This item has the following downloads:
1 CRATERING OF A PARTICLE BED BY A SUBSONIC TURBULENT JET By CASEY Q. LaMARCHE A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2013
2 2013 Casey Q. LaMarche
3 To Rocky, my Mom and Cori
4 ACKNOWLEDGMENTS I would first like to thank my advisor and mentor Prof. Jennifer Curtis, through your patience and guidance I have learned some of my most important life lessons I am indebted to you for all of the knowledge you have provided to me. I would also like to t hank my mother for offering me unconditional love and support through all of my journeys. Cori, you are my rock, I have so much to thank you for that g oes beyond the text that can be contained here. I would also like to thank one of my mentors, Dr. Phil Me tzger for sharing his expertise in this area and teaching me I that if anything ever goes wrong, it was put there. I would like to thank my father for all of his love and help Thomas Lemee, our long discussions of fluid mechanics never stop exciting me I would also like to thank Stephen Arce for being an all around great friend and showing me how aspects of granular flow ca n be used in interesting ways in different fields like biology Jess Clawson and Mrs. Dahl I appreciate all of your help. Braden Sn ook and Gabe Tong have taught me more then they probably know. I hope they continue letting me mentor them. I would also like to acknowledge PIRE NSF and NASA STTR Phase I and II for funding the work that is contained in this dissertation. I am grateful to Prof. Ranga Narayanan and Bungo Shiotani for all of their work for the PIRE program. I would like to acknowledge my officemates for the rich experiences we have shared. Thank you to Prof. Dickinson for the work you have done for the graduate students, you have worked with us on so many occasions to bring the department together as a cohesive unit. To all of my friends thank you for keeping me sane. I owe you all so much.
5 TABLE OF CONTENTS page A CKNOWLEDGMENTS ................................ ................................ ................................ ... 4 LIST OF TABLES ................................ ................................ ................................ ............. 8 LIST OF FIGURES ................................ ................................ ................................ ........... 9 LIST OF ABBREVIATIONS ................................ ................................ ............................ 12 ABSTRACT ................................ ................................ ................................ .................... 13 CHAPTERS 1 INTRODUCTION ................................ ................................ ................................ ..... 15 Future Terrestrial Space Exploration ................................ ................................ ....... 15 Previous Example: Apollo 12 Landing on Lunar Surface (Metzger et al., 2008) ..... 16 Permeability of Lun ar Simulant JSC 1A ................................ ................................ .. 17 Multiphase Flow Modeling ................................ ................................ ....................... 18 Effects of Particle Shape on Cratering ................................ ................................ ..... 19 Cohesion and Fine Particles ................................ ................................ .................... 21 Cratering Mecha nisms ................................ ................................ ............................. 22 Scaling Asymptotic Crater Depth ................................ ................................ ............. 24 Lunar Soil Simulants ................................ ................................ ................................ 26 2 PERMEABILITY OF LUNAR SOIL SIMULANT JSC 1A ................................ .......... 28 Motivation ................................ ................................ ................................ ................ 28 Background and Theory ................................ ................................ .......................... 28 Experimental Hardware and Techniques ................................ ................................ 31 Results and Discussion ................................ ................................ ........................... 33 Summary ................................ ................................ ................................ ................. 39 3 MODELING CRATER GROWTH USING THE TWO FLUID MODEL ..................... 46 Motivation ................................ ................................ ................................ ................ 46 Model Equations ................................ ................................ ................................ ...... 48 Governing Equations ................................ ................................ ......................... 48 Frictional Stress Model ................................ ................................ ...................... 50 Methods ................................ ................................ ................................ ................... 53 Experimental Setup of Metzger et al. (2009a) ................................ ................... 53 Simulation Setup ................................ ................................ ............................... 53 Model Sensitivity ................................ ................................ ............................... 55 Experimental Validation ................................ ................................ ..................... 56 Simulation Predictions ................................ ................................ ....................... 56
6 Results and Discussion ................................ ................................ ........................... 57 Sensitivity Results ................................ ................................ ............................. 57 Frictional Model Study ................................ ................................ ....................... 58 Parametric Study Measured and Predicted Crater Growth ............................ 61 Long Time Simulation ................................ ................................ ........................ 62 Domain Sensitivity Effect of Center Wall ................................ ........................ 64 Simulations of Turning the Jet Off ................................ ................................ ..... 66 Cratering Mechanics ................................ ................................ ......................... 67 Summary ................................ ................................ ................................ ................. 71 4 EFFECT OF PARTICLE PROPERTIES ON CRATER DYNAMICS ........................ 98 Motivation ................................ ................................ ................................ ................ 98 Background ................................ ................................ ................................ .............. 98 Materials and Methods ................................ ................................ .......................... 101 Results and Discussions ................................ ................................ ........................ 105 Cratering Large Glass Spheres ................................ ................................ ....... 105 Effect of Particle Size and Density on Crater Growth ................................ ...... 110 Effect of Particle Shape on Crater Growth ................................ ...................... 115 Scaling Asymptotic Crater Depth ................................ ................................ .... 123 Summary ................................ ................................ ................................ ............... 129 5 CRATERING UNDER LUNAR CONDITIONS ................................ ....................... 151 Motivation ................................ ................................ ................................ .............. 151 Background ................................ ................................ ................................ ............ 151 Experimental Methods ................................ ................................ ........................... 156 HOOSH Based Experiments ................................ ................................ ........... 156 Lunar Regolith Experiments ................................ ................................ ............ 158 JSC 1A Bulk Density Cratering Experiments ................................ .................. 160 Results and Discussion ................................ ................................ ......................... 161 HOOSH V ariable Gravity Experiments ................................ ............................ 161 Cratering of JSC 1A: Effect of Bulk Density and Jet Velocity .......................... 163 Segregation of JSC 1A During Cratering ................................ ........................ 165 Cratering Rates of Sieved JSC 1A versus Glass Spheres .............................. 165 Crater growth for Lunar Soil Simulants Chenobi and JSC 1A ......................... 167 Crater Shape Variations ................................ ................................ .................. 169 Scaling the Asymptotic Crater Depth ................................ .............................. 170 Summary ................................ ................................ ................................ ............... 172 6 CONCLUSIONS AND RECOMMENDATIONS ................................ ..................... 190 Permeability of JSC 1A ................................ ................................ .......................... 190 Two Fluid Modeling ................................ ................................ ............................... 190 Parametric Study of Particle Properties on Crater Growth ................................ .... 192 Experimental Study of Crater Growth in Lunar Conditions ................................ .... 193
7 LIST OF REFERENCES ................................ ................................ .............................. 195 BIOGRAPHICAL SKETCH ................................ ................................ ........................... 205
8 LIST OF TABLES Table page 2 1 Nomenclature and constants used within Chapter 2 ................................ ........... 4 0 2 2 JSC 1A particle properties ................................ ................................ .................. 41 2 3 Permeability ranges for various porous media ................................ .................... 41 3 1 Nomenclature used in this chapter ................................ ................................ ...... 74 3 2 Governing Equations ................................ ................................ ........................... 75 3 3 Constitutive Equations ................................ ................................ ......................... 76 3 4 Turbulence Equations (Fluid Phase) ................................ ................................ ... 78 3 5 Simulation Parameters ................................ ................................ ........................ 82 3 6 Summary of Frictional Models (n=2 for all models) ................................ ............. 83 3 7 Predicted Frictional Viscosity and Pressure Associated with Different Frictional Models ................................ ................................ ................................ 84 4 1 Bulk and Intrinsic Properties of Particles ................................ ........................... 131 5 1 Properties of Particles ................................ ................................ ....................... 175
9 LIST OF FIGURES Figure page 2 1 Apparatus of permeameter system. ................................ ................................ .... 42 2 2 Determination of viscous flow permeability with bulk = 1550 kg m 3 .................. 43 2 3 Permeability of JSC 1A at various bulk densities, bulk ................................ ...... 44 2 4 Determination of the constant in the Carman Kozeny equation. ......................... 45 3 1 Schematic of experimental setup. ................................ ................................ ....... 85 3 2 Axisymmetric formulation computational mesh and simulation domain. ............. 86 3 3 Predicted crater depth with varying frictional models for a 37 m/s Argon jet. ...... 87 3 4 Crater depth predictions with varying Argon jet velocities (Model A). ................. 88 3 5 Crater depth prediction with varying gas type (Model A). ................................ .... 88 3 6 Crater depth prediction with varying nozzle height (Model A). ............................ 89 3 7 Long time crater depth prediction with 37 m/s Argon jet (Model A). .................... 89 3 8 Experimental crater images and corresponding simulations at different times. .. 90 3 9 Simulation domain and computational mesh. ................................ ...................... 91 3 10 Crater depth prediction with varying simulat ion domain (Model A). .................... 92 3 11 Contour plots of solids volume fraction from planar mesh with thin wall simulation for various times. ................................ ................................ ................ 93 3 12 Solid phase mass flux (m/s) vectors and solids volume fraction iso surfaces predicted by Model A. ................................ ................................ .......................... 94 3 13 Gas velocity (m/s) iso surface contours plotted and volume fraction iso surfaces predicted by Model A. ................................ ................................ ........... 96 4 1 Schematic of Experimental Apparatus. ................................ ............................. 132 4 2 Map of particles arranged by particle shape, size and intrinsic density. ........... 133 4 3 Microscope images of 0.25 0.3 mm particles. ................................ ................... 133 4 4 Transient crater depth and width for 2.3 mm glass spheres by jets with velocities 35.36, 67 .18, and 93.74 m/s. ................................ ............................. 134
