This action might not be possible to undo. Are you sure you want to continue?
SPLITTING AND DISPERSION OF BUBBLES BY TURBULENCE
A dissertation submitted in partial satisfaction of the requirements for the degree Doctor of Philosophy in Engineering Sciences (Aerospace Engineering)
Carlos Mart´ ınez
Committee in charge: Professor Professor Professor Professor Professor Professor Professor Juan C. Lasheras, Chairperson Laurence Armi Michael Buckingham Martin Maxey W. Kendall Melville Sutanu Sarkar Forman Williams
Copyright Carlos Mart´ ınez, 1998 All rights reserved.
The dissertation of Carlos Mart´ ınez is approved, and it is acceptable in quality and form for publication on microﬁlm:
University of California, San Diego
To my family: my parents. sisters and Paqui. iv .
. The turbulent break up problem . . . . . . . Experimental Set-up . . . . . . . . . . . . . . Table of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Experimental Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 42 48 48 57 59 v . . . . . . . . . . . A. . . . . . . . . . . . 1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Background and Previous work . . . . Experimental approach . . . . . . . . . . .TABLE OF CONTENTS Signature Page . . . . . . . . . . . . . . . . . . . . . . . . . Experimental results . . . . . B. . . . . . . . . . . . . . . . . . . . . . . . . . . . xviii I Introduction . . . . Measurement of size of bubbles . . . . . . . . . . . . . . . PDPA . . . . 32 A. . . . . . . . . . 1 1 3 10 10 13 17 23 24 27 28 II III Experimental Facility and Flow Conditions . . . . . . . . . xiv Acknowledgments . . . . . . . . . . . iii iv v List of Tables . . 35 IV Break-up Frequency of Bubbles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Image processing . . . . . . . . . xvii Abstract . . . . . . Flow conditions . . . . . . . 2. . . . . . . . . . . . . 1. . . . . . xv Vita. . . . . Introduction . . . . . . . . . . . . . . and Fields of Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A. . . . . . . Dedication . . . . . . . . Particle number density limitations . . . . . . . . . . . . . . . . . . . . . . . Sizing of small particles . . . . . . . . . . . . . . . . C. . . . . . . . . . . . Evolution of the volume-size bubble pdfs . . . . . . . . . . . . . . . . . C. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii List of Figures . . A. . . . . . . A phenomenological model for the bubble break-up frequency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viii list of symbols . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B. . . . . . . . . . . . . . 32 B. . . . . . . . Publications. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2. . . . . . . . . . . . . B. Eﬀect of Gaussian intensity proﬁle of the incident light source 3. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Rate of decay of number of bubbles of a certain class size . . . . . . . . . . . . 4.
. . . . . . . The Poisson random process . . . . . . . . . . . Introduction . . . . . . . . α . . 1 vi . . . . .V P DF of the Daughter Bubbles . . . . . . Comparison with other models D. . . . . . . . . . . . . . . . . . . . . . . . 69 69 71 76 82 100 100 101 105 106 110 113 114 124 127 VI Characterization of the Frozen State . . . . . . . . . . . . . . . . Inter-arrival distance between bubbles . . . . . Deﬁnitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Dependence of the frozen bubble size pdf on ǫ . . . . . . . . . . . . . . . . . . 2. . . . . Model formulation . . . . . . . . . . . . . . . . Statistics of the bubble inter-arrival time . . . . . . . . . . . . . . . . . . . . . . Dependence of Dmax on ǫ . . . . . . . . . . . . . . . . . . . . . 134 B. . . . . . . . 134 A. . . . . 1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Experimental results . . . . . . . . . . . . . . . . . . . . A. . . . . . . . . F. . . . . . . . . . . Eﬀect of the Turbulent Kinetic Energy on the inter-arrival time. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Eﬀect of the Turbulent Kinetic Energy of the ﬂow on the inter-arrival distance. . . . . . . . . . Continuous distribution functions . . . . . . . . . . . E. . . . . . . . . . . . . . . . Introduction . . . . . . . . . . . . . . . . . . . . A. . . . . . . . . . . . . . C. . . . . . . . . . . . . . . . . . . . . . Dependence of the frozen bubble size pdf on the initial void fraction. . . . B. . . . . . . . . . C. . . . . . . . . . . . . . . . . . . . . . . . . VII Conclusion . . . 1. . . . . . . . . 131 A Particle Size Distributions . . . . . . . . . . . . . . . . B. . D. . 137 Bibliography . . . . . . . . . . . . . .
LIST OF TABLES II. . 44 vii . 11 Experimental Conditions.1 Measurement Techniques . . . . .1 IV. . . . DJ . and the open section at the exit of the nozzle. . . . Re has been calculated based on the exit ∗ velocity. . . . . U0 . . . . This open section was kept constant in the three cases of Set 3. . . . . . . . . . . . . . . .
. Flow goes from left to right in each picture. . . . . . . . . . . . . . . . . . . . Standard PDPA Conﬁguration . . .. . . . . . .4 III. . . . . . .1 II. . . . . . . . . . . .13 30 33 34 36 37 III. . . . . . .6 II. . . b) Section 2. . . . Lf indicates the wave number associated with the length of the ﬁlm. . . . . . . Windows 1-5. . . . . . .11 II. . . . Figure taken from Crowe et al. . . . .10 II. . Planar Phase Doppler Anemometry . . . . (•) represents size obtained from the measured area. . . . . . . . . . . . . . . . . . . ψ = 60o . . PDPA Conﬁguration: Plan View . . . . . . . X0 ≈ 4 DJ indicates the virtual origin. . . . . . . Angular intensity distribution of diﬀerent scattering modes of bubbles in water. .3 III. . . . . . . . . . . . . . . . . . . . . . . . . . Lf indicates the wave number associated to the length of the ﬁlm. . . . . . . . . . . . . . . . . ǫ. . . . . . . . . . . . . . . . . . . . . Solid line is reﬂection (p0 ). . . . . . . . . . . . The downstream length of each measuring window indicated by arrows is 7. dashed line is primary refraction (p1 ) and dotted line is secondary refraction (p2 ). .2 III. . . . . . . c) Section 3. . . . . . Windows 11-15. . . . . . Detail of the Jet Nozzle. . . . . .  . . . . . . Particles generated from Radius= 1 pixel to Radius= 25 pixels. . . . . . . . . . . . . . . .1 III. . . . . . Normalized Dissipation Rate of Turbulent Kinetic Energy. Energy Spectra: △ Measured Spectrum.4 II. . . . . . . . . . . . . . . . . Evolution of Dissipation Rate of Turbulent Kinetic Energy.7 III. . . Bubble Size Probability Density Function. Figure taken from Crowe et al. . . . . ǫ. . . . . . . . . . . Similar proﬁles of streamwise velocity. X/DJ = 20. Dimensions of the probe volume . . . . . . . . . ( ) represents the size obtained from the measured perimeter of particle and. . . . . . . . . . . . . D1 D2 and D3 are the three diﬀerent regions of the lens capturing the scattered light and focusing it into the respective detectors. . . . .14 mm. . . . .12 II. . . . . . . . . .8 II. . . . . . . . . . . . . . . .2 II. . . . Energy Spectrum: Inertial Subrange is identiﬁed to measure the dissipation rate. . . Phase-Diameter relation of Bubbles. . . . . . 15 15 16 17 II. . . . . . . . . . . . . . . . . . . Distribution of Photodetector in the Receiving Lens. . . . . . . . . . . . . . . . . . Windows 6-10. . . a) Section 1. . . . . . . . . . . . . . . .9 II. .7 17 18 19 20 21 26 29 30 II. . . . . 45 viii . . . . . . . . . . ◦ Wyngaard Corrected Spectrum. .3 II. . . . . . . . . . . . . .8 IV. .6 III. . . . . . ǫ. . . . . . . . . . . . . . . . . .5 38 III. . . . . . . Centerline Velocity. . . . . . . . . . . . 43 Discretization of the ﬂow region where the bubble break-up takes place. . Experimental Facility. . . . . .LIST OF FIGURES II. . . . . .. . . . . . . . . . . . . . . . . . . . Mie calculations of bubbles Phase-Size relationship for diﬀerent scattering angles.2 Downstream evolution of the dissipation rate of TKE. .1 39 40 40 IV. . . . . . . . . Uo = 12m/s. . . Experimental set 3a.5 Original Image. . . . . . the velocity at the exit of the nozzle is Uo = 17m/s.  . Size of particle measured by the analysis program as function of the size of the particle. . . . . . . . . . . . . . . . Processed Image.
. . . .10 IV.7 IV. . . . . . . . . a) Set 1.394mm. The value refers to the total ﬂux. .IV. .5 IV. . . . .19 IV. . . Evolution of the bubble break-up frequency and break-up velocity with respect to the diameter of the bubble. Evolution of the ﬂux of bubbles per window. .21 IV. Evolution of the ﬂux of bubbles per window. Lw is the length of the measuring window. . . . . . . . . Evolution of the bubble V pdf . . . Evolution of the ﬂux of bubbles per window. Experimental Set 2. . Experimental Set 3. Solid line is the conﬁning force provided by surface tension and the broken line is the one given by the turbulent stresses. . Lw is the length of the measuring window. Evolution of the ﬂux of bubbles per window. . . Characteristic water jet Diameter.584mm. . . obtained in each measuring window. Experimental Set 3c.394mm. Experimental Set 2. . Nt U /Lw . Da = 1.14mm. Da = 0. . . . . . . . . . Experimental Set 3a. experimental Set 3a.394mm. . b) set 3b. experimental Set 2.7 mm. Bubble break-up frequency. . . Da = 1. Da = 0. . . . .2. . Evolution of the ﬂux of bubbles per window. obtained in each measuring window. . Experimental Set 3b. Experimental Set 1. Nt U /Lw . . .15 IV. .20 IV. .16 IV.7. .11 IV. Bubble break-up frequency. . The value refers to the total ﬂux. . . a) Section 1. . .4 IV. Da = 0. Experimental Set 1. . Evolution of the bubble V pdf . . D0 = 2. .0mm. . . Dwj . Nt U /Lw . .12 IV.394mm. . . Bubble break-up frequency. Error bar indicates an estimated maximum ±10% experimental error. D0 = 1. Schematic representation of conditions used for the turbulent break-up measurements. . The value refers to the total ﬂux. . . obtained in each measuring window.The constant kσ = 6 is given in equation IV.13 IV. . . . c) set 2. . . the width and length of all measuring windows were respectively l1 = 16. Evidence of the existence of a frozen V pdf . . . The number shown indicates the total number measured in each window over 1.3. . .394mm. .18 IV. Evolution of the bubble V pdf . Nt U /Lw . . .8 IV. obtained in each measuring window. Da = 0. .9 IV. c) Section 3. . . obtained in each measuring window. . . . . . . . Comparison of experimentally measured bubble break-up frequency with the frequency calculated with model given in equation IV. Da = 0. experimental Set 3b. .20. . . . . Nt U /Lw .67 = mm. . which remains unchanged in the last three measuring windows. Da = 0. . Downstream evolution of number of the largest class-size bubbles. . . . . . In all experiments l1 /Dwj < 0. Lw is the length of the measuring window. Lw is the length of the measuring window.6 IV. Nci = Ni Ui /U1 . .25mm and Lw = 7. . . . . Da = 1. . . . Da = 0.14 IV. b) Section 2. .194mm. Evolution of the bubble V pdf . . Da = 0. experimental Set 1.22 Binary images corresponding to the ﬂow conditions shown in ﬁgure IV. . .17 IV. . .000 images corrected by the ratio of velocities between the measuring window and the ﬁrst window measured. . . D0 = 2. . The value refers to the total ﬂux. Experimental Set 3c. . . Force per unit surface on the bubble. . .584mm. . . .194mm. . ǫ = 2000 m2 /s3 .3 IV. . .394mm. . . 46 47 49 50 50 51 51 52 52 55 55 56 56 57 58 59 60 63 64 66 ix . Evolution of the bubble V pdf . The value refers to the total ﬂux. . . . . . . . experimental Set 3c. . . . . .194mm. Lw is the length of the measuring window. . . . . . . . . . .
b) Downstream evolution of the Sauter Mean Diameter Diameter. .7 V. . . . . . . . . . . downstream evolution of the cumulative volume probability density function. . . . . . D32 = 72 73 75 77 79 80 84 ÈN D ÈN D i i V. Da3 = 1. 000 m2 /s3 at the air injection point. . . .10 V. . . . . . . . . . . . △τt2 = 1 ρ β (ǫD2 )2/3 − 2 0 6σ D0 . blue line represents Konno et al. . . . . . . . . .4 V. . . 86 a) Downstream evolution of the Cumulative Volume Probability Density Function. . . . .11 of TKE was ǫ0 = 1. . . . Left column. . . Initial value of the dissipation rate È È i V. . . . . . In this example. . . . . .5 V. . . . Daughter bubbles pdf predicted by model of Tsouris and Tavlarides 1994. . . . . . . . . . Probability density functions of the daughter bubbles formed from the break-up of a mother one of size D0 . X/DJ = 15. . Right column.3 V. .000 frames per second. . . . The images were taken at 6. . . and ǫ = 1000 m2 s−3 . . b) Downstream evolution of the Sauter Mean Diameter N D3 Diameter.394 mm. . . . . . . . . . . . Experimental Set 2.V. . . . . .32 and the solid symbols are the experimental measurements. . . downstream evolution of the characteristic diameters D32 and Dv90% . . . . . . . . Daughter bubbles pdf predicted by model of Konno 1983 . . .32 and the solid symbols are the experimental measurements. . . . . . 1983 . . . Initial value of the dissipation rate 2 3 i 2 i and Dv90% . . . . . . . . . . .10. In this particular case. . Experimental Set 3a. 88 Initial Bubble Size. the mother bubble is of size D0 = 1 mm. . . . . . . . . The solid lines represent the results obtained from the model integrating equation V. . . . . . △τt1 = 2 ρ β (ǫD1 )2/3 − Dσ and.194 mm. Da2 = 0. . . . . . . . 87 Time evolution of the characteristic break-up of a mother bubble showing a binary bubble splitting mechanism. given in equation V. . . . .8 Diﬀerence of stresses associated to the formation of a bubble and its 6 1 complementary. . . . Diﬀerence of stresses associated to the formation of a bubble and its complementary. . . . . . c) Experimental Set 3c. . . a) Downstream evolution of the Cumulative Volume Probability Density Function. .584 mm. .1 V. . . . . . . . . . . .9 of TKE was ǫ0 = 2. . . . . . Red line represents Tsouris and Tavlarides 1994 . . a) Evolution of the pdf for various values of ǫ and ﬁxed D0 = 3mm. . . . . D32 = Ni Di and Dv90% . . . . . b) Inﬂuence of D0 on the the pf d for a ﬁxed value of ǫ = 1000 m2 s−3 . . . . . . . . X/DJ = 15. and the dissipation rate is ǫ = 1000 m2 s−3 . . . . . .6 V. . . . and black lines represent our present model. . .a) Experimental Set 3a. . . . . . The solid lines represent the results obtained from the model integrating equation V. . . . . . . . b) Experimental Set 3b. . . . . . . . . . . . . .2 V. . 000 m2 /s3 at the air injection point. Da1 = 0. 89 i x . . Comparison of the daughter bubble size pdf predicted by the present model and previously proposed models. . . . . . . . . . . the mother bubble is of size D0 = 3 mm. . .
The solid lines represent the results obtained using our model assuming a tertiary breakup process model to integrate equation V. 1983  model to integrate equation V. △ 120 µm < D < 200 µm. ⋄ 80 µm < D < 100 µm.000. . and the solid symbols are the experimental measurements. . Comparison of the daughter bubbles pdf resulting from a binary and a tertiary splitting. . . and the solid symbols are the experimental measurements. Experimental Set 3c. . b) Downstream evolution of the sauter Mean Diameter and Dv90% . . Error bars are 5% of the value.3 104 VI. . . . 000. . . . Symbols are: ◦ 3 µm < D < 20 µm. . . . Re = 53. Downstream evolution of the Cumulative Volume Probability Density Function. . . . .32.1 VI. . D > 200 µm. .14 V. . Symbols are: ◦ 3 µm < D < 20 µm. . . . . . The solid lines represent the results obtained from the model integrating equation V. . . △ 120 µm < D < 200 µm.5 108 xi . . . . and the solid symbols are the experimental measurements. . . . .32. . The Reynolds Number of the jet has been varied from 32. ⋄ 80 µm < D < 100 µm. . .12 a) Downstream evolution of the Cumulative Volume Probability Density Function. .000. . . . 000.000 to 60. .32 and the solid symbols are the experimental measurements. Time record of bubbles generated by a water jet. . . . . . . . . Re = 53. 40 µm < D < 60 µm. . .13 V. . . . . . ⋄ 80 µm < D < 100 µm. . . .000 to 60. . . .2 103 VI. Experimental Set 3c. . . 40 µm < D < 60 µm. . . . . . . . △ 120 µm < D < 200 µm. . P df s of the normalized Inter-arrival time. . . Experimental Set 2. . ǫ = 1000 m2 /s3 . .32. . . . . . . . . Evolution of Mean Inter-arrival time with the local convective velocity. . 40 µm < D < 60 µm.The solid lines represent the results obtained using Tsouris and Tavlarides 1994  model to integrate equation V. a) Downstream evolution of the Cumulative Volume Probability Density Function. . .4 107 VI. 000. . . . . . △ 120 µm < D < 200 µm. . . b) Downstream evolution of the Sauter Mean Diameter N D3 Diameter. . . Inter-arrival time pdf s of diﬀerent classes of bubbles. . . b) Downstream evolution of the sauter Mean Diameter and Dv90% . . . . . . . . . . . . . . 000 m2 /s3 at the air injection point. . t/tm (tm = mean inter − arrival time) of diﬀerent classes of bubbles. . . . . . 40 µm < D < 60 µm. Symbols are: ◦ 3 µm < D < 20 µm. . The solid lines represent the results obtained using Konno et al. D32 = Ni Di and Dv90% . D > 200 µm. . . U . . . . i 91 92 94 95 96 97 102 VI. . . . .The Reynolds Number of the jet has been varied from 32. . . . Experimental Set 2. . . . .16 V. . . . . D0 = 2 mm . . . . . U .15 V. . X/DJ = 15. Re = 53. . Initial value of the dissipation rate 2 È È i V. Maximum number of bubbles resulting from the break-up of a mother bubble of size D0 . a) Downstream evolution of the Cumulative Volume Probability Density Function (Vpdf). . .17 of TKE was ǫ0 = 1. Evolution of Rms/Mean Inter-arrival time with the local convective velocity. . . . ⋄ 80 µm < D < 100 µm. . . .V. . . Symbols are: ◦ 3 µm < D < 20 µm. . . .
. . . .35 ml/min. . . . . . . . . . . 000. . . . . . . . . . 000. . . .15 VI. . . . . . .000 to 60. . . .25 VI. . . . . 128 Dependence of maximum bubble size on the initial air void fraction. . . . . . . . . . . . . .17 VI. . . . . .24 VI. . . . . . . . . . . 000. Re = 54036. . . . ⋄ Re = 60. . . . .27 VI. . . Re= 93. . 112 Evolution of the normalized Inter-arrival distance pdf of bubbles of size 40 µm < D < 60 µm with the Reynolds Number. . 109 Evolution of the normalized Inter-arrival time pdf of bubbles of size 40 µm < D < 60 µm with the Reynolds Number. . . . . . . . . . . 129 Showing several combinations of initial void fractions and Re which result in identical pdf s. . . . . . . . . . 112 Frozen bubble size pdf for various Reynolds numbers of the submerged water jet. . . . .717 and initial air void fraction α = 2. . 116 Normalization of Gamma pdf s. . . . . . . . . . . . . Re= 54.7 VI. .036 . . . 110 Evolution of Mean Inter-arrival distance with the local convective velocity. . α. U . . . .118 Lognormal pdf . . . 000. . .11 VI. . . . 109 Inter-arrival distance pdf s of diﬀerent classes of bubbles. . . . . . . . . 130 xii . . . . . . . . . ⋄ 80 µm < D < 100 µm. Qa = 34. 40 µm < D < 60 µm. . . . . 122 ˆ Bubble size pdf of D = D/D32 . . . .6 ml/min. . . Qa = 3. . . .9 VI. . . . . . . . . Re = 45. . . . .036 .16 VI. . . . Qa = 7. . 120 Evolution of the bubbles size pdf with the Reynolds number. 121 Evolution of the bubbles size pdf with the Reynolds number. . . . . . . 000. . . . 126 Evolution of the bubbles size pdf with the air void fraction. . . . . . Symbols are: ◦ 3 µm < D < 20 µm. Re = 54036. . . D > 200 µm. . . . 000. 121 ˆ Bubble size pdf of D = D/D32 . . . . . . . . . . . . . Symbols are: ◦ Re = 29. Qa is the ﬂow rate of air injected . . . . . . . . .13 VI. Qa = 7. Re = 53. . 115 Comparison of the experimental bubble size pdf with a Gamma distribution at Re = 53552. . . . . . . . . . . . .8 VI. . .VI.000. 000. . . . . . . . . . . . . . . . . . . . . . 125 Dependence of the frozen bubble size pdf on the initial air void fraction.28 VI. . . . . . . . . . 000. .127 ˆ Bubble size pdf of D = D/D32 . . Symbols are: ◦ Re = 29. . . .26 VI. . . . . . . . Re = 45. . . . . .19 VI. . . . . . . . . . . . . .35 ml/min. . ⋄ Re = 60. . . . . . . . . . Qa = 3. . .14 VI. . . . Symbols are: ◦ Re = 29. . 40 µm < D < 60 µm. . .20 VI. . . . .22 VI. . . . . 000. . . . .6 ml/min. . . . . . △ 120 µm < D < 200 µm. . . 000. . . .6 VI. 123 Variation of the maximum bubble’s size with the jet’s nozzle Weber number. . . . . . . . . . . . . . . 120 ˆ Bubble size pdf of D = D/D32 . . . . . .12 VI. . 117 Comparison of the experimental bubble size pdf at low Reynolds number. . . △ 120 µm < D < 200 µm. . . . . . . . . . . 000. . . . . . . . . . . . . Re = 45. Symbols are: ◦ Re = 29. . . .21 VI. . . Re= 54. 122 Evolution of Drms /Dmean with the water jet reynolds number. . .25 ml/min. . . .10 VI. . . . Symbols are: ◦ 3 µm < D < 20 µm. . . . . . . . . . . .The Reynolds Number of the jet has been varied from 32. 000. . . . ⋄ Re = 60. . .23 VI. . . . . . . . 118 Evolution of the bubbles size pdf with the Reynolds number. . . . . 111 Evolution of the inter-arrival distance pdf of bubbles of size 40 µm < D < 60 µm with the Reynolds Number. ⋄ Re = 60. ⋄ 80 µm < D < 100 µm. . . . Qa = 34.25 ml/min. . . . .18 VI. .29 Evolution of the Inter-arrival time pdf of bubbles of size 40 µm < D < 60 µm with the Reynolds Number. . . . . . Re = 45. . . . . . . . . . . 000. .5 10−4 . .
. . . . . . . . . . . . . . Solid line represents the cumulative probability density function. . . .2 Particle Size Probability Density Functions. . . . . 137 xiii . . 136 Cumulative Probability Density Functions. . .1 A. . broken line is the cumulative volume probability density function. . . . . . broken line is the volume probability density function. . .A. . . . . . . . . . . Solid line represents the number probability density function. . . . . . .
Rate of death of particles.Length of the hot ﬁlm.Interfacial surface tension.Diameter of the water nozzle.Number density of particles of size D. λt .Kolmogorov viscous scale. ˙ De .Width of measuring windows.Integral length scale.Kinematic viscosity. U .Mean value of the local velocity of the water ﬂow. Qw .Taylor microscale.Air ﬂow rate. Da . Dc .Initial air void fraction.Length of the measuring windows.Rate of birth of particles. D0 .Particle diameter. n(D) .Break-up frequency. Lf . Dmin .Mean number of bubbles resulting from the break-up of D0 . Qa . D0 ) . ǫ . xiv . D1 .Critical bubble diameter.Water ﬂow rate.Minimum daughter bubble diameter. η . σ . ν .Width of the water jet. D0 ) . U0 . DJ .Dissipation rate of Turbulent Kinetic Energy per unit mass.Diameter of the air injection needle. m(D0 ) . f (D. Dwj .LIST OF SYMBOLS ˙ Bi . l1 . α = Qa /Qw . g(ǫ.Velocity of the water at the exit of the nozzle. D . Lw . Lx .Daughter bubble pdf resulting from the break-up of D0 .Diameter of the daughter bubble.Diameter of the mother bubble.
Sutanu Sarkar. he has given me the opportunity of meeting the structure engineering group and the beauty of the carbon ﬁber reinforced bridges. experience and patience with me: Geno Pawlak for the unconditional time dedicated to answer all my questions. willing to see and talk with their long-lost brother. Forman Williams. Prof. who manages very well to keep everybody’s attention. I was very excited to meet the youngest member of the family. To all xv . Antonio Sanchez who is always an example of energy and dedication. this period has been harder for them than it has been for me. To Markus Wernli for sharing with me all those afterwork moments. Kurt Keller who carried me through the diﬃculties of the spectra and Antonio Alfaro. have always been a deep part of my development and a source of fruitful ideas. Prof. I also owe thanks to a large number of people who shared their knowledge. Martin Maxey for all his insightful comments and Prof. John Hewson. During my stay in San Diego I have been lucky enough to meet Alexis Hix Lasheras.ACKNOWLEDGMENTS I would like to acknowledge the support and tolerance received from my parents and family during the completion of my graduate studies. whose unconditional help I will never forget and I can only hope to repay her in some way for everything she has done for me. Diego. Charo and Bea. in many ways. cooking skills and late evening conversations. As my professor and as my friend he is always available to share his knowledge and experience. In addition. knowing that. Jos´ Manuel. Professor Juan C. Larri Armi who provided me with a large source of questions and very stimulating discussions. His guidance and support. given not only in the scientiﬁc but also in the personal ﬁeld. I am also in debt to my committee members Prof. Kathy Prestidge. Ludovic Raynal and Paul Rightley. My deepest gratitude is due to my thesis advisor. Lasheras. I am very grateful to Professor Jos´ Luis Monta˜es for all the eﬀort e n he put in this work and for giving me the opportunity of meeting the great person he is. Prof. who exposed me to the fascinating world of the turbulent interaction in two phase ﬂows. I need especially to thank my two sisters. who have e always been close to me. Ken Kiger. and my brother-in-law. Michael Buckingham and Prof. Ken Melville.
Monta˜es. and J. Chapter 5 is comprised. of a paper titled “On the size pdf resulting from the break up of an air bubble immersed into a turbulent liquid ﬂow”. C. Mart´ ınezBaz´n. Finally. Lasheras. currently being considered for publication in Journal n of Fluid Mechanics. C. in part. I also would like to thank Carol and George Hahn who treated me as a member of their family since I landed in San Diego. in full. currently being considered for publication in a n Journal of Fluid Mechanics. her unique understanding and dedication have shown me a new perspective on life. Without her support this thesis would have never been completed. during the development of this work and a research grant from the Consejo Asesor de Investigacion provided by Diputaci´n General o de Arag´n (Spain). J. xvi . J. Lasheras. o Chapter 4 is comprised. friendship and partnership.of them I have to say that I ﬁnally managed to make the bubbles go down. Beyond work and science I need to especially thank the person who has carried me during the last two years. Monta˜es. Mart´ ınez-Baz´n. C. Paqui. C. The dissertation author was the primary investigator and author in each of these publications. and J. L. I would like to acknowledge the support provided by the US Oﬃce of Naval Research. a L. of a paper titled “On the break up frequency of an air bubble injected into a fully developed turbulent ﬂow”. ONR# N00014-96-1-0213.
S. 1998. Under consideration for publication in the Journal of Fluid Mechanics. “Dispersion of Bubbles in Shear Flows. Lyon (France) June 1998.S. C. Sakar Studies in Numerical Methods. Vol. J. 27-12. Professor S.. Williams xvii . FIELDS OF STUDY Major Field: Fluid Mechanics Studies in Fluid Mechanics.VITA 1992 1994-1995 1995 1995-1998 1998 Mechanical Engineering degree. Department of AMES University of California. Professors J. Under consideration for publication in the Journal of Fluid Mechanics. 1998. Professor F.Sarkar Studies in Combustion and Gas Dynamics. Lasheras. Professors C. TSF97. Van Atta Studies in Environmental Fluid Mechanics. 2.C. Pozrikidis.Williams. Seshadri Studies in Mathematics. San Diego Research Assistant.C. Spain Teaching Assistant. University of California. Lasheras. S. Pozrikidis. “On the size P DF resulting from the break up of an air bubble immersed into a turbulent liquid ﬂow”. San Diego Doctor of Philosophy (Aerospace Engineering) University of California. S. 3. J. San Diego M. San Diego PUBLICATIONS “On the break up frequency of an air bubble injected into a fully developed turbulent ﬂow”. Armi Studies in Turbulence. S. Sakar Studies in Computational Fluid Dynamics. Universidad of Zarazoga. Aerospace Engineering. Professors L. Sakar. Professor L.” Third International Conference on Multiphase Flow ICM98. Professor F. “Splitting and Dispersion of Bubbles by Turbulence.” Eleventh Symposium on Turbulent Shear Flows. Lasheras. Armi. C.C.6-17. Department of AMES University of California. Grenoble (France) September 1997.
