NASA Contractor Report 187037
LOX/Hydrogen Coaxial Injector Atomization Test Program
(NASA-CR-I0703/) I_JFCTOR ATOMIZATION Report (Sw_rdrup LOX/HYDROGEN COAXIAL TEST PROGRAM Final Technology) 11 p CSCL 22B NgI-IQI!7

Unclas 0001715

Zaller Inc. Center
Sverdrup Technology, NASA Lewis Research Cleveland, Ohio

October

Prepared for Lewis Research Under Contract

Center NAS3-25266

National Aeronautics and Space Administration
LOX/HYDROGENCOAXIAL INJECTOR ATOHIZATION TEST PROGRAN N. Zaller Sverdrup Technology, Inc. Lewis Research Center Group Brook Park, Ohio 44142 ABSTRACT Quantitative information about the atomization of injector sprays is required to improve the accuracy of computational models that predict the performance and stability margin of liquid propellant rocket engines. To obtain this information, a facility for the study of spray atomization is being established at the NASA Lewis Research Center to determine the drop size and velocity distributions occurring in vaporizing liquid sprays at supercritical pressures. Hardware configuration and test conditions are selected to make the cold flow simutant testing correspond as closely as possible to conditions in liquid oxygen (LOX)/gaseous hydrogen rocket engines. Drop size correlations from the literature, developed for liquid/gas coaxial injector geometries, are used to make drop size predictions for LOX/hydrogen coaxial injectors. The mean drop size predictions for a single element coaxial injector range from.1 to 2000 pm, emphasizing the need for additional studies of the atomization process in LOX/hydrogen engines. Selection of cold flow simulants, measurement techniques, and hardware for LOX/hydrogen atomization simulations are discussed. [NTRO[DJCTIOM Obtaining information about the atomization of injector sprays has been identified by the JANNAFLiquid Rocket Combustion Instability Panel (Ref. 1) and the JANNAFPerformance of Solid and Liquid Rockets Panel (Ref. 2) as critical to improving the accuracy of computational medets that predict the performance and stability margin of liquid propellant rocket engines. The drop size and velocity distributions produced at the completion of atomization are the initial conditions for vaporization, mixing, and combustion stability analyses in liquid propellant combustors. Therefore, atomization information is crucial to the analyst's ability to make hardware performance and stability predictions. If accurate predictions could be made, the expensive testing Performed in engine development programs could be reduced, combustion instabilities could be avoided, and the efficiency of new engines could be optimized. Unfortunately, the physics of atomization are not welt understood, and empirical correlations must be retied on to estimate drop size distributions in spray combustion systems. Computer codes, such as the Coaxial Injection Combustion Nodal (CICM) (Ref. 3), the High-Frequency [njection Coupled Combustion Instability Program (HICCIP) (Ref. 4), and the Rocket Combustor interactive Design Hethedology (ROCCID) (Ref. 5), calculate a spray size distribution to estimate Performance and stability margin. Some codes use drop size correlations derived from cold flow test results. Other codes contain equations with adjustable parameters that have been calibrated by forcing the overall Performance predictions to agree with actual performance measurements. No rocket combustor hot fire data exist that can verify the drop size and velocity predictions of these codes. Drop size and velocity measurements, as welt as local gas velocity measurements, collected In operating combustors using non-lntrusive techniques, are required to validate the atomization models and improve modeling capabilities (Ref. 67). Drop velocity measurements are also required to determine droplet vaporization rates. Since the combustion process in liquid propellant rocket engines is primarily vaporizationtimited (Ref. 8), drop size and velocity information is critical to predicting performance and combustion stability. Size and velocity measurements of vaporizing droplets at supercritica[ condition s are especially needed to validate supercritical vaporization models. in response to the JANNAF Panel recommendations, and the lack of data needed to validate and improve existing atomization models, a spray atomization testing facility is being established at the NASA Lewis Research Center (LeRC). This facility _il[ be used to obtain simultaneous drop size, velocity, and local gas velocity measurements in sprays that simulate the fluid properties occurring in LOX/hydrogen rocket engines. Based on previous studies, the diagnostic techniques, hardware, and non-reacting simutants that can accomplish this task are selected. An evaluation of current drop size predictive Capability ts conducted, by using existing atomization correlations to make drop stze predictions for the test hardware. PROORAN DESC_]PT]ON A test program is being conducted at the NASA Lewis Research Center to obtain hot fire atomization data in liquid oxygen (LOX)/gaseous hydrogen coaxial injector sprays. Before hot fire testing is attempted, nonreacting, supercritica[ pressure, vaporizing sprays will be studied to determine the feasibility of making measurements in such sprays. High speed photography and particle sizing fnterferometry will be used to obtain information about the spray structure, droplet size distributions, droplet velocity distributions, and local gas velocity distributions. An additional series of tests will be conducted, in both cold flow and hot fire sprays, using a high pressure cross flow of gas. This radial gas flow wilt attempt to simulate the effects of high frequency combustion instability pressure waves on the atomization process. Finally, the relationship between the cold flow and hot fire data will be established.