10 4 5 Crater evolution for 2.3 mm glass spheres under jets with varied velocity. ....... 135 4 6 General trajectories of 2.3 mm glass spheres cratered by a 93.74 m/s jet. ...... 136 4 7 Crater depth evolution for glass spheres with various particle diameters and jet velocity 20.90 m/s. ................................ ................................ ........................ 137 4 8 Crater width and volume over time for glass spheres with various diameters and jet velo city 20.90 m/s. ................................ ................................ ................. 138 4 9 Crater depth evolution for small spherical particles with varied intrinsic density and jet velocity 20.90 m/s. ................................ ................................ .... 139 4 10 Crater depth evolution for large spherical particles with varied intrinsic density and jet velocity 31.05 m/s. ................................ ................................ .... 140 4 11 Crater depth evolution for stainless steel particles with equivalent sphere volume diameter of approximately 1 mm with spherical and cylindrical particles and jet velocity 67.18 m/s. ................................ ................................ .. 141 4 12 Crater depth over long time frames for plastic particles with equivalent volume diameter of approximately 1 mm with spherical, cylindrical and cubic particles and jet velocities of 31.05 m/s. ................................ ............................ 142 4 13 Crater evolution of 0.3 mm diameter polystyrene spheres by a 20.90 m/s jet. 143 4 14 Crater d epth evolution of non spherical particles and spherical glass beads with mean sieve diameters between 0.25 0.3 mm and intrinsic densities between 2.5 to 3.9 g/cm 3 by a 20.90 m/s jet. ................................ .................... 144 4 15 Trajectory of particles moving along the outer crater for spherical and non spherical beds. ................................ ................................ ................................ .. 145 4 16 Crater depth evolution for glass spheres with diameters of 1.0 mm, 0.5 mm and a 50% by wt. mixture of 1 mm and 0.5 mm with jet velocity 20.90 m/s. ..... 146 4 17 Image of crater for 50% wt. mix of 1 m m and 0.5 mm glass spheres after the jet was turned off. ................................ ................................ .............................. 147 4 18 The asymptotic crater depth scaled to the original and modified densimetric Froude numbers. ................................ ................................ ............................... 148 4 19 Fitting of crater depth data. ................................ ................................ ............... 149 4 20 Comparison of measured and predicted asymptotic depth from the arctangent fit scaled by the modified densimetric Froude number that accounts for particle shape. ................................ ................................ .............. 150 5 1 Schematic of th e HOOSH. ................................ ................................ ................ 176
11 5 2 Schematic of Experimental Apparatus. ................................ ............................. 176 5 3 Equival ent sphere volume diameter of Chenobi and JSC 1A. .......................... 177 5 4 Microscope images of particles. ................................ ................................ ........ 178 5 5 Crater growth for glass spheres in various gravities. ................................ ........ 178 5 6 Crater images of polystyrene and glass spheres in varied gravity. ................... 179 5 7 Influence of jet Reynolds number and gravity on crater evolution for polystyrene spheres. ................................ ................................ ......................... 179 5 8 The effect of jet velocity and particle density o n crater growth in lunar gravity. 180 5 9 Effect of gravity on crater width for glass spheres. ................................ ............ 181 5 10 Crater growth for JSC 1A at various bulk densities and nitrogen jet velocities. 182 5 11 Segr egation of a bed of JSC 1A by particle size during cratering. .................... 183 5 12 Crater depth and width evolution of JSC 1A, sieved JSC 1A, and glass spheres under air jets with velocity of 25.46 m/s. ................................ .............. 184 5 13 Crater growth for Chenobi and JSC 1A under various velocity air jets. ............ 185 5 14 Crater evolution for Chenobi. ................................ ................................ ............ 186 5 15 Crater shape after 50 sec onds by a 25.46 m/s air jet. ................................ ....... 187 5 16 Crater shape after the jet was turned off. ................................ .......................... 188 5 17 The asym ptotic crater depth scaled to the original and modified densimetric Froude numbers. ................................ ................................ ............................... 189
12 LIST OF ABBREVIATIONS BCF Bearing Capacity Failure D ENSIMETRIC F ROUDE N UMBER (F O ) Ratio of kinetic to potential energy DDS Diffusion Driven Shearing DGE Diffused Gas Eruption DGEE Diffused Gas Explosive Eruption P ERMEABILITY Resistance to flow through a granular material R EYNOLDS N UMBER (R E ) Ratio of ine rtial to viscous forces VE Viscous Erosion
13 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy CRATERING OF A PARTICLE BED BY A SUBSONIC TURBULENT JET By Casey Q. LaMarche May 2013 Chair: Jennifer S. Curtis Major: Chemical Engineering Future space exploration will benefit from a simulation tool that can help mitigate the hazards of rocket exhaust impinging on extraterrestrial regolith covered surfaces and necessitates improved gas particle and particle particle interaction models. Developing models which describe the interaction of rocke t exhaust plume with a particle bed, requires experimental data on cratering behavior specifically on the effects of gravity, jet and particle properties. In order to model cratering of a particle bed by a turbulent subso nic jet an Eulerian framework is employed using locally averaged equations of motion along with closure relations describing the solids phase stress derived from concepts of granular kinetic theory The particle phase frictional stresses, which occur at high solids volume fractions and result from sustained particle particle contact s, are based on critical state theory. A new frictional stress model is developed, which is able to predict the salient trends in crater predictions to changes in jet properties and determine t he cratering mechanisms Next, a parametric study is presented to investigate the effects of particle properties on crater growth. Effects of particle size, shape, and density on asymptotic
14 crater depth and crater growth over time are analyzed. An adjustment to the scaling function between the asymptotic crater depth and jet and particle properties is developed that captures the effects of particle shape. The particle shape is identified to be a non negligible particle property for crater growth. Additionally, c ratering experiments were performed to investigate the effects of cratering in lunar conditions. Variable gravity experiments utilizing particles with different intrinsic densities explored the effects of gravity on crater growth and the dominating crater mechanism s Earth based cratering e xperiments of lunar regolith simulants express the importance of bulk density and the percentage of fine particles on crater growth. The crater growth over time and the asymptotic crater depth are explained in terms of the characterized properties of lunar regolith simulants.
15 CHAPTER 1 INTRODUCTION Future Terrestrial Space Exploration Future exploration of space will necessitate larger rockets with larger rocket exhaust plumes. A thick layer of granular material called regolith covers the surfaces of the moon, Mars some asteroids and Earth Regolith is made up of dust, soil, and broken rocks. The rocket exhaust of the spacecraft will interact with the regolith covered surfaces during landing and takeoff Depending on the properties of the granular material, the rocket exhaust, and the ambient atmosphere environment the jet interaction can have drastically different results including particle liberation, erosion and cratering. The soil that is lofted an d accelerated due to jet plume impingement can cause significant damage and contaminate co landing spacecraft and surrounding structures. Crater formation can be hazardous to a structure's stability on the surface. During the Apollo mission the eroded particle spray obscure d the astronauts' view of the lunar surface (Metzger et al., 2008). The Martian rovers have experimented with several landing techniques including a new model utilized by the rover Curiosity, that has rockets attached to an umbilical structure six feet above the lander in an attempt to mitigate the effects of the plume impingement interacting with the Martian surface (Sengupta et al., 2009). Mitigating the hazards of particle spray and cratering induced by the interaction of rocket exhaust plume with the granular covered surface is paramount to future space exploration. It is extremely expensive to do experiments that mi mic the real conditions on the l unar and Martian surfaces, so to understand the behavior of the particle spray it is beneficial to develop a computational framew ork that can simulate all operating
16 conditions. This software package must be validated with experimental data that explores the important physical regimes. C onducting full scale experiments to capt ure the physics of rocket exhaust blowing lunar or Martian soil is difficult The only feasible way to investigate the physics, which will lead to the design of dust and debris impact mi tigation measures, is to rely on robust computational models. Small sc ale experiments can be performed at low gravity or low pressure, and larger scale experiments at ambient conditions to validate such physics based models. These small scale experiments can be used to calibrate and benchmark simulation tools, which can pred ict the behavior of the large scale, fully lunar or Martian cases. Large scale experiments can then be performed with relevant materials in order to validate the model s ability to account for material specific properties. The interaction of circu lar jets and dense particle laden beds is studied by the use of turbulent subsonic jets impinging on particle beds (Haehnel et al., 2008; Metzger et al., 2009a; Rajaratnam and Beltaos, 1977). When a turbulent subsonic j et impinges onto a particle bed in Earth's atm ospheric conditions, a crater or scour hole, is produced. Scour hole growth is a widely studied area as it is important to hydraulic engineering applications and because scour can play a crucial role in the stability of a foundation's structure ( Aderibig be and Rajaratnam, 1996; Karim and Ali, 200 0; Kobus et al., 197 9 ; Rajaratnam and Beltaos, 1977 ). Previous Example: Apollo 12 Landing on Lunar Surface (Metzger et al., 2008) Due to the low gravity and atmosphere, particles on the lunar surface can be easily ejected. Apollo 12 landed about 183 meters away from the Surveyor 3. The Surveyor appeared to be sandblasted by the dust and debris cast out by the Apollo 12
17 rocket exhaust during landing. The Surveyor's optical mirror was damaged by adhered particles and pitting. The damage done by the dust and debris plume caused by the Apollo 12 rocket exhaust illustrates the importance of effectively protecting equipment and architecture from the soil and particulate plume during lunar landings and take of fs. It is important to note that the rockets for future lunar missions will have much larger landers, which will require more engines with higher thrusts. Permeability of Lunar Simulant JSC 1A Establishing a connection between permeability and other porous media characteristics (e.g., particle size distribution and specific surface area) is important for understanding gas exchange processes at the lunar surface. These processes, in turn, have implications for both lunar surface science and human exploration If ice once occupied regions of the lunar regolith, its long term diffusion history and ultimate fate would be affected by the permeability of regolith (Albert et al. 2000; Albert and Shlutz, 2002). Likewise, permeability strongly influences 1) the form ation of impingement craters during spacecraft ascent and descent (Haehnel et al. 2008; Metzger et al. 2008; Metzger et al. 2009a ; Scott and Ko, 1968); 2) the depth of chemical contamination due to rocket exhaust; 3) the time frame during which exhaust chemicals diffusing out of the regolith can contaminate astronauts and hardware; 4) the ease of pneumatic excavation for targeted collection or movement of regolith; and 5) efficiency of oxygen production during In Situ Resource Utilization (ISRU) (Nemec a nd Levec, 2005). During future manned or robotic missions, understanding these processes in detail will help distinguish naturally existing underground ice from water introduced to the regolith by the penetration of rocket exhaust. Thus, it is important to quantify the range of permeabilities that might be encountered at the lunar surface. Chapter 2
18 reports measurements of the permeability of JSC 1A with bulk density. The bulk density range is similar to what was found on the top layers of the lunar regolit h (Mitchell et al. 1974). The results show that the Carmen Kozeny equation for estimating the permeability of soils was adequate for predicting the permeability of JSC 1A. The permeab ility measurements reported in C hapter 2 show that the permeability of J SC 1A is similar to what was estimated based on Surveyor 5 data by Choate et al. (1968) Multiphase Flow Modeling There are two main modeling frameworks for multiphase flow: Discrete Element Modeling (DEM) and two fluid modeling. DEM solves Newton's laws o f motion for every particle. In the case of DEM, the properties of individual particles can be changed, but simulations are limited to the number of particles that processors can handle, which are generally orders of magnitude less than most realistic appl ications. Two fluid modeling assumes that both the solid and fluid phases are interpenetrating continua, and that ensemble averaged continuity and equations of motion are solved for both phases. Constitutive relations are introduced for two fluid modeling in order to close the equations of motion for the solids phase Nonetheless, two fluid modeling is capable of simulating realistically sized domains. Dense phase solids flows which occur with cratering formation, dissipate momentum due to friction caused by enduring particle particle contacts The two fluid model accounts for friction with a frictional stress in the Newtoni an form constituting a normal and shear component Chapter 3 provides the validation of the two fluid framework for performing cratering simulations. Additionally, Chapter 3 reports the effect of the magnitude of the frictional stress and the relative magnitudes of the frictional pressure and frictional vi scosity in terms of the crater predictions.