A phenomenological model for the break-up frequency is proposed showing that for large bubbles whose sizes are much greater than Dc = 1. W et = ρ △u2 (D0 ) D0 . but also on the value of the dissipation rate of turbulent kinetic energy of the underlying turbulence. in a consistent manner. This simple model predicts detailed experimental measurements of the transient bubble size pdf s performed over a range of bubble sizes and dissipation rates ǫ. Based on energy principles. These measurements were used to determine the break up frequency of the bubbles as a function of their size and the dissipation rate of turbulent kinetic energy of the underlying turbulence. San Diego.ABSTRACT OF THE DISSERTATION Splitting and Dispersion of Bubbles by Turbulence by Carlos Mart´ ınez Doctor of Philosophy in Engineering Sciences (Aerospace Engineering) University of California. a statistical model to describe the bubble size probability density function of the daughter bubbles resulting from the shattering of a mother bubble of size D0 immersed into a fully developed turbulent ﬂow is proposed. Chair The transient evolution of the bubble-size probability density function resulting from the break-up of an air bubble injected into a fully developed turbulent water ﬂow has been measured experimentally using digital image processing techniques. Lasheras. The agreement between the model and the experiments is particularly good for low and moderate values of turbulent Weber number of the bubbles. D0 . it decreases ρ with the bubble size as ǫ1/3 D −2/3 . ǫ. The model shows that the bubble-size pdf depends not only on D0 . ǫ. as well as to determine the bubble size probability density function of the daughter bubbles formed from the break-up of a mother bubble of size. σ where xviii .26 ( σ )3/5 ǫ−2/5 . The model is shown to be in good agreement with the measurements performed over a wide range of bubble sizes and values of ǫ. 1998 Professor Juan C.
xix . and the assumption of tertiary break-up is shown to lead to a better ﬁt of the experimental measurements. of the distribution. The f rozen bubble size distribution. However. has been measured using a Phase Doppler Particle Analyzer (PDPA). D32 . At larger values of W et . it was found that the most probable number of daughter bubbles increases. the data is shown to collapse onto a single distribution. independent of ǫ and α. achieved once the break-up process has concluded.the assumption of the binary break-up is shown to be consistent with the experimental observations. α. It is shown that the shape of the distribution depends on ǫ and on initial void fraction. when the bubble’s diameter is non-dimensionalized with the Sauter Mean Diameter.
Baranaev et al. the chemical engineering community has devoted a considerable amount of work to the study of stirred (or agitated) vessels such as those used to produce emulsions. .Chapter I Introduction I. Central to the development of predictive models of these engineering and natural processes is the description of the break-up of an immiscible ﬂuid immersed into a turbulent ﬂow. . . Although this problem has been the subject of a continued investigation beginning with the early work of Kolmogorov . Although these 1 . depends on the amount of air entrained by the wave action.  and Hinze . and more importantly on the distribution of bubble sizes resulting from the break-up of the entrained air by the underlying turbulence existing under the free surface .A The turbulent break up problem The mass transfer rate occurring in many natural and engineering processes depends on the amount of contact surface which is created between two immiscible ﬂuids. the absorption rate of a certain chemical species depends on the distribution of droplet sizes in which one ﬂuid is dispersed into another. or many other water-soluble trace species. in liquid-liquid or gas-liquid separators. In aeration processes such as the ones which occur naturally during the interaction of the atmosphere and the oceans. and has generated a large bibliography. among many others. . Over the years. For example. . the absorption rate of carbon dioxide. a uniﬁed model capable of predicting the probability density function of the droplets (or bubbles) resulting from the turbulent break up does not yet exit.
This distribution is crucial in the development of models to describe the turbulent break-up . Furthermore. the drop or bubble break-up experiments conducted in turbulent pipe ﬂows have had similar diﬃculties also resulting from the anisotropic nature of the turbulence in the pipe which is composed of free stream turbulence and high shear regions near the walls. by measuring the time evolution of the drop-size pdf s ( . in these vessel experiments. As mentioned above. but more importantly. . These shear regions often cause the preferential migration of the bubbles towards the wall making the characterization of the turbulence under which the break-up takes place very diﬃcult. the transient size pdf s are often measured by withdrawing samples in time which are later diluted. For example. and of the resulting probability density distribution of the size of the daughter droplets. and others). . they have not been able to produce reliable information on the evolution of the droplet sizes during the transient break-up processes. . .2 experiments have provided valuable information on the steady state droplet size distribution. the diﬃculty in the use of stirred vessels is that the turbulence in the vessel is not very well characterized since it is not only inhomogeneous throughout the vessel. nor that they are preserved during the sampling. it is highly anisotropic consisting of high shear regions on the surface of the impeller and strong tip vortices shed by the impeller blades. and often precludes the interpretation of these measurements. due to the complexity introduced by the use of turbine impellers in the agitated tank experiments. there have been numerous experimental studies to determine the drop’s break-up frequency. another important factor in the break process is the knowledge of the distribution of daughter bubbles (droplets) formed after the break-up of the original one. . and others). These sampling techniques neither guarantee that the droplet-size pdf s are frozen. or in turbulent pipe ﬂows. or stabilized. . knowledge which is essential to the development of the models. prior to their measurements. many attempts have been made to determine this frequency in stirred tanks. In the past. Essential to the development of the models is the knowledge of both the probability (or frequency) of the drop’s (or bubble) break-up. . On the other hand. Furthermore. In addition to knowing the frequency of break-up of the bubbles (droplets). the nature of the turbulence existing in the tank is diﬃcult to characterize (.
In the high Reynolds number regime of interest here. σ (I. η η (I. viscous and inertial forces.B Background and Previous work Three basic mechanisms are involved in the break-up of drops or bubbles im- mersed in an immiscible continuous ﬂuid: surface tension. If the Reynolds number of such processes is large. η η (I. µc is the viscosity of the continuous phase. Break-up will happen above a critical value of the Weber number. inertial and capillary forces are dominant while at low Reynolds numbers viscous forces and surface tension are the important eﬀects. η η D D f ( ) ∝ ( )2 f or D << η Dissipation range.1) where D is the diameter of the particle.3 mechanism. if the underlying turbulence is homogeneous and isotropic. the dominant forces acting on the particle are the dynamic pressure produced by the surrounding ﬂuid (characterized by the diﬀerence in the velocity within a distance equal to the diameter of the particle) and the surface tension. He found that the rate of deformation of the dispersed ﬂuid depends on a Weber number deﬁned as: Wes = µc S D . The latter case was studied by Taylor  who conducted experiments to study the deformation and splitting of droplets when one ﬂuid is injected into another.3) . η the Kolmogorov microscale. the average value of the squared velocity diﬀerences at a distance D can be written as: ν D △u2 (D) = ( )2 f ( ). and f is a universal function f( D D ) ∝ ( )2/3 f or D >> η Inertial range. σ is the surface tension.2) where ν is the kinematic viscosity. and S is the maximum shear of the mean ﬂow. Extensive studies on the break-up of droplets and bubbles were made by Kolmogorov  and Hinze  ﬁve decades ago. According to Kolmogorov. I.
t) dD dx. it will disintegrate into smaller droplets.2 and I. viscous forces dominate over the inertial forces and the process can be characterized by We and νd ν . ν and the Weber number used in the present context. Wec . δ δp ˙c ˙ + ▽x (v p) + ▽v (F p) = − (R p) + Q′ + Q′ + Γ .4 If one considers an isotropic turbulent ﬂow and introduces a lump of a second immiscible ﬂuid of interfacial surface tension σ and kinematic viscosity νd . for diameters much smaller than η. Dc . dv is the probable number of particles (bubbles) in the diameter range dD about D located in the spatial range dx about x with velocities in the . is achieved for which the straining and distortion determined by the velocity diﬀerence.4 to explain experimental results where exponents larger than 2/5 have been found. therefore from equations I. the role of the viscosity of the continuous phase can be negligible and only the Weber number characterizes the process .  have extended these ideas by introducing the eﬀect of intermittency in equation I. and. If D >> η. According to Kolmogorov. Kolmogorov showed that the behavior of the droplet can only be dependent on three dimensionless numbers: We = △u2 (D) D ρ σ D νd η. must correspond to a critical Weber number. Dmax . b δt δD (I. as is the case of interest here. the three dimensionless numbers characterize the process. Although the dimensional argument proposed by Kolmogorov provides valuable information on the maximum bubble (droplet) size.3 one obtains: Dmax = ( Wec σ 3/5 η 4 2/5 ) ( 3) f or D >> η ρ ν (I. are small and can no longer overcome the restoring surface tension forces.5) where p(D. △u2 (D). the maximum size of the stable droplet (bubble). These droplets will keep breaking until a critical diameter. However. .4) More recent studies . The most successful and fundamental of these models is based on the population-balance-equation (p-b-e). which describes the drop population history in terms of the particle properties and ﬂow characteristics as they interact among themselves and with the surrounding ﬂuid . v. x. In the case of drops (bubbles) of diameter of the order of the Kolmogorov scale. owing to the surface tension. it does not give any information on the shape of the pdf. Over the last years an important eﬀort has been made by various researchers to develop a model able to predict the shape of the pdfs of the droplet size resulting from the turbulent break-up.
5 range dv about v at a time t. coalescence takes place when there is a collision between particles with a residence time longer than . coalescence and dissolution or evaporation. If dissolution and evaporation eﬀects are negligible. a particle contributes to De when it disappears after it coalesces with another one to create a bigger particle.6) ˙ Q′ dv. .9) v p dv/ p dv . Similarly. v.7) and the bar denotes an average over all velocities v= and R= R p dv/ p dv . t) n(D. The models used to characterize ˙ ˙ the destruction (De ) and formation (Bi ) of droplets are generally based on the particlecollision theory. or it breaks into smaller ˙ ones. (I. caused by collisions with other particles which do not result in coalescence. condensation or dissolution. x. (I. The rate of change of p due to break-up ˙ ˙ or coalescence is given by Q′ and Q′ respectively and ﬁnally Γ represents the rate of c b change of the distribution function. (I.8) Equation I. t) ˙ ˙ + ▽x (v(D. The number of particles per unit volume and per c unit range of diameter is n= p dv . Bi . Eliminating the velocity dependency by integrating equation I. p. In the same manner.5 over the entire velocity space. δ δD (R n) = 0. t)) = De + Bi .  assumed that the break-up of a particle of diameter D occurred when it collided with an eddy of size equal to or smaller than the particle with suﬃciently large energy to overcome surface tension.10) ˙ ˙ where De and Bi respectively represent the rate of death and birth of droplets (bubbles) due to break-up. R = (dD/dt) is the rate of change of the size D of a particle (bubble) due to evaporation.6 can just simply be expressed as δn(D. t) = (dv/dt) is the force per unit mass on the particle. F(D. we get δ δn ˙ ˙ + ▽x (v n) = − (R n) + Qb + Qc . δt δD ˙ where Qb = ˙ ˙ Q′ dv. accounts for the coalescence of smaller particles and for the break-up of bigger droplets to form smaller ones. Qc = b (I. δt (I. Previous investigators.
v ′ )n(v − v ′ )n(v ′ )dv ′ + where λc (v. v ′ ) m(v ′ ) g(v ′ ) n(v ′ ) dv ′ (I. where Sed is the collision cross-section given by Sed = length of the eddy which collides with a particle of diameter D and volume v = ue and ud are the characteristic turn over velocity of the eddy and the velocity of the .13) The collision process is often described by analogy to kinetic theory of gasses whereby. e d π(de +D)2 .6 the time necessary to drain the ﬁlm of ﬂuid conﬁned between the two particles. v ′ )n(v − v ′ )n(v ′ )dv ′ is the number of particles of volume v generated from coalescence. as: gb (v) = (eddy − particle collision f requency) × (breakage ef f iciency) (I. The above birth and death terms are generally written as: De = − n(v) v ∞ 0 λc (v. n(v)gc (v) = n(v) ∞ ′ ′ ′ ′ 0 λc (v. the break-up eﬃciency and the pdf of the daughter droplets. 4 (I. v ′ )hc (v − v ′ . ∞ v v 0 λc (v − v ′ .12) Bi = 0 λc (v − v ′ .  modeled the break-up frequency. In the Bi equation. g(v). g(v) is the break-up frequency of a particle of volume v . v ′ )m(v ′ )g(v ′ )n(v ′ )dv ′ is the number of particles of volume v generated from the break-up of bigger ones. v ′ )hc (v.14) de is the characteristic πD 3 6 . hc (v. The diﬃculty involved in solving equation I.11) f (v. Therefore. v )n(v )dv is the amount of particles of volume v which die due to coalescence and g(v)n(v) is the number of particle which die due to break-up. and f (v. the eddy-particle collision frequency (hb ) is given by: hb (D) = ne Sed (u2 + u2 )1/2 dne . v ′ ) is the coalescence eﬃciency of a collision between a particle of volume v with another of volume v ′ . Prince et al. m(v ′ ) is the number of particles generated once a particle of volume v ′ has been broken. v ′ ) is the probability density function of particles of volume v generated from particles of volume v ′ (this term is typically referred to as the probability density function of daughter particles). namely the coalescence eﬃciency. v ′ )n(v ′ ) represents the collision rate of particles of volume v ′ with a particle of volume v . and f (v. v )hc (v. v ′ )n(v ′ )dv ′ − g(v)n(v) ∞ v (I.10 results from the need to model the diﬀerent terms of the equations. v ′ )hc (v − v ′ .
hc (Di . The above explained model requires the use of non-physical assumptions for the number of eddies and the eddy-particle collision frequency and cross section which make it questionable.18) where Tcij is the coalescence time and τij is the residence time or contact time. Once the particles collide. Therefore. and the model used to account for the eﬃciency of both coalescence and bubble-eddy collision. (I. Dj ) = exp[− Tcij ]. Regarding the resulting size probability density function of the daughter droplets .16) The collision frequency between two particles of diameters Di and Dj .7 particle respectively. which gives the probability of an eddy-particle collision to result in particle break-up can be written as. f (v. c1 e (I. The breakage eﬃciency (λb ). Sed and ne are two ambiguous parameters.18 is based on the accuracy of the assumed probability density function of the daughter particles. 4 (I. and ne is the number density of eddies of size de .17) Di is the diameter of a particle. Ec is an activation energy. coalescence occurs when the residence time is longer than the coalescence time and the eﬃciency can be then estimated as: λc (Di . Dj )nj = SDi Dj (u2 i + u2 j )1/2 nj D D where SDi Dj is the collision cross-section given by SDi Dj = π(Di +Dj )2 . Dj ) is: hc (Di . ¯ D D The reliability of the model given by equations I. : λb (D) = exp − Ec . v ′ ).10-I. c2 τij (I.14. and c1 is a constant of order unity which needs to be modeled. In equation I. uDi is the velocity of the particle of diameter Di and ni is the number of particles of size Di . Coalescence can be modeled in the same way as the breakage. diﬃcult to characterize. Tcij can be estimated calculating the drainage time of the ﬂuid ﬁlm formed between two disks of diameters Di and Dj approaching at a velocity u = (u2 i + u2 j )1/2 .15) where e is the average energy of an eddy. they may stay in contact long enough to coalesce or they may bounce back. where: gc (v) = (particle − particle collision f requency) × (coalescence ef f iciency).
Tsouris and Tavlarides  proposed a model based on surface energy arguments. ﬂow properties (the dissipation rate of turbulent kinetic energy of the underlying turbulence). This knowledge was then used to develop phenomenological models based on dynamic arguments. and its dependence on the dissipation rate of turbulent . . and initial concentration of the bubbles in the ﬂow. which is independent of the intensity of turbulent ﬂow and/or the size of the mother particle. a combination of energy and statistical concepts have also been used to propose more physical models. v ′ ). All these models lack in generality and are not consistent with the experimental observations. Furthermore. we selected to study the turbulent break-up by injecting air bubbles into the fully developed turbulent region along the central axis of a high Reynolds number water jet. we then measured the transient pdf as well as the frozen pdf of the bubbles sizes resulting from the turbulent break-up over a wide range of initial bubble sizes and turbulent conditions characterized by the turbulent kinetic energy (or the dissipation rate) of the underlying turbulence. . These measurements were then used to evaluate the various elements comprising the existing models such as the break frequency g(v) and daughter droplets pdf f (v. .8 (f (v. The objective of this work is to obtain detailed experimental measurements of the probability density function of the droplet sizes resulting from the break-up of an immiscible ﬂuid injected into a turbulent ﬂow of known characteristics using novel PDPA and digital image processing techniques. and to prevent the additional complexity introduced by the use of turbines or any other moving surfaces to generate the turbulence. . viscosity and density ratio between the two ﬂuids). The purpose of this thesis is then to provide the fundamental knowledge of the mechanisms involved in the turbulent break problem and their dependence on the ﬂuid’s properties (surface tension. First a model is presented for the break-up frequency that will be used later to study the shape of the daughter pdf of the bubbles resulting from the break of a mother one. Most of the models proposed a daughter distribution function. v ′ )). from which the daughter droplets have been formed. Some of these distributions are just based on statistical assumptions when no other information is available . . Through the use of Phase Doppler Techniques (PDPA) and high speed image processing. a large number of distributions have been proposed. . In order to isolate the problem.
the shape of the frozen (or unchanged) bubble size probability density function and its dependence on the ﬂow conditions is analyzed.9 kinetic energy and the size of the mother bubble. . Finally. once the break process has ﬁnished.
the particle size distribution has been extensively measured from the mass fraction of particles in a speciﬁc size interval using either Sieving analysis or Sedimentation methods. Most of the techniques used in single phase ﬂows are not applicable when a second phase is present in the ﬂow due to dramatic changes in the physical properties of the ﬂuids. where the properties of the dispersed phase are not modiﬁed by the presence of the sampling probe.Chapter II Experimental Techniques II. For instance. Both methods are very time consuming and require a large amount of measurements to obtain statistically reliable results. In some industrial processes. 18 µ the particle size distribution is calculated from the time dependent concentration of particles at the measuring point. if not impossible. makes it diﬃcult. Since the Sedimentation methods are controlled by the terminal velocity of the particles. These techniques can be divided into diﬀerent groups depending on their performance as shown in table II.A Introduction Detailed measurements of two-phase ﬂows properties are crucial to characterize and control many natural and industrial processes. ut = (ρd −ρ) g D 2 1 .1. Numerous measurement techniques are available for experimental studies of two-phase ﬂows. ρd is the density of the discrete phase and ρ and µ are the density and viscosity of the surrounding ﬂuid respectively. the intrusive methods have been successfully applied. to obtain reliable measurements of the ﬂow characteristics. The presence of these discontinuities. 1 10 . produced by the presence of the discrete phase.
etc. especially in connection with gas-liquid two-phase ﬂows . In this technique. Among the most common on-line. laser diﬀraction. are Flow Visualization (image processing). the suction velocity has to be the same as the local velocity of the ﬂow (isokinetic conditions). optical techniques developed in the last decades. Particle Tracking Velocimetry (PTV) which has evolved to a more advanced Particle Image Velocimetry (PIV). The use of optical probes to measure concentration. Problems such as dispersion of particles in turbulent ﬂows. To avoid theses diﬃculties. A common characteristic of the techniques described above is that all necessarily disturbed the ﬂow because of the presence of the associated probes. Acquisition of a sample and later analysis may modify the properties of the dispersed phase due to the sampling procedure and the presence of the probe. velocity and size distribution has been recently increased in the last decade. Non-intrusive techniques are necessary to study the behavior of particulates subjected to diﬀerent forces produced by the ﬂow. The particles are collected in a bag ﬁlter for an interval of time and analyzed later on. laser at- . atomization.1: Measurement Techniques Measurements of the dispersed phase mass ﬂux and concentration have been performed in the past using isokinetic sampling. optical techniques have been widely used since the development of the laser.11 Property of Particles Size Intrusive Coulter Principle Sedimentation Sieving Fiber Optics Probes Isokinetic Sampling Fiber Optics Probes Conductivity Probes Fiber Optics Probes Non-intrusive Phase Doppler Anemometry Laser Diﬀraction Light Scattering Image Processing Phase Doppler Anemometry Laser Diﬀraction Laser Absorption Light Scattering Phase Doppler Anemometry Particle Image Velocimetry Laser Doppler Velocimetry Concentration Velocity Table II. The principle of this method is based on introducing a sampling probe in the two-phase system and extracting a representative population of particle by suction. require measurements on-line taken in real time.
while sizing techniques based on diﬀraction are limited by the time necessary to process the signal and ﬁt the proper distribution to it. optical techniques. whereas their application to the analysis of axisymmetric ﬂows requires deconvolution techniques to extrapolate the local properties of the ﬂow.. Laser Doppler Velocimetry (LDV) and Phase Doppler Anemometry (PDA). integral methods and local methods.i. The light scattered by the particles moving through a small region of the ﬂow is instantaneously collected by an imaging device. typically generated by a pulsed laser with a very short life time. namely. Theses method are extensively used in two-dimensional ﬂows. One method which is rapidly gaining popularity in the ﬁeld of experimental ﬂuid dynamics is the Particle Image Velocimetry (PIV). On the other hand. This technique utilizes a light sheet. provide time resolved. local measurement techniques allow us the determination of the properties of the ﬂow with a temporal and spatial resolution which depends on the method applied. The research in the ﬁeld of the multiphase ﬂows has advanced considerably with the development of new local. optical techniques are available to obtain detailed ﬂow visualization and quantitative and qualitative measurements. spatially averaged information of the two-phase system along a path deﬁned by the portion of the beam crossing through the ﬂow. For example. A recent development of the PIV techniques is that of Pawlak and Armi . Other non-intrusive. who replaced the pulsed laser by a continuous laser source of light to study the vortex dynamics on arrested ﬂows. . thereby causing the method to be more complicated and less reliable to apply. Typically. laser attenuation can be as fast as the response time of the photodiode used (> 100Khz). Integral methods. The capability of some of these methods allows the simultaneous measurement of the velocity and the size of the particles at a certain spatial location determined by the probe (sampling) volume. it is diﬃcult to obtain an acquisition frequency higher than 1Khz. such as laser attenuation and laser diﬀraction.12 tenuation. usually but not necessarily a CCD (Charged Coupled Device). The time resolution of these methods is determined by their sampling frequency.e. PDA. The instantaneous velocity is obtained from a cross-correlation of two consecutive images obtained from two associated laser pulses with a known time lag. The above optical experimental techniques can be divided in two main groups. whose duration is set depending on the velocity of the ﬂow.
holographic techniques and techniques based on the intensity of the scattered light are under consideration. To overcome these problems. no method has been successfully developed up to date to simultaneously measure the velocity and size of the particles. resulting in incorrect correlations. The non-intrusive techniques described above allow the measurement either of the velocity or the size of the particles. Increasing the shutter speed (exposure time) or the time duration of the illumination system as well as decreasing the aperture of the lens to increase the depth of the ﬁeld will solve these problems at the expense of reducing the illumination. Finally. a more through treatment of these two techniques is now presented in the following sections. These instantaneous images can be obtained . Image processing techniques have also been used to study the transient evolution of the concentration P DF of passive scalars injected in turbulent ﬂows . In these techniques it is especially important that the particle images are in focus. at a given ﬂowﬁeld location. In addition. Phase Doppler Anemometry (PDA) enables simultaneous measurements of the size and velocity of the dispersed phase. together with the velocity of the continuous phase. Image post-processing is used to extrapolate the size of the particles from the images recorded. Particle imaging methods have been commonly utilized to acquire instantaneous information on the particle size. The improvement of the signal processing analysis over the last few years makes this technique one of the most appropriate to characterize multiphase ﬂow systems.B Image processing Image processing is used to determine the spatial instantaneous properties of the ﬂow and the particles dispersed in it.13 Inaccuracies in PIV measurements emerge from particles moving out of the illuminated plane in three dimensional ﬂows. Since both image processing and PDA are key elements in the development of this thesis. which will eventually be translated in erroneous size measurements. although advanced correlation techniques are able to discriminate the signal coming from the seeded particles from that of the disperse phase to develop a two-phase PIV. II. since particles out of focus will produce blurred images. shape and concentration.
In the near ﬁeld. The images were taken by illuminating the ﬂow with a diffused white light and capturing the attenuated light by a Sony XC-77R CCD camera placed in front of the source of light at a short exposure time. the break-up of a single bubble under certain conditions. it was not possible to use PDA techniques which require the particle to be spherical and the evolution of the bubbles size probability density function was characterized by means of analyzing images. so that. since the bubbles are not spherical.25 mm2 area of the ﬂow and was digitalized with a resolution of 200 x 455 pixels. in time. The visualizations generated through a laser sheet gave some very good quantitative results on the eﬀect of the ﬂow on the break-up process but it was not used to quantify the size and number of bubbles produced since this technique is limited to a 2-D cut of the ﬂow.14 using diﬀerent techniques.1. was imported to subtract the background level before applying a threshold. was applied in order to detect the particles contained in the image. a pulsed source of light with a short interval of time or a continuous source of light and a short exposure time.14 x 16. • Pulsed diﬀused white light provided by a Strobe light.1. 2 . a threshold. pixels with an intensity value lower than the threshold were assigned a zero To describe the image processing methodology more clearly. a case where the bubbles are nearly spherical after pinching oﬀ from the air needle was selected as shown in ﬁgure II. shown in ﬁgure II. Each image was divided into windows of equal size for later processing. The 768(H) x 493 (V) image was recorded through a 640 x 480 frame grabber on a computer for later processing. 1/80000 seconds. The images stored on the hard drive of the computer were analyzed as follows:2 • The original image. The light sources used during the development of this thesis are: • Continuous laser light from a 5 Watt Argon Ion Laser. In this particular example the image was divided into four independent windows. Once the background was subtracted. Each window consisted of a rectangular section of approximately 7. close to the injection point. close to a white level to eliminate random noise. that synchronized with a high speed video camera was used to follow. • The portion of the images enclosed in each window was then made binary. • Continuous diﬀused white light from a 1000 Watt lamp was used to study time evolution of the bubble size pdf as described below.
15 Figure II.1: Original Image.2: Processed Image. Figure II. .
Similarly. it was possible to compute the bubble size probability density function at several downstream locations in the ﬂow. In the following development.5.5 Equivalent Diameter (mm) 2 Section 1 Section 2 Section 3 Section 4 Figure II. particles whose areas were less than 3 pixels (≈ 100 µm) were also rejected. The diameter obtained from measuring the area and also from an alternative technique based on measurements of the perimeter of the contour of particle are represented as a function of their real diameter in ﬁgure II. as shown in ﬁgure II. particles that were touching the edges of the windows were rejected to avoid sizing errors. thereby limiting the resolution of the system to particles larger than 100 µm.4. value and pixels with an intensity value higher than the threshold were given a value of one.5 remains always smaller than 15% in both cases and drops to less . The equivalent diameter of the particle was then calculated assuming a spherical shape.2. To analyze the accuracy of the image processing routine used. the area occupied by each particle was easily determined by counting the number of enclosed pixels. As can be seen. corresponding to each section or window. along with their associated errors. From the binary images.16 10 8 PDF(D) 6 4 2 0 0. shown in ﬁgure II. By processing a large number of images as described above and calculating the size of the particles detected in the pictures.5 1 1. the error of the diameter calculated by the processing program.3: Bubble Size Probability Density Function. according to Dp = 2 A π. a set of particle diameters ranging from 71 µm to 2 mm was computationally generated as shown in ﬁgure II.3. detecting in that way any particle contained in the image as shown in ﬁgure II.
optical technique that processes the light scattered by a spherical particle as it crosses through a probe volume .C PDPA Phase Doppler Anemometry (PDA) is a non-intrusive.5: Size of particle measured by the analysis program as function of the size of the particle. (•) represents size obtained from the measured area. ( ) represents the size obtained from the measured perimeter of particle and. II. 25 20 Rm (pixels) 15 10 5 Error (%) 15 10 5 0 0 5 10 R (pixels) 0 -5 -10 -15 15 20 25 0 5 10 15 20 25 Radius (pixels) Figure II.17 Figure II. than 5% for particles of radius larger than 5 pixels (around 350µm) diameter. Since the measurement technique based on the eﬀective area is considerably more accurate. this method was selected to process the results presented in this thesis.4: Particles generated from Radius= 1 pixel to Radius= 25 pixels. resulting in error below 2% for particles of diameter larger than 280 µm.
aligned in the plane of the mean direction of the ﬂow (vertical). The PDA used for this work was a Phase Doppler Particle Analyzer (PDPA) manufactured by Aerometrics. focused two sets of orthogonal beams into a probe volume. A X θ Y Probe Volume dx dy dz Figure II. polarization preserving optical ﬁbers connected the Fiber Drive to the XMT204 transmitter. were used to measure the axial component of the velocity and size of the bubbles.18 formed at the intersection of two monochromatic. all coupled into a Coupler Assembly. provided a measurement of the . perpendicular to the ﬂow. at the exit of the Fiber Drive there were two unshifted beams (green and blue) and two shifted beams (also green and blue). The incoming beam was directed through an acousto-optics modulator (Bragg Cell) which split the beam into two beams of the same intensity. The two green beams. The two beams were directed into a color dispersion prism that produced four monochromatic laser beams. as shown in ﬁgure II. The transmitter had a 50 mm clear aperture and. provided with a 500 mm focal length. Therefore. one with the same frequency as that of the incoming beam and the other (shifted beam) with a frequency 40 MHz higher. The two blue beams. coherent laser beams. The multi-line beam light coming from a 2 Watts Innova 70C Coherent Argon Ion laser was driven into the Fiber Drive (FBD240) of the PDPA system.6.6: Dimensions of the probe volume 10m cable containing 4 single mode. Each individual beam was then picked oﬀ and directed to a diﬀerent exit of the Fiber Drive depending on its wavelength and frequency shift. two blue beams of wavelength 488 nm and two green ones of wavelength 514 nm.
. the probe volume located at the focal point of the lens used in the transmitter is an ellipsoid of the following dimensions: dx = dw = 467 µm. cos(θ/2) dz = dw = 467 µm. The corresponding crossing angle at the probe volume was θ/2 = 1. Figure taken from Crowe et al. dy = dw = 20 mm.7: Angular intensity distribution of diﬀerent scattering modes of bubbles in water. and the focused beam waist dw = 4f λ/πdl ≈ 467 µm. For the above conﬁguration.6. far from the forward direction.7 mm.75. indicated in ﬁgure II.14o . sin(θ/2) The light scattered by particles crossing the probe volume is composed of diﬀraction. The diﬀraction component of the scattered light can be minimized when detected from an angle..  radial component of the velocity. For the case of air bubbles in water.19 Figure II. ψ. dashed line is primary refraction (p1 ) and dotted line is secondary refraction (p2 ). where f is the focal length of the transmitter lens. reﬂection and refraction . λ is the wavelength of the laser beam and dl represents the beam diameter at the exit of the transmitter. Solid line is reﬂection (p0 ). whose relative index of refraction is m = na /nw = 0. The beam separation at the exit of the transmitter lens was 20 mm and the beam diameter 0.