DIAGNOSTIC

TECHN]QUES laser-based diagnostics, often

Non-intrusive

employed to

obtain

quantitative

information,

require optical access to the spray. The high makes providing and maintaining optical access and are not generally built to slto_ application of making drop size exist. Ingebo (Ref. s|zes and velocities
pressure w high temperature environment difficult. Rocket test fscilit|es are of laser diagnostics to the engines.
of rocket combustors expensive to operate, Due to the difficulty data drop a
measurements in hot firing rocket engines, very little rocket contxJstor drop size 9) photographed droplets in a 0.7 NPa (100 psia) LOX/ethanol engine, and obtained from the photographs. George (Ref. 10) used holography to measure drop sizes in
1.1MPa (150 psig) NTO/MHH engine. Conducting photographic studies was extremely time-consuming, since each droplet had to be measured and counted manually. A relatively small number of droplets was counted in both these exper|ments (less than 2000 at each condition), contributing to uncertainty in the droplet size-number distributions. Hot wax freezing atomization |nformation and laser-based about coaxial Line-of-sight techniques, have injector sprays (Ref. 11-18). commonly been used to obtain Neither of these commonly used
techniques obtain drop velocities. Advances in image processing have made photographic techniques easier to use, but photographic techniques only measure the instantaneous concentration of drops. Instantaneous drop concentration measurements have been shown to be tess useful than droplet flux measurements for validating computer codes (Ref. 6,7). Single particle counting techniques are needed to obtain droplet flux information, since these techniques can measure drop size and velocity simultaneously. Particle sizing interferometry (PSI), a single particle counting technique, has been selected for the LOX/hydrogen atomization testing program. PSI can be used to obtain drop sizes, velocities, and local gas velocities. It has been applied successfully to reacting spray flames (Ref. 19,20). PS! must be applied carefully, since it can only measure spherical drops, and is sensitive to alignment. Detailed information about various laser-based drop sizing techniques, including single particle counters, is provided by Hirteman (Ref. 21). Photography will be used to determine the overall spray structure, and to find regions of the spray where PSI could reasonably be applied. HARDWARE A single-element, shear coaxial injector has been fabricated for use in hot fire and cold flow atomization testing. The injector consists of four parts: a LOX inlet, LOX post, gas manifold, end face plate (Fig. 1). By changing out the LOX post and face plate, several injector geometries can be evaluated _ith the same gas manifold, decreasing fabrication cost and down time between test runs. The LOX post has four fins to center it within the gas manifold. Five injector geometries, with varying liquid injection areas and gas injection areas have been selected. These injector geometries are listed in Table I. Different injector geometries w|LL be tested to examine the effect of varying the injector geometry on the atomization process. The injector element, designed for a nominal psia) chamber pressure, is smaller than SSME main chamber injector as RL-IO injector elements. A small injector size was selected to permit application of optical drop sizing diagnostics. 5550 N (75 lbf) thrust at 5.5 Hpa (800 elements, but approximately the same size reduce the spray number density, and

LDX Inla

Figure Table No.

Coaxial Coaxial 2

Injector Injector No. 3
Design Configurations No. 4 No. 5

LeRC Modular No.

D2, D1, 0o,

om (in) om (in) cm (in)

.594.396.132

(.234) (.156) (.052)

.594.318.132

(.23t,) (.125) (.052)

.516.318.132

(.203) (.125) (.052)

.594.396.198

(.234) (.156) (.078)

.437.318.132

(.172) (.125) (.052)

A chamber has been designed for the cold flow and hot fire atomization tests (Fig. 2). The chamber has a maximum working pressure of 6.9 MPe (1000 psia), and diameter of 5 cm (2 in.). Recircu[etion problems, such as Ferrenberg (Ref. 22) encountered when attempting to measure drop sizes in pressurized chambers, are not anticipated, since cryogenic test liquids wilt be used. Small droplets are expected to evaporate quickly, instead of contfnuatty being recirculated beck to the injector face. A cylindrical chamber was chosen over square chamber designs, in order to simulate actual rocket engine recirculation patterns more closely. The chamber wilt be composed of several segments, which could be rearranged to move the window axially, or alter the chamber length. A similar segmented chamber design was used by Burrows (Ref. 23) in a 2.4 HPa (350 psla) LOX/hydrogen rocket engine. Two different windowed chamber segments will be used. One windowed segment wilt be used for the cryogenic temperature cold flow testing, and another for the extremely high temperature, hot fire testing. A small nitrogen purge has been included upstream of each window. The nitrogen purges wilt provide cooling for the hot fire testing chamber, and wilt help keep the windows clear of spray. Another gas part wilt be located at the stde of the injector face. The face port wilt be used only for the cross flow atomization tests.

Figure

High Pressure

Chandler Design

Quartz, sapphire, fused silica, and ptexigtas windows have been used in pressurized chambers where optical access was required. Quartz end sapphire have optical properties that are e function of direction (birefringence), making the application of off-axis particle sizing interferometry complicated. Fused silica is homogeneous and has high transmissibitity. Ptexigtas was discarded as a possible window material due to its to_ melting temperature. Although the strength and temperature resistance of sapphire are superior to those of fused silica, the birefringent properties of sapphire are difficult to overcome for this application, so fused silica windows were selected for the atomization testing. The feasibility of measuring drop sizes with the proposed windo_ configuration and material was examined using particle sizing interferometer (PSI) (Ref. 24) end a Bergh.md-Liu monodisperse droplet generator. The droplet generator was set up to produce a monodisperse stream of 110 /un water droplets. Two 1" thick fused silica windows were placed in the PSl transmitter end detector paths. For these moderately thick pressure chamber windows, due to refraction of the beams by the windows, the probe volume was formed after the minimum diameter of the focussed beams, in the diverging sections of the beam. The interference fringes in the probe volume were no longer parallel, making the measured drop size vary by as much as 30X across the probe volume length (Fig. 3). The optical setup must be altered to allow independent movement of the focussed beam spots (minimum diameters) and the focusing lens (Ref. 25), so that the beams have a minimum diameter at the point of intersection.