19 Effects of Particle Shape on Cratering Often experiments used to validate granular flow models rely on ideal investigations, which are conducted wi th smooth spherical particles and mono disperse or narrow parti cle size ranges (B ttega et al., 2009; Boemer et al, 1997; Kuang et al., 201 3 ). However, experiments performed with non ideal particles, like sand, have also been used to validate granular flow models, but the effect of particle shape is either ignored (Ac osta Iborra et al., 2011) or dealt with in terms of empirical parameters (McKeen and Pugsley, 2003). Studies of particle shape effects on fluidized bed and hopper flow have been performed with DEM modeling (Cleary and Sawley, 2002; Hilton et al., 2010), wh ich generally incorporates idealized non spherical particles (i.e. cylinders and ellipsoids) and simulation results largely are not compared to experimental data. Metzger et al. (2009a) and Metzger et al. (2009b) provide some of the few reports of the transient crater growth. Metzger et al. (2009a) performed a parametric study on the effects of jet properties and report that the crater growth rate at low velocities increased with increased jet velocity. The only study that investigated a change in the crater growth rate over time for a series of particles with a single propert y specifically particle size, change is Metzger et al. (2009b) in which they report the cr ater growth for four sieved ranges of sand particles within 200 600 microns Metzger et al. (2009b) found that the long term transient crater depth and width growth increased with decreased particle size. This is the first work to investigate the effects of particle shape on cratering. While previous studies have ignored the effects of particle shape on scour hole or cratering behavior, the effect has been more widely studied on fluidized beds (Hilton et
20 al ., 2010; Sau et al., 2010; Zhou et al., 2011). For non spherical particles in fluidized beds, t he effect of particle shape on por e size and projected area dominate the deviation from theory based on spherical particles (Hilton et al., 2010). The ratio of t he surface area of the particle to the surface area of a sphere with the same volume, or sphericity, is a measure of the particle shape. The sphericity is often multiplied by the particle diameter in various models to account for the shape of the parti cle ( Kunii and Levenspiel, 1997 ). Hilton et al. (2010) imple mented sphericity dependence in the coefficient of drag, which improved the friction factor and pressure drop correlation predictions when compared to their DEM results. Particle shape affects the permeability of the granular material, which in turn effects the penetration of the jet into the bed (Carman, 1937; Niven, 2002). The gas penetration from the jet into the bulk of the particle bed is important for the various cratering mechanisms, as the gas pressure on the top layers of particles as well as deep into the bed have implica tions on the relative magnitude of forces and therefore the cratering rates and dominating cratering mechanisms (Metzger et al 2009a) Friction and drag forces are also important to the crater growth rate and cratering mechanism s Particles erode along the surface due t o the fluid particle drag force and the gas shear stress along the surface, while the frictional force between particles impedes the particle movement. Al so, it was shown in Metzger et al. (2009a) that the particles erode off of the inner crater edge and deposit on the outer crater, the surface of the bed or outside of the box containing the particle bed Particle beds are able to form piles, and the steep est angle that the bed can withstand is referred to as the angle of failure. If the angle of the pile exceeds the angle of failure, then the sides of the pile
21 fail, material avalanches off of the wall, and the angle reduces to the angle of repose. Often, t he reported angle of repose is measured somewhere between the actual angle of repose and the angle of failure. The outer crater then periodically avalanches when the angle exceeds the angle of failure Metzger et al. (2009a) identified that there is a cert ain amount of particle recirculation, which subsequently has a large impact on the crater volumetric growth rate. The rate at which particles avalanche from the outer crater into the inner crater results from the difference in angle of the outer crater and the angle of repose of the material (Metzger et al., 2009a), which is dominated by the friction between particles A ngular, rough, and non spherical particles have higher friction than smooth, spherical particles (Anthony and Marone, 2005; Cho et al., 200 6; Shinohara et al., 2000). While Chapter 4 is the first to investigate the effects of particle shape on cratering, it additionally provides crater growth data for spherical particles, which is also lacking in the literature. Cohesion and Fine Particles Previously, the effects of cohesion on crater growth were largely ignored. Cohesion is the property of a material to stick ing together as opposed to adhesion in which materials stick to other materials In terms of particles, cohesiveness is the st ickines s of particles with each other. Particle size and shape, electrostatics and moisture content contribute to the cohesiveness of a particle bed (Rhodes, 2008 ) Mechanical interlocking of particles can increase the effects cohesion as rough surfaced particle s can become jammed together With reduced particle size, van der Waals forces increase relative to gravity forces, which tend to increase cohesion. Additionally, cohesion is increased by w ater adhered to the surface of particles that can induce
22 capillary forces in the form of liquid bridges between particles Van der Waals forces are a function of particle separation distance and Royer et al. (2009) found for approximately 100 micron diameter particles that cohesion forces caused by van der Waals and capil lary forces are orders of magnitude higher for smooth spherical particles than the same particles with nanometer sized defects added to the particles. The cohesiveness of the bed tends to increase with decreasing particle size, and fine particles describe the particle sizes when cohesion becomes significant This work defines fine particles as particles below approximately 50 microns. Chapter 4 and 5 investigate the effects of cohesion on crater growth. Cratering Mechanisms To date, five different craterin g mechanisms have been identified that have been identified in cratering of particle beds by jets : Viscous Erosion (VE) (Roberts, 1963), Bearing Capacity Failure (BCF) (Alexander et al., 1966), Diffused Gas Eruption (DGE) (Scott and Ko, 1968), Diffusion Dr iven Shearing ( DDS ) ( Metzger et al., 2009a) and Diffusive Gas Explosive Erosion (DGEE) (Mehta et al., 2011). The details of all five mechanisms can be found elsewhere ( Alexander et al., 1966 ; Mehta et al., 2011; Metzger et al., 2009 a; Metzger et al., 2010 ; Roberts, 1963 ; S cott and Ko, 196 8). This work mainly discusses the three mechanisms that are expected to be relevant: Viscous Erosion (VE), Diffusion Driven Shearing (DDS), and Bearing Capacity Failure (BCF). VE (Roberts, 1963) ensues when the shear forc e of the jet, proportional to the dynamic pressure, on the surface overcomes the shear strength of the surface particles, causing them to erode into the flow. Shearing that occurs from DDS is caused by gas that diffuses through the pore spaces of the granu lar material and drags the particles through the subsurface layers. DDS can be identified by particles shearing tangentially
23 around the crater (Metzger et al., 2009 a ) When the stagnation pressure of the jet exceeds the bearing capacity of the soil the cr ater forms as a shallow cup due to BCF, and particles move perpendicularly away from the jet impingement region (Alexander et al., 1966 ; Metzger et al., 2009 a ). The VE and DDS mechanisms can be differentiated by how deep shearing occurs into the bed, where VE presents with erosion of only the top layers as opposed to a thick band of subsurface partic le shearing with DDS (Metzger et al., 2010). The following mechanisms are not expected to be relevant to this work, due to the circumstances in which they occur under. Mehta et al. (2011) discovered that under supersonic pulsed jets high pressure gradient to particle gravity forces arise that result in explosive erosion or DGEE DGE occurs when the jet pressurizes the fluid in the pores of the soil in a radiall y expanding subsurface region, causing an eruption of particles in an annular ring around the jet (Scott and Ko, 1968), but is thought to be a secondary cratering mechanism (Metzger et al, 2009a). In Chapter 2 the results of the permeability of JSC 1A are described in relation to DGE. Metzger et al. (2009a) studied cratering of sand at high and low velocities. At low jet velocities VE dominated cratering the crater depth was reported to grow logarithmically with time, and the crater split into an inner p arabolic crater and an outer conical crater However, Metzger et al. (2009a) reported that for their high velocity experiments, in which DDS or BCF dominated, crater s formed that did not grow logarithmically with time but rather a deep crater formed that quickly reached its maximum size. Metzger et al. (2009b) also found that the cratering when BCF dominates, and likely similar for DDS, is mostly controlled by the plume parameters (i.e. jet extinction length) by performing experiments on loose and compacted beds of JSC
24 Mars 1A (a Martian soil simulant). Additionally, Metzger et al. (2009b) reports that when BCF or DDS dominate, the final crater depth reached is independent of bed bulk density, but more compressed beds crater slower. Berger et al. (2013) identified particle particle collisions as a signif icant factor for surface erosion rates far down wind of the stagnation region from rocket exhausted impinging on in lunar regolith. In Chapter 4, two new particle induced cratering mechanisms are identified and discussed. Scaling Asymptotic Crater Depth T he crater depth has often been reported to grow logarithmically with time (Kobus et al. 1979; Metzger et al., 2009a; Rajaratnam and Beltaos, 1977; Westrich and K obus, 1973). After growing logarithmically or arctangently with time for ma ny generations, the crater will reach or approach an equilibrium or steady state size, which is often referred to as the asympt otic size (Haehnel et al., 2008 ; Rajaratnam and Beltaos, 1977; Sarma, 1967; Westrich and Kobus, 1973). Sarma (1967) discussed the crater depth growt h behavior and functions to describe the growth over time, most of which asymptote. Rajaratnam and Beltaos (1977) investigated the growth of scour holes under vertically impinging jets impacting beds of sand and polystyrene particles and found that the asy mptotic depth, D asym scales linearly with the erosion parameter, E c as shown in equation (1 1) below. (1 1) w here H is the height of the nozzle exit above the bed, m 1 is the slope and m 2 is the intercept. Rajaratnam a nd Beltaos (1977) propose that m 2 i s weakly dependent on the ratio of the pipe diameter, d pipe and H. D as y m H = m 1 E c + m 2
25 The erosion parameter that has been u sed widely, given as equation (1 2) below, is proportional to the densimetric Froude number, Fo (Aderibigbe and Rajaratnam 1996; Hae hnel et al., 2008; Rajaratnam and Beltaos, 1977; Rajaratnam 1982). Rajaratnam and Beltaos (1977) did not account for the dependence of crater depth on particle shape Additionally, it is not clear whether they report the static (jet was turned off for the crater measurement) or dynamic (jet remained on during crater measurement) asymptotic crater depth. (1 2) w here v is the jet velocity, "!= p f is the difference of the intrinsic particle density, p and the fluid density, f and d p is the p article diameter. There are many different ways of measuring the particle diameter. Rajaratnam and Beltaos (1977) did not explicitly express which mean di ameter they used for equation (1 2), but it was most likely the sieve diameter as Aderibigbe and Rajar atnam (1996) report using the median sieve diameter for scaling their experiments. However, the sieve diameter does not account for particle shape and can under report the size of elongated particles. Rajaratnam (1982) performed crater, or scour hole, exp eriments with water jets and found the asymptotic crater depth to scale with equation (1 2), but the results of their experiments did not scale with Rajaratnam and Beltaos' (1977) experiments (Haehnel et al., 2008). Haehnel et al. (2008) investigated the a symptotic crater depth for three different materials and found that it is weakly influenced by the permeability, a measure of the resistance of flow through a porous medium, of the bed. Haehnel et al.'s (2008) scaling relationship included dependence on th e permeability, which may have E c = v g f # $ % & ( d p d pi pe H = F o d pi pe H
26 provided an account for particle shape. They were able to get better agreement between the data of Rajaratnam and Beltaos (1977) and Rajaratnam (1982) by including the fluid viscosity and granular material's permeability in t heir modified scaling term. The particle density, size, or shape of the particles tested by Haehnel et al. (2008) did not overlap but they were still able to find decent agreement with their scaling parameter. Chapters 4 and 5 investigate the effects of pa rticle properties on the densimetric Froude number. Lunar Soil Simulants JSC 1A is a low titanium lunar mare soil simulant that was sieved and milled from volcanic ash to retain a particle size distribution similar to the samples returned from the Apollo m issions, of particles with diameter 1 mm and lower (Alshibli and Hasan, 2009, He, 2010; McKay et al, 1994). Alshibli and Hasan (2009), Zeng et al. (2010) and LaMarche et al. (2011) found that many of the geotechnical properties of JSC 1A are close to lunar soil. Chenobi, also referred to as CHENOB1, is chemically enhanced OB 1 in which plasma arc techniques are used to make glass and agglutinates from the anorthosite (Battler and Spray, 2009) OB 1 is a Shawmere ano rthosite that was crushed to match the particle size distribution of lunar highland regol ith, specifically Apollo sample 64 500, and additionally contains crushed olivine slag mixed to produce a particle size distribution of the agglutinate and glass compo nents (Battler and Spray, 2009) To the author s knowledge very little work has yet to be performed on characterizing Chenobi as it is a very new simulant. P article shape and bulk compressibility are reported elsewhere (Rickman et al., 2012; Rickman et al., 2013 ; Schrader et al, 2010; Rahmatian and Metzger, 2012). Schrader et al.'s (2010) guide to lunar simulants provides chemical
27 composition and shape descriptions for JSC 1A, OB 1 and various other lunar simulants as well as r ecommendations of use for lunar simulants includin g JSC 1A and Chenobi. Chapter 5 reports crater growth investigations for both Chenobi and JSC 1A.