In the case of experimental facilities where there is an interface between two diﬀerent media separated by a wall. In order to have accurate measurements.20 Figure II.8: Mie calculations of bubbles Phase-Size relationship for diﬀerent scattering angles. it is very important that the receiver be installed perpendicular to the wall. as shown in ﬁgure II.8. show that the relationship between the phase shift and the diameter of the bubble remains linear for scattering angles as small as 55o . Figure taken from Crowe et al.  the most important scattering mode is reﬂection when acquiring the scattered light at an angle ψ in the range 50o and 80o . Bachalo. as seen in ﬁgure II. . and to satisfy at the same time the restrictions imposed by the experimental facility. as described in the following sections. we chose to work with the 60o scattered light. as described in detail in . the optimum scattering angle to capture the reﬂective light seems to be between 70o and 80o . Although. Mie calculations included for completeness in ﬁgure II.. the RCV208 receiver unit was then placed at an angle of 60o from the plane formed by the set of the green beams (vertical plane) and in the plane of . Therefore. as is the case of a water tank.9. an angle that guarantees a linear relationship phase-diameter.7. showed that the diameter of the particle is proportional to the phase shift between the incident and scattered light when one of the scattering modes is dominant.
Three of the photodetectors capture the green component of the scattered light collected on the three distinct regions. For particles moving at velocities much slower than the speed of light (U ≪ c). δf (II. fD = U . which provides a higher quality signal with a higher Signal to Noise Ratio (SNR) than smaller aperture receivers such as that documented in previous work by Kiger . the light detected from three. crossing at angle . δf = λ/2sin(θ/2) is the fringe spacing of the interference pattern formed by the intersection of the two laser beams. the velocity component perpendicular to the bisector of the two incident beams is given by.21 350 Phase (degrees) 300 250 200 150 100 5 0 0 0 100 200 300 400 500 D i a m e t e r (µm ) Phase 12 Phase 13 Figure II. thereby reducing the length of the probe volume to dzef f = 150 × 500/250/sin(60o ) = 346 µm. well-deﬁned. Similarly the fourth photodetector captures the blue component of the light to determine the radial component of the velocity. The scattered light by a bubble crossing through the probe volume is acquired by the 72-mm clear aperture receiver with 500-mm lens.1) where U is the velocity of the particle. ψ = 60o . ). After being color ﬁltered. Subsequent processing enables the calculation of one component of the velocity (streamwise component) and also of the size of the bubble (. The signal is focused through an internal 250 mm lens into 150 µm spatial ﬁlter.9: Phase-Diameter relation of Bubbles. incidence of the two blue beams. as in a LDV system. diﬀerent regions of the receiving lens is sent to four photodetectors. A brief description of the measurements of velocities and size of particles will be presented in the following paragraphs.
Both processors have a burst detector in the analog and frequency domain (FTBD) to be able to validate signal with SNR lower than 0dB while rejecting possible coherent noise. Therefore. the scattering mode (p = 0 reﬂection. depending on the burst length. p = 1 ﬁrst refraction. geometry) . λ (II. As described by Ingrahain et al. . A summary of the similarities and diﬀerences between the DSA and the newer RSA processors is described below.9. the RSA performs a large number of DFTs ( Direct Fourier Transforms) at once. .22 θ at the probe volume as seen in ﬁgure II. The signal from the photodetector is sent to the analog processor and processed as described in . D. The phase shift between the Doppler signals detected by two diﬀerent detectors provides a measure of the size. . A two-photodetector system enables only the measurements of phase shifts between zero and 2π. of the moving droplet (bubble). from the comparison of the measurements of pairs of detectors having diﬀerent separations. The ﬁrst approximation of the diameter is obtained from the phase shift of two detectors (φ12 ).2) The parameter χ(m. p = 2 second refraction. providing two phase diﬀerences. the error increases for a certain range of SNR (Signal to Noise Ratio) of the burst. A more reﬁned measure of the diameter of the particle is calculated from the shift of a second pair of detectors (φ13 ). The processor used is a RSA (Real time Signal Analyzer).6. m. An example of such measurements is shown in ﬁgure II. The main diﬀerence between the DSA and the RSA processor is that. younger brother of the DSA (Doppler Signal Analyzer) whose operation has been very well documented elsewhere . and fD is the frequency Doppler detected by the photodetector. as described in : △φab = 2πD χ(m. geometry) depends on the ratio of index of refraction of the two media. Therefore a three-detector system is commonly used. the number of samples used to perform the DFT and compute the Doppler frequency is . Comparison between the diameter of the particle extrapolated from (φ12 ) and that from (φ13 ) gives a ﬁnal measure of the size of the droplet if both values are within a certain tolerance or the measure is rejected otherwise. . p. p. The quadrature downmixing performed by the RSA processor is also sampled using 1-bit sampling to accelerate the processing time. φ12 and φ13 . while the DSA is basically a sampler and relies on a computer to do all the data processing. ) and the geometrical conﬁguration of the receiver.
adequately satisﬁed for the bubble diameter range present in this work. Bachalo and Sankar. . Those oscillations can be minimized and even eliminated . thereby invalidating the present method. Furthermore. for the Argon ion laser source used in this study of λ = 514 nm.C.1 Sizing of small particles The principle of geometrical optics used to compute the light scattered by par- ticles and its applications in this thesis is based on the assumption that the particle diameter is larger than the wavelength of the laser light source λ. D. Special attention has been given to obtaining reliable measurements on air-water systems II. it is not an error-free technique and it could produce misleading results when improperly used. and also that one of the modes is the dominant component of the light scattered. in the RSA all the processing is done by the processor and sent to the computer for display afterwards. . The following sections summarize some of the inherent problems faced when using Phase Doppler Anemometry and suggestions to overcome such problems. the linear relationship between the phase shift and the bubble diameter is. The applicability of the principle of geometric optics is restricted to relatively large particles of diameter. such that 1 ≪ α = πD/λ. The RSA processor normally performs four DFT in velocity measurements and up to ﬁfteen DFT in sizing measurements for each burst. For instance. The algorithm uses the four greatest number of samples obtained while the DSA uses a ﬁxed number of samples chosen by the user to perform the DFT.23 optimized in real time. have shown that the interference between the reﬂecting and refracting modes gives rise to oscillations in the phase-diameter calibration curve of the PDPA. Since the diameter of the smallest bubbles measured in our experiments is 10 µm. This approach implies that no interference of the scattering modes is considered. therefore. Unlike the DSA. providing an averaged value if the variance of the collected data is within a certain tolerance. the minimum particle size that can be accurately measured is approximately 2 µm. Although PDA has been accepted as one of the most reliable techniques to simultaneously measure the size and velocity of particles in turbulent ﬂows. while for smaller particles diﬀraction becomes an important component of the scattered light. resulting in greater accuracy and resolution in every burst.
using PSL (polystyrene latex) ﬁne particles of diameter smaller than 10 µm. However. very large particles were erroneously reported as even larger ones. an incorrect.C. and eﬀectively .. They showed that the strong oscillations appearing in the calibration curves for certain conﬁgurations were the result of erroneous calculations based on calculating the phase at single points on the receiver surface instead of integrating along the entire surface. Sankar et al. since the incident beams have a gaussian intensity proﬁle. depending on the particle trajectory across the probe volume and on the probe volume size. farther away from the receiver lens. In particular they proved that increasing the beam intersection angle changed the phase relationship between reﬂected and refracted modes. The receiver lens readily removes these oscillations by performing a spatial integration of the scattered light collected in the selected area.20. diluted in water. . and . unexpected scattering mode may become dominant and lead to erroneous diameter measurements. solving the Lorenz-Mie theory and performing spatial integration of the light scattered over the three diﬀerent collection areas occupied by the three detectors. demonstrated that the uncertainty of the measurements decreases from ±1 µm to ±0. Sankar and Bachalo  analyzed the eﬀect of the particle trajectory by using a geometrical optics approach and taking into account the nonuniform illumination experienced by the particles as they move through the measuring volume. They showed how errors were minimized and even eliminated by choosing the proper optical conﬁguration and also by increasing the ratio of beam diameter to maximum particle size. of relative index of refraction m = 1. Similarly. and therefore simulating the light detected by the three distinct regions of the collecting lens.24 by choosing a proper PDPA conﬁguration. They showed that water drops of diameter comparable to that of the focused beam were reported as smaller particles when they crossed the probe volume through its outer part. II.4 µm when the mean collective angle is increased from 20o to 70o . . as described in references .2 Eﬀect of Gaussian intensity proﬁle of the incident light source As described in the previous section the correct application of the PDPA is based on using intensity of the dominant scattering mode to calculate the appropriate phase-diameter calibration curve. Calibration curves were generated by Sankar and Bachalo.
8o to 5. similar to the one proposed by Gr´han.25 overcame the nonuniform illumination eﬀect. with the eﬀect being less pronounced in the case of bubbles. reﬂection is the dominant mode. where the intensity of the beam is higher due to its gaussian proﬁle. the sizing error was decreased from 50% to 10% for particles crossing through the edge of the measuring volume as presented in . • Using Intensity Validation to reject particles which pass through the edge of the probe .75). φ13 . dominated by reﬂection (refractive index of 0. As a solution that produces the smallest trajectory eﬀects. the refracted component becomes dominant in this case. in contrast with the case of water droplets moving in air where it underestimates their size. the refracted light emerges from the inner portion of the volume. The measurement is validated if the values of both diameters agree. φ12 .4o . For instance. the ambiguity error in our experimental facility was minimized by combining the following three methods: • Optimizing the crossing angle of the incident laser beams and using an optical conﬁguration of 60o oﬀ-axis. e • Using two set of photodetectors to perform phase validation. Similarly. they proposed an optical conﬁguration of the receiver unit of 70o oﬀ-axis. studied the trajectory-ambiguity eﬀect in air bubbles moving e in water conﬁgurations. When a bubble crosses the probe volume through the edge closer to the receiver. These errors caused by inhomogeneous illumination are relatively small for small bubbles and also when the size of the probe volume is large compared with that of the particles. A combination of this method with the one described in the following point has been proved to reduce the measurement errors. by increasing the beam crossing angle from 1. This method is not completely eﬀective by itself as reported by Bachalo . In the case of bubbles for a collecting angle of 60o used in this work. After reviewing the published literature. Since the phase-diameter correlation has been calculated assuming a reﬂected dominant mode. Therefore. . Gr´han et al. They showed that trajectory ambiguity overestimates the bubble diameter. The diameter is measured from the phase shift of two detectors. This value is compared to the diameter measurement obtained from the phase shift of another pair of detector.. resulting in errors smaller than 10% in all case. the reﬂected component comes from the outer portion of the volume where the light intensity is lower. the associated measurements would be erroneous.
.10: Planar Phase Doppler Anemometry volume producing wrong measurements. the DPDA is shown in ﬁgure II. The Dual Phase Doppler Anemometry (DPDA) developed by Tropea et al. D4 Modified Standard PDA : D1. it is very instructive to describe one of the most recent systems developed to eliminate trajectory errors. A reported large bubble with a very low intensity could be a erroneous measure of a small bubble crossing through the edge of the probe volume. consists of a set of four photodetectors integrated in a single receiver unit combining a Planar PDA (PPDA) with an Modiﬁed Standard PDA (MSPDA) conﬁguration. Therefore. their contributions come with time-delay and the refractive component has a higher amplitude. Although both the reﬂective and refractive modes of the scattered light are collected. Combination of the size measured by the two conﬁgurations produces a trajectory error-free measurement validating only particles . the signal from the refractive mode can be distinguished from that of the reﬂective one so that sizing error can be eliminated. Although it has not been used in the present work. The other two detectors are positioned in a matter similar to that of a standard PDA although the beam intersection plane is perpendicular to the main direction of the ﬂow in contrast to standard PDA where the main direction of the ﬂow is parallel to the beam intersection plane.10. Two of the detectors are located in the plane of the intersecting laser beam containing the axis of the main direction of the ﬂow. D2 Figure II.26 Z Flow Direction Plane XOY O Plane ZOY D1 D3 D4 X D2 Receiver Y Planar PDA: D3.
1 . Vm (II. Vm . if a characteristic volume mean diameter of the bubble .05. II. If the probability of ﬁnding n particles in the probe volume follows a Poisson distribution given by: Pn = N n e−N .3 Particle number density limitations The Phase Doppler method requires that only one particle be detected in the probe volume at a time to avoid overlapping of doppler signal from diﬀerent particles. the maximum number of par- ticles has to be N= 0. The diameter of the particle is determined from the reﬂective mode since it is independent of the particle index of refraction. therefore.5) Considering the worst case of the experiments reported in this thesis where the ﬂow rate ˙ of air injected is Qa = 70ml/min. n! (II.27 for which the size obtained using the MSPDA agrees with the one determined using the PPDA within a certain tolerance. for a given concentration of particles. the optimal concentration of particle in the ﬂow must be: cpopt = 0. P2 = 0. to minimize the probability of having more than one particle in the sampling volume at the same time.C. The relative probability of ﬁnding two particles is: ∗ P2 = P2 N = . and cp is the concentration of particles in the ﬂow.3) where N = cp Vm is the number of particles found in the probe volume. this system has been used to measure particle index of refraction combining the phase measurements detected from reﬂection and refraction in SPDA conﬁguration . the particle index of refraction can be obtained from the phase of the refractive mode. Knowing the diameter and the optical parameters. P1 2 (II. The size of the probe volume has to be deﬁned. If two particles of diﬀerent size are crossing the measuring volume at the same time. the signal processed to calculate the diameter will have contributions from both particles and must be rejected.4) ∗ For a maximum coincidence tolerance of 5%.1 and. In addition.
32( X )Qo . the amount of liquid entrained by the water jet can be calculated by Qe ≈ 0. the optimal concentration of bubbles obtained from equation II. the number of bubbles generated per unit time is: ˙ Qa 6 4 particles ˙ . velocity and concentration has been obtained using high speed imaging techniques. Bre˜a de la Rosa et al. therefore.28 distribution is D30 = 200 µm.4 Measurement of size of bubbles The PDA technique has been successfully used to size water droplets in air for many years. developed a theoretical and n experimental study of the characterization of bubbles using light scattering interferometry. where the bubble size distribution is measured. the mean concentration of particles in the ﬂow is ˙ N Q = 0.5×10−12 m3 . the Doppler signal has been visualized and controlled on an Oscilloscope in order to obtain a single burst and good visibility to guarantee reliable results. included in our system.61×109 particles/m3 . Not much work has been dedicated to light particles like air bubbles in water using PDA. With the new processors. providing that the burst emitted by two diﬀerent particles do not overlap. Qo = 63 × 10−6 m3 /s. For the D case of the minimum ﬂow rate utilized. the total ﬂow rate at 20 diameters is Q = 441 × 10−6 m3 /s and. Therefore cpopt ≥ ˙ N Q in all the cases tested and the probability of ﬁnding more than one particle in the probe volume is kept smaller than 5% having a very low rejection rate in our measurements.5 is cpopt = 1. it has been possible to measure the size of particles even when more than one are present in the sampling volume by analyzing diﬀerent portions of signal. Additionally.6) At twenty jet diameters. In all the experiments reported here. . more recent developments in signal processing have been developed to measure the signal at extremely high rates using fourier transform. This study was devoted to develop calibration curves for diﬀerent conﬁgurations of .33× 109 particles/m3 . On the other hand.C. Until now. most of the data related to bubble size. N= 3 = 27 × 10 second πD30 (II. a technique which is very limited and time consuming. and for a characteristic size of the probe volume of Vm = 75. II..
among experimental measurements concerning microbubbles of diameter from a few microns to several hundred microns are those by Rightley. .  and 75o . was used in a posterior study n by Bre˜a de la Rosa et al. although.. and by Gr´han at al. The later shows that the trajectory errors.11: Standard PDPA Conﬁguration the Phase Doppler Particle Analyzer (PDPA) in order to ﬁnd an optimal conﬁguration. . to analyze the morphology of spheroidal bubbles using n light scattering interferometry. Theoretical studies show that scattering collecting angles higher than 55o can be used to measure the size of bubbles using a PDPA. discussed earlier. Since the reﬂective component of the light scattered by air bubbles in water is dominant at some scattering angles (Figure II. The conﬁguration proposed by Bre˜a de la Rosa et al. the diﬀerence of phase between the incident light and the reﬂected light is directly proportional to the spherical scatterer as shown by Bachalo and Houser. The fore mer compares the measurements obtained with a PDPA with those done using a Malvern diﬀraction system resulting in a good agreement when there is coincidence between the probe volume formed by the two sets of laser beams (blue and green) used by the PDPA. .29 Z Z' in plane YZ Flow Direction D2 D1 D3 X Beam 1 O θ ψ= 60 o Beam 2 Y X Figure II. are eliminated when mea- . and plotted in Figure II. some experimental studies often use a conﬁguration with a receiver unit placed at 70o .8.7). Most of the above work is related to size of bubbles of the order millimeter. . .
11 and II.12 (plan view) and whose description is presented in the following paragraph. D1 D2 and D3 are the three diﬀerent regions of the lens capturing the scattered light and focusing it into the respective detectors.13: Distribution of Photodetector in the Receiving Lens. Therefore. after an extensive review of the bibliography and analysis of the most common problems associated with the measurement of the size of bubbles (described in the previous sections) the conﬁguration chosen for this study was the one sketched in ﬁgures II. suring bubbles at a receiving angle of 70o .30 ψ= 60 o Y Laser Beams Z Receiver Transmitter Figure II.12: PDPA Conﬁguration: Plan View Z' X D1 D2 D3 Figure II. containing the . The vertical plane YX. respecting at the same time the optimal optical conﬁguration described above. The water tank was designed in an hexagonal cross-section in order to have both the transmitter and the receiver perpendicular to walls and avoid any undesired scattering eﬀect due to the presence of the interface.
the value of the receiver focal length was corrected using Snell’s law to take into account the path of the scattered light changes due to the changes in the index of refraction of the two propagating media. The receiver unit was placed on the horizontal plane YZ at an angle ψ = 60o oﬀ-axis from the transmitter axis. must be parallel to the plane of the fringes formed in the probe volume (which lay in the YZ plane) to obtain a linear calibration curve between the phase shift and the size of the bubble. The scattered light collected on three diﬀerent areas of the receiver lens was sent to three photodetectors for later processing as described earlier. was formed by the intersection of the two green beams. The three areas were parallel to each other as shown in ﬁgure II. . as seen in ﬁgure II.13 and.11.31 main direction of the ﬂow. Finally.
isotropic and nearly in equilibrium turbulent ﬂow. This type of ﬂow.1. which makes it more suitable to produce a better understanding of the physical problem.e. Most of the experiments have been done in stirred tanks. Their experimental results have produced very useful data. To study the particle’s break-up into a homogeneous. commonly used within the chemical engineering community. free of solid boundaries. but they have been very diﬃcult to understand due to the complicated type of turbulence produced in this type of experiments. consisted of a submerged water jet where air was injected at a given position along its centerline. In order to maximize 32 . knowledge of the turbulent characteristics is still very poor. i. mainly for their practical applications. a more controlled and much better understood type of ﬂow was chosen: an axisymmetric turbulent water jet. shown in ﬁgure III. III. The ﬂow is highly anisotropic consisting of strong shear regions near the blades where the break-up mainly takes place.Chapter III Experimental Facility and Flow Conditions The turbulent break-up problem has been extensively studied in various experimental conﬁgurations. Although several correlations have been developed to account for this eﬀect. the blades used in the stirred tanks.A Experimental Set-up The experimental facility. does not have the interaction of the particles with walls or any turbulence generator.
the accuracy of the phase Doppler and other optical measurements. the water was allowed to overﬂow from the top of the tank through a set of gutters placed on each side. U0 . Re = U0 DJ /ν. Uniform velocity was achieved at the exit of the nozzle in the submerged water jet by the use of two perforated plates located upstream of a high-contraction-ratio (250:1) nozzle.2.1: Experimental Facility. Although diﬀerent nozzle exit diameters could be used.1 mm. The water jet nozzle was located at the bottom and the jet discharged vertically upwards into the tank.8 m PDPA Transmitter Receiver Figure III. as seen in ﬁgure III. in all the experiments reported here the nozzle exit diameter was 3. on the diameter of the nozzle DJ and on . The jet Reynolds number. based on the velocity at the exit of the water jet.33 PDPA 10 mm 60 0 Receiver Transmitter 710 mm 10 mm 1. the tank in which the jet discharged was designed with an hexagonal cross-section. In order to minimize the recirculating ﬂow produced in the tank by the entrainment of the high momentum water jet.
2: Detail of the Jet Nozzle. could be varied along the axis of the water jet from 10 to 50 jet diameters by moving the needle vertically. which determines the value of the 3. could be systematically varied up to 105 .1 mm Needle Water Input Water Input 50 mm Figure III. turbulent kinetic energy of the underlying turbulence where the bubble break-up process starts. The bubble injection point. These positions were several diameters downstream from the end of the potential core region of the jet in all Reynolds number cases studied and the turbulence was fully developed in the scales of interest to our problem. To avoid any undesirable vibration eﬀects at the air injection point. ν. however. Air was injected coaxially on a selected downstream location at the axis of the submerged turbulent water jet through a small hypodermic needle.34 the kinematic viscosity of the water. All the experiments presented in this work. the needle was supported at the crossing points with perforated plates. corresponded to injection points located between 10 and 25 jet diameters downstream from the nozzle exit section. 130 mm Perforated Plate 40 mm .
the dissipation rate of turbulent kinetic energy.4 m/s. From the one-dimensional spectrum of the ﬂuctuating component of the axial velocity. could be increased from 5. Under this conditions the Kolmogorov scale is η = (ν 3 /ǫ)1/4 ≈ 6 µm.35 The water ﬂow rate. III.B Flow conditions The axisymmetric jet has been extensively studied experimentally. analytically as well as computationally during the past years.1 mm diameter nozzle.1 is λt = 277 µm and the corresponding Reynolds number is Rλt = 443. and Townsend . Rλt . longitudinal Taylor microscale. Lx . measured at X/DJ = 20. estimated from equation III.88 (Re )−1/2 X ≈ 240 µm.0 10−2 . ν (III.6 m/s and ǫ ≈ 1000 m2 /s3 . the ﬂuctuating velocity is approximately u′ = 1. the Taylor microscale. ′ 30 u¯2 ν u′ λt . Qw . could be varied from 6. resulting in the initial void fraction of air at the point of injection α = Qa /Qw . Qa .13 × 10−5 m3 s−1 to 2. λt . can be estimated as: Lx = π E11 (k1 = 0) . the longitudinal integral scale. to 1. The ﬂow rate of air.5 10−4 to 2. for a local velocity value of 6. ǫ. Uo . for the 3. and considering the turbulence to be locally homogeneous and isotropic. The work of Wygnanski and Fiedler  has become an standard reference for quantitative descriptions of .25 × 10−6 m3 s−1 . For example. These values are consistent with those calculated using the correlations proposed by . 23].8 ×10−8 m3 s−1 ǫ = 15 ν 1 λ2 t Rλt where u = ′ = = ǫ . ′ 2 u¯2 ∞ 0 2 k1 E11 (k1 )dk1 . Important information has been collected in the reviews by Monin . Hinze . 25. a range of exit velocities. λt = 0. ranging from 2.3 × 10−4 m3 s−1 providing. and the Reynolds number based in the Taylor scale. and E11 is the one-dimensional spectrum of energy [29. from 8 m/s to 30 m/s.1) u′2 is the rms of the streamwise component of the ﬂuctuating veloc- ity.
14 12 10 U o/Uc 8 6 4 2 0 0 10 20 30 X/D 40 50 60 Uo/Uc Bu=4.08, xo=3.78
Figure III.3: Centerline Velocity. the mean velocity proﬁles and turbulent properties of jets. The centerline axial velocity for a self-preserving jet is given by : 1 X X0 U0 = ( − ), Uc Bu DJ DJ (III.2)
where U0 is the velocity at the exit of the nozzle, Uc is the centerline velocity at a distance downstream X/DJ , Bu is an empirical constant which determines the rate of decaying of the axial velocity and X0 is the virtual origin. The values of Bu and X0 may depend on the exit conditions as pointed out by Hussein et al.. The measured centerline velocity has been plotted in ﬁgure III.3 as a function of the axial position, X/DJ . The axial centerline velocity, Uc , has been normalized with the exit velocity, U0 and the downstream coordinate by the diameter of the nozzle, DJ . The experimental measurements have been compared with the equation III.2. A downstream rate of decaying of the centerline velocity with a constant Bu = 4.08 and a virtual origin, deﬁned as the distance between the origin and the interception of the straight line with the x-axis, X0 /DJ = 3.78 has been found. These results are in a good agreement with the constants presented by previous investigators ,. The
1 0.8 U/Uc 0.6 0.4 0.2 0 0 0.05 0.1 r /x 0.15 0.2
6D 10D 15D 20D 30D
Figure III.4: Similar proﬁles of streamwise velocity. centerline velocity velocity follows the characteristic 1/X decaying law of round jets for downstream distances X/DJ > 10. The mean axial velocity normalized by the centerline velocity, U/Uc , is represented versus the non-dimensional radial distance, r/x, in ﬁgure III.4. The mean velocity proﬁles collapse at downstream distances of X/DJ > 10, giving a radial position for which the axial velocity has decayed to 50% of its value at the centerline, (r/x)1/2 ≈ 0.1. Since the mean rate of dissipation, ǫ, will be necessary in the following chapters to characterize the bubbles break-up, it was measured for several ﬂow conditions at diﬀerent downstream positions along the axis of the jet to study its evolution. The energy spectrum, shown in ﬁgure III.5, was calculated from hot ﬁlm measurements applying the Taylor hypothesis to convert from the temporal domain (frequency) to the spatial domain (wave number), k = 2 π f /U . Notice that the length scale associated with the length of the ﬁlm, Lf , falls within the inertial subrange of the energy spectrum. Therefore, the spectrum drops oﬀ at a wave number slightly larger than that associated with the length of the probe and, it was not possible to resolve smaller scales of the ﬂow. In order to overcome this problem, a ﬁrst attempt to correct the measured one-
0.01 0.001 -5/3 0.0001 E (m3/s2) 10- 5 10- 6 10- 7 10
Lw Energy Spectrum Corrected Energy 1 10 100 1000 k=2πf/U (1/m) 104 105
Figure III.5: Energy Spectra: △ Measured Spectrum, ◦ Wyngaard Corrected Spectrum. Lf indicates the wave number associated with the length of the ﬁlm. X/DJ = 20, Uo = 12m/s. dimensional spectra was done by using Wyngaard’s correction, , represented in ﬁgure III.5. Although Wyngaard’s correction has been successfully used in many cases, the values of ǫ obtained using equation III.1 with E11 being the corrected spectra were much smaller than expected. An alternative procedure, described below, was applied. To calculate the dissipation rate of turbulent kinetic energy, ǫ, an alternative method was applied. Since the inertial subrange can still be identiﬁed in the measured energy spectra as shown by the -5/3 line in ﬁgure III.5, it can be used to infer the dissipation rate, ǫ. At high enough Reynolds numbers, the inertial range obeys the following power law : E11 (k1 ) = 18 8 2/3 2/3 −5/3 ( ) ǫ k1 , 55 9α (III.3)
where E11 (k1 ) is the one-dimensional spectrum, k1 is the wave number in the direction of the ﬂow and α = 0.45 is an empirical constant given by Gibson in a high Reynolds number turbulent jet . Once the inertial range has been identiﬁed, we can ﬁt an energy spectrum given by equation III.3 only into the region of the measured spectrum which
follows the -5/3 power law. l0 (III.3 is an equation with only one free parameter.4) .001 0. Figure III.01 0. ǫ. used to extrapolate the value of the dissipation.39 0. is eventually removed by viscous dissipation.7 for two diﬀerent exit velocities.5 10 -6 Lf 10. Lf indicates the wave number associated to the length of the ﬁlm. Measurements of ǫ along the axis of the jet are presented in ﬁgure III.0001 E (m3/s2) 10. Since at high Reynolds numbers the turbulent kinetic energy.8 10. 000 and Re = 53. which can be adjusted until the error obtained during the ﬁtting process is minimized.5 calculated averaging a larger number of blocks to get a higher resolution in the inertial subrange. The dark line indicates the best ﬁtted spectrum using equation III. U0 = 12m/s and U0 = 17m/s.3. Re = 38. Equation III. which correspond to Reynolds number.7 10. 000.6: Energy Spectrum: Inertial Subrange is identiﬁed to measure the dissipation rate. Therefore: ǫ∝ u3 0 .6 shows the same energy spectrum presented in ﬁgure III. 000 respectively.9 10 100 1000 k=2πf/U (1/m) 104 105 Figure III. Notice that the repetitiveness in the calculation of ǫ is very good when comparing the two set of experiments at Re = 38. ǫ. which cascades down from the large scales to the smaller ones. ǫ can be estimated by measuring the energy ﬂux across the large scales .