Effect

of Refract|on

on Interference

Fringes
LITERATURE REVIE'W Many atomization correlations have been derived from experiments using cold flowing simutents with properties that are very different from the reactants under consideration. Empirical drop size correction factors relating the properties of the simuLent to the actual propellent properties are occasionally employed, such as the property correlations attributed to Ingebo (Ref. 13) and Wolfe and Anderson (Ref. 6). A literature survey was conducted to identify atomization correlations applicable to coaxial injectors. Any
correlation developed for injectors employing liquid jet breakup by high vetocity, co-ftouing gas streams was considered applicable. These correlations are presented in Table !I. Detailed atomization literature reviews are given by Ferrenberg (Ref. 22) and Lefebvre (Ref. 26). Several researchers based their correlations on experimental data for which injection parameters, such as the liquid properties, were widely varied. Most of the data for these correlations vere collected using either laser diffraction techniques or hot wax freezing techniques. Table Reference Nuktyama, Tanasawa (Ref. 27)

D3= " 585. [

Atomization

Correlations Correlation

Applicable

Coaxial

Injectors

-_-._ p,

Weiss, Worsham (Ref. 11)

-+,,7/

/ oV_-_=) _

"* / (looo 4

t_S _1S
,.=., 1+10oo v'"4 (JLl/ "If

P=;_ J

PiT P L a p g|*/*="

Ip,v:)l

28) DvO. s = 9 x _Vtr_ - B I P L 01/2 9, =--_/= 121a I Pg gvL !

Marshall

_'"
Rizkalla, Lefebrve (Ref. 14)

<v_=p,)"A_'p_ '

-,.,,e,

. 1,.f,/

v;"_op=/

Lorenzetto,

Lefebvre

i o.,,el(

_,=.o22/--:_.=/ /1 + w_". 14.3zo-' + Do=I._. P,.__v_./ (____y*" _, o,. J _ v:)

"J ,4

Jasuja

Ingebo

Hautman
According to the atomization breakup of a Liquid jet by a high

theory velocity

proposed by Mayer (Ref. 28), gas stream are a fLmction
the droplet of the liquid
sizes resulting surface tension,
viscosity, and density, as welt as orifice diameter and atomizing gas velocity. The relative influence of each of these parameters on the drop size distribution is not known. Numerous correlations relating injection parameters to drop sizes produced at the completion of atomization have been proposed. These correlations, often developed using a variety of non-reacting fluids, are applied to reacting sprays. Data are usually taken at _nbient pressure and ten_oerature, instead of the high ten_)erature and high pressure environment of operating co_bustors. The gas and liquid injection velocities are used as input for these correlations: the actual velocity field do,stream of the injection plane is ignored. To assess the agreement among the corre[ations fn predictions for the LeRC modular coaxial injector of the LeRC modular coaxial injector configurations were selected for were substituted which the into each Table ii, these correlations were used to make drop (Fig. 1). The hot fire injection parameters for were calculated (Table Ill). These three injector varied a moan over drop a wide range. These size was predicted hot

size three

configurations fire parameters (Table IV).
relative gas/liquid velocity atomization correlation, and
So_e_corretations diameter (D32), exactly, since distributions.

predict to

median in

diameter

(Dvo_5) , while

others

predict

Sauter

contributing the atomization Simmons (Ref.
variation correlations 29) compared
the drop size prC=_ictions, provide no information about the Sauter mean diameter and
this variation cannot be calculated the shape of the drop size the mass median diameter of 200 drop
size distributions obtained from tests of various fuet nozztes. Simmons found that themess median diameter of the distributions was 1.2 times the Sauter mean diameter to within 5%. The drop sizes predicted by the three mess median diameter correlations (Weiss and Worsham, Hayer, Kim and Marshall) were reduced by a factor of 1.2. Only the adjusted Sauter mean diameter predictions for the correlations in Table II are reported in Table IV.

lOX/Rydrogen

injection
ParAmeters Configuration No. 3.132.0710.0898.0159 (.052) 4.0110) (.198) (.0350)
LeRC Nodular Injection Do, cm (in) Ag, cm2 (in 2) (lb/s) (tb/s) Parameters No. 1.132.198.0838.0210 (.052) 4.0307) 4.185) (.0462)
Injector No. 2.132.130.0873 (.052) (.0201) (.192) (.0395)

WL, kg/s Wg, kg/s

PL' kg/m3 (lb/ft3) pg, kg/m3 (lb/ft APL' MPa (psi) VL, Ws Vg, Ws Vr, a, #L' #g, (it/s) (it/s) 3)
1080 (67.2) 4.47 1.75 57.0 (.279) (254) (187)
1080 467.2) 4.47 1.90 59.4 (.279) (275) (195)
1080 (67.2) 4.47 2.01 61.1 (.279) (291) (200)
237 4777) 180 (591) 9.72 E-3 E-4 46.66 (9.84 E-4) E-5) E-6) 9.72 i.46 8.99
308 (1010) 249 (817) E-3 (6.66 E-4) E-5) E-6) 9.72 1.46 8.99
502 (1650) 441 (1450) E-3 E-4 E-6 (6.66 (9.84 (6.04 E-4) E-5) E-6)
m/s (it/s) N/m ([b/ft) kg/m-s kg/m.s ([b/ft.s) ([b/it.s)

1.46 8.99

E-4 (9.84 E-6 (6.04

E-6 (6.04

Drop Size Predictions
LOK/HydrogenTestlng Nean Diameter, No. 1.6.7.6 #m No. 0.8.5.2
Sauter No. 1 Nukiyame, Weiss, Hayer Kim, Tanasawa (Ref. 11) 27) 1300 2.5.26 12) (Ref. (Ref. 14) 15) 47 8.2
Worsham (Ref. (Ref. Harsha[[ 28) (Ref.
Rizkal[a, Lorenzetto, Jasuja lngebo
Lefebvre Lefebvre 16) 17) 18)

(Ref. (Ref.

