity between size classes is much Jess than it is at Z = 25 mm. At the centerline, variation in velocity among size classes results in the increase in the gas phase fluctuating velocities as shown in Figure 5. In the reacting case, the relative velocities remain quite high at Z = 75 mm. This occurs because a 68 µm drop at Z = 75 mm in the reacting case must have been much larger at Z = 25mm. Figure 13b presents radial profiles of V(D). Note that the largest drops have the highest radial velocities. Hence, only the largest drops possess enough momentum to overcome the inward force imposed by the atomizing air and entrained air. Interestingly, the 15 µm drops do not reflect the large velocities away from the centerline shown by the gas phase until well downstream of the atomizer. Further examination of the droplet size velocity correlation at r = 4 and 8 mm at Z = 25 mm reveals that, already higher levels of radial momentum of the largest drops. Similar results were observed by Hardalupus et al. (1990), despite the differences in the flow geometry. In their case, however, stronger dependence of W upon D is observed, which is associated with the presence of aerodynamic swirl in their case. They did not, however, provide W(D) for the reacting case. McDonell and Samuelsen (199lb) do not show W(D), but state that no strong dependence of Won D is observed, which is consistent with the present study. The role of atomizing air swirl vs aerodynamic swirl appears differently in the azimuthal velocity dependence upon D based on the comparisons of the present case and that of Hardalupus et al. ( 1990).
Droplet fluctuating velocities are not presented here for brevity, but are available (McDonell and Samuelsen, 1990). Essentially, these results indicate that the fluctuat- ing velocities of the l<!rge drops are less than those of the small drops at most locations: This is con~istent with observations of Hardalupus et al. ( 1990). The ftucLUating velocities of the gas phase is not necessarily less than or greater than that of the droplets. This is difficult to interpret because the fluctuating velocities of the drops are due to sup~rp9sition o f velocity variations due to (I) variation in drop origination, and (2) v~ri~tion imparted by the gas phase turbulence. Jn the former case, the variation in the a~rodynamic effects on drops arriving•from different locations can cause a relatively large variation in the measured velocity at a given point.

Droplet dynamics
The time averaged structure of the spray has been presented thus far. However, t~e instantaneous behavior is also important to understand. cspcciully in terms of loca I mixing effectiveness. Figure 14 presents maps of the droplet size vs time measured ut different points in the spray. The axis labels shown in the lower left plot are the same for each location. The sections of the time series shown are representative of the droplet dynamics at each of the points. Figure 14a presents the map for the non-reacting case. In this case, the abscissa scale is condensed by a factor of 10 from that of"the reacting case ( Figure I 4b). FFTs of these time series showed no strongly dominating frequency, indicating that the drops arrive in essentially random fashion . Figure 14a indicates that, near the centerline, the drops arrive in a consistent manner, with no large time gaps between droplets.
Farther from the centerline (e.g., Z = 50mm, r = 16mm; Z = 75mm , r = 24mm), more variation in the time between drops is observed with more pronounced voids and clusters of drops. Figure I 4b shows the same type of map for the reacting case. Here, the impact of reaction is evident, especially when comparing the Z = 50 mm locations to the Z = 7 5 mm locations. It is difficult to see any structure at the centerline due to the time scale compression, but like the non-reacting case, drop arrival is more consistent than it is at the edge of the spray. At Z = 75 mm, the voids present in the reacting case arc increased compared to those in lhe non-reacting case, with even more time when no drops are present. This type of droplet arrival leads to local variation in stoichiometry which is not desirable for either stability or emissions characteristics. Time resolved measurements are important to have when examining the local structure of the spray. More work is required to correlate the clustering with either atomization or aerodynamics or other factors . Recent numerical studies indicate that the clustering c<in be directly caused by the aerodynamics of the flow (e.g ., Squires and Eaton. 1990).

SUMMARY
A detailed characterization of a swirling air-assisted spray has been conducted. This study reflects one part of a study which removes fuel type, geometry, and stoichiometry from the list of variable, and concentrates on the role of the atomizer type. Where possible, results from this study are compared to the findings of other studies which provide spatially resolved measurements in reacting sprays. The comparison of results indicate that (I) only a small number of data sets are available against which to compare the current results, and (2) consistency between data sets has not yet been reached. The majority of the studies do find that (I) droplets modulate the gas phase mean and fluctuating properties in a complex fashion , (2) the drops have less impact on the swirling component of velocity than on the radial or axial components, (3) reaction causes large increases in axial and radial velocities, and (4) the presence of reaction tends to increase the local drop size distribution means at the locations where measurements were obtained. Also, it is evident that the complexity of spray Rames will require additional studies such as the present one to separate system specific effects.
For the present spray. the following observations have been made: • The present of drops alters the structure of the turbulence of the gas phase, including the degree of isotropy. • The presence of reaction accelerates the gas phase in the axial and radial directions while having little impact on the amount of swirl in the Row. The presence of reaction causes the values of 11'/v' and r//w' to increase which is attributed in part to unsteadiness associated with the downfired orientation. immediately downstream of the atomizer for the non-reacting case. which is due to the rapid evaporation of small drops and the intense mixing in this region. Entrainment dilutes the peak levels of vapor at locations farther downstream. • Reaction rapidly consumes the vapor, but the presence of intermittent large drops combined with the convection of hot products upstream cause local pockets of vapor at rhe edge of the spray. This behavior is consistent with photographs which show drops persisting at radial locations which are beyond the location of the apparent reaction zone. The peak levels of hydrot:arbons in the reacting case reflect a lean reaction. • The presence of reacrion reduces the mean size of the drops at most locations, indicating that the rapid vaporization of small drops which would initially give rise to an increase in the mean size has been completed, and that significant vaporization has occurred for even the largest drops. air upon the liquid sheet, and the lack of swirl present in the fuel stream. In general. it is observed that the velocities of the drops exceeds tha t of the gas phase in the axial and radial direction. and that the velocity of the gas exceeds that of the drops in the azimuthal direction. • Detailed examination of the local time resolved structure shows local clustering of drops for both non-reacting and reacting cases. More work is needed to determine if the clustering is due to a tomization or associated with aerodynamics.