40 104 Dissipation Uo=17 m/s Uo=12 m/s Uo=12 m/s 1000 ε (m2/s3) 100 10 1 10 X/DJ 100 Figure III. .6 10 X/DJ-Xo/DJ 100 Figure III.001 ε DJ /U03 0.5 10.01 Dissipation Uo=17 m/s Uo=12 m/s Uo=12 m/s 0.7: Evolution of Dissipation Rate of Turbulent Kinetic Energy. ǫ.0001 10. 0.8: Normalized Dissipation Rate of Turbulent Kinetic Energy. ǫ.
and the characteristic length. equation III.3 and III. DJ DJ (III. .7. . since Uc /U0 ∝ (X/DJ − X0 /DJ )−1 and r1/2 ∝ (X − X0 ) as shown in ﬁgures III. The solid line in ﬁgure III.4 can be rewritten as: 3 ǫ DJ /U0 ∝ ( X0 −4 X − ) . In the case of round jets.4. Uc . (r1/2 ).5) This equation was derived by Friehe et al.7 represents a correlation of ǫ measurements based on the scaling argument given by equation III.4.8 will be used later on in this thesis to estimate ǫ for the various ﬂow conditions. The results presented in ﬁgure III.5 for the experimental measurements shown in ﬁgure III. and compared with experimental data. Figure III. if the characteristic velocity is deﬁned as the centerline velocity.41 where u0 and l0 are the characteristic velocity and length of the turbulence respectively.. as the radial distance at which the velocity is Ur1/2 = Uc /2.8 shows the evolution of ǫ normalized as in equation III.
buoyancy eﬀects can be shown to be negligible since the terminal velocities of all the bubbles were always an order of magnitude smaller than the mean velocity of the jet. i. we conducted a number of experiments in which we varied systematically both the initial bubble size and the value of the TKE of the underlying turbulence at the location of the air injection point. In all cases. we followed the evolution of their break-up as they were convected downstream by the mean axial velocity of the jet to regions of lower and lower turbulent kinetic energy. 23]). agreed with those reported in the literature. performed for all ﬂow conditions. The desired initial value of the turbulence kinetic energy was selected by positioning the air injection needle at a given downstream distance from the nozzle of the water jet.A Experimental approach In order to measure the break-up frequency of the bubbles as a function of their size and of the turbulent kinetic energy (TKE) of the underlying turbulence. All experiments were performed in the following way. since we know both.e.1 (our measurements. Notice that. After injecting the air bubbles. the selected downstream distances were greater than or equal to 15DJ to ensure a region of fully developed turbulence. we performed measurements of the transient bubble-size pdf s at 15 discrete downstream regions (or windows) as indicated in ﬁgure IV.2. During this evolution. 42 . ǫ) decays with the downstream distance as shown in ﬁgure IV. Since in the water jet the turbulence kinetic energy (or the dissipation rate. [1. the residence time and the value of the underlying turbulent kinetic energy in each window.Chapter IV Break-up Frequency of Bubbles IV.
Since bubble sphericity is a necessary condition for the use of phase Doppler techniques. it could be assumed to be nearly constant. the velocity at the exit of the nozzle is Uo = 17m/s. X0 ≈ 4 DJ indicates the virtual origin. The length of each window along the axial direction was chosen such that ǫ varied a maximum of 10% along its length and. where bubbles are not spherical. After measuring the bubble’s size at each downstream location. they were then classiﬁed into 10 diﬀerent size-bins. Phase Doppler . can measure all the needed parameters in each experiment in order to be able to calculate the bubble break-up frequency as a function of the bubble’s size and ǫ. we also varied both the diameter of the injection needle. A summary of the various ﬂow conditions selected for our measurements is given in table IV. with their turbulent Weber number less than unity. and the bubbles are nearly spherical. resulting in a variation of the initial bubble diameter Do .1. This discretization allowed us to measure the rate of decay of the number of bubbles of a certain class size as a function of their size and of the local value of ǫ. therefore. and the air injection velocity.43 105 104 1000 ε (m2/s3) 100 10 1 0. ǫ. Da . the evolution of the bubble-size probability density function in the region of interest in our experiments. we. was always measured by means of digital image analysis. Only when the break-up is ﬁnished. in addition to varying ǫ by changing the location of the injection point. Note that in our experiments. Furthermore. Ua . thus.1 10 X/DJ-Xo/DJ 100 Figure IV.1: Downstream evolution of the dissipation rate of TKE. the cross-stream width of the window was always selected to be large enough to include all the region where the bubbles resulting from the break-up were dispersed by the turbulence.
From this. Each image was then divided into ﬁve windows of 7. which resulted in a minimum measurable bubble diameter of 80 µm. diﬀused white light and capturing the image with a Sony XC-77R CCD camera placed in front of the light source using a short exposure time of 1/80.500 25.25 mm and was digitalized with a resolution of 200 x 455 pixels.84 9.2.584 1.000 seconds. To increase the spatial resolution of the discretization.500 Injection point X/DJ 25 15 15 15 15 Table IV. a bubble-size histogram was measured from each of the 15 .000 frames.3 cm x 1.88 9. binary images as shown in ﬁgure IV. To maximize the resolution of our measurements.000 25.48 1.194 U0 (m/s) 17.500 25.84 4. and the open section at the exit of the nozzle. we recorded 1. the camera was focused in a small region of 2. thus allowing for the calculation of an equivalent bubble diameter.1: Experimental Conditions. The images were taken by illuminating the ﬂow with uniform.394 0.07 Set1 Set2 Set3 a) b) c) 51. This open section was kept constant in the three cases of Set 3. Each frame was ﬁrst subtracted from the background illumination and subsequently an edge detection threshold operation was applied.7 cm. From the binary images.394 0.0 17.394 0. Anemometry (PDA) was used to measure the bubble-size pdf s. The 768(H) x 493 (V) images were captured with a 640 x 480 pixel resolution frame grabber and stored in a computer for later processing.0 17.0 Re = U0 D∗ J ν Ua (m/s) 8.3 cm apart. the mean shape of the projection of the bubble on any given plane was assumed to be statistically the same and independent of its orientation.14 mm x 16. Since the underlying turbulence was isotropic.000 51. Using this method.44 Da (mm) 0. Re has been calculated based on the exit velocity.3 were produced. At all vertical positions of the CCD camera.0 17. the camera was traversed vertically to three positions 2. the measuring windows were overlapped 50% of their length as indicated in ﬁgure IV. ∗ U0 . DJ . This digitalization allowed for a pixel-size resolution of 30 µm. we computed the projected area of the bubbles. and to resolve the entire breakup region while maintaining the desired resolution.0 17.
a) Section 1. Experimental set 3a. Flow goes from left to right in each picture. The downstream length of each measuring window indicated by arrows is 7.45 W2 W4 W1 W3 W5 a) W7 W9 W6 W8 W 10 b) W 12 W 14 W 11 W 13 W 15 c) Figure IV. Windows 1115.14 mm.2: Discretization of the ﬂow region where the bubble break-up takes place. Windows 1-5. b) Section 2. Windows 6-10. c) Section 3. .
3: Binary images corresponding to the ﬂow conditions shown in ﬁgure IV. . c) Section 3. b) Section 2. a) Section 1.2.46 a) b) c) Figure IV.
For the very low air void fraction generated in our experiments. the width and length of all measuring windows were respectively l1 = 16. see chapter II. which is spatially nearly uniform. was then calculated. The air was always injected at the jet’s center axis. and the small depth of the ﬁeld used to acquire our images. In all cases.47 X l1 High Re Submerged Water Jet Dwj l2 Lw Da Air Injection Neddle Figure IV. the bubble-volume probability density function. isotropic turbulence. A schematic of the ﬂow conditions under which all the break-up experiments were conducted. the bubbles remained at the center of the jet. showing all the important dimensions of interest in our experiments. The above described digital image processing method was ﬁrst calibrated against well known bubble geometries in the range of bubble diameters produced in our experiments and found to be accurate within 10%.4.25mm and Lw = 7.14mm. and during their break-up. Dwj . windows or measuring locations. the characteristic width . we are conﬁdent that the equivalent size measured is within the above estimated accuracy. being transported laterally by the action of the turbulence to radial distances always smaller than 30% of the width of the jet.3. Characteristic water jet Diameter.4: Schematic representation of conditions used for the turbulent break-up measurements. From each histogram. Before beginning the discussion of the measurements. V pdf . it is important to emphasize that in all our experiments the bubbles broke up under the action of fully developed. In all experiments l1 /Dwj < 0. is given in ﬁgure IV.
v ′ . a critical capillary length. ǫ. Da1 = 0. IV. The evolution of the volume-size bubble pdf resulting from the turbulent break-up of the air as it is convected to regions of decaying dissipation rate. Within this central region of the jet.1).394 mm. which in turn was always larger than the maximum radial distance where the bubbles were dispersed by the turbulence. D.B IV. Da3 = 1. the measured rms of the axial and transversal component of the jet velocity. the velocity ﬂuctuations of the underlying turbulence of the jet results in deformation forces that are much greater than the conﬁnement forces due to surface tension and. on the . In each of the three diameter cases.e. the turbulent kinetic energy was measured to be nearly uniform. . i. the air injection velocity was selected to be equal to the mean velocity of the water jet at the point of injection. For each value of ǫ.B. is shown in ﬁgure IV. When the air is injected into the water.1 Experimental results Evolution of the volume-size bubble pdfs Let us begin considering the cases of air injection at X/DJ = 15 and Uo = 17m/s.48 of the jet in the region of interest. For the largest needle diameter. l1 . the injection velocity of the air was adjusted to give always the same ﬂow rate. exists such that. was always three to ﬁve times larger than the lateral dimension of the measuring window.194 mm.584 mm. In agreement with measurements of high Reynolds number axisymmetric jets. we investigated three diﬀerent air injection diameters. Da2 = 0. the bubbles only see the nearly homogeneous. the bubble breaks. a result consistent with early measurements . ǫ) of the underlying turbulence.. and on the value of the turbulent kinetic energy (or dissipation rate.5. which correspond to a Reynolds number of the water jet Re = 25. However. . isotropic turbulent ﬂow existing on the jet’s centerline at the injection point. In other words. (set 3 in table IV. Under these conditions. u′ . Dc . For clarity we have only plotted four measuring locations (windows). Dwj . Thus. were always nearly uniform throughout the volume of interest within 3 to 5%. 500. when air exits the needle it is only exposed to the turbulent stresses resulting from the ﬂuctuating component of the water velocity. Qa = 72ml/min. the probability of the break-up depends on the characteristic size of the bubble.
since.4 0.8 0. Da = 1. This critical capillary diameter. In this particular experiment the value of the critical capillary length was Dc ≈ 120 µm at the point of injection of air.5: Evolution of the bubble V pdf . and at the ﬁrst measuring location. As the bubbles are transported downstream by the mean motion of the water jet to the next measuring location.44 X/DJ = 23. Dc is given by Dc ∝ (σ/ρ)3/5 ǫ−2/5 . as the bubbles are broken by the turbulence stresses from the liquid.49 1.194mm. they continue breaking at a rate determined by the local value of the intensity of the turbulent kinetic energy (or dissipation rate). since initially D ≫ Dc . the bubble begins to break at a certain frequency right at the injection point. at X/DJ = 19.7 mm with relatively wide tails.2 1 Vpdf (D) 0.6. Note that. ǫ.5 D (mm) 2 2. In each case the rate of . This break-up process continues while the bubbles are convected downstream to regions of lower and lower dissipation rate. their size monotonically decreases with the downstream distance. X/DJ = 16.07 Figure IV. the turbulent stresses are equal to the surface tension forces.2 0 0 0. until eventually we observed that the V pdf reaches a frozen or unchanged shape. The existence of a frozen state is well expected. shown in ﬁgure IV.4 1. the measured volume-size bubble pdf is observed to have evolved considerably to a new shape which has resulted from the rapid decay in the number of large sized bubbles and the associated increase in the number of the smaller ones. Thus. the volume pdf already shows a large peak at D ≈ 1.44 (window number 4). Experimental Set 3c.6 0.83 X/DJ = 34.06.5 1 1.5 3 X/DJ = 16. mean.06 X/DJ = 19.
Da = 1.6 0. Lw is the length of the measuring window.5 1 1.5 D (mm) 2 2. obtained in each measuring window. Da = 1.4 0.4 1. Experimental Set 3c.2 1 Vpdf (D) 0. which remains unchanged in the last three measuring windows. Nt U /Lw . . experimental Set 3c.194mm.70 X/DJ = 32.07 Figure IV.194mm.7: Evolution of the ﬂux of bubbles per window.88 X/DJ = 34. The value refers to the total ﬂux.8 0.6: Evidence of the existence of a frozen V pdf .2 0 0 0.5 3 X/DJ = 31. 5 104 N t (x) U(x) / Lw (# bubbles / s ) 4 104 3 104 2 104 1 104 0 15 20 25 X/DJ 30 35 Figure IV.50 1.
49 X/DJ= 23.2 1 Vpdf(D) 0. Da = 0.5 1 1.9: Evolution of the ﬂux of bubbles per window.394mm.5 104 2 104 1.83 X/DJ= 34. Experimental Set 3a. Nt U /Lw .2 0 0 0. .10 1.4 X/DJ= 16. 3 104 N t (x) U(x) / Lw (# bubbles / s ) 2. The value refers to the total ﬂux.07 2.8: Evolution of the bubble V pdf .4 0. Lw is the length of the measuring window. Da = 0. experimental Set 3a.51 1.6 0.5 104 1 104 5000 15 20 25 X/DJ 30 35 Figure IV.5 3 Figure IV.394mm. obtained in each measuring window.5 D (mm) 2 X/DJ= 19.8 0.
5 1 1.8 0.07 Figure IV. Da = 0.11: Evolution of the ﬂux of bubbles per window.5 104 1 104 5000 15 20 25 X/DJ 30 35 Figure IV.584mm.2 0 0 0. The value refers to the total ﬂux. experimental Set 3b.5 D (mm) 2 2. Da = 0.5 3 X/DJ= X/DJ= X/DJ= X/DJ= 16.15 19.4 1.83 34.60 23.6 0. Lw is the length of the measuring window.52 1. Experimental Set 3b.10: Evolution of the bubble V pdf . obtained in each measuring window. Nt U /Lw .584mm. 3 104 N t (x) U(x) / Lw (# bubbles / s ) 2. .2 1 Vpdf (D) 0.4 0.5 104 2 104 1.
also increases as (X/DJ )8/5 . dispersion eﬀects are not inﬂuencing our measurements. and Bi and De stand for the rate of birth and death of the bubbles respectively due to break-up. and coalescence eﬀects can also be neglected. U is the convective velocity (which in our experiments was measured to be the same for all the bubbles regardless of ˙ ˙ their size and equal to the local velocity of the water jet).7 shows the downstream evolution of the total ﬂux of bubbles of all sizes. an asymptotic value of about 4 × 104 bubbles/s. Since we are looking at the steady state problem (δNt /δt = 0). In the above described experiment. at which point no more break-up occurs. the location where Nt (x) U (x) becomes constant. δt (IV. Nt (x) U (x)/Lw sharply increases until reaching. indeed. and the pdf will remain frozen from that point on. The rate of change of the total number of bubbles of all sizes per unit volume. or equivalently. a downstream distance exists where all the bubbles become smaller than Dc . Thus.1.1) where Nt is the total number of bubbles measured in the entire volume of the measuring window which has a length Lw and across-sectional area A. coalescence and dissolution. Figure IV. When ˙ ˙ the bubble break-up process is ﬁnished. this frozen pdf s is achieved at X/DJ ≈ 26. while the bubbles reside in regions of high dissipation rate of turbulent kinetic energy. can be calculated as : δnt ˙ ˙ + ▽[nt U ] = Bi + De . indicating the existence of a frozen state. Here no further break up takes place. further downstream. the end of the break-up is marked by the downstream location where ▽[Nt (x) U (x)] = 0. Furthermore. As ǫ decays with the downstream distance (ǫ ∝ (X/DJ )−4 ). nt = Nt A Lw . see ﬁgure IV. In our experiments the residence time of the bubbles in the measuring region is so small that dissolution eﬀects can be shown to be negligible. Dc . the bubble void fraction is always very small. and. Bi = De = 0.6 that the volume-size bubble pdf measured in the last three measuring stations are identical. the critical diameter. Note in ﬁgure IV. near the injection point. Observe that. As a result of these combined eﬀects.53 decay of the bubbles’ size is obviously associated with the local value of ǫ at the injection point and the convective velocity at which they are transported to regions of decaying ǫ. Nt (x) U (x)/Lw . It is important to recall that the width of our measuring window spans the entire small radial region where the bubbles are dispersed by the turbulence. we are measuring all the bubbles .
as was the case with the largest diameter needle.7 × 104 bubbles/s for the case of Da1 = 0.394 mm and Da2 = 0. the air injection velocity was increased to 4. To maintain the injection rate of air equal to the previous case while varying the needle’s diameter.584 mm are shown in ﬁgures IV.13. IV. we conducted additional experiments by injecting the air at X/DJ = 25. Nt U/Lw .394 mm). Thus. However. reached at the location of the frozen pdf has decreased from 4 × 104 bubbles/s in the case of Da3 = 1. the measured evolution of the volume-size bubble pdf in these two cases appears to be similar to the one described for the largest injection needle. the bubble break-up frequency appears to be much smaller. Re = 51. the evolution of the V pdf is also similar to the one described above.54 resulting from the break-up at each downstream location. corresponding to Uo = 17m/s.394 mm. since the values of the dissipation rate are now much smaller. In these cases.05 × 104 bubbles/s.48 m/s and 9. These results are shown in ﬁgures IV. From the shape of the initial volume-size bubble pdf . Note that the asymptotic value of the bubbles’ ﬂux. 1. it can be seen that this case resulted in a much broader initial distribution with tails extending up to 5 mm. Thus. Qualitatively. the air exiting the needle is quickly decelerated to the mean velocity of the water jet in approximately 1 mm.194 mm to 2. 000. even though we begin with much larger bubble-sizes.12 and IV. a length which is almost an order of magnitude smaller than the dimension of the ﬁrst measurement window. Qualitatively.8. the bubbles break-up only under the eﬀect of the underlying fully developed turbulence existing on the axis of the water jet. The measurements corresponding to the same ﬂow conditions but with a needle injection diameter of Da1 = 0.10 and IV. However. This also results in a decrease in the total number of bubbles measured at the location of the frozen pdf . the maximum asymptotic total ﬂux of bubbles measured in this case. . indicating that in the largest injection needle the bubble break-up has produced a larger number of smaller bubbles. This decay is a consequence of the apparent increase of the position of the peak in the frozen V pdf . but into a water jet of a much larger Reynolds number. it should be pointed out that the initial bubble pdf (measured at the ﬁrst window) is narrower as the needle diameter is decreased.84 m/s respectively.9.11 respectively. is the smallest of all the cases studied. Using the smallest air injection needle (Da1 = 0. IV.
55 0. X/DJ= 27. The value refers to the total ﬂux.4 0.36 3 4 5 1.7 0. Da = 0.6 Vpdf(D) 0.5 0.10 X/DJ= 38.394mm.30 X/DJ= 43. obtained in each measuring window. Lw is the length of the measuring window.12: Evolution of the bubble V pdf . Nt U /Lw .3 0. Da = 0. Experimental Set 1.2 0.8 0.37 X/DJ= 32. .394mm. experimental Set 1.01 X/DJ= 45.2 104 N t (x) U(x) / Lw (# bubbles / s ) 1 104 8000 6000 4000 2000 25 30 35 X/DJ 40 45 Figure IV.13: Evolution of the ﬂux of bubbles per window.1 0 0 1 2 D (mm) Figure IV.
14: Evolution of the bubble V pdf .394mm.394mm. Da = 0.56 2 X/DJ= 16. Experimental Set 2.5 3 Figure IV. experimental Set 2. The value refers to the total ﬂux.5 Vpdf (D) 1 0.31 X/DJ= 27.15: Evolution of the ﬂux of bubbles per window.5 0 0 0.5 D (mm) 2 2.46 X/DJ= 23. Nt U /Lw .23 1. Da = 0. Lw is the length of the measuring window.11 X/DJ= 19. .5 1 1. 8 104 N t (x) U(x) / Lw (# bubbles / s ) 7 104 6 104 5 104 4 104 3 104 2 104 1 104 0 15 20 25 X/DJ 30 35 Figure IV. obtained in each measuring window.06 X/DJ= 33.
and Bi (D) and De (D) account for the birth and death rate of bubbles of this class-size due to break-up. 35 40 45 50 IV. this case resulted in the maximum asymptotic total ﬂux of bubbles. as shown in ﬁgure IV. Re = 51. and Da = 0. To obtain additional information on the dependency of the bubble break-up frequency on ǫ and on the bubble size D. The number shown indicates the total number measured in each window over 1.15. 000. This corresponds to the case of U0 = 17m/s.000 images corrected by the ratio of velocities between the measuring window and the ﬁrst window measured. ≈ 7. we conducted a third set of experiments injecting the air at the maximum ǫ allowed by our experimental techniques.16: Downstream evolution of number of the largest class-size bubbles. injection point at X/DJ = 15. ﬁgure IV.2) ˙ ˙ where n(D) is the number of bubbles of size D per unit volume. As expected from the fact that the dissipation rate of the underlying turbulence is the largest of all our experiments. δt (IV. coa- .15.14 and IV.B. Nci = Ni Ui /U1 .2 Rate of decay of number of bubbles of a certain class size The rate of change of the number of bubbles of a certain bubble-size bin is given by: δn(D) ˙ ˙ + ▽[n(D) U ] = Bi (D) + De (D) .57 200 Set 1 Set 2 Set 3-1 Set 3-2 Set 3-3 150 N m U/U1 100 50 0 15 20 25 30 X/DJ Figure IV.394mm (set 2 in table IV.4×104 bubbles/s.1).
17: Bubble break-up frequency. the birth rate is zero since there are no larger bubbles to be broken. Dmax ): g(ǫ.3 allows a straightforward determination of g(ǫ. As mentioned above. the number of bubbles corresponding to the largest size bin only changed due to bubble break-up.3894) R= 0. lescence and dissolution1 . and equation IV.87 * x^(0. where N (D) is the number of bubbles of size D measured in the entire volume of our window of length Lw and cross-sectional area A: g(ǫ. (IV. To calculate the break-up frequency of a bubble as a function of its size and of the dissipation rate of turbulent kinetic energy of the underlying turbulence. we only used the data corresponding to the largest bubble-size bin. In considering the largest size-class bin only. Dmax ) n(Dmax ) . Dmax ) is the break-up frequency of bubbles of size Dmax .4) Since.3) where g(ǫ. and in the steady state situation of interest here. n(D) = N (D)/(A Lw ). Dmax ) = − ▽[n(Dmax ) U ] . Dmax ) = − 1 ▽[N (Dmax ) U ] . N (Dmax ) (IV.5) ˙ ˙ The expressions of Bi (D) and De (D) are given in equations I.D) (1/s) 180 160 140 120 100 80 y = 21.9958 60 0 50 100 150 ε ( m2/s3 ) 200 250 Figure IV.2 simpliﬁes to: ▽[n(Dmax ) U ] = −g(ǫ.58 200 Break-up frequency g(ε. since coalescence and dissolution were negligible in our experiments. equation IV.11 and I.12 . Experimental Set 1. n(Dmax ) (IV.
12 * x^(0. Furthermore.3934) R= 0. IV. IV.19.18 and IV. In each case. In the following section a phenomenological model for the bubble’s break-up frequency. D. D0 . the initial size of the bubbles.19.37-0. will be discussed later.17. The dependency with D as indicated by the factor in front of the power is not readily apparent and.4 (from 0. is assumed to be in the . D). with the exponent being approximately constant and equal to 0. ǫ. it is assumed that the bubbles are injected into a turbulent water ﬂow which is locally homogeneous. to represent all cases in just one plot. Note that in all cases shown in ﬁgures IV. From the measurements of the dissipation rate. in all the experiments discussed above. and the bubble size.18: Bubble break-up frequency. the data has been normalized by the velocity at the ﬁrst measuring location.59 800 Break-up frequency g(ε . ﬁgures IV.39).9822 100 0 500 1000 1500 2000 ε ( m2/s3 ) 2500 3000 Figure IV.17-IV. isotropic and nearly in equilibrium. along the jet’s axis.16. Experimental Set 2. The measured downstream evolution of N (Dmax ) U corresponding to the largest sizeclass bin. ǫ. is discussed and compared with the above results. g(ǫ. the break-up frequency increases as a power function of the dissipation.C A phenomenological model for the bubble break-up frequency Consistent with the experimental evidence. U1 . is given in ﬁgure IV.D) (1/s) 700 600 500 400 300 200 y = 34. we then calculated the break frequency as a function of the dissipation rate.
6) Thus.60 500 Break-up frequency g(ε. η < D0 < Lx . Experimental Set 3. where η is the Kolmogorov microscale of viscous dissipation of the underlying turbulence. and that local isotropy can be applied as the best approximation to describe the underlying turbulence under which the break-up takes place. (IV.e.95 * x^(0.7) . we will also assume that the turbulence is fully developed.3798) R= 0. The basic premise for our model is that for the bubble to break. the surface restoring pressure. 3 πD D (IV.02 * x^(0.3 * x^(0. deﬁned as the surface energy per unit volume is: τs (D) = σ 6 Es = 6 . This assumption is strictly valid for the very small void fraction used in our experiments. i.9949 y = 31.3769) R= 0. We will further assume that the bubble void fraction is always very small.19: Bubble break-up frequency.9932 400 600 ε ( m2/s3 ) 800 1000 Figure IV. and this deformation energy is provided by the turbulent stresses produced by the surrounding water. inertial subrange. its surface has to deform. The minimum necessary energy to deform a bubble of size D is: Es (D) = π σ D 2 .9932 y = 33. k. the presence of the air bubbles do not aﬀect the evolution of the turbulence in the water. Since the one-dimensional energy spectra measured at the bubble injection point show in all cases a k−5/3 dependence on the wavenumber. and that there is no two-way coupling between the two phases.3796) R= 0.D) (1/s) 450 400 350 300 250 200 150 100 0 200 y = 28.
l. When τt (D) > τs (D). . t) − u(x. we will assume that the conﬁnement force is given only by equation IV. the mean value of the velocity ﬂuctuations between two points separated a characteristics distance D can be estimated as : △u2 (D) = |u(x + D. Equation IV. the bubble deforms and eventually breaks up. The mean value of the square of the diﬀerence of velocities at two neighboring points under isotropic and homogeneous turbulent conditions was calculated by Batchelor and given by: |u(x + r. in the homogeneous and isotropic turbulence conditions of interest here. One can estimate the average deformation energy per unit volume produced by the turbulent stresses resulting from the velocity ﬂuctuations existing in the liquid between two points separated a distance D as: τt (D) = where ρ is the density of the water. t) − u(x. t) |2 = β (ǫ D)2/3 . such that bubbles with D < Dc are stable and will never break.7. to the small. The equality. t) |2 = 6 u2 − 2 Rii (r.61 In the case of an air bubble submerged in a turbulent water ﬂow. (IV.10) . deﬁnes a critical diameter. Before going any further it is worthwhile to show the derivation of equation IV. (IV.8) where D is within the inertial subrange. t) . kolmogorov scale. η. thus. τt (D) = τs (D). from the large.9 is obtained integrating over the whole range of turbulent scales. integral scale. Dc . and the internal viscous deformation forces are negligible compared to the surface tension forces. Following Kolmogorov’s universal theory. 2 (IV. the Ohnesorge number is very small (notice that Oh = √ µa ρa σ D = 2 × 10−2 for D = 10µm).9 due to its relevance in this model. It is important to notice that the stresses produced by the gradients of velocity between two points separated a distance D are the consequence of the action of all the scales present in the turbulent spectrum. A bubble of size D > Dc has a surface energy smaller than the deformation energy and. it will break. Thus.9) 1 ρ △u2 (D) .
t) sin kr dk . where α = 1. t) are the single scalar energy and correlation functions respectively.11) and. t) uj (x + r. t) |2 = 4 ∞ 0 of the integrand for values such that 1/l ≪ k ≪ 1/η. Since.62 where u is the root-mean-square of any component of the velocity and Rij (r.12) where E(k. (IV. The critical diameter. Similarly it is suppressed when E(k.20. 2 D Thus. t) − u(x. The value of Dmin can be calculated by simply equating the surface energy to the deformation energy between points apart a distance Dmin : σ 1 2/3 ρ β ǫ2/3 Dmin = 6 . Substituting equations IV. t) |2 = 4 ∞ 0 E(k. t) − u(x.11 and IV. sin kr 9 ) dk = Γ(1/3) α (ǫ r)2/3 . kr (IV. t) dk . If the energy spectrum. substituting D in equation IV. Rii (r.13 tends to zero when k is of the order of 1/l since the term (I − sin kr kr ) becomes very small. any two points on the surface of the bubble separated a distance D ′ such as Dmin < D′ < D will experience stresses from the surrounding turbulence with suﬃcient energy to produce the break-up of the bubble.16) (IV. t). t) = α ǫ2/3 k−5/3 .12 in equation IV.13) Provided that l ≫ r ≫ η. Dmin = ( 12 σ 3/2 −1 ) ǫ . is then deﬁned by the cross-point βρ of the two curves shown in ﬁgure IV. t) = ui (x.9 where β= 9 5 Γ(1/3) α = 8. therefore.14 we obtain the mean value of the velocity ﬂuctuations given in equation IV. t) (I − sin kr ) dk . Note that for a bubble of size D > Dc . E(k. dominated by the behavior is deﬁned as E(k.2 . Dc = ( 12σ )3/5 ǫ−2/5 .15) . t) and R(r. t) is the velocity correlation tensor. equation IV.7 is a universal constant: |u(x + r.10: |u(x + r. t) = 2 R(r. : 3/2 u2 = 0 ∞ E(k.14) α ǫ2/3 k−5/3 (I − If now the stresses are calculated in a distance r = D. kr 5 (IV. t) = 0 ∞ E(k. The value of the integral is. β ρD (IV. kr (IV. t) → 0 as k → η.