Hautmen (Ref.
drop size predictions for the LeRC coaxial hardware vary from 0.1 to 2000 _. No correlation predicts mean drop sizes within IOX of any other correlation for all three hardware configurations that were examined. This wide range of predictions emphasizes the current lack of understanding of the atomization process, and the need for data that can be used for atomization model validation. Some of these correlations were developed using a variety of non-reacting fluids, flow rates, and geometries, with the goal of making the correlation applicable to a wide range of hot fire conditions. However, the injection conditions in LOX/hydrogen engines are quite different from the injection conditions previously studied, especially the high relative gas/liquid velocity, high chamber density, and low liquid surface tension. Therefore, additional studies for validation of LOX/hydrogen atomization models should better simulate the conditions encountered in LOX/hydrogen engines. SELECTIOM OF MOM-REACTING SINtJLAMTS Since the conditions in LOX/hydrogen rocket engines are very different from any encompassed in previousty conducted atomization studies, atomization testing is required that simulates LOX/hydrogen engines more closely. A non-reacting liquid simutant is needed that is safer to use than LOX, can be vaporized, and has a relatively tow critical pressure. Liquid nitrogen satisfies these requirements. The properties of LOX and liquid nitrogen, along with the properties of other liquids that have been previously used as LOX simulants, are listed in Table V. To assess the ability of these liquids to simulate LOX atomization, the liquid properties were substituted into the atomization correlations in Table If, and a mean drop size was predicted for the second LeRC coaxial injector configuration. The drop size predictions for the various liquids are presented in Table V%. By coltq)ar|ng the drop size predictions for all the liquids to the LOX drop size predictions, it can be seen that liqu|d nitrogen simulates LOX more closely than any of the other liquids.

Liquid

Properties Pressure MPa(psia) 5.51 4.14.101.101 (800) (600) (14.7) (14.7)
of CommnlyUsedLO0( Surface Tension N/m(tb/ft).0097.0074.019.026.017 (6.7 E-4)
Siautants D_nsity kg/m'(lb/ft 1080 (67,2) 788 (49.2) 1565 (97.7) 806 (50.3) 764 (47.7) 997 (62.2) 7 _) Viscosity kg/m.s(lb/ft.s) 1.5 E-4 1.4 E-4 6.8 1.5 1.8 8.9 E-4 E-3 E-3 E-4 (9.8 (9.7 (4.6 (1.0 (2.7 (6.0 E-5) E-5) E-4) E-3) E-3) E-4)
Temperature K(=R) Liquid Liquid Oxygen Nitrogen 106 (190) 83.3 (150)
(5.1E-4) (1.3 (1.8 (1.2 (4.9 E-3) E-3) E-3) E-3)
Freon 113 Jet A (Ref. Shellwax Water 18) 13)
298 (537) 298 (537). 298 (537)

270 (Ref.

(14.7)
PredictedlCeanDropSizes Liquid Oxygen Liquid Nitrogen 1900 2.2.25 ]28 6.2

Different

Liquids Shettwax 9.6 2.9.0 Water

Freon 113

Nukiyama, Weiss, Hayer

Tanasawa (Ref. 11)

1700 1.9.22

2700 3.3.11

4100 7.9 1.11

2500 8.6 1.18

Worsham (Ref. (Ref. 28) (Ref.
K|m, Marshall Rizkatta, Lorenzetto, Jasuja lngebe (Ref. (Ref.

12) (Ref. (Ref. 14) 15)

32 7.6
Lefebvre Lefebvre 16) i7) 18)

Hautman (Ref.

The Liquid properties were also substituted into two property correlations (6,1]) that have been used to "correct" the predicted drop size in hot wax experiments. The correction factors relate the properties of different liquids to the properties of LOX. Both of these correlations predict that liquid nitrogen properties are so similar to LOX properties, that almost no drop size correction would be required. The predicted correction factors are included in Table VII.

Drop Size Liquid Oxygen

Correction Liquid Nitrogen.98.99

Factors

Different Jet A
Liquids Shettwax 270.21.35 _ater

Wolfe, Ingebo

Anderson (Ref. 13)