28 CHAPTER 2 PERMEABILITY OF LUNAR SOIL SIMULANT JSC 1A Motivation JSC 1A is a lunar soil simulant that attempts to match lunar soil mechanically and chemically. Lunar soil simulant is used to study excavation and ISRU methods and help design pneumatic conveyers and chemical reactors. Specific knowledge of the permeability of JSC 1A as a functi on of bulk density is necessary for these methods and designs to apply to the full range of lunar conditions. Since permeability is a measure of a porous medium's resistance to fluid flow through its pores, it should only depend on the characteristics of t he media, not on the fluid properties. However, measuring permeability with gas, rather than liqu id, has many benefits the major one being experimental time. In addition, permeability testing in situ on planets and celestial bodies with low atmospheres ( i.e., Moon, Mars, asteroid, Mercury) necessitates using a gas to probe the soil. Hence, the present work utilizes gas phase permeability. Background and Theory For the unidirectional viscous flow, the pressure drop per unit length of a porous bed ( dp/dx ) i s given by Darcy's Law u = K dp dx (2 1) where is the fluid viscosity, u is the superficial fluid velocity, and p is the pressure. Table 2 1 provides the nomenclature used in this chapter.
29 The viscous permeability, K has units of area, is material specific and independent of fluid properties. For the flow of compressible ideal gases of constant mass flux u Darcy's law is given as up = 1 2 K dp 2 dx (2 2) After integration, equation ( 2 2) becomes u 1 p 1 = K p p L (2 3) where u 1 and p 1 are the ups tream velocity and pressure, respectively, L is the total bed length, and p is the mean pressure within the porous medium. It should be noted that, for the same material, permeability measured with a gas is often slightly higher than when measured with a l iquid. The enhanced permeability is attributed to molecular slip at the capillary wall when measuring the permeability with a gas. Darcy's law is only valid for purely viscous flow through porous media. In the case of the gas flow, partially viscous (slip flow) or free molecular flow can often occur (Bear, 1988; Carman, 1937; Carman, 1950; Scheidegger, 1974). In partially viscous or slip flow, the pore or capillary diameter approaches the molecular mean free path of the gas, increasing the frequency of coll isions of molecules with the pore side walls relative to molecule molecule collisions. In free molecular flow, or Knudsen flow, the pore diameter is smaller than the molecular mean free path of the gas, so gas molecules collide with the walls more frequent ly than with each other. First Adzumi (1937), and later Klinkenberg (1941), attributed these enhanced permeabilities with a gas to slip effects.
30 The effect of the increased wall flux and permeability of the gas in the slip flow regime can be described via analysis of flow through a capillary. Poiseulle's law for viscous flow of gas through a capillary is given as u 1 p 1 = d c 2 32 p p L c (2 4) where d c is the capillary diameter and L c is the length of the capillary. The similarity between equations (2 3) and (2 4) pres ents the basis for modeling flow in porous media as flow through bundles of capillaries. The enhancement of the permeability due to slip effects depends on the capillary diameter, gas density, thermal molecular velocity v and the fraction of gas molecules that undergo diffuse reflection at the capillary walls, f 1 (Carman 1956). Flow through a capillary with slip effects can be expressed (Carman, 1956) u 1 p 1 = p p L c d c 2 32 + d c 2 f 1 ( ) 2 # v f 1 $ % & ( ) where v = 8 R T M (2 5) or equivalently u 1 p 1 p = d c 2 p 32 + d c 4 L c R T 2 M 2 # f 1 f 1 (2 6) The first term on the right hand side of equation (2 6) is the viscous flow term while the second term is the molecular slip term (Carman, 1956) The thermal molecular velocity v is a function of molecular weight M the universal gas constant R and the absolute temperature T Using the assumption that the porous medium is made up of a bundle of parallel capillaries, and by analogy with equation (2 6) equation (2 3) can be modified to account for slip as ( Adzumi, 1937; Carman, 1956; Klinkenberg 1941)
31 u 1 p 1 L p = K p + C (2 7) For partially viscous flow through porous media, a straight line, with intercept C and slope K result s when u 1 p 1 L p is plotted against the mean pressure p for large p As p is reduced, there is a transition from partially viscous flow towards free molecular flow; this transition is associated with a minimum on the plot of u 1 p 1 L p versus p (Carman, 1956). For uniform pore size in the bed, the viscous permeability K can be expressed as a function of the bed porosity the tortuosity of the pores the shape of the capillary cross section (through a shape factor k o ), and the particle specific surface S o via the Carman Kozeny equation (Carman 1937 ; Carman, 1956). K = 1 2 k o 3 S o 2 1 # ( ) 2 (2 8) For unconsolidated media that contain random pore shapes and sizes, the relationship expressed in equation (2 8) is still valid provided k o and are not significantly affect ed by the varying pore configurations ( Carman, 1956). And, in many cases, 1 2 k o 0.2 (Mauran et al. 2001; Chapuis and Aubertin, 2003). Experimental Hardware and Techniques Gas phase permeability measurements were obtained in a Humboldt permeameter H 4146 by flowin g compressed nitrogen gas through a sample of JSC 1A. For various flow rates, the steady state axial pressure drop was measured across the
32 entire system using a Dwyer series 477A 6 digital manometer with a 0 344.8 kPa range and an accuracy of 0.1% F.S. On e pressure tap was placed at the inlet of the system while the other was placed at the outlet. No hysteresis in the pressure drop with varying flow rate was observed. The flow rate, ranging from 5 to 45 L min 1 was controlled using a Sierra MKS 0 50 SLPM mass flow controller with an accuracy of 2% F.S. and was monitored upstream of the permeameter. A porous stone with a 0.154 m diameter and width of 0.005 m was fixed to the base of the permeameter using silicone caulk. The cylindrical body of the standar d' Humboldt permeameter is made up of two sections that do not produce an airtight seal. Hence, in the present work, a single hollow metal cylinder 0.154 m inner diameter and 0.170 m long, fixed to the base plate of the apparatus, was used instead. Figure 2 1 provides a schematic of the experimental apparatus. The lunar soil simulant JSC 1A was poured into the apparatus using an ASTM standard funnel. The samples were densified inside of the permeameter so that the bulk density could be determined without subsequent disturbance. An overburden weight was placed on top of the sample, and the table was vibrated. Given this procedure, the bed densification should be fairly uniform b ecause the overburden weight was much heavier than the soil sample. Nine different bed compactions, with bulk densities ranging from 1550 and 2000 kg m 3 were obtained by vibrating the table at variable speeds for different times. The beds were densified only in the axial direction. After a sample was prepared, another porous stone with a diameter of 0.154 m and width of 0.003 m was carefully placed on the top of the soil. Pressure drop experiments at varying flow rates were first performed with no JSC 1A present in the system in order to
33 determine the background pressure drop. Background values for porous stones and fittings were subtracted from the pressure drops measured across the system and the bed. The JSC 1A was characterized in terms of its particl e density and size distribution, as well as its specific surface. Particle density was measured using a Quantachrome Ultrapyc 1000 Gas Pycnometer (accuracy 0.03%). The average particle specific surface and the particle size distribution were obtained usin g the laser diffraction based Coulter LS13320. The Coulter LS 13320 calculates the average specific surface under the assumption that the particles are spherical Table 2 2 shows the relevant particle properties of JSC 1A. The poured, untapped, bulk densit y of JSC 1A is 1550 kg m 3 Results and Discussion Figure 2 2 presents an example of a permeability measurement at a bed bulk density of 1550 kg m 3 When the pressure drop flow rate data are plotted using the relationship given in equation (2 7) the line ar region at higher p is associated with partially viscous flow. The slope of this linear region is associated with the viscous permeability K For all the bulk densities studied, a linear fit to the partially viscous flow region yields an average norm of residuals on the order of 10 4 As p is reduced, there is transition towards free molecular flow and the linear region is no longer present. The measured viscous flow permeabilities as a function of the bed bulk density are shown in Figure 2 3 The permea bility decreases as the bulk density increases. The permeability of JSC 1A for the range of bulk densities investigated is on the order of 10 12 m 2
34 This range of permeabilities for JSC 1A is similar to measurements made by Sizemore et al. (2008) for a variety of Martian analogs. Sizemore et al. (2008) measured the permeability of JSC Mars 1, spherical particles, and soils associated with shallow ground ice from the Antarctic Dry Valleys. The lowest porosity for the Martian simulant in the Sizemore et al. study was greater than the highest porosity for the JSC 1A measured here. When the permeability ranges are compared considering the porosity ranges, JSC Mars 1 is less permeable than JSC 1A. This difference in permeability between JSC 1A an d JSC Mars 1 may be due to the presence of additional terminating pores in the JSC Mars 1, so the actual pore space available for gas flow in the Martian simulant is reduced. As a reference, Table 2 3 contains permeability values of various materials for c omparison to the permeability of JSC 1A. When bulk density and porosity are provided in the publications, they are also reported in Table 2 3 The permeability of JSC 1A is comparable to that of peat (Ball et al. 1997 ; Bear 1988 ), dune sand and soil from Linnaeus Terrace (Sizemore et al. 2008). The bulk density of the lunar surface regolith varies with geographical location as well as soil depth. A variety of bulk density and soil depth experiments on the top 60 cm of lunar soil estimate its bulk density between 1500 kg m 3 and 1740 kg m 3 (Mitchell et al. 1974). However, based on drive tube data from Station 4 of the Apollo 17 mission, the bulk density was shown to be as high as 2290 kg m 3 at a depth of 100 cm below the lunar surface (Mitchell et al. 1973). The permeability measurements conducted in the present study are consistent with this bulk density range, so they are relevant for lunar regolith under realistic packing conditions.
35 The permeability values measured for JSC 1A are also consistent wi th estimates based on Surveyor 5 data (Choate et al. 1968). Using crater diameter, exhaust gas, and firing time, Choate et al. (1968) deduced that the permeability of the top 25 cm lunar soil, with an assumed porosity in the range of 0.3 to 0.5, would be between 1x10 12 and 7x10 12 m 2 Hence, JSC 1A is also a good simulant for lunar soil in terms of its permeability characteristics. Unfortunately, there are only a few studies that have investigated the possible effects of soil permeability on different lun ar events. Scott and Ko (1968) studied the flow of rocket engine exhaust into the lunar soil by performing simulations of axisymmetric, isothermal transient flow of an ideal gas through a porous medium following Darcy's law ignoring slip flow, inertial flo w effects and changes in gas dynamics with surface geometry. They did not consider any specific values for the rocket exhaust viscosity, firing time, permeability or porosity, as their simulations were performed using scaled parameters. Nevertheless, their results describe gas penetration into the soil, and the development of pressure gradients in the lateral and radial directions. As time progresses, the pressure gradients advance and become steeper in the radial direction. Scott and Ko (1968) also studied diffusion of pressure gradients when the jet was turned off. Their results identified diffuse gas eruption as one mechanism of cratering and describe the distance into the soil the gas will penetrate, as well as areas of possible crater formation by locat ing regions of unstable soil. Scott and Ko (1968) relate permeability to the time it takes for pressure gradients to develop; these pressure gradients determine the direction of gas flow. For the same jet, a lower permeability soil will reach steady state conditions much slo wer than a more
36 permeable soil. Toroidal cratering occurs when pressure gradients become steep in the radial direction, and the pressure becomes high enough away from the center to lift the soil. Similarly, a spike of soil under the jet occurs when the jet is turned off, as the pressure gradients are steep directly under the jet. Less permeable soil produces less toroidal soil erosion and is less likely to produce a spike of soil under the jet after shutdown than a more permeable soil. A more permeable soil will lead to contamination at a greater depth and radial distance than a less permeable soil. Results of the present study on the permeability of JSC 1A could be combined with a simulation like the one presented by Scott and Ko (1968) to estimate the depth of contamination in the lunar regolith by rocket exhaust, the time scale of the contamination to diffuse out of the soil, as well as regions of unstable soil during rocket firing. Results from the present experiments also imply that t he enhanced flux of gas due to slip at the walls should be included in simulations like Scott and Ko's. Recently, Colaprete et al. (2010) presented the results of the LCROSS mission showing an abundance of H 2 O in one of the Moon's polar craters. Water is present as both ice deposits as well as (chemically or physically) adsorbed on regolith grains (Colaprete et al. 2010). Lunar soil permeability affects the presence, age, and persistence of ice (Andreas 2007 ; Carruba and Coradini, 1999; Vasavada et al. 1999). The flux of water vapor through the regolith via Knudsen diffusion directly depends on soil properties like average pore size, porosity and tortuosity (Salvail and Fanale, 1994). Water can also migrate through the lunar soil as adsorbed molecules hoping (desorbing and re adsorbing) from grain to grain and depends on the pore size (Cocks et al. 2002; Hodges, 2002; Schorghofer and Taylor, 2007).