7.5 4 D ∆τ τs = k σ σ/D τt= 1/2 ρ β (εD)2/3 Figure IV. Solid line is the conﬁning force provided by surface tension and the broken line is the one given by the turbulent stresses. The determination of the pdf of the daughter bubbles is a complex issue which until today has precluded the development of a uniﬁed theory (model) to describe the turbulent break-up. leads to the fact that when it breaks. The fact that a range of dimensions exists between D and Dmin along which the bubble may break.63 3000 2500 Pressure (Pa ) 2000 1500 1000 500 Dmin 0 0 0. As is the case in any mechanical process. it may result in a wide distribution of daughter bubble sizes. Therefore.5 D/Dc 3 3. 6 σ/D. the larger the probability that the bubble will break in a certain time.5 1 1. at which a bubble of size D will break under a turbulence characterized by a dissipation rate. we postulate that the rate at which the break-up process takes place is inversely proportional to the diﬀerence between the non-inertial forces which produce the motion. we postulate that the larger the diﬀerence between the gradient of pressure produced by the turbulent ﬂuctuations 1 on the surface of the bubble ( 2 ρ △u2 (D)). the break-up frequency should decrease to a zero limit value as this diﬀerence of pressures vanishes. In other words. D).The constant kσ = 6 is given in equation IV. and the restoring pressures caused by surface tension. ǫ.20: Force per unit surface on the bubble. However. one can determine the frequency. without addressing this complex problem of determining the pdf of the daughter bubbles. On the other hand. g(ǫ.5 2 2. the equation of the bubble’s deformation can .
equation IV. D) is then given by: 1 g(ǫ.20) 0 0 Dc 500 1000 1500 D (µm) 2000 Figure IV.17 can be expressed as: σ D 1 △u2 (D) −6 .17) (IV.64 2500 Break-up frequency.4 0.6 0. . D) (1/s) 2000 1500 1000 500 D gmax 1. g(ε.18) (IV.2 1 0. ub (m/s) (IV. Since ub ∝ D/tb . D) = = Kg tb △u2 (D) − 12 ρσ D D β(ǫD)2/3 − 12 ρσ D D = Kg . be written as: ab = ub Fe 1 △u2 (D) σ = = −6 . the characteristic velocity of deformation would be proportional to the diﬀerence of stresses acting on the surface of the bubble and. The break-up frequency g(ǫ. velocity and time of the bubble break-up (deformation) process and Fe are the forces acting on the surface of a bubble of size D. ǫ = 2000 m2 /s3 . ub and tb are the characteristic acceleration. = 2 2 D ρ D2 tb Thus. tb M ass 2 D ρ D2 where ab . the bubble break-up time can be estimated as: tb ∝ D △u2 (D) − 12 ρσ D .8 0.2 2500 Break-up velocity.19) (IV.21: Evolution of the bubble break-up frequency and break-up velocity with respect to the diameter of the bubble.
the experimentally measured dependence of the break-up frequency on the bubble size also appears to be in excellent qualitative agreement with the model. The agreement is within the 10% maximum experimental error shown by the error bars in ﬁgure IV. 2 .25 have been found experimentally 2 . it agrees remarkably well with the measured frequencies.25 has been obtained by best ﬁtting the transient V pdf s while solving the inverse problem of calculating the daughter pdf . contains bubbles of sizes between two values. the break-up frequency decreases as the bubble diameter is increased since in all cases D0 > Dgmax . Dm1 and Dm2 . corresponding to the largest diameters. The dependence of the break-up frequency. the calculated values of g(ǫ.22. consistent with the model. the break-up frequency decreases monotonically with the bubble size. in order to test the accuracy of our model given in equation IV. given by equation IV. As discussed earlier. is given by: gmax (ǫ) ∝ (σ/ρ)−2/5 ǫ3/5 .22a to D0 = 1. The maximum break-up frequency.22. Notice that. Due to the discretization needed in our experiments. the break-up frequency is zero for bubbles of size D ≤ Dc . it is important to note that after reaching a maximum at Dgmax = 1.21. Observe that the break-up The value of Kg = 0. but more importantly.7 mm in ﬁgure IV. the break-up frequencies have been calculated from the data corresponding to the largest bubble-size bin which contains enough number of samples to ensure statistically meaningful results.20. Although in our experiments we were only able to vary slightly the bubble’s diameter at the point of injection. this bubble-size bin.63 Dc . jointly with the experimental data. the characteristic bubble size D0 varied from D0 = 2. and it increases rapidly for D > Dc . Dm2 ) corresponding to the break-up frequencies of the upper and lower boundaries of the largest bubble-size bin.21) The measured bubble break-up frequency corresponding to the three set of experiments discussed in the previous section are plotted in ﬁgure IV. (IV. Thus. Dm1 ) and g(ǫ. It should be noted that the model not only describes the qualitative trends. In the experiments shown in ﬁgure IV. on the bubble diameter is shown in ﬁgure IV.22c. However. achieved at Dgmax .2 was given by Batchelor (1956)  and Kg = 0. As expected.67 mm in ﬁgure IV.22.20.65 where the constant β = 8. we have plotted.
66 220 200 180 g(ε.20. c) set 2. D0 = 2. D0 = 2.67 = mm. . b) set 3b.0mm.D) (1/s) 500 400 300 200 100 c) 0 500 1000 1500 2000 ε (m2 / s3 ) 2500 3000 Figure IV.D) (1/s) 160 140 120 100 80 a) 0 50 100 150 ε (m2 / s3 ) 200 250 450 400 350 g(ε.22: Comparison of experimentally measured bubble break-up frequency with the frequency calculated with model given in equation IV. Error bar indicates an estimated maximum ±10% experimental error.7 mm. a) Set 1.D) (1/s) 300 250 200 150 100 b) 0 200 400 2 3600 ε (m / s ) 800 1000 800 700 600 g(ε. D0 = 1.
D) ≈ ǫ1/3 D−2/3 . ρ (IV. Dc < D < Dgmax . (IV. Similarly. the regime in which most of the previous investigations may have been conducted. D) valid for D/Dc ≈ 1 can be written as: σ g(ǫ. C. Dgmax = 1. Furthermore.67 mm to 450 s−1 for D0 = 2. D) ∝ ( )−2/3 ǫ3/5 ρ D −1 . by the way.  among others. Previous work conducted in turbine mixers generally were done at low ǫ.63 Dc = 1. for small bubbles of sizes smaller than Dgmax but comparable to the critical diameter. comprised of a paper titled “On the break up frequency of an air bubble injected into a fully developed turbulent ﬂow”.0 mm. The dependence of the break-up frequency on the bubble diameter is worthy of further discussion since it appears to have been a source of controversy with previous investigators. In the limit of very large bubbles. .24) On the other hand. J. their reported increase of the break-up frequency with the diameter is also consistent with our present model. σ gmax (ǫ) ∝ ( )−2/5 ǫ3/5 .23) and that the maximum break-up frequency increases proportionally to ǫ3/5 and decays with the interfacial tension as σ −2/5 . the break-up frequency decreased from 540 s−1 for D0 = 1.63 ( 12σ 3/5 −2/5 ) ǫ .7 mm when ǫ is 230 m2 s−3 .25) This is. it is important to point out that the diameter for which the break-up frequency is maximum depends only on Dc . D/Dc ≫ 1.63. and the asymptotic dependence of g(ǫ. the surface tension forces become very small and the break-up frequency can be approximated by: g(ǫ. Mart´ ınez-Baz´n. Dc (IV.22) This is indeed the dependence that we have measured since in all our experiments D/Dc was always greater than 1. the turbulent stresses dominate. βρ (IV. Thus. This chapter is. a . for ǫ = 950 m2 s−3 .67 frequency decreased from 250 s−1 for D0 = 2.0 mm to 200 s−1 for D0 = 2. in part. and in the size range Dc < D < Dgmax .
C. Monta˜es. The dissertation author was the primary investigator and author of this publication. and J.68 L. Lasheras currently being considered for publication in Journal n of Fluid Mechanics. .
A Introduction The dispersion of an immiscible ﬂuid into a turbulent one is commonly found in many chemical engineering processes as well as natural processes.Chapter V P DF of the Daughter Bubbles V. For example. among others). the absorption rate of a given chemical compound depends not only on the dynamics of the motion between the two immiscible ﬂuids. The frame work of the ”population balance equation” has been widely used by the chemical engineering community and others to study these processes (. ). . and has generated a large bibliography (. 69 .e. Resulting from its widespread use. but more importantly on the size pdf of the drops (or bubbles) in which one ﬂuid is dispersed into the other. . . . . in liquid-liquid or gas-liquid chemical separators. i. the problem of the break-up of an immiscible ﬂuid immersed into a turbulent ﬂow has been the subject of a continuing investigation. . The size distribution of the drops (or bubbles) resulting from the turbulent break-up and the dynamics of their interactions with the underlying turbulence plays a determinant role in the overall performance of these processes. air bubbles entrained by the air-sea interaction contribute to the exchange of gases and water vapor between the oceans and the atmosphere (. . . . Surface-dominated diﬀusion processes of this nature are also found extensively in naturally occurring phenomena. Quantitative understanding of the drop (or bubble) break-up and their coalescence is essential to the development of predictive models for the behavior of theses processes. . .
δt (V. equation V. ǫ) of the underlying turbulence.3 requires the solution of the two closure problems.4) . and f (D. very small volume fraction of the ˙ dispersed phase. etc). the ﬁrst one involves the break-up frequency. g(D0 ) is the break-up frequency of a droplet (bubble) of size D0 . Equation V. : δn δ ˙ ˙ + ▽x (v n) = − (R n) + Qb + Qc . D0 ). δt δD (V. D0 ) = Kg D0 . Although equation V.e. . in this study.2(ǫD0 )2/3 − 12 ρD0 g(ǫ. Generally.1) where v is the mean velocity of all the particles of size D. the coalescence eﬀects can be neglected (Qc = 0). we will restrict our analysis to only the case of bubbles breaking into a turbulent water ﬂow. This equation describes the time rate of change of the number of drops (or bubbles) of a certain size Di existing in a time t.70 .3) where m(D0 ) is the mean number of droplets (or bubbles) resulting from the break-up of a mother droplet (or bubbles) of size D0 .1 is formulated as. x. i. Qb and Qc are the rate of change of n due to the break-up and coalescence respectively. t). D0 ) is the size distribution of daughter droplets (bubbles) formed from the breakage of a mother drop (bubble) of size D0 . D0 )g(D0 )n(D0 )dD0 − g(D)n .3 applies to either drops or bubbles. (V. σ 8. in the absence of dissolution (or evaporation) eﬀects. the bubble’s break-up frequency is a function of both its size and the turbulent kinetic energy (or dissipation rate. (V.2) In the simpliﬁed case of very diluted systems.. at a given position x. and the time evolution of the number density of particles can simply be written as the Death of particles of size Di due to their break-up into smaller ones and their Birth resulting from the break-up of larger particles: δn + ▽x (v n) = δt ∞ D m(D0 )f (D. R = dD/dt is the rate of ˙ ˙ change of size D of a particle due condensation. and the second one the pdf of the daughter particles f (D. In the previous chapter we have shown that under the dilute conditions of air bubbles immersed into a turbulent water ﬂow whose turbulence is homogeneous and nearly in equilibrium. ni (Di . g(D). δn ˙ ˙ + ▽x (v n) = Qb + Qc . evaporation and dissolution.
In the present chapter. These measurements.25. to test a new phenomenological model for f (D. and Lx is the streamwise integral length scale. performed over a wide range of bubble sizes and turbulent kinetic energy of the underlying turbulence of the water. D0 ). At time t = t0 . The initial bubble diameter is assumed to be in the inertial subrange of the underlying turbulence. will produce its break-up. When the air bubble is injected into the turbulent water jet. was found to be in excellent agreement with experimental measurements. Rather than attempting to solve the otherwise intractable inverse problem of calculating f (D. V. In this chapter we will use a detailed set of measurements of n. The above break up frequency. were conducted by measuring the transient bubble-size pdf resulting from the injection of a volume of air of a certain size into the fully developed region along the axis of a high Reynolds number water jet. D0 ) from equation V. where η is the Kolmogorov microscale of the viscous dissipation. the bubble is immersed into a stationary water ﬂow whose turbulence is nearly homogeneous and isotropic. the velocity ﬂuctuations of the underlying turbulence result in pressure deformation forces acting on the bubble’s surface that. which is based on the simple premise that the break-up time of the bubbles is inversely proportional to the sum of the non-inertial forces acting on the surface of the bubble (producing its deformation and conﬁnement).71 where Kg is a constant found experimentally to be equal to 0. when greater than the conﬁnement forces due to surface tension. we will use this bubble break-up frequency model to solve the second closure and to provide a model for f (D.B Model formulation Consider a mother or parent bubble (mother or parent shall refer to a bubble before break-up) to be spherical with an initial diameter D0 .3. presented in chapter IV. This model is then contrasted to previous ones and compared to experimental measurements.3. Since the Ohnesorge numbers of the bubbles of interest here . D0 ) in equation V. η < D0 < Lx . we are proposing a new phenomenological model for the daughter bubble-size pdf based on a weighted probability model.
πD0 0 (V.25 Dc 0. In this example. △τt1 = 1 ρ β (ǫD1 )2/3 − Dσ and. △τt2 = 1 ρ β (ǫD2 )2/3 − Dσ .25 1. . ub is the deformation velocity.1: Diﬀerence of stresses associated to the formation of a bubble and its comple6 6 mentary. ab = ub /tb . 2 2 0 0 the mother bubble is of size D0 = 1 mm.2 is a constant obtained integrating the diﬀerence between the velocity ﬂuctuations. The conﬁnement energy per unit volume is simply.75 D2 1 D0 1.72 Turbulent. and tb is the bubble’s . D0 ) = 1/tb .5) where β = 8. over the whole range of turbulence scales. △u2 (D0 ).6) When τt > τs the bubble will break in a certain time tb . with a frequency g(ǫ. τt = 1 2/3 ρ β ǫ2/3 D0 . Here. of two points separated by a distance D0 . In chapter IV we have shown that the bubble break-up time can be estimated from the deformation acceleration. are always very small (Oh = √ µa ρa σD < 10−3 ).5 D (mm) ∆ τ2 τs= 6 σ/D τt= 1/2 ρ β ε 2/3 D2/3 Figure V.5 D1 0. the internal viscous deformation forces are negligible compared to the surface tension forces. Confinement stresses (K Pa ) 5 4 3 2 ∆ τ1 1 0 0 Dmin 0. τs = 2 σ 6 σ D0 3 =6D . produced by the non-inertial forces acting on the surface of the bubble. 2 (V. and ǫ = 1000 m2 s−3 . The average deformation energy per unit volume acting on the surface of the bubbles is. and thus will be ignored.
. a bubble of size D0 only breaks into two bubbles of complementary masses with diameters D1 and D2 (an important point to be addressed later on).2 0. the mother bubble is of size D0 = 3 mm. are not uniform with the distance D. the bubble is shattered into an array of smaller bubbles whose size pdf is denoted by f (D. tb ∝ D0 β (ǫD0 )2/3 − 12 Dσ ρ 0 .6 D*= D/D0 0. given in equation V.7) As shown in the above equation. g(ǫ. In this particular case.2 0 0 0.4 0.8 0. ǫ) of the underlying turbulence. Dmin . our bubble splitting phenomenon is not a purely random process. Upon break-up. breakup time given by. D0 ).10. If the bubble splitting process were a purely random event. D0 ) = 1/tb . The ﬁrst premise of our bubble shattering model is that upon breakage. In fact. since the values of the pressure deformation forces. (V. [(1. and the dissipation rate is ǫ = 1000 m2 s−3 . the probability of this event would always be p(D1 /D0 ) = 1.4 0.8 1 Figure V. there is a minimum distance. τt . for a spherical bubble of size D0 .6 0.δ5/3 ].D*3)2/9 . depends on both the size of the bubble and on the value of the turbulent kinetic energy (or dissipation rate. the break-up frequency. such as the single division of a segment of length D0 into two segments of sizes D1 and D0 − D1 .73 1 [D*2/3 . such that the turbulent stresses acting between two points separated .2: Diﬀerence of stresses associated to the formation of a bubble and its complementary. However.δ5/3] 0.
and Dmax = D0 1 − taken to be equal to η. from a bubble of size D0 should be zero. P (D1 /D0 ) ∝ 6σ 1 ρ β (ǫD1 )2/3 − 2 D0 1 6σ ρ β (ǫD2 )2/3 − 2 D0 . associated to the generation of a bubble of size D1 .9) (V. the probability for the splitting of a fraction of size D1 < Dmin = ( β12σ 0 )3/2 ǫ−1 . On the other ρD hand.1. 1/2ρ β (ǫDmin )2/3 . or similarly. D/D0 . △τt1 = 1 2 ρ β (ǫD1 )2/3 − 1 2 6σ D0 . From a simple mass balance. D2 = D0 1 − P (D1 /D0 ) ∝ 1 ρβ(ǫD0 )2/3 2 2 D1 D0 3 1/3 . are just equal to the conﬁnement forces due to the surface tension. Our model implies that Dmin ≤ D1 ≤ Dmax . associated with the formation of a bubble of size D1 . the daughter f (D/D0 ) represents the probability density function of the daughter bubbles resulting from the break-up of a mother bubble of size D0 using a nondimensional diameter of the daughter bubble. where Dmin = IV). (V. we postulate that the probability of a certain size pair to form should be weighted by the product of the excess stresses associated with the length scales corresponding to each bubble size.10 is shown in ﬁgure V.10) where D ∗ = D1 /D0 . Otherwise.74 by this distance. (V. δ = Dc /D0 = (Dmin /D0 )2/5 . Thus. 6σ/D0 .8) In other words. involving a diﬀerence of stresses △τt2 = 6 ρ β (ǫD2 )2/3 − Dσ . there is also the formation of a complementary bubble of size D2 . The evolution of the diﬀerence of stresses in equation V. and D∗2/3 − δ5/3 (1 − D ∗3 )2/9 − δ5/3 12σ βρ 3/5 . Thus. P (D1 /D0 ) ∝ p(D1 /D0 ) [△τ1 ] [△τ2 ] . Dmin is P (D1 /D0 ) d(D1 /D0 ) = 1. 1 ∗ ∗ Dmax ∗ Dmin ǫ−1 (as deﬁned in Chapter Dmin D0 3 1/3 provided that Dmin > η. we postulate that the probability 0 of the formation of a pair of bubbles of sizes D1 and D2 from the shattering of a mother bubble of size D0 is weighted by the product of the two surplus stresses1 .2. However. the probability of the splitting of any portion of size Dmin < D1 < D0 should be a function of the diﬀerence in the stresses. and Dc = 12σ βρD0 3/2 ǫ−2/5 is a critical diameter deﬁned by the crossing point of the curves τt and τs in ﬁgure V. . Using the normalization condition. and neglecting compressibility eﬀect in the air inside the bubbles.
3 0 0 0.4 mm D0= 0.2 0.4 D/D0 Figure V.8 1 D0= 0. a) Evolution of the pdf for various values of ǫ and ﬁxed D0 = 3mm.4 0.5 0 0 2.3: Probability density functions of the daughter bubbles formed from the break-up of a mother one of size D0 .6 0. 0.8 1 .2 0.6 D*= D/D0 0. b) Inﬂuence of D0 on the the pf d for a ﬁxed value of ǫ = 1000 m2 s−3 .5 2 1.5 f*(D/D0) D0 = 3 mm 1 0.5 1 0.5 ε= 1000 m2 s.5 mm D0= 1 mm D0= 3 mm f*(D/D0) 0.75 2 e=10 m2/s3 e=100 m2/s3 e=1000 m2/s3 1.
C Comparison with other models Three main approaches have been used in the past to model f (D. they also assumed the existence of a minimum droplet size. (V. is given by : D1 D1 2 2 2 2 2 e(D1 ) = πσD1 + πσD2 − πσD0 = πσD0 + 1− D0 D0 3 2/3 − 1 . The dependence of f ∗ (D1 /D0 ) on ǫ and D0 is shown in ﬁgure V. Dmin . This gives a minimum probability for the formation of two bubbles of the same size (since their surface energy is maximum).. which correspond to the case of two daughter bubbles with the same volume. is then calculated as f (D1 .12) where the energy to form a bubble of size D1 . D0 . (V. e(D1 ). Note that the peak of the distribution is always located at D∗ = (1/2)1/3 = 0. . .8. m(D0 ) = 2. They proposed that upon break-up. . D0 ) = f ∗ (D1 /D0 )/D0 . . statistical models.. (V. Among the most widely used phenomenological models based on surface energy considerations is the one proposed by Tsouris and Tavlarides (1994). and a maximum probability for the formation of a pair made up of a very large bubble and its complementary very small one. Their model for the daughter bubble pdf gives f (D1 . f ∗ (D1 /D0 ) can then be written as: f ∗ (D1 /D0 ) = ∗ Dmax ∗ Dmin D∗2/3 − δ5/3 D∗2/3 − δ5/3 (1 − D ∗3 )2/9 − δ5/3 (1 − D ∗3 )2/9 − δ5/3 d(D ∗ ) . and that the pdf s become wider as ǫ or D0 are increased. D0 ) = emin D0 0 (emin + [emax − e(D1 )])dD1 + [emax − e(D1 )] . V.11) The probability density function of the daughter bubbles resulting from the break-up of a mother bubble. . D0 ): phe- nomenological models based on surface energy considerations.3.13) . and only two daughter droplets of size D1 and D2 are formed whose most probable sizes are inversely proportional to the amount of surface energy created in the break-up process.76 probability density function of the ratio of diameters D ∗ = D1 /D0 . and hybrid models based on a combination of both. To avoid the singularity present in their model. .
6 0. 0. emax (D0 ).2 0.15) . are formed.5 1 0.12 has a U shape. Pure statistical distributions have also been used by Coulaloglou and Tavlarides (1977) . . The The daughter bubble’s size pdf given by equation V. and minimum energy.4).8 1 (V. (1966). Prince and Blanch (1990) .77 2. and others. . .5 0 0 0. . and others. Coulaloglou and Tavlarides assumed a normal distribution previously proposed by Valentas et al.13 reaches a minimum value when a bubble of minimum diameter.14) Equation V. and has been shown to lead to results which are radically diﬀerent from the experimental distributions measured in stirred tanks and other turbulent ﬂows. Dmax = (D0 − Dmin )1/3 . The maximum energy. Dmin .4: Daughter bubbles pdf predicted by model of Tsouris and Tavlarides 1994. (ﬁgure V. An important unresolved issue in the above model is the need to deﬁne a criterion for the value of Dmin . 3 3 and a complementary one of maximum size. (V.5 Tavlarides 1994 2 f*(D/Do) 1. emin .4 D/Do Figure V. corresponding to the formation of two bubbles of diameter D1 = D2 = D0 /21/3 is: 2 emax = πσD0 [21/3 − 1] . is given by: Dmin 2 emin = πσD0 D0 2 + 1− Dmin D0 3 2/3 − 1 .
(1983) . Longuet-Higgins proposed a simple mechanism for the break-up process by simulating it through a series of a random divisions of a cubical block of size unity by a number of planes which are parallel to the faces of the block. Km . The number of elements of each daughter droplet formed. and that the probability distribution function of daughter droplets (or bubbles) is well represented by a normal distribution.4 (2v − v0 )2 exp −4. and three dimensions.78 Prince and Blanch a random distribution with the same probability for each size. v0 . .5 2 v0 v0 . . K2 . v0 ) = 2. Novikov (1997) . is divided into J units of volume elements ve . where vi is the volume of the droplet i. All combinations are then derived from the diﬀerent arrangements of K1 . Coulaloglou and Tavlarides showed that this distribution could be adjusted to predict some of the experimental results obtained in stirred tanks.16) where v0 and v are the volume of the mother and daughter droplets respectively. In studying the pdf resulting from cuts performed in one.6% of the daughter droplets were in the range of volumes 3 between 0 and v0 = πD0 /6. is given by Ki = vi /ve . . he obtained an inﬁnite number of possible distributions depending on the number of random partitions performed in the initial dimension. (V. By adjusting the combination of the number of cuts performed in each dimension. they obtained the daughter bubble size pdf as f (v. so that J = v0 /ve . They calculated the probability of obtaining a certain combination of three droplets and weighted their probability by a factor proportional to the energy contained in the turbulent scales of size equal to the size of the daughter droplets. and others. . They assumed that m(D0 ) = 2. m(D0 ) = 3. They assumed that the volume of a mother droplet. it is worth noting the model proposed by Konno et al. numbered by the index i. two. (1980)  . This model is based on their limited experimental observations where three daughter droplets are produced from the break-up of a mother one. By selecting the variance so that more than 99. Purely statistical studies have been performed by Longuet-Higgins (1992) . he showed that the model could ﬁt various experimental results Concerning hybrid models. Their statistical model then states that the parent droplet breaks into m daughter droplets whose volume is given by a random combination of elements ve . which satisfy the .
8 1 All the combinations of the J elements.17.19) where α = 0. .5 f*(D/Do) 2 1.4 D/Do Figure V. by giving a maximum probability for combinations of particles of similar size. one needs to use a large value for J. .51. .5 Konno 1983 3 2. . (V. This model has the deﬁciency that the distribution of the daughter droplets is basically neither dependent on the size . and Konno et al.2 0.6 0. The probability of a certain combination of Ki elements is then given by P ∝ E(K1 )E(K2 )E(K3 ) . K1 + K2 + .5 0 0 0.17) 0. gives all the possible values of the non-dimensional combinations of volumes Ki that can be formed. taken in m groups satisfying equation V.5: Daughter bubbles pdf predicted by model of Konno 1983 . (V.79 3. E(Km ) . used J = 100. (V. conservations of volume. + Km = J . and ν is the kinematic viscosity of the continuous phase.18) where E(Ki ) is the spectrum function of the turbulent kinetic energy estimated from the Heisenberg energy spectrum as: E(Ki ) = 8 9α 2/3 ǫ2/3 π 6ve Ki −5/3 1+ 8ν 3 (π/6ve Ki )4 3α2 −4/3 .5 1 0. To obtain a continuous distribution with this model. Konno’s distribution produces a low probability for combinations of very large and very small particles.
D0) 2.6 0. Red line represents Tsouris and Tavlarides 1994 .21) .6. Konno’s distribution. of the mother droplets. A comparison of the daughter bubble pdf resulting from the above previous models and the model presented here is shown in ﬁgure V.2 0. 1983.8 1 e= 4. notice that as the value of the dissipation rate of TKE.5 2 1.5 0 0 0.80 3.5 3 f*(D/D0) = D0 f (D.5 m2/s3 e=10 m2/s3 e=1000 m2/s3 Tavlarides 1994 Konno 1983 Figure V. blue line represents Konno et al. : D0 f (D. nor on the turbulent kinetic energy of the underlying turbulence. In fact. and black lines represent our present model. from conservation of energy. To conclude this section. D0 ) = f ∗ (D/D0 ) = Γ(12) Γ(9)Γ(3) D D0 8 1− D D0 2 . D/D0 . a statistical model based on energy and entropy principles has been proposed by Cohen (1991) . (V. He states that. the energies of the system before and after the breakage are related by: 2 ET + σ π D0 = ES + ER .6: Comparison of the daughter bubble size pdf predicted by the present model and previously proposed models.20) where f ∗ (D/D0 ) represents the probability density function of the daughter bubbles resulting from the break-up of a mother bubble of size D0 using a nondimensional diameter of the daughter bubble.5 1 0. two facts which are contrary to experimental observations. ǫ. shown in ﬁgure V.4 D/D0 0. (V. is decreased our pdf takes a shape similar to that predicted by Konno et al.5 is practically universal and can be approximated by a Beta function. 1983 .
ESmax 2 σ π D0 3 . ER = 0. and ER is the remaining absorbed energy such kinetic energy. (V. Ki ): i= Di Dmin 3 .25) 3 .23) and consequently the maximum number of bubbles created is given by.22) ˜ where ET = imax i=1 ET 2 σπ D0 ˜ and ES = ES 2 σπ D0 are the non-dimensional breakage and resulting surface energies respectively. 3 3 ˜ Ni Di = D0 .29) . the maximum number of resulting bubbles is given by Nmax = i Ni . ES is the total surface energy of the resulting bubbles.26) Cohen then. (V. therefore.27) imax i=1 Therefore.81 where ET is the breakage energy. ESmax (V. . (V. Dmin . imax is the largest bubble size found in the resulting group of daughter bubbles. ˜ Nmax = ET + 1 −1 . (V. from conservation of volume. oscillations.28) i Z i /i! . imax ˜ ˜ ET + 1 = ES = i=1 Ni Di D0 2 . Ni = i! where Nmax = i=1 imax (V. Dmin = 3 σ π D0 .as the number of primary blocks of size Dmin within a larger bubble of size Di (similar to the volume elements deﬁned by Konno et al. If all the energy results into the breakage.24) Dmin ˜ = ET + 1 D0 and. The maximum value of ES is stated for a monodisperse size distribution producing. etc. “i” represents the class size and. (V. By using entropy arguments he ﬁnds that the most probable distribution resulting from shattering of mother bubble is: Zi ˜ . deﬁnes the bubbles (droplet) size class . and equation V.21 can be rewritten as. Nmax = 2 Since ESmax = ET + σ πD0 . the minimum diameter. ”i”. (V.
82 and provided that, Z imax /imax ! = 1 . ˜ From equation V.28, the total number of daughter bubbles is N = and the daughter bubbles pdf is deﬁne as, f ∗ (i) = ˜ Ni
imax i=1 imax i=1
(V.30) ˜ Ni =
imax Z i i=1 i! ,
Zi i! imax Z i i=1 i!
2 + σ π D0 and,
In a turbulent break-up, of interest here ESmax =
1 ET = 2 ρ β (ǫ D0 )2/3
3 π D0 6 .
3 1 2/3 π D0 2 ρ β (ǫ D0 ) 6
The daughter bubbles pdf , f ∗ (i), is then obtained combining
equations V.26, V.29, V.30 and V.31. The shape of the pdf , f ∗ (D/D0 ) resulting from Cohen’s model is somehow similar to the Beta function proposed by Konno (1983) . On the other hand, this type of pdf has incorporated the dependency of ǫ in the model to account for the minimum diameter possible. As a instructive example, choosing a characteristics value of the dissipation, ǫ = 1, 000 m2 s−3 and a diameter of the mother ˜ bubble, D0 = 1mm, we obtain ET = 9.36 which results in a maximum number of bubbles Nmax = 1112. It will produce an explosive break-up which has not been observed in our experiments.
From the set of experiments presented in the previous chapter, the evolution
of the bubble size pdf , as the bubbles are broken by the turbulence and transported downstream by the mean convective motion of the jet, was measured using digital image processing techniques. These bubble size pdf measurements were then discretized into a family of size bins and used to calculate the rate of change of the number of bubbles of a certain size per unit volume as they are convected into regions of decreasing values of ǫ. In the various experiments performed both, the initial bubble size and the initial values of ǫ at the injection point were systematically changed. A summary of the test conditions used for the measurements is given in table IV.1, the downstream evolution of the Cumulative volume pdf , and Dv90% and D32 of this experimental measurements are shown in ﬁgure V.7. The qualitative trend of all cases is similar, we will begin by discussing ﬁrst the
83 results of Set 2 in table IV.1. In this case, the air was injected on the jets’ centerline at a location of 15 jet’s diameters downstream from the jet’s exit nozzle. The initial jet
∗ velocity was U0 = 17 m/s, and the jet’s Reynolds number of Re = U0 DJ /ν = 51, 000.