1.0 1.0

.52.63

.24.41

.19.38
Two criteria are used for the gaseous hydrogen simutant selection. A gaseous simulant is needed that is relatively non-hazardous, and could be used to match the high injection velocities of hydrogen. Gaseous helium was selected, since it is inert and has a high sonic velocity. The sonic velocity of higher molecular weight gases, such as nitrogen, is Lower than the hydrogen injection velocity predicted for the LeRC n_xlular coaxial injector. The use of these heavier gases would prevent gas velocity matching between the cold flow and hot fire cases. COMCLUDINGRENARKS There is wide disagreement among drop size correlations currently avaiLabLe for coaxial types of injectors. Additional studies of the atomization of supercritical pressure, vaporizing sprays are required to increase our understanding of the liquid breakup process, and to obtain data useful for validation of computer models that predict performance and stability margin of LOX/hydrogen engines. To accomplish this task, a facility is being established at the NASA Lewis Research Center to examine the atomization of high pressure cryogenic sprays. Based on the results of numerous atomization studies, liquid nitrogen and gaseous helium are shown to closely simulate the properties of liquid oxygen and gaseous hydrogen. Xt is anticipated that cold flow and reacting spray studies, using liquid nitrogen/gaseous helium and liquid oxygen/gaseous hydrogen, respectively, will result in similar spray distributions. To test this hypothesis, particle sizing interferoe_try and high speed photography will be applied to non-reacting sprays to obtain information about spray structure, droplet size and velocity distributions, and Local gas velocity. Future plans include LOX/hydrogen testing emptoy]ng the same diagnostic techniques and hardware as the cold flow testing. The data obtained from this program wit[ be useful for validating existing atomization models, assessing the accuracy of previously developed drop-s|ze correlations, and establishing benchmark data for computational codes that attempt to model liquid breakup from first principles. ACJOi(31_.EDGi_IENTS This direction work was supported by NASA Lewis Research Center provided by Nark Klem. under contract NAS3'25266 with technical

MONENCLATURE A AR B DO D1 Injection Area, cm2 (in 2) Area Ratio (0.3), om (in) Ref. 28 (0.74), Ref. 18
Tangential-SLot-to-Inner-Tube Jet Stripping Orifice Parameter Diameter, cm (in) cm (in) #m /u,
LOX Post Diameter, Gas Annulus

Diameter,

D32 OVO.5 g LOX NTO/HHH PSl O SSHE V

Rean Diameter,

Volume Nedian Diameter, Acceleration Liquid Nitrogen Particle Volumetric Oxygen due to

Gravity,

m/s 2 (ft/s
Tetroxide/Ronomethyt Sizing Interferometry m]/s

Hydrazine

Flow Rate,

(ft]/s)

Space Shuttle Velocity,

Rain Engine

m/s (it/s)

w AP 3, # p o'

Mass Fto_ Rate, Pressure Molecular Viscosity, Density, Surface Drop,

Lib/s)

MPa (psi) m (it)
Mean Free Path, kg/m.s (tb/ft.s) 5)

kg/m 3 (tb/ft Tension,

N/m (tb/ft)

Subscripts

g L m r T
Gas Liquid MoLecuLar Relative Total REFERENCES
1. Jensen, R. J.: A Summary of the JANNAF Workshop on Liquid Rocket Engine Combustion Driven Instability Mechanisms. 26th JANNAF Combustion Meeting, Vot. 2, D. L. Becker, ed., CPIA-PUBL-529-VOL-2, Chemical Propulsion Information Agency, Laurel S MD, 1989, pp. 273-288. 2. Gross, K. W.: D. L. Backer, ed., pp. 351-353. Liquid Engine Jet Atomization CPIA-PUBL-476-VOL-2, Chemical Workshop Report. 24th Propulsion Information JANNAF Combust|on Meeting, Agency, Laurel, MD, 1987, Vol. 2,
3. Sutton, R. D.; Schuman, M. D.; Model. NASA CR-129031, 1974.

and Chadwlck,

W. D.:

Operating

Manual

Injection

Combustion
4. Prim, R. J.; and Breisecher, K. J.: Dimensions. NASA TM-I02381, 1989. 5. Muss, J. Methodology. A.; and Johnson, NASA CR-185187, C. W.: 1990.

Nonlinear

CombustiOn Instability

Model in Two- to Three-

User's

Rocket Combustor

Interactive

Design

(ROCCID)

6. Ferrenberg, A. J.; 85-1316, July 1985.

and Varma, M. S.:

Data for

Spray Combustion

Modeling.

AIAA Paper

7. Dodge, L. G.; and Schwatb, J. A.: Fuel Spray Evolution: Nonevaporattng Spray. ASME Paper 88-GT-27, June 1988. 8. Prim, R. J.; and Heidmann, M. F.: Propellant Combustion Chambers. NASA TR R-67, 1960. 9. Ingebo, R. D.: Photomicrographic Liquid Oxygen. NASA TM-D-290, 1960. 10. George, O. J.: Rocket Injector Tracking Vapor|zation

Comparison

of Experiment

and CFD Simulation

as a Design

Criterion

Rocket-Engine

of Ethanol

Drops tn a Rocket

Chamber Burning

Ethanol

Hot Firing

and Cold FLow Spray Fields. tn High Velocity
AIAA Paper 73-1192, ARS J., vot.

Nov. 29, no.

1973. 4,
11. Weiss, Apr. 1959, 12. Kim, rot. 17,
M. A.; and Worsham, pp. 252-259. end Marshall, May 1971, pp. and Burick,

C. H.:

Airstreams.

K. Y.; no. 3,

W. R., Jr.: 575-584. R. J.:

Drop-Size

Distributions

from Pneumatic

Atomizers.

AIChE J-e

13. Fa[k, A. Y.; CR-120968, 1973. 14. Rizkatta, Characteristics.

Injector

Design Guidelines

Gas/Liquid

Propellant

Systems.