37 The diffusivity through JSC 1A can be estimated using an exp ression like the one proposed by Millington and Quark (1964), K = 1 32 D D o d p 2 where D/D o is the ratio of diffusivity of the gas in a porous medium to the diffusivity of the gas in air and is the mean squared pore diameter. Using the permeability measurements reported here, more realistic estimates of the flux of water vapor through the regolith pores can be made. In addition, information about the pore diameter based on the permeability measurements can aid in the estimation of the migration of adsorbed water molecules (e.g., slip effects influenced the permeability measurements reported here indicating the pore diameter for JSC 1A is on the order of Nitrogen's mean free path at standard temperature and pressure). Finally, the permeability measurements reported here can help differentiate between naturally occurring water and water introduced by rocket exhaust during takeoff and landing. Soil contaminated by rocket exhaust affects sampling results when testing for ice in shadowed craters (Hodges 2002). In order to ensure clean samples, a model similar to Scott and Ko (1968), combined with the soil permeability information provided here, can estimate the distance from the landing site that is needed to ensure clean samples. No study to date has considered the per meability of the lunar soil when analyzing the availability of ice on the moon below the regolith surface. Hindmarsh et al. (1998) compared sublimation rates of ice through a thick layer of sediment in Beacon Valley, Antarctica considering vapor pressure g radients determined by (1) conditions at the ice surface and (2) by atmospheric conditions at the sediment surface. The former method is similar to that employed by Salvail and Fanale (1994), Vasavada et al. (1999) and others. The results of Hindmarsh et a l. showed that the latter method (vapor pressure
38 gradients determined by atmospheric conditions at the sediment surface) produced sublimation rates three orders of magnitude greater than the former. Hindmarsh et al. (1998) considered the increased sublima tion rate due to Darcian flow and assigned a permeability value of 10 15 m 2 for the Beacon Valley soil, consistent with the measurements of Sizemore et al. ( 2008). The permeability value used by Hindmarsh et al. is a lower limit where Darcian flow time sca les compete with diffusion time scales. Since the permeability value used by Hindmarsh et al. is orders of magnitude lower than what is reported here for JSC 1A, Darcian flow should not be ignored when estimating the deposition of ice (or other contaminant s) from rocket exhaust, or the sublimation and movement of ice in lunar regolith. The permeability measurements for JSC 1A provide a link between particle characteristics and permeability that can be applied to other soils. The data reported here, as well as the results of other studies (Koponen et al. 1997; Mauran et al. 2001), imply that the Carman Kozeny relationship (equation 2 8) using effective particle properties (i.e. specific surface measur ed with light scatter technique ) works well in estimating the permeability of soils with a wide particle size distribution and irregular particle shapes. Figure 2 4 shows that JSC 1A follows the Carman Kozeny relationship even though the stimulant has an extremely wide range of particle sizes/shapes. F rom the slope of this curve (linear fit with a norm of residuals of 3x10 13 ) and the measured average particle specific surface, the calculated value for 1 2 k o = 0.18, which very close to the 0.2 value estimated by Carman (1937)
39 Summary The permeability of JSC 1A was measured for a range of bulk densities, and the viscous flow permeability was found to be on the order of 10 12 m 2 Permeability decreases as bulk density is increased. The viscous permeability is linearly proportional to # 3 /(1 # ) 2 and is well desc ribed using a Carman Kozeny type model even though the particle size distribution is extremely broad. Hence, these results provide evidence that the Carman Kozeny model could be applicable to other naturally occurring soils if effective soil properties are considered. To date, Darcian flow through lunar soil has been ignored in assessing the sublimation rates of ice under layers of regolith; permeability measurements of JSC 1A reported here show that Darcian flow needs to be included in such rate estimates The permeability measurements reported here are also needed to estimate regions of contamination by rocket exhaust and areas of unstable soil due to diffuse gas eruption. Finally, the methods described here provide a reference for testing the permeabilit y of lunar soil in situ Future exploration of the lunar soil could measure the permeability in order to test the porosity/bulk density of the lunar soil at a location.
40 Table 2 1. Nomenclature and constants used within Chapter 2 Term Unit Definition D m 2 s 1 Knudsen diffusion coefficient d c m Capillary diameter D o m 2 s 1 Diffusion coefficient of gas in air m Mean square pore diameter f 1 Fraction of molecules that undergo diffusion reflection at capillary walls K m 2 Viscous flow permeability k o Shape factor L m Length or porous medium/capillary length L c m Length of capillary M kg kmol 1 Molecular weight p Kg m 1 s 2 (Pa) Pressure p 1 Kg m 1 s 2 (Pa) Pressure at the inlet of the porous medium/capillary Kg m 1 s 2 (Pa) Mean bed pressure p Kg m 1 s 2 Pressure drop across the length of bed/capillary R kg m 2 s 1 kmol 1 K 1 Universal gas constant = 8.315 x 10 3 J kmol 1 k 1 S o m 1 Particle specific surface T K Temperature of the gas u m s 1 Superficial fluid velocity u 1 m s 1 Superficial fluid velocity at the inlet of the porous medium/capillary Greek symbols $ Kg m 1 s 1 (Pa s) Gas viscosity kg m 3 Gas density p kg m 3 Particle d ensity bulk kg m 3 Bulk density # Porosity = 1 p / bulk Tortuosity = (ave) Length of capillary/length of bed m s 1 Thermal molecular velocity
41 Table 2 2 JSC 1A particle properties Particle Density (kg m 3 ) Specific Surface (m 1 ) Arithmetic number mean particle diameter (m) Median number particle diameter (m) Arithmetic volume mean particle diameter (m) Median volume number particle diameter (m) 2928 1026x10 2 0.17 0.12 236 137.7 Table 2 3 Permeability ranges for various porous media Material Bulk Density (kg m 3 ) Porosity Permeability (x10 12 m 2 ) Reference Clean Gravel 1000 100000 Bear (1988) Snow 300 400 500 8000 Albert et al. (2000) Limestone 0.422 0.531 178 704 Rice et al. (1970) Sandy Clay 10 100 Moldrup et al. (2001) Peat 10 0.1 Bear (1988) Sand 0.381 0.477 7.6 78 Rice et al. (1970) Great Sand Dunes 1730 1590 0.331 0.586 1.33 2.4 Sizemore et al. (2008) Lunar Soil 0.3 0.5 1 7 Choate et al. (1968) JSC Mars 1 945 795 0.606 0.669 0.989 4.71 Sizemore et al. (2008) JSC 1A 1990 1550 0.322 0.470 0.981 6.21 This work Linnaeus Terrace 1860 1580 0.312 0.415 0.67 2.37 Sizemore et al. (2008) Loess 1060 693 0.586 0.729 0.0727 0.691 Sizemore et al. (2008) Beacon Valley 1950 1500 0.276 0.436 0.0199 2.5 Sizemore et al. (2008) Unweathered Clay 0.0001 0.00000001 Bear (1988)