The air injection velocity was Ua = 9.84m/s, and equal to the local mean velocity of the water at the air injection point. Thus, upon injection into the water, the air bubbles were only subjected to the velocity ﬂuctuations of the underlying turbulence of the water ﬂow. These turbulent stresses resulted in deformation forces that were much greater than the conﬁnement forces due to surface tension, and the bubble was broken into a set of daughter ones. The characteristic break-up time of the bubble depends on its size, D0 , and on the value of the turbulent kinetic energy (or dissipation rate, ǫ) of the underlying turbulence. This break-up process continued while the resulting array of bubbles were convected downstream to regions of lower and lower dissipation rate, ǫ, until eventually it was observed that the bubble size pdf reached a frozen or unchanged shape, shown in chapter IV. The downstream evolution of the cumulative bubble volume pdf corresponding to these experiments is shown in ﬁgure V.7b where, for clarity, we have presented four measuring locations only. Note that the values of the characteristics diameters, Dv90% , and D32 2 decay with the downstream distance, X/DJ , due to the break-up process until they all reach asymptotic values at a downstream distance of approximately 28DJ , at which point the cumulative volume pdf is no longer observed changing. It is important to emphasize that in all our experiments, the bubbles broke up under the action of a fully developed, spatially nearly uniform, isotropic turbulence, and in the absence of any solid surfaces. The air was always injected at the jet’s center axis. During their break-up, the bubbles remained at the center of the jet, and were transported laterally by the action of the turbulence to radial distances always smaller than 30% of the width of the jet. In ﬁgure IV.4, a schematic of the ﬂow conditions under which all the break-up experiments were conducted shows all the important dimensions of interest in our experiments. It should be noted that in all cases, the characteristic width of the jet in the region of interest, Dwj , was always larger than the lateral dimen2
ÈN D ÈN D
3 i 2 i
where Ni is the number of bubbles of size Di measured.
Dv90% = Diameter of a bubble such that 90% of the total volume of air is contained in bubbles of smaller diameter.
1 Cumulative Volume Pdf 0.8 Diameter (mm) 0.6 0.4 0.2 0 0 1 2 D (mm) 3 4 5 X/D= 27.37 X/D= 32.08 X/D= 38.30 X/D= 45.36
4 Dv90% 3.5 3 2.5 2 1.5 1 0.5 0 25 30 35 X/DJ 2 Dv90% D32 1.5 Diameter (mm) 40 45 50 D32
1 Cumulative Volume Pdf 0.8 0.6 0.4 0.2 0 0 0.5 1 1.5 D (mm) 2
Experimental Set 1
X/D= 16.11 X/D= 19.46 X/D= 24.64 X/D= 33.23 2.5
0 15 20 25 X/DJ 30 35
1 Cumulative Volume Pdf 0.8
Experimental Set 2
2.5 Dv90% 2 Diameter (mm) 1.5 1 0.5 0 3 15 20 25 X/DJ 30 35 D32
0.6 0.4 0.2 0 0 0.5 1 1.5 D (mm) 2 2.5 X/D= 16.10 X/D= 19.49 X/D= 25.02 X/D= 34.07
1 Cumulative Volume Pdf 0.8
Experimental Set 3a
3 Dv90% 2.5 Diameter (mm) 2 1.5 1 0.5 0 3 15 20 25 X/DJ 30 35 D32
0.6 0.4 0.2 0 0 0.5 1 1.5 D (mm) 2 2.5 X/D= 16.07 X/D= 19.44 X/D= 25.05 X/D= 34.07
Experimental Set 3c
Figure V.7: Right column, downstream evolution of the cumulative volume probability density function. Left column, downstream evolution of the characteristic diameters D32 and Dv90% .
32. When integrating the right hand side of equation V.85 sion of the measuring window. D0 ) calculated from equation V. X/DJ = 16. Within this central region of the jet. D0 )g(D0 )n(D0 )dD0 − g(D)n(D) . the value of g(ǫ. The phenomenological model for the daughter pdf presented in section V.32 were taken to be equal to the pdf measured at the ﬁrst measurement location. v ′ .1. .32 is simply expressed as δ(U (X) n(D))/δX. The initial conditions used to integrate equation V. and we measured only the bubble size pdf resulting from the break-up and transport processes at each downstream location. the process is one-dimensional. integrating equation V.32) where U (D) is the mean convective velocity of a bubble of size D. u′ . the term δ(U (D) n(D))/δX in equation V. a result consistent with early measurements . D0 ) which depends on D0 and ǫ.. i. is given in ﬁgures . Thus.1). In the steady state experiments described by our measurements show that all the bubbles are convected at the same mean velocity regardless of their size by the mean motion of the water jet.e.4 and the experimentally measured values of ǫ (ﬁgure IV.e. which in turn was always larger than the maximum radial distance where the bubbles were dispersed by the turbulence.3 takes the following form. at each downstream location. (V. Thus. were always nearly uniform throughout the volume of interest within 3 to 5%. The downstream evolution of the calculated cumulative bubble volume pdf . l1 . δ(U (D) n(D)) = δX ∞ D m(D0 )f (D. Since the problem of interest here is a steady state process. δn/δt = 0.3 in order to obtain the evolution of the bubble size distribution assuming a one-dimensional process. the coalescence eﬀects were negligibly small in our experiments since the bubble void fraction was always very small. i. As mentioned above. the turbulent kinetic energy was measured to be nearly uniform. equation V. the measured rms of the axial and transversal component of the jet velocity.B was then used to solve equation V. where U (X) is now the corresponding mean velocity of the water jet measured at each measurement location and X is the downstream location along the axis of the jet . In agreement with measurements of high Reynolds number axisymmetric jets. radial dispersion eﬀects did not aﬀect the measurements. .32 we used. it represents the spatial downstream evolution of the number density of the bubbles of size D as they are transported downstream by the convective velocity of the water jet and broken by the water turbulence with a frequency g(ǫ.
4 0.23 0. The solid lines represent the results obtained from the model integrating equation V. b) Downstream evolution of the Sauter Mean Diameter Diameter.5 1.5 D (mm) 2 2.5 Diameter (mm) Dv90% D32 Dv90% D32 1 0.86 1 Cumulative Volume Pdf 0.8 0. 000 m /s Ni D i at the air injection point. Experimental Set 2. D32 = 3 Ni D i 2 3 2 and Dv90% .11 X/D= 20.64 X/D= 33.6 0. X/DJ = 15.32 and the solid symbols are the experimental measurements.5 1 1. 15 20 25 X/DJ 30 35 40 È È .2 0 0 a) 2 X/D= 15 X/D= 20 X/D= 25 X/D= 35 X/D= 16.8: a) Downstream evolution of the Cumulative Volume Probability Density Function. Initial value of the dissipation rate of TKE was ǫ0 = 2.58 X/D= 24.5 0 10 b) Figure V.
X/DJ = 15. Initial value of the dissipation rate of TKE was ǫ0 = 1.5 Dv90% D32 Dv90% D32 15 20 25 X/DJ 30 35 40 0 b) Figure V.5 D (mm) 2 2.5 Diameter (mm) 1 0.4 0.07 0. i . The solid lines represent the results obtained from the model integrating equation V.39 X/D= 34.8 0. D32 = 3 Ni D i and Dv90% .32 and the solid symbols are the experimental measurements.65 X/D= 27.6 0.2 X/D= 22.5 3 1.5 1 1. 000 m2 /s3 N D2 È È i at the air injection point.9: a) Downstream evolution of the Cumulative Volume Probability Density Function.2 0 0 a) 2 X/D= 17 X/D= 22 X/D= 27 X/D= 37 X/D= 17.87 1 Cumulative Volume Pdf 0. Experimental Set 3a. b) Downstream evolution of the Sauter Mean Diameter Diameter.
Our model is based on the assumption that upon breaking.8a). W et = ρ △u2 (D0 ) D0 . This assumption appears to be consistent with all of our measurements performed at low and moderate values of the turbulent Weber number. m(D0 ) = 2. where they are compared to the experimental measurements. Note that the agreement with the experimental data is very good.88 Figure V. The images were taken at 6.8a and V.9b corresponding to case Set 3a). The same excellent agreement shown between the calculated and the measured pdf was also found for the other cases studied (see ﬁgures V. between the shape of the measured and the calculated cumulative volume pdf s. the bubble is split into two daughter ones.10: Time evolution of the characteristic break-up of a mother bubble showing a binary bubble splitting mechanism. (ﬁgure V. σ in which the surface of the mother bubbles is observed to .000 frames per second.8b).8b. not only concerning the various moments of the bubble size distribution (ﬁgure V. but more importantly. V.9a and V.
Da3 = 1.394 mm.584 mm. c) Experimental Set 3c.89 a) b) c) Figure V.11: Initial Bubble Size. Da1 = 0. b) Experimental Set 3b.194 mm. . Da2 = 0.a) Experimental Set 3a.
upon break. ﬁgure V. between two points separated a distance equal to the size of the bubbles were the only one producing the changes in the energy associated to surface tension. the surface of the bubbles are fairly smooth and we can even observe several events where a binary break-up is about to happen. increases. Thus. Da3 = 1.194 mm. A typical sequence of the evolution of a bubble during its break-up is shown in ﬁgure V. where we have presented the experimental measurements obtained in Set 3c and compared them with the predicted by our model assuming a binary break-up. This sequence was taken at 6. Observe that the bubble marked A1 breaks into two labeled A11 and A12 . However. To address the point of the possible formation of a large number of daughter bubbles. It is clear that the assumption of binary break-up may not hold in this case. we can estimate their maximum number with a simple energy argument. If the energy associated to the diﬀerence of velocities.12.1. This binary splitting process appears to continue until the break-up is ﬁnished.10. indicated as Set 3 in table IV. the amount of daughter bubbles. may result in larger production of small bubbles. we have observed that at larger values of the turbulent Weber number.11c. although diﬃcult to characterize. On the other hand.394 mm. This fact is very apparent in ﬁgure V. Figure V.394 mm to Da3 = 1.194 mm.90 be relatively smooth. the total energy associated to the diﬀerences of velocity and .11a. Notice that in ﬁgure V. the assumption of binary break-up appears to be well justiﬁed only at low and moderate W et .11 shows three instantaneous images of the initial bubble breakup process corresponding to the three diﬀerent air injection needles tested. at which the surface of the mother bubble is smooth and the formation of bubbles produced due to the local deformations on the surface of the bubble is negligible. produced by the turbulent ﬁeld. an issue that must be further investigated. To analyze the eﬀect of the initial size of the bubble in number of bubbles formed after its break-up and the shape of the resulting daughter bubbles pdf we conducted a set of experiments where the diameter of the air injection needle was varied from Da1 = 0. shows that the bubbles are more irregular with more protuberances on their surface which. m(D0 ). Da1 = 0. Subsequently A11 breaks into A111 and A112 . and neglecting viscous and compressibility eﬀects. and later on the bubble A12 breaks into A121 and A122 .000 frames per second and shows consecutive steps in the bubble break-up process.
8 0. Dv90% D32 Dv90% D32 20 25 X/DJ 30 35 40 È È . b) Downstream evolution of the Sauter Mean Diameter Diameter. X/DJ = 15. The solid lines represent the results obtained from the model integrating equation V.4 0.5 0 15 b) Figure V.5 3 3.32 and the solid symbols are the experimental measurements.07 0. 000 m /s Ni D i at the air injection point.91 1 Cumulative Volume Pdf 0.2 0 0 a) 2.12: a) Downstream evolution of the Cumulative Volume Probability Density Function.5 1 1.65 X/D=27. Initial value of the dissipation rate of TKE was ǫ0 = 1.30 X/D=22.6 0. Experimental Set 3c.5 1 0. D32 = 3 Ni D i 2 3 2 and Dv90% .39 X/D=34.5 2 Diameter (mm) 1.5 X/D=17 X/D=22 X/D=27 X/D=37 X/D=17.5 2 D (mm) 2.
(V.35) From conservation of volume. (V. f ∗ (D∗ ): 1 1 + A = A Nt 0 f ∗ (D ∗ ) D∗11/3 dD ∗ + Nt 0 1 f ∗ (D ∗ ) D∗2 dD ∗ .36) Equations V. and be rewritten as: 1+A =A i=1 ∗11/3 3 = D0 . Expressing equation V. . (V.272 * x^(3.32) R= 1 8 10 4 6 104 N to 4 104 2 104 0 0 10 20 30 A= ρ β ε2/3 Do5/3 / 12σ 40 50 Figure V. f ∗ (D ∗ ).13: Maximum number of bubbles resulting from the break-up of a mother bubble of size D0 .35 and V.33 can Ni Di + i=1 ∗2 Ni Di . 1 = Nt 1 0 ∗3 f ∗ (D∗ ) D ∗3 dD ∗ = D30 .33) i=1 where Ni is the number of bubbles of size Di . Thus.34 as a function of the daughter bubble pdf . i=1 (V. π 3 1 2 ρ β (ǫD0 )2/3 D0 + π σD0 = 2 6 π 3 1 ρ β Ni (ǫDi )2/3 Di + π σ 2 6 m 3 i=1 Ni Di 2 Ni Di . surface tension should remain constant. and D∗ = D/D0 is a nondimensional diameter of the resulting daughter bubble.92 1 105 y = 0. Equation V.34) where A = ρ 2/3 5/3 D0 = 12 σ ǫ δ−5/3 is proportional to the bubble’s turbulent Weber num- ber.36 could be solved to obtain the number of bubbles resulting from the break-up of a mother bubble D0 and the daughter bubbles pdf .
Now. V. now although the agreement is reasonable.7. a number which is very much diﬀerent from the maximum possible number of daughter bubbles.12). 000 bubbles of diameter Db = 72 µm. the number of daughter bubbles obtained.9). gives.35 and V.36 in conjunction with the equation V. small values of W et and that at large values of W et the number of daughter bubbles could be greater than two. This increase in the number of daughter bubbles with W et could explain why we obtained a poorer agreement between the pdf calculated with our model and the experimental measurements for the case of large W et (see ﬁgure V. and closer to reality. it is reasonable to expect that our assumption of m(D0 ) = 2 will hold only for small values of A or. Nt0 . shown in ﬁgure V.38) Solving equation V.39) For the particular case of ǫ = 1.53 . In fact.37 for the real. Thus. and equations V. Nt0 ≈ 0.93 For example. equivalently. f ∗ (D/D0 ).272 δ−5. as opposed to the cases where the deformations on the surface of the bubbles were relatively small and the agreement was excellent (ﬁgures V.11 for f ∗ (D/D0 ) are satisﬁed. (V. To consider . assuming that all of them are equal. the maximum number of daughter bubbles. non-trivial solution of Nt0 . 000 m2 /s3 and D0 = 2 mm. Note that.35 and V. there is a noticeable diﬀerence between the calculated pdf and the measured ones. as in the experimental set 3c.8.36.37 is only a result of the energy arguments used in its derivation. substituting the daughter bubbles pdf . predicted by our model into equations V. the daughter bubble size pdf would be a delta function at D = Db .37) 1/3 .37 is Nt0 ≈ 21. we obtain A = 29. (V.36 would simplify to: Nt0 and Db = 1 Nt0 1/3 A Nt0 5/9 + 1 =1 +A . and obtain a new number of daughter bubbles until equations V. we could use this number to recalculate the daughter bubbles pdf . given by equation V. For such value of A.32 = 0. Obviously this is not the case observed in our experimental results. (V. the large amount of dauther bubbles given by equation V.13.272 A3. for the same value of A. f ∗ (D/D0 ).35 and V. if all the daughter bubbles were of equal size with a diameter Db . is approximately 3.
42) Equation V. ǫ = 1000 m2 /s3 . Dmin ≤ D1 ≤ Dmax and Dmax is now given by Dmax = D0 − 2 12 σ β ρ D0 9/2 1/3 .41 can be used as an extension of our bubbles break model for cases where the estimated number of bubbles resulting in the break-up is larger than two.94 1. (V.5 Binary Tertiary 1 f*(D/D0) 0. (V.6 0.5 0 0 0.2 0. D0 = 2 mm a tertiary break process.8 1 Figure V. although a more elaborated model contemplating all the possible combinations of three bubbles of diameters D1 . d(D ∗ ) (V.40) the daughter bubbles pdf resulting from the break-up of a mother bubble of size D0 is obtained as: D ∗2/3 − δ5/3 ∗ Dmax ∗ Dmin f ∗ (D1 /D0 ) = 1−D ∗3 2 1−D ∗3 2 2/9 2 − δ5/3 2 D ∗2/3 − δ5/3 2/9 .41) − δ5/3 where D∗ = D1 /D0 . The daughter bubbles pdf .14: Comparison of the daughter bubbles pdf resulting from a binary and a tertiary splitting. D2 and D3 is being presently considered. if for simplicity we assume that the bubbles are broken in a bubble of diameter D1 and two bubbles of equal diameter D2 .4 D/D0 0. such that 3 3 D2 = D0 − 2 D1 1/3 .
The solid lines represent the results obtained using our model assuming a tertiary breakup process model to integrate equation V. along with the experimental results. in ﬁgure V. It is important to emphasize that. in the break-up process.20 X/D= 34.5 X/D= 17 X/D= 20 X/D= 26 X/D= 35 X/D= 17.5 2 D (mm) 2. it is expected to result in a more rapid break process than that considering only two shattering bubbles.32.15: Downstream evolution of the Cumulative Volume Probability Density Function.44 X/D= 26.6.07 Figure V. as opposed to previously proposed models.2 0 0 0. Notice that the maximum of the pdf has been shifted to D/D0 = 0. Experimental Set 3c.41 is shown.15 for the experimental set 3c.5 3 3. The agreement between the model and the experimental results is excellent conﬁrming that. The downstream evolution of the Cumulative volume pdf obtained using the model presented in equation V.6 0. A comparison of the daughter bubble size pdf predicted by our model and those predicted by previously proposed models is shown in ﬁgure V. resulting from equation V.30 X/D= 19. the probability of formation of large bubbles has decreased considerable and therefore.4 0.95 1 Cumulative Volume Pdf 0. and the solid symbols are the experimental measurements.5 1 1.41 is shown in ﬁgure V.7 which corresponds to three bubbles of the same size.14 compared to that obtained using a binary break-up process for the same ﬂow conditions.8 0. the probability density function f ∗ (D/D0 ) given by our model is a function of not only the size of the mother . the number of bubbles created and the shape of the duaghter bubbles pdf is strongly dominated by the magnitude of the turbulent stresses.
5 Diameter (mm) 1 0.16: a) Downstream evolution of the Cumulative Volume Probability Density Function.11 X/D= 20.The solid lines represent the results obtained using Tsouris and Tavlarides 1994  model to integrate equation V. and the solid symbols are the experimental measurements.58 X/D= 24.96 1 Cumulative Volume Pdf 0.2 0 0 a) 2 X/D=20 X/D=25 X/D=35 X/D= 16. b) Downstream evolution of the sauter Mean Diameter and Dv90% .5 3 1.5 D (mm) 2 2.32. Experimental Set 2.23 0.4 0.5 Dv90% D32 Dv90% D32 10 15 20 25 X/DJ 30 35 40 0 b) Figure V. .8 X/D=15 0.5 1 1.6 0.64 X/D= 33.
5 0 10 b) Figure V.8 0. 1983  model to integrate equation V. 15 20 25 X/DJ 30 35 40 .5 1 1. and the solid symbols are the experimental measurements.64 X/D= 33. b) Downstream evolution of the sauter Mean Diameter and Dv90% .4 0. Experimental Set 2.2 0 0 a) 2 X/D=15 X/D=20 X/D=25 X/D=35 X/D= 16.5 D (mm) 2 2.11 X/D= 20.97 1 Cumulative Volume Pdf 0.58 X/D= 24.5 Diameter (mm) Dv90% D32 Dv90% D32 1 0.5 1.17: a) Downstream evolution of the Cumulative Volume Probability Density Function (Vpdf).23 0.6 0. The solid lines represent the results obtained using Konno et al.32.
but also of ǫ. Notice that in ﬁgure V. To illustrate this point we have plotted f ∗ (D/D0 ) corresponding to three diﬀerent values of ǫ (ǫ = 4. where we compare them to the experimental measurements.6. As the bubbles are broken. It is clear that due to the fact that the daughter bubble pdf is independent of ǫ.3D0 < D1 < D0 and. it is more probable to form bubbles of similar size which translates into a faster evolution of Cumulative Volume pdf shown in ﬁgure V. As the value of ǫ increases.5. The resulting evolution of the cumulative volume pdf is shown in ﬁgure V. this model grossly underpredicts the evolution of the volume pdf . using the daughter bubble pdf proposed by Konno et al. Although the frozen pdf . the spectrum of bubble sizes resulting from the turbulent break-up of a mother bubble D0 is wider. D0 . and therefore. as shown in ﬁgure V. the temporal evolution is not very well described. The downstream evolution of the Cumulative Volume pdf obtained by integration equation V.32. is shown in ﬁgure V. 10 m2 s−3 .32 using Tsouris and Tavlarides’ model in the same manner as we did for our model. The large discrepancy between the measured and calculated evolutions of the D32 is due to the fact that this model gives the highest probability to the formation of a very large and very small pair of bubbles.3D0 after the break-up of a mother bubble of size D0 is negligibly small and independent of the value of both ǫ and D0 . Therefore. The discrepancy between the experimental results and the results obtained with this model is very apparent. resulting in a larger amount of small bubbles in the early stages of the break process (close to the point of injection) where the value of ǫ is the largest. and 100 m2 s−3 ).16.. obtained with this model seems to be in excellent agreement with the experimental results. the probability of producing bubbles smaller than 0. .17. reached at X/DJ = 35.98 bubble. Note that the model given by Tsouris and Tavlarides 1994 has a U shape giving a minimum probability for the formation of two daughter bubbles of equal volume (D/D0 = 0. a fact which brings to question the validity of the basic assumptions made in the model. and a maximum probability for the formation of widely diﬀerent volume pairs. the size of the daughter bubbles created with this model are always in the range 0. f ∗ (D/D0 ) widens out toward smaller sizes.17. We have integrated equation V. This discrepancy is not surprising since.5 m2 s−3 . producing smaller bubbles. the probability of forming small bubbles predicted by our model increases as the value of ǫ is increased. after being injected in the ﬂow.8).
Monta˜es. . C. Lasheras currently being considered for publication a n in Journal of Fluid Mechanics. This chapter is. and J. In this evolution. L. This explains why although initially the calculated cumulative Volume pdf obtained using our model are diﬀerent from that obtained using Konno’s model. in part.99 by the turbulent stresses of the water jet they are transported further downstream in a continuous break-up process. comprised of a paper titled “On the size pdf resulting from the break-up of an air bubble immersed into a turbulent liquid ﬂow”. to locations of lower value of ǫ. the shape of our daughter bubble pdf becomes closer to that proposed by Konno et al. This decrease of ǫ and D0 as the bubbles are convected downstream in the ﬂow produces an evolution on the daughter bubbles pdf increasing the value of the peak and therefore increasing the probability of forming bubbles of equal size. Mart´ ınezBaz´n. C. J. the ﬁnal stage of the frozen pdf is well reproduced by both models. The dissertation author was the primary researcher and author listed in this publication. by the mean motion of the ﬂow.
and the bubble size distribution can be measured using a Phase Doppler Particle Analyzer (PDPA).A Introduction In chapters IV and V we studied the time evolution of the bubble size pdf and proposed phenomenological models for the bubble break-up frequency and the daughter bubble size pdf . and on the initial value of the air void fraction (deﬁned here as the ratio between the air ﬂow rate and the water ﬂow rate. 100 . reached once the break-up has concluded and an unchanged bubble size pdf has been achieved. their shape is nearly spherical. we will analyze the characteristic of the inter-arrival time/distance (the time/distance between two consecutive bubbles of the same size). on the turbulent kinetic energy of the underlying turbulence at the air injection point.Chapter VI Characterization of the Frozen State VI. is the shape of the frozen bubble size pdf s and their dependence. and in order to study the eﬀect of breakup and later dispersion by the turbulent water jet on particles of diﬀerent diameters. Another important issue to address. due to its relevance in mass and heat transfer processes. Due to optical lim1 The frozen state refers to the state of the pdf at which no more changes due to break-up are observed. on this type of ﬂows (turbulent jet). α = Qa /Qw ) introduced in the ﬂow. Particularly. In this chapter we will focus our attention on the frozen state1 of the bubble’s break-up process. Since the frozen pdf s are achieved at a location where the bubbles are in equilibrium with the turbulence.
the characteristics of the pdf s of the time and distance between two consecutive bubbles of the same size (inter-arrival time and inter-arrival distance respectively) will be analyzed.101 itations of this instrument. This chapter concludes with the study of the eﬀect of the initial air void fraction on the frozen bubble size pdf . Based on detailed measurements of the axial and radial variation of the bubble pdf s. consequently. The measured probability density functions of the size of the resulting bubbles were found to quickly evolve downstream from the injection point until they eventually reached asymptotic. the turbulent kinetic energy of the underlying turbulence was maximum. and the injected gas began to break under the eﬀect of the turbulent stresses exerted by the surrounding water. we can compute the time and. frozen shapes. remain unchanged as one moves downstream. the bubble pdf s.1). it can be concluded that in all the experiments presented here the frozen state was reached at a downstream distance smaller than 20 jet diameters. the air was injected at 10 diameters downstream from the nozzle exit section. the diameter of the frozen bubbles generated was always selected to be between 10 µm and 500 µm in order to be able to measure the size of the bubbles accurately.B Statistics of the bubble inter-arrival time In addition to study the characterization of the break-up process and the time- averaged bubble size pdf we have also analyzed the time series of the bubbles measured by the PDPA. This process continued downstream from the injection point until a location was reached where an frozen shape of the bubble pdf was achieved. In the following. Since the turbulent kinetic energy of the underlying turbulence of the submerged water jet decays with the downstream position (see ﬁgure IV. 10 jet diameters from the air injection point. VI. As mentioned before. upon reaching a frozen shape. and no further break-up takes place. at the injection point. the distance separating two consecutive bubbles of the . Since we are able to measure not only the size of the bubbles crossing throughout the probe volume but also their velocity and the arrival time. To achieve these conditions. Subsequently the variation of the frozen bubble’s size pdf s with the value of the turbulent kinetic energy existing at the point of air injection will be discussed.
Here we want to present some experimental measurements of the time series to establish if the break-up of bubbles by a turbulent ﬂow contains some sort .1 time (s) 0. and therefore analyze the ﬂuctuations of the concentration of bubbles and its dependence on the bubble size. dissipation rate of TKE and initial void fraction. 000.1 by the concentration of a large number of bubbles within some intervals of time may indicate the presence of some kind of time correlation or it may be just the consequence of a random occurrence. deﬁned as St = D 2 u′ 36 ν δ (VI. the eﬀect of any clustering in the time series will be only the consequence of the break-up process. the time dependence of the process is missing.1 shows a plot of a time series data of the bubble size diameter.1) is small in the range of bubble size of interest here. This information can be very useful in order to study clustering events of bubbles of diﬀerent sizes or infer the formation frequencies of bubbles if they are responding to certain frequencies of the ﬂow. Figure VI.1: Time record of bubbles generated by a water jet. In the analysis of the bubble-size probability density functions.15 Figure VI.102 300 250 Diameter (µm) 200 150 100 50 0 0 0. Re = 53. It is expected that bubbles of similar diameter will behave similarly in terms of formation and dispersion throughout the spray.05 0. same size. If the bubble’s Stokes number. It looks as if could be some clustering of occurrence times of the bubbles. The possible clustering shown in ﬁgure VI.
is not represented due to the very small number of bubbles collected in this particular bin. 200 µm > D.04 t (s) Figure VI. This provided information on the temporal variation of concentration of bubbles of diﬀerent sizes as explained above. For this purpose. the size distribution was divided in ﬁve size-bins. we should subdivide the bubble sizes in classes which might behave equally. In order to look for any possible time correlation in the break-up of bubbles. The time separating the arrival of two consecutive bubbles which belong to the same class was recorded. of time-correlated behavior.025 0. Symbols are: ◦ 3 µm < D < 20 µm.005 0. 3 µm < D < 20 µm.015 0.103 1000 3 < D < 20 40 < D < 60 80 < D < 100 120 < D < 200 100 PDF(t) 10 1 0 0. 000. 40 µm < D < 60 µm. The inter-arrival time pdf s have been plotted in a log-linear scale to highlight the fact that they are exponentially distributed. 40 µm < D < 60 µm.2: Inter-arrival time pdf s of diﬀerent classes of bubbles.02 0. The probability density functions of the inter-arrival time corresponding to four of the size bins are given in ﬁgure VI. ⋄ 80 µm < D < 100 µm.03 0.01 0. 120 µm < D < 200 µm. the probability of the inter-arrival time between two consecutive bubbles increases as the inter-arrival time decreases. Therefore the inter-arrival time pdf s do not show . There is not a predominant frequency (or time) at which bubbles are produced. Re = 53.2 for the case of Re=53. Notice that for the four sizebins presented.035 0.000 and ﬁxed ﬂow rate of air. and 200 µm > D. △ 120 µm < D < 200 µm. 80 µm < D < 100 µm. The distribution corresponding to the ﬁfth sizebin.