A. A.; and Lefebvre, A. H.: J. Eng. Power, rot. 97,
Influence of Liquid Properties no. 2, Apr. 1975, pp. 173-179. of Drop Size

on Airbtast

Atomizer
15. Lorenzetto, G. E.; and Lefebvre, AIAA J., vol. 15, no. 7, July 1977, 16. Air
A. H.: Measurements pp. 1006-1010.

on a PLain-Jet

Airblast

Atomizer.

Jasuja, A. K.: PLafn-Jet Airblast Atomization of ALternative Pressure Conditions. ASNE Paper 82-0T-32, Apr. 1982.

Petroleum

Under High Ambient
17. Ingebo, R. D.: Gas Density (Also, NASA TN-102013).

on Dropstze

Simulated

Sprays.

89-2322,
18. Hautman, D. J.: Spray Characterization of Liquid/Gas 25th JANNAF Combustion Meeting, Vo[. 3, D. L. Secker, ed., Information Agency, Laune[, ND, 1988, pp. 327-339. 19. Mao, C.-P.; Wang, Twenty-first Symposium 20. Presser, C.; Hethano[/Dodecane 21. Hir[eman, Techniques, pp. 35-60. 22. Ferrenberg, NASA Contract 23. Burrows, TM-X-52023, G.; and Chig|en, (International) G.; N.: An Experimental on Combustion, The
Coaxial Injectors CPIA-PUBL-498-VOL-3,
the Center Liquid Chemical Propulsion

Swirled.

Study of Combustion C. T.:

Air-Assist Institute, Combustion
Atomizer Spray Flames. 1986, pp. 665-673. of Methanol and
Sernerjian, H. Spray Flames.
Gupta, A. K.; and AIAA Paper 90-2446,

Avedisian, July 1990.

E. D.: Particle Sizing by Optical, Nonimaging Techniques. R. Tishkoff, eta[, eds., ASTH STP 848, American Society
Liquid Particle Testing and
Size Measurement Haterials, 1984,
A.; Hunt, NAS8-34504), M. C.: 1963.
K.; and Duesberg, NASA CR-178751, Processes

J.: 1985. Related

Nixing

Study.

(RI/RD85-312,

Rocketdyne;

Radiation

Oxygen-Hydrogen

Pressures.
24. Bacha[o, W. D.; and Houser, Size and Velocity Distributions. 25. Dur[o, D. and F. G.; and in
J.: Phase/Doppler Spray J. Optical Eng., rot. 23, J. F. G. M. C.: et Laser at,

Analyzer no. 5,

for 1984, for
Simultaneous pp. 583-590. Combustion. Publishers, 6, pp.

Heasurements

Hendes-Lopes, D.

Ve[ocimetry eds., Ktuwer

Instrumentation 1989, 233-261, pp. 1980.

for 151-177.

Combustion 26. Lefebvre,

Flow A. H.:

Engines, Airbtast

Dur_o,

Academic Sci., vo[.

Atemization.

Energy

Combust.

27. Nukiyama, S.; of Liquid Properties pp. 68-75. 28. Mayer, Dec. 1961, E.: pp.
Tanasawa, Y.: on the Size of
An Experiment Drops. Trans.
on the Atomization of Jpn. Soc. Nech. Eng.,

Liquid. vol. 5,

Fourth no. 18,

Report, The Feb. 1939,

Theory of 1783-1785. C.:

Velocity

Streams.

ARS J.,

29. Simmons, H. Size/Volume-Fraction
The Correlation Distribution.
of Drop-Size J. Eng. Power,

Distributions vo[. 99,

Fuel July

Nozzle 1977,

Sprays. Part pp. 309-314.
National Aeronautics and Space Administration 1. Report No,

Report

Documentation

Accession No.

3. Recipient's Catalog No.

2. Government

CR- 187037

5. Report Date

4. Title and Subtitle

LOX/Hydrogen

Program

6. Performing

Organization Code

7. Author(s)

8. Performing

Organization

Report No.

M. Zaller

10. Work Unit No.

(E-5849)

506-42
9. Performing Organization Name and Address 11. Contract or Grant No.
Sverdrup Technology, Inc. Lewis Research Center Group 2001 Aerospace Parkway Brook Park, Ohio 44142
12, Sponsoring Agency Name and Address

NAS3-25266

13. Type of Report and Period Covered

Contractor Final

14. Sponsoring

Agency Code

National Aeronautics and Space Administration Lewis Research Center Cleveland, Ohio 44135-3191

15. Supplementary Notes

Project Manager, Mark Klem, Space Propulsion Technology Division, NASA Lewis Research Center. Prepared for the 27th JANNAF Combustion Meeting, Cheyenne, Wyoming, November 5-9, 1990.

16. Abstract

Quantitative information about the atomization of injector sprays is required to improve the accuracy of computational models that predict the performance and stability margin of liquid propellant rocket engines. To obtain this information, a facility for the study of spray atomization is being established at the NASA Lewis Research Center to determine the drop size and velocity distributions occurring in vaporizing liquid sprays at supercritical pressures. Hardware configuration and test conditions are selected to make the cold flow simulant testing correspond as closely as possible to conditions in liquid oxygen (LOX)/gaseous hydrogen rocket engines. Drop size correlations from the literature, developed for liquid/gas coaxial injector geometries, are used to make drop size predictions for LOX/hydrogen coaxial injectors. The mean drop size predictions for a single element coaxial injector range from. 1 to 2000/zm, emphasizing the need for additional studies of the atomization process in LOX/hydrogen engines. Selection of cold flow simulants, measurement techniques, and hardware for LOX/hydrogen atomization simulations are discussed.