42 Figure 2 1. Apparatus of permeameter system.
43 Figure 2 2. Determination of viscous flow permeability with bulk = 1550 kg m 3
44 Figure 2 3. Permeability of JSC 1A at various bulk densities, bulk
45 Figure 2 4 Determination of the constant in the Carman Kozeny equation
46 CHAPTER 3 MODELING CRATER GROWTH USING THE TWO FLUID MODEL Motivation The work of Metzger et al. (2009a) provides measurements of transient cratering behavior such as crater depth and width th rough the cratering experiment They give a rigorous parametric study of the effect of the gas jet properties on the cratering of the particle bed, which also offers a benchmark opportunity of a two fluid model for simulating the jet interaction with the particle laden surface. Metzger et al. ( 2009a) investigated the cratering mechanisms dominating in beds cratered by both high velocity jets and low velocity jets. They also identified DDS BCF, and VE to be possible mechanisms that dominate in t he experiments they performed. The two fluid (Eulerian Eulerian) framework is employed to simulate the interaction of a turbulent subsonic jet with a particle bed and is derived by averaging the equations of mass and motion for each phase and using constitutive relations to close the averaged equations. The two fluid approach is one of two methods for modeling granular materials. The second is Direct Element Modeling (DEM), or th e Lagrangian approach, in which Newton's equations of motion are solved for each particle throughout the domain. T he properties of individual particles can be changed easily with DEM, but the number of particles that current processors can simulate limits the method Most published works in DEM typically involve under a million particles. Two fluid modeling has the disadvantage of needing constitutive relations to close the balance equations, but has the advantage of modeling realistically sized domains. T he granular phase equation of motion has collisional, kinetic and frictional stresses. The collisional and kinetic stresses dominate in dilute solids flows and are
47 caused by inelastic collisions and particles crossing imaginary shear planes, respectively. The frictional stress contribution is orders of magnitude higher than the collisional and kinetic stress contributions in higher solids volume fraction flows (dense phase flows) and the frictional stresses are caused by enduring contacts where momentum i s dissipated by friction. The frictional model is empirically based and contains material specific constants, which should be adjusted for modeling a particular material. Studies have tested the various frictional models in the past (Benyahia, 2008; Passa lacqua and Marmo, 2009; Srivastava and Sundaresan, 2003; van Wachem et al., 2001) and results consistently show the importance of implementing the correct frictional stress in order for the model to accurately predict the physical event. The results from p revious studies also caution against the ad hoc use of the frictio nal models due to their inherent empiricism (Benyahia, 2008; Passalacqua and Marmo, 2009; Srivastava and Sundaresan, 2003; van Wachem et al., 2001) The frictional stress is modeled as a pre ssure and a viscosity, which are derived using heuristic arguments. Others have shown that two fluid modeling is a viable approach to modeling complex systems, as long as the appropriate models are implemented correctly. The two fluid model has been used e xtensively in the past to model spouting and jet fluidized beds, which are jet impinging from below (Boemer et al., 1997; Dan et al., 2010; Duarte et al., 2009; Gryczka et al., 2009; Hong et al., 1996; Hosseini et al., 2010; Peng et al., 2010; Sa ntos et al ., 2009; Utikar and Ranade 2007; van Wachem et al., 2001). Single phase models have also been used in the past to study the prediction of
48 flow patterns in horizontal scour holes by modeling a jet impinging into a crater with solid walls in the shape of a scour hole (Gunal and Guven, 2006; Karim and Ali, 2000 ). Metzger et al. ( 2009a) performed a single phase simulation of a jet impinging inside of a well with solid walls in order to investigate the properties of the jet inside of the crater. The previously performed scour and crater simulations were single phase and therefore did not account for the interaction and interpenetration of the fluid with the particle phase. This work is the first to model a jet impinging from above that utilizes the two fluid fra mework. The importance of correctly using the frictional models is shown in this work It is also the first to present the effect of the empirical values used in the frictional model on the overall model's prediction. The results of Metzger et al. ( 2009a) parametric study were used to validate two fluid modeling as a tool to predict crater formation. This work provides a rigorous analysis of the frictional model form used for modeling dense particle phase two fluid modeling. Finally, it presents a frictional model that predicts the behavior of sand well, which has not been shown in the past. Model Equations Governing Equations The two fluid approach models both the fluid and solid phases as interacting and interpenetrating continua. The ensemble av eraged equations of motion of Anderson and Jackson (1967) and Jackson (1997; 1998) are provided in Table 3 2 and were used to describe the fluid (gas) and particle phases. Constitutive equations are needed to close the averaged solids phase momentum equati on (solids phase stress). The collisional and kinetic particle phase stresses depend on the magnitude of the particle velocity fluctuations. A granular temperature can be introduced, which is similar to the
49 thermodynamic temperature of a gas, and is a meas ure of the particle velocity fluctuations. ( 3 1) A balance of granular energy, is used to close the solids phase governing equations. Granular energy is generated by shear in the particle phase, diffuses along gradients of granular energy, dissipates due to inelastic particle particle collisions and is generated or dissipated from interactions of the fluid phase with the particle phase. The collisional and kinetic contributions to the solids phase stresses depend on the fluctuating velocity. Lun et al. (1984) derived expressions for these relations based on dense gas kinetic theory (Chapman and Cowling, 1970) allowing for inelasti c particle particle collisions. Only the collisional and kinetic contributions of the solids phase stress are take n into account for the granular energy balance as frictional stress is considered to dissipate to true thermal energy ( Johnson and Jackson, 1987) An algebraic assumption is employed for the granular energy balance which assumes that the granular en ergy i s conserved locally and neglecting convection and diffusion T he lift force was neglected, as it is small compared to the drag and gravitational force s (Kok and Renno, 2009; White and Schulz, 1977; Zou et al., 2007). A list of nomenclature used in this cha pter is provided in Table 3 1. The governing equations and constitutive relations are found in Tables 3 1 and 3 2, respectively The granular temperature equation describes the random motion of the particle phase due to particle particle interactions as we ll as with the fluctuating gas phase. The Reynolds Stress Model (RSM) with standard wall functions was used to model the gas = 1 3 v s 2 3 2
50 phase turbulence. The RSM used here is of the form described in Cokljat et al. (2006) and assumes that the solids phase is dilute. It is assumed, for the case of a jet impinging on a dense particle bed, for regions that the turbulence is important the solids phase is dilute. For the RSM employed, transport equations (provided in Table 3 4 ) for the Reynolds stresses and turbulence diss ipation rate, % g are only solved for the gas phase. The term & Rg in the Reynolds stress transport equation accounts for the interaction of the gas phase and solids phase turbulence. The interaction term & Rg is the contribution of the solids drag, the particle fluid covariance and volume fraction flux is simplified by neglecting anisotropy to the form presented in Table 3 4 where & kg is a modified version of the model originally described by Simonin and Viol let (1990) (Cokljat et al., 2006). The turbulent energy dissipation rate, % g utilizes the Elghobasi and Abou Arab (1983) form of the dissipation term due to the interaction of the solids phase with the gas phase, & g and is presented in Table 3 4 (Coklja t et al., 2006). Frictional Stress Model At higher solids volume fraction s, sustained contacts between particles occur. The dissipation of momentum due to these enduring contacts is accounted for by the frictional stress, which is considered in Newtonian form and can be expressed in terms of a pressure (normal stress) and viscosity (shear stress). The frictional stress is added to the kinetic and collisional stress terms (Savage, 1998). The frictional stress models are based on critical state theory (John son and Jackson, 1987; Roscoe et al., 1958; Roscoe, 1970; Schofield and Wroth, 1968; Srivastava and Sundaresan, 2003) which is based in soil mechanics and the resulting stress equations are empirical. Johnson and Jackson (1987) proposed a form of the fric tional stress components (normal and shear) based on the critical state assumption,
51 in which the medium does not dilate or contract while deforming. The frictional pressure controls the compressibility of the bed while the viscosity controls how liquid lik e the bed behaves. Johnson et al. (1990) further developed the frictional pressure to the for m that is widely used to date The frictional stress of the Johnson et al. model has been used extensively in the literature to describe beds made up of spherical particles (Johnson et al., 1990; Srivastava and Sundaresan, 2003). Srivastava and Sundaresan (2003) found that the critical state assumption holds well for hopper discharge with a frictional model based on the Johnson and Jackson (1987) model. For increasi ng solids volume fraction, the frictional stress is zero until a critical solids volume fraction, # s,min is achieved T he frictional stress increases with the solids volume fraction, # s until the maximum random solids packing volume fraction, # s,max is reached. Generally, # s,min is considered to be around 0. 5. When # s,min is too high, the frictional stress will not include dissipation due to enduring contacts and the bed will act too much like a liquid. However, if # s,min is taken at a value too low, the particles cannot be assumed to endure these long contacts. A version of the frictional model by Syamlal et al. (1993) uses a value of 0.63 to turn on the frictional stress ( # s,min =0.63), which allows for particle packing near the packing limit and above w hich the frictional stress is so high that further packing is extremely difficult. The frictional stress in their model is much lower and has been shown to not work as well as models with # s,min =0.5 (Benyahia, 2008; Passalacqua and Marmo, 2009). The form of Johnson and Jackson (1987) and Johnson et al. (1990) frictional pressure model is given in equation (3 2).
52 ( 3 2) where Fr, n and p are material specific empirical constants. Johnson and Jackson (1987) and Johnson et al. (1990) relate the frictional shear stress to the frictional normal stress through Coulomb's law. The first model analyzed in this paper, Equation ( 3 3) below and in Table 3 3 is the frictional stress relationship, based on Coulomb's law, proposed by Schaeffer (1987) and is used to model the frictional viscosity. s f r = P s f r s i n( F ) 2 I 2 D ( 3 3) I 2D is the second invariant of the deviatoric stress tensor and is the internal angle of friction. A minimum value (1x10 12 ) for I 2D was applied to avoid singularity in calculating the frictional viscosity, which did not affect the results, as the frictional viscosity never reached a high enough value to be a result of th e limit. The variables Fr, n and p can be adjusted separately for P s,fr and s,fr Srivastava and Sundaresan (2003) propose a model for the frictional viscosity that contains an ad hoc account for fluctuations in the strain rate associated with the formati on of shear layers (Savage, 1998). The f rictional viscosity model proposed by Srivastava and Sundaresan (2003) is given as, s f r = P s f r s i n( F ) 2 I 2 D + d s 2 ( 3 4) The Johnson et al. (1990) frictional stress model with the frictional viscosity of Schaeffer (1987), Equation ( 3 3) has been shown to predict dense beds of glass beads P s f r = 0 s s m i n F r s # s m i n ( ) n s m a x # s ( ) p s > s m i n $ % & & & &
53 with too high frictional stress (Benyahia, 2008; Srivastava and Sundaresan, 2003). By accounting for the granular temperature as in equation ( 3 4), the frictional viscosity is effectively lowered. Joh nson et al. (1990) originally used the Coulomb frictional viscosity model, which does not include the I 2D term, to validate their chute experiments with glass beads. Perhaps the Johnson et al. (1990) frictional stress model predict s a bed with too high fri ctional stress for glass spheres when Equation ( 3 3) is used since the I 2D term increases the frictional viscosity. Methods Experimental Setu p of Metzger et al. ( 2009a) The experimental set up is explained in detail in Metzger et al. ( 2009a) and Figure 3 1 provides a schematic of the experimental apparatus. The gas traveled down a long straight pipe directed into a particle bed. A circular jet was produced by flow expanding out of the exit of a pipe. The circular jet straddled an outwardly beveled clear wall which contained the bed of particles. The bed of particles was composed of Jetty Park Beach sand that had been sieved to a particle size of 100 to 180 microns which is mono disperse relative to JSC 1A and Chenobi beds described in Chapters 1 and 5 The beveled window was placed so that the jet was split and the crater evolution was monitored by, and recorded with, video acquisition. Simulation Setup The differential equations were discretized with the finite volume method and second order implicit time stepping was used. The SIMPLE (Patankar, 1980) pressure correction algorithm calculate d the flow field. The algebraic form of the granular energy balance (neglecting convection and diffusion) was found to be accep table by comparing the results to those pre dicted by the full partial differential equation balance.