000.3. P DF (t) = λ e−λ t .3: P df s of the normalized Inter-arrival time. a ﬂow in which bubbles of size D1 are formed at a ﬁxed frequency of 100 Hz. for example. △ 120 µm < D < 200 µm.3) The probability density functions of the inter-arrival time. Imagine.001 0 1 2 3 t/tm 4 5 6 Figure VI.104 1 3 < D < 20 40 < D < 60 80 < D < 100 120 < D < 200 PDF(t/tm) 0. Note that we are able to represent with a unique distribution the pdf of the dimensionless time t∗ = λ t. 000. The slopes of the tails represent the mean rate of arrival of bubbles of each size class. 40 µm < D < 60 µm.1 0. ∗ (VI. for the above cases has been represented in ﬁgure VI. If we calculated the pdf of the time between two consecutive bubbles of size D1 we would ﬁnd a strong peak in the distribution for a value of t equal to 10 ms. normalized by the mean value. Re = 53. ⋄ 80 µm < D < 100 µm. As .01 0.2 correspond to an intermediate Re = 53. t/tm (tm = mean inter − arrival time) of diﬀerent classes of bubbles. a high concentration of bubbles at that characteristic frequency. An exponential distribution is given by the following equation. it can be clearly seen that the pdf s follow exponential distributions of diﬀerent mean value. t∗ = t/tm = λt. Symbols are: ◦ 3 µm < D < 20 µm. tm = 1/λ. (VI.2) where 1/λ is the mean as well as the root-mean-square value of the distribution. Although the data presented in ﬁgure VI. P DF (t∗ ) = e−t .
It provides information on the number of particles crossing the probe volume during a certain data acquisition time. n! (VI.3.02) = 0. as one would expect. λ. λ. 2. we will start by giving its expression and analyzing the assumptions needed in the process. .1 The Poisson random process The Poisson distribution is commonly used in waiting-line evaluations. To study the characteristics of Poisson processes. t) represents the probability of n occurrences at a rate λ during an interval of time t. 1. VI. This result would indicate that individual bubbles within the turbulent jet act independently of each other and. see . The average . Similarly the exponential distribution is used to evaluate the time between arrivals. All Poisson processes share the following properties: • The underlying rate under which the events occur.105 indicated by equation VI. making it a centerpiece in the theory of queues or waiting lines. t) = (λ t)n e−λ t n = 0. when using the non-dimensional variable. the probability of counting n=10 bubbles crossing through the measuring point during t=0. 0. p(n. the exponential approximation may not apply. A new application of both distributions has been found lately in the ﬁeld of turbulent sprays . the movement of one of them is not conditioned by the neighboring ones. 500. In a more dense case. .B. In our particular case the Poisson distribution concerns only the arrival time of bubbles at the measuring point. . it appears that the process of formation on bubbles by a high intensity homogeneous turbulence is a random process which can be well represented by a Poisson distribution. t∗ . provid- ing probabilities of the number of arrivals at a certain point.4) where p(n. it is very important to introduce at this point some of the basic assumptions governing the Poisson random process and their similarities to the bubble break-up problem. where the particle’s movement may inﬂuence the others. is constant in time. the pdf s(t∗ ) follow the same exponential distribution with slope -1. for example. Before pursuing any further. λ. .02 seconds when the mean crossing frequency is λ = 500 is p(10.125. Since the exponential distribution provides the probability for the time between successive events occurring in a Poisson process. therefore.
Events in nonoverlapping intervals of time occur independently. All the above assumptions hold true for the exponential distributions since they are related to the Poisson process. Tn are disjoint subset of T. λT2 . examples exist wherein the assumptions don’t strictly apply. . the inter-arrival time. . • A Poisson process has no memory. T3 . the number of events in these subsets. The practical value lies in how closely the theory reproduces our experimental data. . • There are no multiple events. . it implies that at least two bubbles have to be generated at the same time. n1 . Although in a Poisson process two events cannot happen exactly at the same time. n3 . on the inter-arrival time of the various bubble size-bins. . VI. λT3 . n2 . The next section focusses on the eﬀect of the turbulent kinetic energy. When a bubble breaks. So far we have shown the probability density function of the inter-arrival time of bubbles of the diﬀerent size-bins and how their production is similar to a Poisson random process.106 number of bubbles crossing the probe volume per unit time is constant and only depends on the ﬂow conditions and the amount of air injected into the ﬂow. Similarly. . Ci Ui A (VI. This property would not be realistic if we think in the formation mechanism of bubbles from the break-up of a mother bubble by the turbulent ﬂow. Consequently. the number of bubbles measured during an interval of time T2 is independent of the number measured during a previous interval T1 . nn are independent random variables that follow Poisson distributions with parameters λT1 . . the average number of events in a subset is proportional to its size. Therefore. The events in disjoint subset are independent of each other and the Poisson parameter λTi is proportional to the subset size. Therefore. at the injection point of air.5) where Ci = number of particles / V olume and Ui are the concentration and velocity of a bubble sized in class i.2 Eﬀect of the Turbulent Kinetic Energy on the inter-arrival time. The time between two consecutive particles which belong to the same size-bin can be deﬁned as: △ti = 1 . .B. we only measure one bubble crossing the probe volume at a time. and A is the jet cross-section. . If T1 . λTn . . . . . T2 .
( ). these type of bubbles are broken as ǫ is increased.107 0.5 6 Velocity (m/s) 6. while large bubbles within class 5. in addition to increasing the velocity of the bubbles. increases as the concentration.5 7 7. and the velocity. 40 µm < D < 60 µm.000. although their velocity increases. △ti . Ci . In fact. is unity in all cases as it would be expected if the inter-arrival pdf s . U .5 shows that the ratio of the rms to the mean value. Symbols are: ◦ 3 µm < D < 20 µm. (◦). As the value of the mean velocity of the jet.4. ﬁgure VI.01 T1 T2 T3 T4 T5 0. △ 120 µm < D < 200 µm. ǫ.1 Time (s) 0. decreases. and class 3. lessen their inter-arrival time until a point where the break-up eﬀect is dominant and it starts increasing due to fact that the number bubbles of this class diminishes. Therefore. The behavior of the standard deviation of the inter-arrival time is identical to that of the mean value. becomes larger.4: Evolution of Mean Inter-arrival time with the local convective velocity. (△). equation III. The Reynolds Number of the jet has been varied from 32. the Reynolds number of the jet increases and consequently so does the value of the turbulent kinetic energy.4 shows the evolution of the mean inter-arrival time of the ﬁve bubble size-bins deﬁned above for various ﬂow conditions. ⋄ 80 µm < D < 100 µm. Figure VI. the number of bubbles may increase as well due to the break-up since the turbulent kinetic energy of the ﬂow is larger. (⋄). t′ /tm . Ui .000 to 60.5 5 5. increase their inter-arrival time since. D > 200 µm. Ui . ( ). This is true for bubbles of small size which belong to class 1. class 2. Bubbles of size class 4. U .5 Figure VI.001 4 4.
5 6 Velocity (m/s) 6. U . Selecting a non-dimensional variable of time.5 Figure VI. The results are consistent with those shown in ﬁgure VI. has a self-similar behavior that can be perfectly represented by an exponential distribution.108 1. 40 µm < D < 60 µm.3 1. ⋄ 80 µm < D < 100 µm. therefore. or similarly with respect to ǫ. Error bars are 5% of the value. To study the variation of the inter-arrival time with respect to the Reynolds number of the water jet.6 4 T1'/T1 T2'/T2 T3'/T3 T4'/T4 T5'/T5 4. . the product of a random process. and the possible clustering (if any) mentioned at the beginning of this section is.The Reynolds Number of the jet has been varied from 32.7 0. D > 200 µm.4 1. the probability density function.7. 40 µm < D < 60 µm.5 7 7.9 0. The self-similar behavior is independent of the Reynolds number and diameter of the bubble.000. △ 120 µm < D < 200 µm.8 0.4. Symbols are: ◦ 3 µm < D < 20 µm. t/tm (tm is the inter-arrival mean value). as shown in ﬁgure VI. which is the same in this case since the Re is varied by changing the velocity at the exit of the nozzle. the inter-arrival time of the class 2 bubbles decreases since their convective velocity is faster and more and more bubbles are produced due to the break-up of larger ones.5: Evolution of Rms/Mean Inter-arrival time with the local convective velocity. As Re is increased. pdf (t/tm ). we have plotted in ﬁgure VI.000 to 60.6 the pdf s of the inter-arrival time of bubbles of size class 2.1 t'/tm 1 0. pdf (t/tm ) = exp [−t/tm ]. were in fact exponential distributions. at three diﬀerent values of the Reynolds number.5 5 5.2 1.
000. 000. ⋄ Re = 60. 000.000 Re= 60.06 Figure VI.000 Re= 60. 000.02 0.01 0. .000 Re= 45. 1 Re= 29.1 PDF(t/tm) 0.05 0. Re = 45. Re = 45.001 0 1 2 3 t/tm 4 5 6 Figure VI. 000.109 1000 Re= 29.7: Evolution of the normalized Inter-arrival time pdf of bubbles of size 40 µm < D < 60 µm with the Reynolds Number.000 0. ⋄ Re = 60.01 0.000 100 PDF(t) 10 1 0 0. Symbols are: ◦ Re = 29. 000.000 Re= 45.6: Evolution of the Inter-arrival time pdf of bubbles of size 40 µm < D < 60 µm with the Reynolds Number.04 0. Symbols are: ◦ Re = 29.03 t (s) 0.
000. Re = 53. are very informative.8 for the same bubble classes and experimental conditions as those presented in ﬁgure VI.2. The distance between two consecutive bubbles of the same class is calculated from their corresponding inter-arrival time. 40 µm < D < 60 µm. as: △ li = △ ti Ui = 1 .C Inter-arrival distance between bubbles The properties of the time between two consecutive bubbles. Figure VI. Ci A (VI. 0. and their convective velocity. Another important aspect of the process is the knowledge of the concentration of bubbles.6) The probability density function of the distance between two consecutive bubbles of the same size is shown in ﬁgure VI. Ui .25 VI. λ. in other words. shows that the steepest slope of the inter-arrival time pdf s .8: Inter-arrival distance pdf s of diﬀerent classes of bubbles. △ ti . the knowledge of the distance between two consecutive bubbles which belong to the same bubble size-class. Ci . the pdf of the inter-arrival distance of bubbles within the same class follows an exponential distribution with an intensity parameter. As would be expected. in the previous section. Symbols are: ◦ 3 µm < D < 20 µm.05 0. explained in the previous section.1 0 0. △ 120 µm < D < 200 µm.15 0.2 0.2. which only depends on the number (concentration) of particles of a certain class “i”.110 100 3 < D < 20 40 < D < 60 80 < D < 100 120 < D < 200 10 PDF(l) 1 0. ⋄ 80 µm < D < 100 µm. since bubbles of the same size move at the same velocity.1 l (m) Figure VI.
The Reynolds Number of the jet has been varied from 32. 40 µm < D < 60 µm. . bubble size-class 4 (120 µm < D < 200 µm) and.5 6 Velocity (m/s) 6. Symbols are: ◦ 3 µm < D < 20 µm. indicating that the convective velocities of all bubble-size classes. D > 200 µm.1 Distance (m) 0. △ li = △ ti Ui . Ui .000. are the same and equal to that of the main ﬂow. followed by bubble-size class 3 (80 µm < D < 100 µm). △ 120 µm < D < 200 µm. is the inverse of the mean value of the inter-arrival distance. the next section will discuss the evolution of the pdf s(△ li ) with the Reynolds number or. the least steep one corresponds to bubble-size class 1 (3 µm < D < 20 µm).000 to 60. The same trend is appreciable in ﬁgure VI.5 Figure VI. lm .111 1 L1 L2 L3 L4 L5 0.01 4 4.8 for the pdf s of the inter-arrival distance.5 5 5. U . ǫ. similarly. corresponds to the bubble-size class 2 (40 µm < D < 60 µm). of the underlying turbulence at the air injection point. Since the pdf (△ li ) are exponential distributions whose intensity factor. the evolution of the pdf s(△ li ) with the dissipation rate of turbulent kinetic energy. λ = 1/lm .5 7 7.9: Evolution of Mean Inter-arrival distance with the local convective velocity. ⋄ 80 µm < D < 100 µm. ﬁnally.
10: Evolution of the inter-arrival distance pdf of bubbles of size 40 µm < D < 60 µm with the Reynolds Number.05 0.01 0. Symbols are: ◦ Re = 29.000 0.1 0.1 PDF(l/lm) 0.000 Re= 60. 000. Symbols are: ◦ Re = 29.000 Re= 45. Re = 45.000 Re= 60.15 l (m) 0.000 Re= 45. . 000. 1 Re= 29.3 Figure VI.1 0 0.001 0 1 2 3 l/lm 4 5 6 Figure VI.112 100 Re= 29. Re = 45.000 10 PDF(l) 1 0. ⋄ Re = 60. 000. 000. 000. ⋄ Re = 60.2 0.25 0. 000.11: Evolution of the normalized Inter-arrival distance pdf of bubbles of size 40 µm < D < 60 µm with the Reynolds Number.
Therefore. each one characterized by a continuous intensity function.65 m/s. while the concentration of bubbles between 100µ m and 120µ m (△) remains practically unchanged until eventually they start breaking and. Edwards and Marx  developed a theoretical framework for the analysis of the time-based statistics of sprays. On the other hand. all the pdf s of their inter-arrival distance collapse in a self-similar exponential function when the inter-arrival distance. therefore. As ǫ increases the concentration. The number of bubbles per unit volume measured at a ﬁxed location depends on the turbulent kinetic energy at the location where the air is initially injected. they are created and dispersed in a random way.1 Eﬀect of the Turbulent Kinetic Energy of the ﬂow on the interarrival distance. the measured bubbly jets behaves as an ideal spray driven by many superposed Poisson processes (of bubble sizes). instead. and establishing an analogy with the properties of an ideal spray .10 representing the pdf of the inter-arrival distance of bubbles of diameter D such that 40 µm < D < 60 µm. their concentration decreases. as shown in ﬁgure VI. Analogous to the pdf of the time between to consecutive bubbles of the same size. the concentration of bubbles larger than 200 µm ( ) diminishes with ǫ since these are the bubbles most likely to break. The tails of the pdf s become steeper as the Reynolds number increases. l.113 VI. or number of particles per unit volume. the characteristics of the family of bubbles formed by the break-up of a volume of air introduced into a turbulent water ﬂow are the following: . Based on their approach. L2 and L3 in ﬁgure VI. lm . The same type of results are shown in ﬁgure VI. The mean inter-arrival distance of bubbles of diameters smaller than 100 µm (L1 . shown at U ≈ 5. we observe that the bubbles are not formed in a predictable fashion but. indicating the growth in the number of bubbles of this class with ǫ. of large sized bubbles is expected to decrease while the concentration of smaller ones will increase.9.11. The mean value of the inter-arrival distance between two consecutive bubbles of various sizes is presented in ﬁgure VI. is normalized by their corresponding mean. From the above results on the inter-arrival time an inter-arrival distance.9) decreases.C. as ǫ is increased. This is a consequence of the augmentation of the number of bubbles smaller than 100µm in the break-up process due to the increase of ǫ.
The bubble ﬁeld is not highly ordered. VI. etc. The pdf s of the inter-arrival time or distance between two consecutive bubbles of the same class follow the exponential distribution. all bubbles. X/DJ = 10. Similar type of measurements have been presented by Edwards and Marx . These mea- surements were obtained by ﬁxing the injection ﬂow rate of the air while systematically increasing the Reynolds number of the water jet.12. size. developed by . which follow the ﬂow. in the range of bubble size and jet Reynolds numbers presented here. characteristic of steady state case. Measurements at the edge of the water jet would indicate a preferential concentration of those bubbles transported by large scales. This means that the probability of ﬁnding a bubble at a certain location at a ﬁxed time is only a function of the location and the time and not whether others bubbles exist nearby. DJ . have been kept constant. the Reynolds number of the submerged jet. since the diameter of the water nozzle. For convenience.114 .The statistics of the bubble ﬁeld are not aﬀected by events in the past or in the future. feel the inﬂuence of the water jet which. These results would be shown in the above pdf s as a strong deviation from their exponential distribution.e. measured in the internal shear layer of a kerosene spray ﬂame. Therefore it is in some sense random. Therefore. . . for very small particles. produced by the break-up of an air mass injected into a turbulent water jet. at its axis.Each of the particles carries a set of marks that represent the characteristic of the bubble being considered. . only by the present.D Dependence of the frozen bubble size pdf on ǫ The measured frozen bubble size pdf s are shown in ﬁgure VI. which in turn represents a certain value of ǫ at the injection point. and the air injection point. All bubbles. . are created and transported downstream in the same way. Individual behavior of diﬀerent bubble-size classes is independent on each other and the entire ﬁeld may be composed by superposition of single-class behaviors.The bubbles are modeled as non-interacting particles. can be characterized as isotropic and homogeneous turbulent ﬂow.. The eﬀect of bubbles collision is negligible. velocity. Each one corresponds to a given jet Reynolds number. i.
0.03 0.025 0.02
Re=26776 Re=37487 Re=48197 Re=53552 Re=66940 Re=93717
0.015 0.01 0.005 0 0 50 100 150 200 250 300 350 400
Figure VI.12: Frozen bubble size pdf for various Reynolds numbers of the submerged water jet. Re =
U 0 DJ ν ,
will be used as a parameter to indicate increasing levels of turbulent ki-
netic energy existing at the bubble’s injection point. The ﬁrst important conclusion, evident from these results, is that the shape of the bubble size pdf strongly depends on the turbulent kinetic energy (or dissipation rate ǫ) of the underlying turbulence. As the value of ǫ at the air injection point increases, the lump of air injected into the turbulent jet is subjected to stronger turbulent stresses. Consequently the value of the critical bubble’s diameter, Dc , decreases. This result is apparent in ﬁgure VI.12 where it is seen that as the Reynolds number is increased the pdf becomes narrower, indicating that the large bubbles have been broken, generating, a large number of small ones. In order to correlate the experimental bubble size distributions with the magnitude of the turbulent kinetic energy of the underlying turbulence, we considered various known two-parameter distribution functions2 . Inspection of these pdf s revealed that for a wide range of values of the turbulent kinetic energies, their shape appeared to be very well ﬁtted by Gamma
A review of the distribution functions most commonly used is given in Appendix A.
0.016 0.014 0.012
Re=53552 Gamma Re=48197 Gamma
0.01 0.008 0.006 0.004 0.002 0 0 50 100 150
Figure VI.13: Comparison of the experimental bubble size pdf with a Gamma distribution at Re = 53552. distributions (ﬁgure VI.13) given by: P DFg (D) = cb D b−1 e−cD , Γ(b) (VI.7)
where Γ(b) is the gamma function of b, D diameter of the bubble, b = ( Dmean )2 , and Drms c=
Dmean 2 Drms .
Dmean and Drms are the mean and the standard deviation of the probability
density function P DFg (D) respectively. If the pdf s follow a Gamma distribution and D ∗ = c D, P DFg (D ∗ ) = D ∗(b−1) −D∗ . e Γ(b)
Notice that if b = ( Dmean )2 reached a constant value within a range of Reynolds number, Drms P DFg (D ∗ ) would be independent of Re , a point that will be addressed later on in this chapter. Equation VI.8 can be expressed as: Γ(b) P DFg (D∗ ) ∗ = e−D , ∗(b−1) D (VI.9)
showing that the term on the lhs of equation VI.9 is an exponential function of D ∗ . Thus, rescaling the bubble size pdf with Γ(b) and D∗(b−1) as shown in equation VI.9, the experimentally measured pdf should follow an exponential distribution if indeed their
PDF(D*) Γ(b) D* -(b -1)
0.01 0.001 0.0001 10- 5 10- 6 0 2 4 6
Re=37487 Re=48197 Re=53552 Re=66940 Exponential
Figure VI.14: Normalization of Gamma pdf s. shapes corresponded to Gamma distributions. Figure VI.14 shows the good agreement of the Gamma distribution with the experimental measurements for a wide range of Re . At low levels of the turbulent kinetic energy, the resulting shape of the bubble pdf s, shown in ﬁgure VI.15, were found to greatly deviate from Gamma distributions. Furthermore, we have also observed that for very large values of the turbulent kinetic energy the bubble size pdf s are best represented by log-normal distributions, P DFlog (D) = √ (ln(D) − µs )2 1 , exp − 2 2 σs 2π D σs (VI.10)
where µs and σs are the mean and root-mean-square of ln(D). This result agrees well with statistical models based on the Central Limit Theorem, . If now, D ∗ = (ln(D)− µs )/σs , the probability density function of D ∗ , P DFlog (D∗ ) = D σs P DFlog (D), is then given by a normal distribution: D ∗2 1 . P DFlog (D ∗ ) = √ exp − 2 2π (VI.11)
The experimental pdf (D ∗ ) measured at Re = 93717 is given in ﬁgure VI.16, showing the good agreement with a log-normal distribution at this high value of Re ( very large values of the turbulent kinetic energy).
006 0.118 0. .5 10−4 . Re= 93.15: Comparison of the experimental bubble size pdf at low Reynolds number.012 0.1 PDF(D) D σ 0.16: Lognormal pdf .717 and initial air void fraction α = 2.0001 -3 -2 -1 0 1 2 3 4 (Ln(D)-µ)/σ Figure VI.01 0.01 0.004 0. 1 0.002 0 0 50 100 150 200 250 300 350 400 Diameter (µm) Figure VI.001 Re=93717 Normal 0.008 Re=28034 Gamma PDF(D) 0.
At this point the parameter b = ( Dmean )2 in equations VI.25 ml/min. The water jet Reynolds number was increased from Re = 28. It is very important to remark that all the pdf s(D) collapse on a single curve.23.17 where for simplicity we have represented the bubble size pdf for only three values of the Re .25 ml/min.5 × 10−4 to 1. At high Reynolds numbers. we have found three diﬀerent shapes of the bubble size pdf depending on the level of the turbulent kinetic energy of the underlying turbulence. 000 for three diﬀerent values of the ﬂow rate of air. characterized in these cases by D32 .35 ml/min. The same type of behavior has been observed for the other cases tested at Qa = 7. we chose D32 ˆ as the normalization bubble’s diameter and we deﬁned D = D/D32 . for the same experiments previously shown in ﬁgure VI. and since b = ( Dmean )2 .18.17 ˆ are given in ﬁgure VI.8 is independent of the Reynolds number. high ǫ.56. which are shown in ﬁgures VI. The probability ˆ ˆ density function of D. Drms /Dmean .54 − 0. The slope of the exponential tails of the pdf s increases with the Re . pdf (D). No simple model has been found for very low values of the Reynolds number. while as the value of the Reynolds number decreases (decreasing ǫ). Therefore. The above described shapes of the pdf were found for all the cases investigated here where the initial air void fraction was varied by almost two orders of magnitude (from 2.6 ml/min. a consequence of the decrease of the largest bubble size with Re .8. with the Reynolds number of the water jet. and Qa = 34. Notice that Drms /Dmean rapidly decays as the Reynolds number is increased until it reaches an asymptotic value at approximately Drms /Dmean ≈ 0. shown in ﬁgure VI. the pdf are best ﬁtted by Gamma distributions. Qa = 7.19-VI. the shape of the pdfs is very well represented by a log-normal distribution. we conducted a series of experiments where ǫ was varied while keeping constant the ﬂow rate of air injected into the ﬂow. showing that the slope of the tails of the bubbles size distribution is just a function of the maximum bubble’s diameter.119 Summarizing the above results. Qa = 3. In order to represent the bubble size pdf in non-dimensional variables. the probability density . To study the evolution of the shape of the bubble size pdf with the turbulent stresses acting on the surface of the bubbles.7 Drms and VI.22. and Qa = 34. The experimental measurements corresponding to the lowest ﬂow rate of air are presented in ﬁgure VI. 000 to Re = 93. we have studied the evolution of the Drms ratio between the mean diameter and the rms.6 × 10−2 ).35 ml/min. Due to its relevance in equation VI.
1 0.1 0.001 0 0.120 0.17: Evolution of the bubbles size pdf with the Reynolds number.5 1 D/D32 1. 10 Re= 41000 Re= 51436 Re= 93038 1 PDF (D/D32 ) 0.18: Bubble size pdf of D = D/D32 .0001 1 0-5 0 50 100 150 200 250 300 350 400 Diameter (µm) Figure VI.5 2 2. .5 3 ˆ Figure VI. Qa = 3.6 ml/min.001 0. Qa = 3.6 ml/min.01 Re= 41000 Re= 51436 Re= 93038 PDF (D) 0.01 0.
Qa = 7.5 0 50 100 150 200 250 300 350 400 Diameter (µm) Figure VI.1 0. .25 ml/min.19: Evolution of the bubbles size pdf with the Reynolds number.25 ml/min.5 3 D/D32 ˆ Figure VI. Qa = 7.1 0. 10 Re= 41035 Re= 51436 Re= 80037 1 PDF (D/D32 ) 0.001 0.121 0.5 2 2.001 0 0.0001 10.5 1 1.01 0.20: Bubble size pdf of D = D/D32 .01 Re= 41035 Re= 51436 Re= 80037 PDF (D ) 0.
.35 ml/min.5 1 1.0001 10.01 Re= 41035 Re= 54036 Re= 93038 PDF (D) 0.35 ml/min.01 0.5 3 D/D32 ˆ Figure VI.1 0. 10 Re= 41035 Re= 54036 Re= 93038 1 PDF(D/D32) 0.5 0 50 100 150 200 250 300 350 400 Diameter (µm) Figure VI.001 0. Qa = 34.122 0.21: Evolution of the bubbles size pdf with the Reynolds number.001 0 0. Qa = 34.22: Bubble size pdf of D = D/D32 .5 2 2.1 0.
25 ml/min Qa= 34.7. D32 depends on b and c. Since. the dependence of the maximum sized . given by equation VI.5 < b < 4.2 20 30 40 50 60 70 80 90 Re 100 x 1000 Figure VI.4 0.5 ml/min Qa= 7.8 Qa= 3. then. D32 = cb b+2 e−c D Γ(b) D cb b+1 e−c D Γ(b) D dD dD = b+2 . c (VI. D32 . equivalently. only depends on c = Dmean 2 Drms or. at high Reynolds number b is constant and independent on the air ﬂow rate and Re .14) is almost constant for all the values of b reported here.34 = D32 b+2 (VI.23: Evolution of Drms /Dmean with the water jet reynolds number. it only depends on one parameter. the Sauter Mean Diameter. which deﬁnes the slope of the tails of size pdf . the slope of the exponential tails of the bubble size pdf . 1. pdf (D). Due to its importance in determining the slope of the tails.7 0. can be simply determined by knowing the dependence of D32 on the Reynolds number and on the air void fraction.5 0. function.35 ml/min Drms/Dmean 0.6 0.123 1 0. is deﬁned as. = c (VI.7. given by the parameter c in equation VI.13) √ b Drms ≈ 0.9 0. Notice that since Drms the ratio √ b . If the bubble size pdf s are well described by Gamma distributions.3 0. Drms .12) and therefore.
Observe in ﬁgure VI. on the value of dissipation rate of kinetic energy at the air injection point is analyzed in the following section. and ǫ equation VI.17) is the Weber number based on the water nozzle’s diameter.6 predicted value. on the local value of the rate of dissipation of turbulent kinetic energy of the underlying turbulence.24 that. U0 . we measured a noticeable decay in the value of the exponent of the power law which deviates considerably from the −0. an averaged value of ǫ can be used to take into account its downstream decaying until the break-up process 3 concludes. represented by D32 . ǫ = ν3 . the diameter of water nozzle and the air injection point were kept constant (DJ = 3. the maximum droplet size decays with the Weber number with a power law almost equal to the one predicted by the simple dimensional argument given in equation VI. DJ . Since in all the experiments reported here. Therefore. and the velocity of the water jet at the exit of the nozzle. Dmax ∝ Wen −3/5 . and in equilibrium turbulence is given by . and Dmax the diameter of the largest stable bubble. presented in chapters IV and V. .1 mm and X/DJ = 10). (VI. at larger initial void fractions. Dmax . homogeneous.16. DJ where Wen = ρ U0 2 DJ σ (VI. : Dmax ∝ where σ is the surface tension.15 can then be expressed as. At this moment. the averaged dissipation rate can be estimated as ¯ ∝ U0 /DJ .16) (VI. it still remains unresolved if this is an eﬀect of the 3 Dmax has been deﬁned to be proportional to the critical diameter.E Dependence of Dmax on ǫ The dependence of the maximum bubble size3 .15) the dissipation rate of turbulent kinetic energy.124 bubble. Dc . predicted by the classical model of Kolmogorov based on local isotropic. However. η4 σ ρ 3/5 ǫ−2/5 . ρ the density of the continuous phase. for very small injection ﬂow rates of air (small initial air void fractions). VI.
is to incorporate the eﬀect of intermittency into equation VI.1 0. (VI.06 Qa=3.44) R= 1 Dmax/DJ 0.94 * x^(-0. values of the dissipation rate of turbulent kinetic energy weaker than the average value ǫ are . DJ (VI.24: Variation of the maximum bubble’s size with the jet’s nozzle Weber number. (VI. .02 0 0 1 104 2 104 3 104 4 104 5 104 We n Figure VI.19) 3 If we deﬁne the integral length scale Lx with the nozzle’s diameter DJ and ǫ ∝ U0 /DJ .99 y = 4. equation VI.20.04 0. Wen .5 ml/min Qa=7.08 y = 4. .125 0.18) ρ ǫ2/3 Lx σ 5/3 .15.18 can be expressed as a function of the Weber number.16 * x^(-0.53) R= 1 y = 3. and Novikov and Dommermuth. or if it is simply the eﬀect of coalescence. as 3 Dmax ∝ Wen − 5 + η .35 ml/min 0. A possible argument proposed by Baldyga and Bourne.20) The increases of Dmax due to the additional intermittency eﬀect introduced by the increase of air void fraction can be understood from equation VI. Therefore.3 * x^(-0. Novikov and Dommermuth  then found that Dmax can be expressed by Dmax ∝ where W = σ ρ 3/5 ǫ−2/5 W η .53) R= 0. Qa is the ﬂow rate of air injected air void fraction on the turbulent spectrum.25 ml/min Qa=34.