17. Key Words

(Suggested

by Author(s))

18. Distribution

Statement

Atomization Coaxial injectors Sprays Drop sizing
19. Security Classif. (of this report) 20. Security Classif.

Unclassified Subject

- Unlimited 15

Category

(of this page)

No. of pages

Price*
Unclassified NASA FORU 162S oc'r_

Unclassified

*For sale by the National Technical Information Service, Springfield, Virginia 22161

Atomizer: A Dynamic Atomicity Checker For Multithreaded Programs (Summary)
Cormac Flanagan Department of Computer Science University of California at Santa Cruz Santa Cruz, CA 95064 Stephen N. Freund Department of Computer Science Williams College Williamstown, MA 01267

Abstract

Ensuring the correctness of multithreaded programs is difcult, due to the potential for unexpected interactions between concurrent threads. We focus on the fundamental non-interference property of atomicity and present a dynamic analysis for detecting atomicity violations. This analysis combines ideas from both Liptons theory of reduction and earlier dynamic race detectors such as Eraser. Experimental results demonstrate that this dynamic atomicity analysis is effective for detecting errors due to unintended interactions between threads. In addition, the majority of methods in our benchmarks are atomic, supporting our hypothesis that atomicity is a standard methodology in multithreaded programming.
this limitation, consider the following excerpt from the class java.lang.StringBuffer. All elds of a StringBuffer object are protected by the implicit lock associated with the object, and all StringBuffer methods should be safe for concurrent use by multiple threads. Excerpt from java.lang.StringBuffer
public final class StringBuffer { public synchronized StringBuffer append(StringBuffer sb) { int len = sb.length();. // other threads may change sb.length(). // len does not reflect length of sb sb.getChars(0, len, value, count);. } public synchronized int length() {. } public synchronized void getChars(.) {.}. }

1 The Need for Atomicity

Multiple threads of control are widely used in software development because they help reduce latency and provide better utilization of multiprocessor machines. However, reasoning about the correctness of multithreaded code is complicated by the nondeterministic interleaving of threads and the potential for unexpected interference between concurrent threads. Since exploring all possible interleavings of the executions of the various threads is clearly impractical, methods for specifying and controlling the interference between concurrent threads are crucial for the development of reliable multithreaded software. Much previous work on controlling thread interference has focused on race conditions, which occur when two threads simultaneously access the same data variable, and at least one of the accesses is a write [1]. Unfortunately, the absence of race conditions is not sufcient to ensure the absence of errors due to unexpected interference between threads. As a concrete illustration of
The append method shown above rst calls sb.length(), which acquires the lock sb, retrieves the length of sb, and releases the lock. The length of sb is stored in the variable len. At this point, a second thread could remove characters from sb. In this situation, len is now stale and no longer reects the current length of sb, and so the getChars method is called with an invalid len argument, and may throw an exception. Thus, StringBuffer objects cannot be safely used by multiple threads, even though the implementation is race-free. To catch errors like this, we focus on a widely-applicable non-interference property called atomicity. A procedure (or code block) is atomic if for every (arbitrarily interleaved) program execution, there is an equivalent execution with the same overall behavior where the atomic procedure is executed serially, that is, the procedures execution is not inter-

Proceedings of the 18th International Parallel and Distributed Processing Symposium (IPDPS04)
0-7695-2132-0/04/$17.00 (C) 2004 IEEE
leaved with actions of other threads. This non-interference guarantee reduces the challenging problem of reasoning about an atomic procedures behavior in a multithreaded context to the simpler problem of reasoning about the procedures sequential behavior. The latter problem is signicantly more amenable to standard techniques such as manual code inspection, dynamic testing, and static analysis. In addition, atomicity is a natural methodology for multithreaded programming, and experimental results indicate that many existing procedures and library interfaces already follow this methodology [3]. Finally, many synchronization errors can be detected as atomicity violations. Although atomicity is a widely-applicable and fundamental correctness property of multithreaded code, standard testing techniques are inadequate to verify atomicity. Testing may discover a particular interleaving on which an atomicity violation results in erroneous behavior, but the exponentially-large number of possible interleavings makes obtaining adequate test coverage essentially impossible.
2.2 The Theory of Reduction
Our analysis leverages Liptons theory of reduction to dynamically detect atomicity violations. The theory of reduction is based on the notion of right-mover and left-mover actions [4]. An action b is a right-mover if, for any execution where the action b performed by one thread is immediately followed by an action c of a concurrent thread, the actions b and c can be swapped without changing the resulting state. Conversely, an action c is a left-mover if whenever c immediately follows an action b of a different thread, the actions b and c can be swapped, again without changing the resulting state. We classify operations performed by a thread as (left or right) movers as follows: Operation lock acquire lock release access to protected data access to unprotected data Mover Status right-mover left-mover both-mover non-mover
2 Checking Atomicity Dynamically
We present a dynamic analysis for detecting atomicity violations. For each code block annotated as being atomic, our analysis veries that every execution of that code block does not interfere with other threads. Intuitively, this approach increases the coverage of traditional dynamic testing. Instead of waiting for a particular interleaving on which an atomicity violation causes erroneous behavior, such as a program crash, the checker actively looks for evidence of atomicity violations that may cause errors under other interleavings. Our approach synthesizes ideas from dynamic race detectors (such as Erasers Lockset algorithm) and Liptons theory of reduction [4], as described below. We refer the reader to the extended version of this paper for the complete details of our approach, as well as a discussion of related race condition and atomicity checking tools [3].