54 Additionally, van Wachem et al. (2001) found the algebraic model to be sui table for dense particle beds. The frictional stress models did not contain a maximum limiting value. The simulation was pr oven to be independent to various aspects of the mesh (i.e. wall placement, mesh size, domain, etc.). The results of the sensitivity analysis showed that mesh size of the order 1 mm by 1 mm near the region of interaction between the jet and the bed was necessary to resolve all of the gradients for the simulation. When the mesh element size is too large in the impingement region, the crater starts to form much later and does not happen smoo thly; the particle phase jumps between positions. In order to reduce the overall number of cells a hanging node adaption was used, which split s cell size in half only in desired areas by placing a node half way in between two nodes on the edge of a mesh el ement. Meshes were created using the program Gambit v2.4.6. Simulations were carried out with CFD solver Fluent and all frictional models were added using the User Defined Function (UDF) capability, where the models were written in C and compiled into the solver. Figure 3 2 presents the (a) computational mesh and (b) simulation domain for the axisymmetric simulations. For all domains, the meshes were separated into two zones, where the region that held the particles is referred to as the bed, as illustrate d in Figure 3 2. The solids volume fraction was initialized to 0.58 in the bed and 1x10 5 everywhere else in order to avoid a stiff matrix and to aid in convergence. Cratering predictions were not sensitive to the initial value of 1x10 5 which was checked by lowering the value by two orders of magnitude. The jet inlet was designated as a velocity inlet and the gas velocity was set to the experimentally published value ( Metzger et al., 2009a) in the axial direction, the
55 turbulence conditions were set using turbulence intensity and length scale. The particle phase inlet conditions were set to have solids velocity magnitude of 0 and solids volume fraction of 1x10 5 Table 3 5 provides all of the values used for the simulations. For convergence, the time step w as set very low (1x10 7 s) for at the beginning of the simulation and slowly increased to 1x10 6 s by 0.1 seconds. All times reported, associated with the simulations are in seconds of real time and from here on will be referred to with the unit of second s, to be concise. The residuals were set to 1x10 5 in order to reduce error in converging with a strict tolerance on all variables changing between consecutive iterations. Convergence was checked for every time step and the number of iterations per time st ep varied depending on the amount of interaction occurring on each time step. Trajectory of the solution changed if the convergence criteria were not met even for only a few time steps Model Sensitivity Sensitivity to mesh size as well as time step were tested to ensure that accurate resolution of gradients were captured. The effect of the wall boundary on the crater predictions was investigated by implementing a mesh geometry with walls twice as far as the original mesh geometry. The effect of the turbu lence model on the model prediction was explored by comparing simulation predictions employing the k % model to the RSM. The RSM was much less sensitive to time step size and also predicted better resolution of gas phase vortical structures. The accuracy o f the RSM and k % model were verified by simulating a jet hitting an orthogonally placed wall and compared to benchmark data of Cooper et al. (1993) and Craft et al. (1993). T he k % model tends to overestimate the production of turbulent kinetic energy ( k g ) in regions where streamline curvature is high (high strain
56 rate). Both models have problems modeling stagnation regions, where strain rates are often high, but in general the RSM is more accurate than the k % model. A thorough discussion of the turbulenc e model stagnation regi on limitations are described in Craft et al. (1993). The turbulence models predicted results similar to th os e of Craft et al. (1993). The RSM was used to simulate the low velocity jets (all conditions presented in Figure 3 4 to Figur e 3 6 ), but the k % model was used to simulate the high velocity jets (near sonic flow) for cratering mechanism validation. Experimental V alidation Results of the frictional model study were evaluated alongside of the experimental measurements of Metzger et al. (2009a) and Metzger (personal communication, 2009) One of Metzger et al. (2009a) experiments was used to test all frictional stress models which will be referred to as the base case. Only the lowest and highest conditions of the parametric study performed by Metzger et al. ( 2009a) were simulated (e.g. the lowest and highest jet velocity) All experimental conditions simulated, unless otherwise noted, are for a 1.02 cm diameter nozzle with height of 7.62 cm above the bed. The base case utilizes a 3 7 m/s Argon jet Simulation Predictions Once the empirical frictional stress model was set for the base case by comparing the crater depth predictions to the experimental data and selecting the best model, it was used to predict crater formation for other experimental conditions. The results of Metzger et al. ( 2009a) and Metzger (personal communication, 2009) provided a large amount of parametric data to compare with the predictions of the models presented here as well as models currently used in the liter ature (Johnson et al., 1990; Ocone et al., 1993; Srivastava and Sundaresan, 2003). The frictional m odel was not
57 changed when simulating cases other than the base case; only the simulated jet operating conditions were changed (e.g. gas velocity, gas species etc.). All simulations were carried out to 1 second. The surface of the bed was tracked and the crater depth is represented with a solids volume fraction of 0.02. The predicted crater depth was not sensitive to the solids volume fraction value used to tr ack the surface of the bed. Results and Discussion The frictio nal stress formulation dominates the model s ability to predict the cratering of the particle bed. The effect of the overall magni tude of the frictional stresses as well as the relative magnitu de of the frictional pressure and frictional viscosity are investigated and the results are presented with respect to the cratering phenomena. A model is presented, Model A, which accurately predicts the salient features of crater growth measured by Metzge r et al. ( 2009a) Additionally, the presented model is employed to explore the cratering mechanism for both the low and high jet velocity regimes. Sensitivity Results The investigation of the crater predictions to the size of the domain (the distance of t he side walls ) show that the normal mesh is sufficient and the crater predictions are not affected by doubling the distance of the walls. The increased wall distances were tested for two separate frictional models to make sure that the crater predictions a re not affected for a higher frictional stress model. The pipe length is slightly below the length necessary for the flow to become fully developed. As long as the mesh was resolved enough at the interface of the bed and the gas, the longer pipe did not af fect the crater prediction.
58 The results of a simulation where the full partial differential equation for the granular temperature balance, found in Table 3 2 showed that the algebraic assumption for determining the granular temperature is sufficiently ac curate. Frictional Model Study Adjustments to the Johnson and Jackson (1987) and Schaeffer (1987) frictional stress model s were made in order to investigate and optimize the frictional model's predictions of the crater depth and width. The frictional pressure of Johnson et al. (1990), Johnson et al. model in Table 3 6 is generally taken as the standard model in the literature (Benyahia, 2008; Passalacqua and Marmo, 2009; Srivastava and Sundaresan, 2003). Other published frictional models based on the Johnson and Jackson (1987) model were also tested (i.e. Ocone et al. (1993) and Srivastava and Sundaresan (2003)). Table 3 6 shows the main frictional models Model A, in Table 3 6 has an increased frictional viscosity relative to the frictional pressure compared to the Johnson et al. model The frictional viscosity and the frictional pressure in Model B are increased compared to the Johnson et al. model and Model A Model C allow s for investigation of the effect of increasing the frictional pressure rela tive to the frictional viscosity with respect to Model A The Ocone et al. model (Ocone et al., 1993) reduces the overall frictional stress by two orders of magnitude with respect to the Johnson et al. model. The last model in Table 3 6 the Srivastava and Sundaresan (2003) model, accounts for strain rate fluctuation s in the frictional viscosity. Figure 3 3 shows the prediction of the crater depth over time to 1 second for the base case by the frictional models in Table 3 6 Th e Johnson et al. model predic ts the crater depth to grow too quickly when compared to the measured depth Since sand is
59 more angular and rough than spherical beads it is not surprising t hat the frictional model requires a higher frictional viscosity than one for smooth spherical beads Model A is the most effective in predicting the crater growth for sand and was produced by increasing the frictional viscosity by a factor of 2 relative to the Johnson et al. model. Since Model A better predicts the crater growth for sand the model adjus tments shown in Table 3 6 for Model B and Model C are based on Model A as opposed to the Johnson et al. model. There is little change between the crater predictions of Model A and Model B. Simulations were also performed with other frictional models that simultaneously increased the frictional stress for P s,fr and s,fr further from the magnitude of Model A (i.e. Fr= 0.1 and p=5.9 ; Fr = 1.0 and p=5), and predict a negligible difference in crater depth and width over time as compared to Model B. Additionally, a simulation was performed with a model that increased the frictional viscosity further relative to the frictional pressure with regards to Model A, but the crater prediction is again negligibly different than that of Mod el B. The similar bed prediction of increased frictional models indicates that once the frictional stress is of a certain order, or the frictional viscosity is relatively large compared to the frictional pressure, the cratering predictions become insensiti ve to increases. Decreasing the frictional stress by two orders of magnitude, for the Ocone et al. model, causes large changes in the cratering predictions compared to the Johnson et al. model. The O cone et al. model over pred icts the crater depth, as thi s model significantly underestima tes the frictional stresses. When the frictional stresses are too low, as seen for instance with the Ocone et al. model, the bed behaves too liquid like
60 and the jet drives deep into the bed followed by the bed sloshing back The splashing of the bed predicted by the Ocone et al. model can be identified in Figure 3 3 where the crater depth quickly decreases after reaching a maximum at around 0.5 seconds, which corresponded to a jump in the crater width after the crater slosh es. The Ocone et al. model's predict ed crater width behaves similar to the depth; the bed splashes to a width too wide, over predicting the crater width, followed by an inward sloshing of the crater sidewalls. Frictional m odel with increased P s,fr only (M odel C) Srivastava and Sundaresan model and the Johnson et al. model demonstrate increased crater depth growth over time compared to Model A The three former models have reduced frictional viscosity with respect to the frictional pressure relative to the latter The way in which the Srivastava and Sundaresan model accounts for fluctuations in the strain rate results in prediction s of a bed that craters faster than the Johnson et al model. The Srivastava and Sundaresan model has the same form of the fricti onal pressure as the Johnson et al. model but the frictional viscosity, Equation ( 3 4) is decreased relative to that of the Johnson et al. model Equation ( 3 3) Table 3 7 provides the orders of magnitude of the frictional pressure and viscosity present in the bed at 0.5 seconds for the models presented in Table 3 6 Additionally, Table 3 7 gives the maximum predicted solids volume fraction found in the bed by the various frictional models, which allows for examination the bed compressibility and should not be confused with the maximum random packing solids volume fraction, # s,max from the frictional pressure model
61 Model C predicts the second fastest crater growth, even though the frictional viscosity form is the same as Model A. When the frictional pressure is increased, at the expense of the frictional viscosity (Model C, Johnson et al. model and Srivastava and Sundaresan model), the model predicts a much faster cratering rate one that exceeds the experime ntal observations. For the case of increased P s,fr relative to s,fr the bed flows much like a liquid because the frictional viscosity is too low given the high level of pressure in the bed. In addition, for very short times, the bed expands due to the ve ry large frictional pressure, which is expressed by the dip in the crater depth in Figure 3 3. U nder the condition of increased P s,fr relative to s,fr the bed behaves similar to the model predictions employing the Ocone et al. model, which predicts liqui d like behavior because the frictional stresses are too low. As indicated in Figure 3 3, the cratering prediction can be greatly affected by adjusting the frictional model magnitude and the relative value of the frictional viscosity to the frictional press ure. The predictions of Model A's configuration are in satisfactory agreement with the results presented by Metzger et al. ( 2009a) Parametric Study Measured and Predicted Crater Growth Model A was used to predict crater formation to 1 second with condi tions reported by Metzger et al. ( 2009a) for various parametric simulations. Figure 3 4 shows the predictions of Model A for the case of Ar gon jets with velocities 37 m/s and 56 m/s. The predictions of Model A and experimental results for crater depth as a function of gas species are plotted in Figure 3 5. In Figure 3 5 the simulation and experimental data are for 56 m/s jets with gas species Argon, Nitrogen and Helium. The effect of pipe height on M odel A's predictions of crater depth is presented in Figur e 3 6. The conditions of the experiment and simulation in Figure 3 6 was a 40 m/s Nitrogen jet with
62 nozzle exit heights of 7.62 cm and 10.16 cm. All experimental data presented in Figure s 3 4 through 3 6 were taken from Metzger et al. ( 2009a) and personal communication with Metzger (2009) The transient crater growth scales with jet Reynolds number. Analysis of Metzger et al.'s (2009a) results indicate that the scaling with changing gas is proportional to g p g ( ) p u s t 2 where u s,t is the terminal velocity of the particle (accounting for sphericity) indicating that the crater growth is proportional to the ratio of the drag to the inertia. The simulation predictions illustrated in Figure s 3 4 through 3 6 were computed in an axisymmetric domain. In Figure s 3 4 and 3 5, Model A predicts the crater depth to 1 second well for a variety of gas species and jet velocities. The correlation coefficient between the measured and predicted crater depth is between 0.947 and 0.997 for all cases in Figures 3 5 to 3 6 with the exce pt ion of the 56 m/s jet of Nitrogen, which had a correlation coefficient of 0.772. The predicted crater depth by Model A, Figure 3 6, is more sensitive to the pipe height than what was found experimentally. However, the experimental data does not have erro r bars, so it is difficult to tell if the predictions are within the experimental error. In conclusion, the simulations performed with Model A are able to predict the salient features of Metzger et al.'s (2009a) parametric study. Long Time Simulation Metz ger et al. ( 2009a) describes the balance of competing forces during the sand cratering experiments, which led to complex crater formation. Initially one parabolic shaped crater forms, which later collapses into a dual crater shape consisting of an inner an d outer crater. The inner crater maintains the parabolic shape while the outer
63 crater becomes conical with walls that oscillate between the angle of failure and the angle of repose of the sand. The time it takes for the crater to split into two regions de pen ds on the properties of the jet. However, for the relevant experiential cases, the crater split occurs well after 1 second. Figure 3 7 compares the long time crater depth growth of the measured depth of Metzger et al.'s (2009a) base case to the predicte d depth of Model A. In Figure 3 7, Model A's predicted crater depth diverges from the experimental results at around 4.2 seconds Experimentally the split of the crater into a dual crater occurs at about 3.7 seconds, which is close to when the simulation predicts the crater depth to begin to diverge from the experimentally measured depth. There are a few possible reasons for the under predicted depth at later times. T he simulation does not properly predict the formation of the outer crater, Figure 3 8, wh ich could affect the recirculation rate of the solids phase. In Figure 3 8 the column of images on the left are experimental snapshots of the crater over time under the base case conditions, in the middle and on the right are the predicted solids volume fr action contours for corresponding times by Model A and the Ocone et al. model, respectively. A nother possible explanation is due to t he axisymmetric boundary predicts particles moving too fast along the boundary, as the jet velocity reduces due to the incr eased distance from the pipe, the effect of the discretized axisymmetric boundary formulation along the boundary becomes more pronounced. The Ocone et al. model predicts a crater that is too deep at 0.1 04 seconds and demonstrates a collapsing crater by 1 s econd, after which the simulation was terminated. At 1 second the Ocone et al. model is showing compaction near the surface of the crater, due to the frictional pressure being too low and allowing the solids phase
64 to compress under the opposing forces of t he jet and the spl ashing solids. Model A predicts the crater growth well for both 0.104 and 1 seconds. Close inspection of the top of the crater prediction at 3.77 seconds by Model A exhibits a drop in the position of the more dilute solids volume fraction s to about halfway down the side of the crater, compared to above the top of the crater found at 1 second. At 3.77 seconds, the image from the experiment shows the outer crater just starting to form. Well defined inner and outer craters are found experimen tally at 38.02 seconds. However, at 38.02 seconds, Model A predicts a minor formation of an outer crater, where there is a slight angle to the top of the crater. Model A does not correctly predict the outer crater size or sidewall angles. It is explainable that the two fluid model with Equation ( 3 3) would not correctly predict the outer crater since the model is only valid for continuously deforming regions (Schaeffer, 1987). Domain Sensitivity Effect of Center W all The effect of the viewing window in t he center of the flow on the jet in experiments, Figure 3 1, must be understood in order to properly validate the model. In order to test the effect of a wall in the center of the domain, 2D planar meshes were used as opposed to 3D meshes in order to reduc e computational time. Reuge et al. (2008) found that the 2D planar results were similar to the 3D results when modeling fluidized beds. Figure 3 9 (a) and (c) provide the simulation domain and computational domains, respectively, for the 2D planar mesh wit hout a wall. T he simulation domain of 2D planar mesh with a wall is depicted in Figure 3 9 (b) and the computational mesh in (d). T he wall was placed directly along the centerline of the jet, splitting the mesh in half with the bed only to one side of the wall. The top of the wall in Figure 3 9 (b) and (d) is 2 cm above the top of the particle bed.