126 0. ρ.005 0 0 50 100 150 200 250 300 350 400 Diameter (µm) Figure VI. The results presented in this section can be used to predict the slope of the tails in the bubble size pdf s. (VI. ǫ.25: Dependence of the frozen bubble size pdf on the initial air void fraction. and surface tension. At high Reynolds numbers and low values of air void fraction. and since b is constant at large Re . Dmax .15 and VI.5 e-4 Lognormal alpha=2.12. .15.025 0. σ) and the dissipation rate of turbulent kinetic energy. the slope of the tails (given by c) can be determined from c∝ 1 ∝ D32 ρ σ 3/5 ǫ2/5 .21) which is a function of the ﬂuid properties (density.02 PDF(D) 0.01 0. from equations VI. Re= 54.03 e-3 Gamma Alpha=7. Therefore.015 Aplha=3. is given by equation VI. which is proportional to D32 .45 e-3 Gamma 0.036 acting on the break-up process and larger values of Dmax are able to hold the turbulent stresses.
alpha= 3.5 e-4 alpha= 1.6 e-3 alpha= 7.45 e-3
10- 5 0 50 100 150 200 250 300 350 400
Figure VI.26: Evolution of the bubbles size pdf with the air void fraction, α. Re = 54036.
Dependence of the frozen bubble size pdf on the initial void fraction, α
The eﬀect of the initial air void fraction, α = Qa /Qw , on the shape of the frozen
pdf is shown in ﬁgure VI.25. These pdf s correspond to a representative intermediate jet Reynolds number of Re = 54, 036, varying the void fraction by nearly two orders of magnitude. Qualitatively, the observed eﬀect of increasing the void fraction appears to be very similar to the eﬀect of decreasing the value of the dissipation rate. While the pdf s are log-normal at low void fractions as was the case for very high values of ǫ, they are better ﬁtted by Gamma functions at larger values of the air void fraction. The evolution of the frozen bubbles pdf is shown in ﬁgure VI.26 for three diﬀerent values of α. The slope of the tails of the pdf s increase as α decreases. Although the eﬀect of α on the shape of the pdf s is noticeable, it is not as strong as the eﬀect of ǫ discussed earlier. ˆ ˆ As was the case in the evolution of pdf with ǫ, the pdf s(D), where D = D/D32 , also collapse in a single, self-similar curve showing that D32 is the only parameter needed to describe the behavior of the frozen bubble’s size pdf , see ﬁgure VI.27. The possible
alpha= 3.5 e-4 alpha= 1.6 e-3 alpha= 7.45 e-3
PDF (D32 )
0.001 0 0.5 1 1.5 2 2.5 3
ˆ Figure VI.27: Bubble size pdf of D = D/D32 . Re = 54036. causes of the changes in the pdf s with α are coalescence, modiﬁcations in the turbulent energy spectra due to the presence of the bubbles which translate into lower values of the ǫ, increasing therefore the value of the critical diameter, Dc ∝ (σ/ρ)3/5 ǫ−2/5 , and consequently the value of Dmax , or a combination of both eﬀects. To account for the eﬀect due to changes in ǫ the energy spectrum was measured using hot ﬁlm anemometry, at a point downstream in the jet where the break-up takes place, radially displaced a distance close to the jet’s centerline but far enough to avoid any contact of the bubbles with the probe. Therefore, if the energy spectrum was aﬀected by the bubbles, an indirect eﬀect at the measuring point would have been detected. Both spectra, measured in the ﬂow with and without bubbles, turned out to be identical. Therefore, at this moment it is not possible to determine if it was due to the fact that the changes in the spectrum at local eﬀects and only the areas occupied by the bubbles are aﬀected or because the energy spectrum is not aﬀected by the presence of the bubbles at all. A possible explanation of the increased value of Dmax with the air void fraction is that, consistent with the above simple isotropy model given in equation VI.15, one would expect that the turbulent kinetic energy existing in the liquid per unit total volume
y = (1 + m1 * α )m2
m1 m2 Value 3.52e+03 2.109e-01
2 Dmax/Do max 1 0.9 0.8
Figure VI.28: Dependence of maximum bubble size on the initial air void fraction. Re= 54,036 would decrease with increasing air void fraction (a simple geometrical argument due to the volume occupied by the air). Therefore, the dissipation rate would be given by , ǫef f = ǫ 1+α (VI.22)
where ǫef f is an eﬀective dissipation rate acting in a volume which has been increased due to the increased volume of air. The maximum bubble size should increase with α, Dmax ∝ σ ρ
ǫef f −2/5 =
ǫ−2/5 (1 + α)2/5 .
Since we have deﬁned the void fraction, α = Qa /Qw , as the ratio between the air ﬂow rate and the water ﬂow rate, α is a global parameter whose value may be diﬀerent from the local void fractions at the points where the break-up occurs. To account for this eﬀect we have deﬁned a more general dependence on Dmax , Dmax ∝ and therefore, Dmax = (1 + m1 α)m2 , Dmax (α → 0) (VI.25) σ ρ
ǫ−2/5 (1 + m1 α)m2 ,
This demonstrates that there is in fact a single parameter.012 Re=53552. Alpha= 1. . α. which controls the maximum bubble’s diameter.002 0 0 50 100 150 200 250 Diameter (µm) 300 350 400 Re= 41035.01 0. and therefore the shape of the frozen bubble’s size pdf s. Alpha= 4.004 0.29 we present evidence that such a function should exist.29: Showing several combinations of initial void fractions and Re which result in identical pdf s. The evolution of Dmax with α at a ﬁxed value of Re = 54036 is shown in ﬁgure VI.15 which have been proposed to account for void fraction eﬀects . .28 where Dmax (α → 0) has been selected to be the measured value of Dmax at the lowest void fraction.130 0.36 e-3 0. ǫef f . α = 3.008 PDF(D) 0. in ﬁgure VI. Although the formulation of the dependence of the pdf on the void fraction and the characteristics of the underlying turbulence is still unresolved.006 0. Alpha= 6. Dmax . Note that we were able to generate identical pdf s with diﬀerent pairs of values of α and Re. approaches to zero.21 e-3 Re= 35835. This measured dependence seems to be much stronger than the one predicted by the various modiﬁcations of the correlation given in equation VI.5 × 10−4 . where Dmax (α → 0) corresponds to the maximum bubble’s diameter when the initial void fraction.87 e-4 Figure VI.
have the draw back that they require the use of physically questionable closures for the collisions between the particles and the “eddies”. In the past. Experiments performed over a wide range of bubble diameters and turbulent intensities show that the break-up frequency always increases with the value of the dissipation rate of turbulent kinetic energy. These models. ǫ. which are derived from an extension of the classical kinetic theory of gases . . . as a function of their size and of the value of the turbulent kinetic energy of the underlying turbulence.Chapter VII Conclusion We have conducted a series of well controlled experiments where the transient evolution of the bubble size pdf resulting from the break-up of air bubbles of known diameters injected into a fully developed turbulent water ﬂow has been systematically measured using non-intrusive optical techniques. numerous models have been proposed for the breakage frequency which assumed that the break up of the particle occurs. the number of “eddies” with a certain energy. These measurements have been used to calculate the bubble break-up frequency and the daughter bubble size pdf in nearly homogeneous and isotropic turbulent conditions. involving assumptions for the eddy-particle collision cross section. Tsouris and Tavlarides  assumed that the particle’s break up frequency is equal to the product of a “drop−eddy” collision rate and a breakage eﬃciency.  among others. they also show a non-monotonic dependence of the break-up frequency on the bubble size. all these models predict only a 131 . For example. through the interaction of the particle with an imaginary array of turbulent “eddies” which are assumed to make up the turbulence. Furthermore. However. etc.
Furthermore. ǫ. We have also proposed a model for the size pdf of the daughter bubbles produced by the shattering of a mother bubble immersed into a fully developed turbulent water ﬂow. This model is shown not only to describe the observed qualitative trends.132 monotonic increase of the break up frequency with both the bubble size and the dissipation rate of turbulent kinetic energy. the surface of the bubble presents more local deformations which eventually break into small bubbles. The model. while for large bubbles. For small size bubbles. we have shown the existence of a characteristic bubble size. ǫ.Our model is based on the premise that the break-up frequency is a function of the sum of all the forces acting on the surface of the bubble. the experimental measurements of the transient bubble size pdf performed over a range of bubble sizes and dissipations rates. which produce its deformation and conﬁnement. the break-up frequency is shown to increase with the bubble size as D Dc − 1. but also to be in excellent agreement with the measured values. it decreases as D−2/3 . The model is based on the premise that the probability of a given pair to form is proportional to the product of the surplus energy corresponding to two size scales formed. At larger Weber numbers. Dgmax . and does not require invoking the use of an imaginary array of “eddies” of unknown number and collision cross-section. a model based on the assumption of tertiary break up was found to lead to better agreement with the experiments. This maximum frequency increases as ǫ3/5 and decreases with the interfacial surface tension as σ −2/5 . predicts. . in a consistent manner. Contrary to previously existing models. In this work we have proposed a phenomenological model for the bubble breakup purely based on dynamic arguments whose premises are radically diﬀerent from those used in the past. although simple and straightforward. In these cases. The agreement was found to be good at small and moderate values of the mother bubble’s turbulent Weber number under which conditions a binary break up was found to be consistent with experimental observation. our phenomenological model predicts a dependence of the daughter bubble size pdf on both the bubble size and the value of the dissipation rate of TKE of the underlying turbulence. This model has been shown to be in very good agreement with the experimental data of the transient bubble size distribution. for which the break-up frequency is maximum. whose diameters are comparable to the critical capillary length.
Furthermore. once the break up process has concluded. ǫ. has also been studied in a turbulent. The frozen pdf s are shown to be strongly dependent on the dissipation rate of turbulent kinetic energy of the underlying turbulence. the shape of the bubble size pdf . .133 Finally. axisymmetric water jet. independent of ǫ and of α. and on the initial air void fraction. it has been found that when the bubble’s diameter is made dimensionless with the Sauter Mean Diameter of the distribution. α. D32 . the pdf s collapse into a single. universal function.
NT (A. The sum of the entire distribution NT Nj j=1 NT Nb i=1 P df (Di )△D = is unity where Nb = (Dmax − Dmin )/△D is the number of bins in which the density function has been discretized. The interval △D must be chosen in such way that it is large enough to collect enough particles and suﬃciently small to obtain good resolution. Similarly the variance of the distribution is expressed as: NT σ = j=1 2 ¯ (Dj − D)2 = NT Nb i=1 ¯ (Di − D)2 Ni .3) 134 . NT (A. NT (A. The number density function is then deﬁned as: P df (Di )△D = Ni . The mean value of the distribution is calculated as follows: ¯ D= j=1 NT Dj = NT Nb i=1 Di Ni .2) where j indicates the particle number counter and i is the bin number in the distribution.1) where Ni is the number of particles between Di −△D/2 and Di +△D/2 and NT is the total number of particles measured.A Deﬁnitions Particle-size distributions can be constructed from measures of the size of a large amount of particles and counting the number of particles which fall between the diameters Di − △D/2 and Di + △D/2.Appendix A Particle Size Distributions A.
7) The integral of the Probability Density Function. △D→0 △DNT (A. is unity: Dmax P df (D)dD = 1 . deﬁned in equation A. If the interval of size is inﬁnitely small. Cpdf (D). Similarly the .135 Some of the most commonly used characteristic diameters are: Surface Mean Diameter: D20 = NT j=1 Volume Mean Diameter: 1/2 2 Dj = NT 3 Dj Nb i=1 2 Di Ni NT 1/2 . Therefore the normalized V pdf (D) writes as: V pdf (D) = D3 P df (D) Dmax Dmin D3 P df (D)dD .5) D32 = NT 3 j=1 Dj NT 2 j=1 Dj = Nb 3 i=1 Di Ni Nb 2 i=1 Di Ni . if it has been properly normalized.1 versus the diameter.8) Another very useful characteristic Probability Density Function is the so-called Volume Probability Density Function. it is very instructive to represent the cumulative density function. Dmin (A.1 is obtained as: P df (Di ) = lim Ni . it is appreciable the weighting eﬀect of larger particles on the volume pdf . The broken line represents the Volume pdf of the same distribution. (A.9) A typical probability density function is plotted in ﬁgure A. which represents the volume occupied by particles of size D compared to that of the entire distribution. (A. V pdf (D).6) where D32 is the diameter of a particle whose ratio of volume to area is equal to that of the entire distribution.4) NT j=1 D30 Sauter Mean Diameter: = NT 1/3 Nb = i=1 3 Di Ni NT 1/3 . Cpdf (D) represents the probability of ﬁnding a particle whose diameter is smaller than D and therefore Cpdf (Dmin ) = 0 and Cpdf (Dmax) = 1. (A. the probability density function. In addition to the above probability density functions. pdf . (A.
136 0.008 0. broken line is the volume probability density function. is deﬁned as the fraction of volume occupied by particles of diameter smaller than or equal to D. cumulative volume probability density function. Some of the commonly used continuous size distributions are described below.1: Particle Size Probability Density Functions. A new characteristic diameter can be deﬁned as the diameter of a particle whose size is bigger than 90% of the particles in the distribution. The characteristic parameters describing the P df s(D) are determined from the experimental measurements.002 0 0 50 100 150 200 250 Diameter (µm) 300 350 400 Figure A. D90% .006 0. Cvpdf (D). Solid line represents the number probability density function.2 versus the diameter of the particle. and similarly. the diameter of a particle such that 90% of the total volume is contained within particles of smaller diameter. . they have usually been replaced by mathematical expressions which are functions of a limited number of parameter. Dv90% . The number and volume cumulative density functions are shown in ﬁgure A.004 0.01 Probability density function 0. Due to the complexity of representing the experimental size distributions. D90% and Dv90% are a measure of the largest particles encountered in the distribution.
9 0. a task that could be endless. Log-normal Distribution Function The log-normal distribution is a mathematical distribution which has been frequently used to describe many of the experimental particle distributions found in natural and industrial processes.2: Cumulative Probability Density Functions.137 Cumulative probability density function 1 0. based on empirical arguments. . The log-normal function can be deﬁned as the normal function . A description of the derivation of log-normal distributions to represent size density functions can be found somewhere else .3 0.8 0.1 0 0 50 100 150 200 250 Diameter (µm) 300 350 400 D90% Dv90% Figure A. The section will only cover some of the distributions which appear in this thesis.2 0.4 0. The purpose of this section is not to review all of them. broken line is the cumulative volume probability density function. describing their similarities and diﬀerences. have been historically ﬁtted to experimental measurements. but to give a brief description of some of them.B Continuous distribution functions A wide number of size distributions. Solid line represents the cumulative probability density function.5 0. A. speciﬁcally those occurring in atomization. A more detailed description of the most common distributions used to characterize sprays can be found in .7 0.6 0.
(A. Rosin-Rammler Function Probably the function most frequently used to for droplets size distribution in sprays is the Rosin-Rammler distribution.14) where Cvpdfr (D) is the fraction of volume contained in particles of diameter less than D.10) where µl and σl are deﬁned as the mean and standard deviation of the distribution on the logarithm of D variable. (A. Although its use has been based in empirical approaches. Brown at al.11) and Dmax σl = Dmin (ln(D) − µl )2 P dflog (D)dD . (A. (A. A complete description of the various statistical formulas to describe the diﬀerent moments of the distributions is summarized by Mugele et al. (A.  proposed a physical derivation of the distribution using fractal analysis. (A. A Rosin-Rammler function can be described as: Cvpdfr (D) = 1 − exp − D δ n (ln(D) − µvl )2 1 exp − 2σl2 2πDσl . : P dfn (D) = n1 Dn2 exp(−n3 D)n4 . ln(D). Nukiyama and Tanasawa Function One last function of interest to explain here is that described by Nukiyama and Tanasawa. Dmax µl = Dmin ln(D)P dflog (D)dD .12) Similarly the volume density function can be written as: V pdflog (D) = √ where µvl = µl + 3σl2 . .13) ..138 when the logarithm of the diameter is used as independent variable: P dflog (D) = √ (ln(D) − µl )2 1 exp − 2σl2 2πDσl . .15) . and δ and n are two empirical parameters.
. . . . . n2 = α − 1. Many other empirical distributions such the Log-hyperbolic.15.15 for n1 = λα /Γ(α). If a new.18) (A. α which is the squared ratio between the mean value and the standard deviation of the distribution. Γ(α) (A. (A. the Volume probability density function from a Gamma distribution. A more simpliﬁed and handy density function derived from equation A. non-dimensional. Γ(α + 3) (A.16) obtained from equation A.15 is the Gamma probability density function: P dfg (D) = λα α−1 D exp(−λD) . Finally. diameter is deﬁned as D∗ = λD.20 has been reduced to an equation of only one parameter. σ2 (A. deﬁned as in equation A.16 with a new shape parameter α∗ = α + 3 and identical scale parameter λ. can be found in the literature. Modiﬁed RosinRammler.16 can be rewritten as: P dfg (D ∗ ) = D ∗α−1 exp(−D∗ ) .19) ¯ being D and σ the mean and root-mean-square values of the distribution. ¯ D . The interested reader is referred to Lefebvre. the Gamma probability density function of equation A.  for more detailed information. can be expressed as: V pdfg (D) = λα+3 D α+2 exp(−λD) . P dfn (D) is function of four empirical parameters. Γ(α) (A. a fact that does not make it very convenient to ﬁt experimental measurements. n3 = λ and n4 = 1. Upper-limit.139 As shown in equation A. Γ(α) is the Gamma function deﬁned as: Γ(α) = 0 ∞ xα−1 e−x dx .9.20) Equation A.17) The parameters α and λ are the shape and scale parameters respectively deﬁned in the form: α= ¯ D σ λ= 2 .21) which is an expression similar to that in equation A.  and Crowe et a.
The theory of Homogeneous Turbulence. Sankar. Lisbon. International Symposium on cavitation Inception. Meudon. R. F. Bourne. Analysis of the light scattering interferometry for spheres larger than the light wavelength. Berkman and R. M. Drop breakup and intermittent turbulence. J.N.. Cambridge at the University Press.  J. B.1 Bibliography  R. ASME Winter Annual Meeting. D. V. Journal of Chemical Engineering of Japan. International Journal of Multiphase Flow. Antonia. Baldyga and J. D. V. D. 23(4):695–700. Tevenovskiy.  W. L. Wang. V. No. A n study on the morphology of spheroidal bubbles using light scattering interferometry. D. J. and E. Measurements of dissipation rate and some other characteristics of turbulent plane and circular jets. L. Experimental methods in multiphase ﬂows. Calabrese. Sankar. 1993. 1989. and W. Ye. No. Satyaprakash B. 20:261–295.4:602–609. 2nd Congres Francophone de Velocimetrie Laser. Portugal. D. . No. AIChE Journal. Bachalo and S. D. V. SSSR. Bachalo.5:583–590. n A theoretical and experimental study of the light characterization of bubbles using light scattering interferometry. Batchelor. K. N.  A. S. Dispersion of viscous liquids by turbulent ﬂow in a static mixer.France. Navk. Bachalo. Bre˜a de la Rosa. G. 1988. Bachalo. Bachalo and M. On the measure of minimal ﬂuctuations in a turbulent ﬂow.  W. 23. Optical Engineering. 1956. K.  W. and Hussain A. Sankar. J. Houser. Cambridge. 1980. Proceedings of the Fourth International Symposium on the Applications of Laser Anemometry to Fluids Mechanics. Phys. K. 1994. 1984. Dokl. Phase/Doppler spray analyzer for simultaneous measurements of drop size and velocity distributions. A. 1988. Smith. Fluids. 66:5:821–824.  A. Akad. 26.  M. Carr. Bre˜a de la Rosa. D.  P. 1949. 89:63–72.  G.6:738–741. Weber. Baranaev. R. and W. 1990. S. Tregubova. 34.
CRC Press. D. 12:34. part II: Evaluating steadiness using the interparticle time distribution. No. ICMF 98. C. Sommerfeld M. Wohletz. Phys. Geophys. Br¨der. Atomization and Sprays.. In Third International Conference on Multiphase Flows. Crowe. Sommerfeld and S.  C. bubble velocity. 1998. France. Wang. and P. Drop break-up in turbulent stirred-tank contactors. Simultaneous void fraction measurements. Shattering of a liquid drop due to impact. Tavlarides. V. Edwards and K. 93:18–1. Brown and K. .5:457 –505. M. M. Trans.  R. K. Multipoint statistical structure of the ideal spray. A. L. 435:483–503. D. A. 93:3591–3600. 1966. Marx.  C. Cutter. C. K.No 4:2758 –2763. and N. Flow and turbulence in stirred tanks. Analysis of the hydrodynamics in a bubo ble column by phase-doppler anemometry. Edwards and K. Chemical Engineering Sciences. 63:11:5442–5453.  R. AGARD Conference Proceedings.  D. Gibson. P. AIChE Journal. 15:161–173. M. Coulaloglou and L. 1988. P. Part I. 1963.. Marx. Atomization and Sprays. Derivation of the Weibull distribution based on physical principles and its connection to the Rosin-Rammler and lognormal distributions. and C. Drop break-up in liquid stirred-tank contactors.2  M. 1995. Gibson. Part III. 1962. Lyon.. D. Res. 1991. Appl. Calabrese. AIChE Journal. H. Jet turbulence: Dissipation rate measurements and correlations. Vagle. Spectra of turbulence in a round jet.4:677–681.5:435 –455. London Ser. Description of interaction processes in agited liquid-liquid dispersions.4:657–666. AIChE Journal. F. Droplets. Spectra of turbulence at high reynolds number.  M. Dang. Fluid Mech. Vol.  C.78. and size estimate using a single optical probe in gas-liquid two-phase ﬂows. Farmer and S. R. Turbulent Shear Flows. 1992. Vol. Multiphase Flows with Particles and  L. Van Atta. F.  R. Sci. Soc. 32. 1998. 1986. D. 1995. W. On the determination of breaking surface wave distribution using ambient sound. A.  M. 1972. J. Y. Friehe. Tisler. T. 32.  W. H. Gibson. Cohen. part I: Fundamental concepts and the realization density. 32:1289–1297. Cartellier.A. Vol. 195:1281– 1983. Phil. Chang.  C. Bryner. 1995. Multipoint statistical structure of the ideal spray. 1986. and Tsuji Y. Calabrese. V. No. J.  A. J. Nature. T. 1977. Instrum.  C. Rev..
A. Lefebvre.  A. and S. N. Navk. Kolmogorov.O. Arai K.. Gr´han. Probability and statistics for modern engineering.  K. Konno. Lisbon. 1983. Kolmogorov. Capp. and S. J. 13.No. Gouesbet. 258:31–75. San Diego. Matsunaga. Dokl.  A.  M. Proceedings of the Seventh International Symposium on the Applications of Laser Anemometry to Fluids Mechanics. axisymmetric. W. 1994. Akad.  L. 1994. Lapin.1:67– 73. Russell. Innocenti. Ibrahim. Hesketh. 1980. cinetique et role de la diﬀusion. and T. Journal of AIChE.  M. New York. PhD thesis. University of California. 3:289–295. Durst. 1998..M. P. Two-Phase Shear Layer. Chemical Engineering Science. Etchells. 16. Onofri. Journal of Chemical Engineering of Japan. e 1997. Fluid Mech. Hemisphere Publishing Corporation. No 3:505–513. Logaritmically normal distributions of fragmentory particle sizes. 1983. N.  M. F. 1941. K. 1975. Tropea. P. PhD thesis. 1995. Simulation model for breakup process in an agitated tank. Aoki. Grenoble (France). Scale-eﬀect on breakup process in liquid-liquid agitated tanks. J. On the attainment of steady state in turbulent pipe ﬂow of dilute dispersions. MacGraw Hill Book C. T. G. 1990. and Bachalo W. Fundamentals of the hydrodynamics mechanisms of splitting in dispersion process. Navk. F.  H. Turbulence. G. Signal processing considerations for laser doppler and phase doppler applications. turbulent jet. W. No 1:1–9. Chemical Engineering Science. 53. In 5th International Symposium on the Application of Laser Techniques of Fluid Mechanics. L. D. George. H. Velocity measurements in a highReynolds number. O. 46. No. 66:5:825–828. and C. 1989. SSSR. Monterey. J. M. Hussein. S. and W. Melange turbulent a partir d’une source isolee: geometrie.  J. 1949. California.  C. Bubble breakage in pipeline ﬂow.  K. Girasole. second edition. Brooks/Cole Publising Company. Journal of Chemical Engineering of Japan. T. O. momentum-conseving. Kiger. Akad.4:312–319. F. 1991. Werthimer. On the breakage of drops in a turbulent ﬂow. SSSR. Konno. Saito.  A. Y.3  G. 31:99–101. Dokl. . D. 1. 1955.  J. Hinze. Measuree ment of bubbles by phase doppler techniques and trajectory ambiguity. Particle Disersion and Inter-Phase Kinetic Energy Transfer in a Turbulent. Saito. Hinze. Kostoglou and Karabelas A.Portugal.. Portugal. Lisbon. No.  R. Atomization and Sprays. Universit´ Joseph Fourier.
Prince and H. 1951.  G.  G. Xu. Longuet-Higgins. Ind. Chemical Engineering Science. K. and D. 1979. H. Kumar. Tropea. University of California. Narsimhan.  G. third edition.G. AIChE Journal.  P. Gandhi.  E. G. and K. and C. Das. San Diego. . 49:2194–2198. R. MA. Blanch.No. B.  A. Distribution of droplets in a turbulent spray. Fluid Mech. London Ser. MIT Press. T. Belmont. California. Soc. 1997. Mugele and H. Gupta. Ricou and D. Chemical Engineering Science. Trans. Portugal. 36. Measurements of entrainment by axisymmetrical turbulent jets. Inst. 1995. Rev. 30:457–467. Vol. Nambiar. S. Evans. A two zone model of breakage frequency of drops in stirred dispersions. Yaglom.  F.4  M. 1987. A new model for the breakage frequency of drops in turbulent stirred dispersions. J. R. 1995. Gandhi. volume 1. R. Das. R. Cambridge.. T. Bubble Dispersion and Interphase Coupling in a Free Shear Flow.  G. 5:5479–5482.10:1485–1499.43.  J. Ramkrishna. 11:21–32. D.  R. Nambiar. The role of surface-wave breaking in air-sea interaction. The laws gorverning the ﬁneness of powdered coal. Narsimhan. and D. R. 1933. J. P. 1992.  W. 1961. San Diego. Ann. 1992. M. Spalding. University of California. Dommermuth. Lisbon. R. Chem. R. K.  M. PhD thesis. 47:2989–3002. Fluid Mech. Mixing and Stability in Two-Layer Arrested Wedge Flows. Kumar. Phasee doppler anemometry with dual burst technique for particle refractive index measurements. Rightley. Novikov and D. P. 1990. 1994. Melville. Phil. 18:279–321. 439:611–626. second edition. Mechanics of Turbulence.No 6:1317–1324. Brenn G. Gouesbet.  P. Onofri. A. S. S. J. A. W. 56. . S. G. K. Duxbury Press.. Vol.  D. Drop size distributions in sprays. Rammler. Hydraulics. Eng. Chemical Engineering Science. M. T. and K. G. Fuel. No. Rosin and E. Breakage functions for droplets in agitated liquid-liquid dispersions.No 31:29–36. 1984. 1997.A. 7. Gr´han. 1996. Ramkrishna. AIChE Journal. J. Bubble coalescence and break-up in air-sparged bubble columns. 34:257–265. The crushing of air cavities in a liquid. A model for transitional break-up probability of droplets in agitated lean liquid-liquid dispersions. Mathematical Statistics and Data Analysis. PhD thesis. Physical Review E.  D. Rice. Nejfelt. Pawlak. Statistical Fluid Mechanics. 1994. Proceedings of the Seventh International Symposium on the Applications of Laser Anemometry to Fluids Mechanics. Monin and A.
D. 3:395–406. Applied Optics. K. Trans. Taylor. 1992. Journal of Scientiﬁc Instrumensts (Journal of Physics E). Cambridge University Press. Kamemoto. No. J. Williams.  S. and P. and W.Inc. Onofri. A.. No. V. 304:155–210. Fundam. Sankar and W. G. Applied Optics. Tavlarides. Droplets breakage in stirred dispersions. Sankar. S. Soc. Inenaga.9:1377–1391. V. 30.146:501–523. Light Scattering by Small Particles. Sankar. L. Weber. 5:533–542. Proceedings of the Sixth International Symposium on the Applications of Laser Techniques to Fluids Mechanics.  A.  A. Combustion Theory. Cambridge. T.  J. California. and W. Y. A. Some measurements in the self-preserving jet. Trajectory dependent scattering in phase doppler interferometry: minimizing and eliminating sizing errors.. van de Hulst. 1976. C. No. A. 1934. 40. Series A.  K.33:4914– 4920. New York. Portugal. Phil. .  A.  H. The Structure of Turbulent Shear Flows. 19:437–451. Inc. Wygnanski and H. Chem. A. Computational Chemical Engineering. D. Dover Publications. H. Menlo Park.. 258:577–612. 1966. 1996. On the clouds of bubbles formed by breaking wind waves in deep water. Narsimhan. Proceedings of the Seventh International Symposium on the Applications of Laser Anemometry to Fluids Mechanics. Portugal. J. Breakage and coalescence in dispersed phase systems. Bachalo. J. G. N. Proceedings of the Royal Society (London). Haugen. 1994. Ramkrishna. Tsouris and L. Breakage and coalescence models for drops in turbulent dispersions. London Ser. Wyngaard. Thorpe. R.  S. The Benjamin/Cummings Publishing Company. MA. and their role in air-sea gas transfer. and G. 1981.. Tropea. Bachalo. Series 2 Volume 1:1105– 1108. The formation of emulsion in deﬁnable ﬁeld of ﬂow. V. Lisbon. second edition. C. 1994. Lisbon. 1994. Valentas and Amudson N. 1982. Chemical Engineering Science. Xu.5  S. 1969. Engng. I. 1991. Sathyagal and D. Fiedler. Breakage functions from experimental drop-size distributions.12:1487–1496. N. Solution of inverse problems in population balances II: particle break-up. Townsend. R. D.  I. Measurements of small-scale turbulence structure with hot wires. Response characteristics of the phase doppler particle analyzer for sizing spherical particles larger than the light wavelength. 30. D. B.  S. A.  C. Bachalo. D. Fluid Mech.  C. No.  F. 1985. 1991. Sathyagal. Ramkrishna. Sizing ﬁne particles with the phase doppler interferometric technique.  G. Gr´han. 1968. 51. Dual mode phase e doppler anemometry. AIChE Journal.
This action might not be possible to undo. Are you sure you want to continue?