2.1 The Lockset Algorithm
In order to identify atomicity violations, our tool rst identies race conditions using a variant of Erasers Lockset algorithm [5]. This algorithm tracks a lockset for each shared variable. This lockset contains all locks that have been consistently held on all accesses to that variable. Each lock set initially contains all locks, and on each shared variable access, the Lockset algorithm removes from the corresponding lockset all locks not held by the current thread. If the lockset for a variable becomes empty, then no lock consistently protects all accesses to that variable, and our analysis assumes that there may be a race condition on that variable. This information regarding race conditions is then used by the following reduction algorithm.
Once a thread acquires a lock, no other thread may acquire or release it. Hence the acquire operation can be moved to the right of a step by a concurrent thread without changing the resulting state. Similarly, a release operation commutes to the left. An access (read or write) to a shared variable that is protected by a lock is a both-mover since no other thread can simultaneously access that variable. In contrast, an access to a variable on which there may be race conditions is a non-mover since other threads may concurrently access the same variable. To illustrate how the classication of actions as various kinds of movers enables us to verify atomicity, consider the rst execution trace in the diagram below. In this trace, a thread (1) acquires a lock m, (2) reads a variable x protected by that lock, (3) updates x, and then (4) releases m. The execution path of this thread is interleaved with arbitrary actions b1 , b2 , b3 of other threads. Because the acquire operation is a right-mover and the write and release operations are left-movers, there exists an equivalent serial execution (with the same nal state 7 ) in which the operations of this path are not interleaved with operations of other threads, as illustrated by the following diagram. Thus the execution path is atomic. Reduced execution sequence

acq(m)

x = t+1

rel(m)

More generally, suppose an execution path through a method contains a sequence of right-movers, followed by at most one non-mover action and then a sequence of leftmovers. Then this path can be reduced to an equivalent serial execution, with the same resulting state, where the path is executed without any interleaved actions by other threads. Our dynamic analysis veries that every executed trace through each atomic method is reducible, and it reports warnings when irreducible paths are observed.

4 Conclusions

Developing reliable multithreaded software is notoriously difcult, because concurrent threads often interact in unexpected and erroneous ways. Clearly, the cost-effective development of reliable multithreaded systems requires the development and application of methods for controlling the interference between concurrent threads. The notion of atomicity provides a strong (indeed maximal) and widelyapplicable non-interference guarantee. This paper presents a dynamic analysis designed to catch atomicity violations would be missed by traditional testing or (static or dynamic) race-detection techniques. We suggest that the wider adoption and emphasis on atomicity in multithreaded software could provide many benets, which may include: simpler procedure specications; better static analyses; decreased testing cost; easier code inspection; and detecting more scheduler-dependent bugs. In additon, whereas this work has focused on multithreaded systems, an interesting avenue for future work is to study what notions of non-interference similar to atomicity are appropriate for distributed systems. Acknowledgments. This work was partly supported by the National Science Foundation under Grants CCR-0341179 and CCR-0341387, and by faculty research funds granted by the University of California at Santa Cruz and by Williams College.
3 Implementation and Evaluation
We have developed an implementation, called the Atomizer, of the dynamic analysis outlined above. The Atomizer takes as input a multithreaded Java program and rewrites the program to include additional instrumentation code. This instrumentation code calls appropriate methods of the Atomizer run-time library that implement the Lockset and reduction algorithms and issue warning messages when atomicity violations are detected. For the StringBuffer class, the Atomizer detects that append contains a window of vulnerability between where the lock sb is released inside length and then re-acquired inside getChars, and produces the following warning, even on executions where this window of vulnerability is not exploited to produce an observable error. Error report
StringBuffer.append is not atomic: Atomic block entered at StringBuffer.append(StringBuffer.java:445) at BreakStringBuffer.main(Test.java:21) Atomic block commits at lock release: at StringBuffer.length(StringBuffer.java:144) at StringBuffer.append(StringBuffer.java:451) at Test.main(Test.java:21) Atomicity violation at lock acquire: at StringBuffer.getChars(StringBuffer.java:326) at StringBuffer.append(StringBuffer.java:455) at Test.main(Test.java:21)

References

[1] A. D. Birrell. An introduction to programming with threads. Research Report 35, Digital Equipment Corporation Systems Research Center, 1989. [2] C. Flanagan and S. N. Freund. Atomizer: A dynamic atomicity checker for multithreaded programs. In Proceedings of the ACM Symposium on the Principles of Programming Languages, pages 256267, 2004. [3] R. J. Lipton. Reduction: A method of proving properties of parallel programs. Communications of the ACM, 18(12):717 721, 1975. [4] S. Savage, M. Burrows, G. Nelson, P. Sobalvarro, and T. E. Anderson. Eraser: A dynamic data race detector for multithreaded programs. ACM Transactions on Computer Systems, 15(4):391411, 1997.
The application of the Atomizer to over 100,000 lines of Java code demonstrates that it detects defects in multithreaded programs that would be missed by existing racedetection tools, and it produces fewer false alarms on benign races that do not cause atomicity violations. In addition, the Atomizer found no atomicity violations in over 90% of the methods annotated as atomic that were exercised during our test runs. While certainly sensitive to the coverage of our testing, this statistic suggest that atomicity is a fundamental design principle in many multithreaded systems, especially library classes and reusable application components.