65 In Figure 3 10 the experimental crater depth measurements for the base case ( Metzger et al., 2009a) are compared to the predictions by the base case with Model A's in the axisymmetric domain planar mesh without a wall, and a planar mesh with a wall The 2D planar and axisymmetric meshes similarly predict the crater growth to 1 second, but the axisymmetric domain predicts a deeper crater than experimental measur ed at 1 second. In the absence of the wall, t he 2D planar mesh initially under predic ts the crater depth and by about 0.5 seconds starts to accurately predict the crater depth to 1 second. Reuge et al. (2008) describes p ossible reasons for the 2D planar mesh's ability to better predict the crater depth than the axisymmetric mesh. Comparing the crater prediction of the planar mesh with and without a wall in the center of the domain provides interesting physically relevant results about the effect of the wall on the crater depth prediction. In the presence of the wall, the planar mesh over predicts the crater depth until about 0.5 seconds, when it started to under predict the crater depth to 1 second. By considering the effe ct of the wall on the model's prediction, the results are congruent with the experimental results. In other words, the simulation with a wall in the center of the domain adjusts the predicted crater growth in a way that is closer to the experimental result s. The simulation predictions indicate that the presence of the wall is minor, but that it slightly incre ases the initial cratering rate and reduces the erosion rate over time. Figure 3 11 provides solids volume fraction contour images for Model A's predi ction of the crater formation in the presence of a wall for various simulated times. The simulation performed on the planar mesh with a wall initially predicts a drop in the crater depth around 0.1 seconds, as expressed in Figure 3 10. Contour plots of th e
66 solids volume fraction in Figure 3 11 demonstrate that the initial drop in the crater depth along the centerline is due to the erosion of parti cles when the jet first impinges on the bed. In the presence of the wall, the jet is initially de flected by the wall and impinges slightly away from the wall, causing erosion of the bed between the wall and the impingement region. The p ocket of eroding solids splashes against the wall, resulting in a temporary increase in the bed depth. As the jet develops over tim e it expands towards the wall eroding the particles awa y. The jet then continues to penetrate the bed near the wall. In the case of the jet impinging in the presence of a wall, t he prediction of decreased crater growth rate is expected as the presence of the wall reduces the momentum of the jet by dragging along the wall. Simulations were also performed to investigate the effect of the wall thickness and the angle of the bevel. Increasing the angle of the bevel and th e thickness of the wall results in redu ced erosion. An analysis of the boundary layer thickness of laminar flow over a semi infinite flat plate (Bird et al., 2002) indicate s that the boundary layer s for the jets studied in this work are on the order of 0.7 mm and below (reducing with increased jet velocity), 3 cm from the tip of the plate. However, this is a rough approximation of the boundary layer thickness as it assumed to be laminar and a thin flat plate. Additionally, w all boundary conditions and effects of stagnation flow on the turbulence model effected the prediction of the crater depth along the wall (Craft et al., 1993). Simulations of Turning the Jet Off In cratering experiments when the jet is extinguished after running for various le ngths of time, a crater remains with walls at the angle of repose. Experiments performed with analogues particles in a similar experimental apparatus show that a crater remains when the jet is terminated even at approximately 1 second Proper
67 prediction of the model keeping the crater from filling in has implications that can be applied to two fluid modeling for many applications. For example, for two fluid modeling of hopper discharge, the discharged granular media will form a heap wi th sides of the heap self supported at the angle of repose. The model's ability to predict the response of the crater to turning the jet off was investigated by turning the gas velocity to a very small number and modeling the flow as laminar at 1 second. The jet off simulations were performed in the axisymmetric domain implementing Model A. Tracking the surface of the crater with a high enough solids volume fraction, i.e. s =0.58, show that Model A is able to hold the crater at a nea rly constant size after the extinguishing the jet The angle of the top portion of the remaining crater is slightly less than 20 degrees, which is lower than the approximately 35 degree angle of repose of the sand shown in Metzger et al. ( 2009a) The lack of Model A to predict t he correct angle of the crater indicates that more development is needed for the frictional viscosity. Cratering Mechanics Metzger et al. ( 2009a) reported results of both high and low velocity cratering experiments. The low velocity experiments, like the base case and the parametric study, provided quantitative results that were used to validate Model A. Metzger et al. ( 2009a) observed that the cratering mechanisms for the low velocity regime are dominated by VE along the edges of the inner crater. They di d not report the velocity for the high velocity experiments, but in a private communication with Metzger (2009) it was described as very close to the speed of sound. The results from the high velocity experiments provide qualitative information about the d ominating cratering mechanisms
68 for a compressible jet. The cratering mechanisms were explored with Model A's predictions for both the base case, with a 37 m/s Argon jet, and a compressible near sonic Nitrogen jet impinging on beds of sand. The high velocity study was performed in the axisymmetric domain and computation mesh that are both provided in Figure 3 2. The Mach number of the inlet of the pipe was set to 0.75 and the resulting Mach number at the exit of the pipe was 0.9. The simulati on conditions were set to avoid shoc ks. Regardless, the density based solver was used in order to account for the compressibility of the Nitrogen jet. The k % turbulence model was used to describe the gas phase only for the high velocity simulations; the s imulations reported previously all utilized the RSM turbulence model. The k % turbulence model did effect the quantitative predictions of the two fluid model for the low velocity base case. Nonetheless, the k % model was expected to be adequate for investi gating the cratering mechanism of the high velocity regime as it was assumed that the reduced accuracy of the k % model would only affect the accuracy of the predicted erosion rate T here is a correlation between the turbulent shear stress and the erosion rate (Haehnel, 2011). However, an accurate prediction of the erosion rate is not necessary for identifying the cratering mechanism especially for the case of DDF or BCF, as neither is governed by surface erosion. Predictions of the mass flux and gas velocity in the domain, Figure s 3 12 and 3 13, provide evidence of the cratering mechanisms. Model A's simulation of the base case shows VE to be the dominating mechanism. There is no bulk shearing in the bed (i.e. the shearing occurred in solids volume fr actions less than 0.5). In Figure 3 12 (a) the solids mass flux (solids volume fraction solids velocity) vectors overlaid on top of
69 solids volume fraction iso surfaces for Model A's prediction of a 37 m/s Argon jet after 0.5 seconds. T he mass flux vector s for Model A's prediction of a near sonic jet after 0.067 seconds is presented in Figure 3 12 (b) The dark gray mass flux vectors represent lower mass flux magnitude and the light gray mass flux vectors represent higher magnitude. In Figure 3 12 (a) ther e are no mass flux vectors below the solids volume fraction iso surface of 0.5. For both Figure s 3 12 (a) and (b) the mass flux vectors in the center of the shearing region are generally the lightest in color, indicating the mass flux is the highest. Howe ver, in reality the particles at lower solids volume fractions are actually moving with the highest velocity. In Figure 3 12 (a) the movement of solids is in the thin shearing layer, which dominates for experiments in the low jet velocity range and is cons istent with Metzger et al. 's (2009a) description of VE The mass flux vectors in Figure 3 12 (b) express the movement of the solids phase in the bulk, much below solids volume fraction of 0.5. The movement of solids beyond the thin shear layer of solids vo lume fractions 0.02 to 0.5 indicates that mechanisms other than VE are dominating for the high velocity regime For the near sonic case, the DDS mechanism is identified due to the trajectory of the mass flux vectors in Figure 3 12 (b) The mass flux indica tes the solids phase is moving tangentially to the crater, even in the stagnation region, indicating the DDS mechanism and not the BCF mechanism occurs Figure 3 13 provides gas velocity magnitude iso surface contours plotted on top of solids volume frac tion iso surfaces for Model A's predictions of (a) 37 m/s Argon jet after 0.5 seconds and (b) a near sonic compressible Nitrogen jet. The gas velocity iso
70 surface contours in Figure 3 13 are scaled from 0 to 1 in order to show the gas velocity gradients in side of the bed. The DDS mechanism occurs when the gas diffuses into the bed faster than the solids can move and the gas drags the particles away from the crater ( Metzger et al., 2009a) The limited velocity gradients inside of the bed in Figure 3 13 (a) indicate that there is little gas penetration into the bed and the gas erodes the top layers of sand without diffus ing deep into the bed. Figure 3 13 (b) illustrates deep penetration of the gas into the particle bed in the form of high gas velocity magnitudes Comparing Figure 3 13 (a) to Figure 3 13 (b) e xpresses that the gas penetrates much further into the bed and the velocity decays much slower for the near sonic case than the base case. In Figure 3 13 (a) the gas velocity reduces to 0.05 m/s app roximately 11 mm below the solids volume fraction 0.5. For the near sonic case in Figure 3 13 (b) the gas velocity reduces to only 0.8 m/s by about 11 mm below the solids volume fraction of 0.5 and, not shown in the figure reduces to 0.05 m/s about 58.2 m m below solids volume fraction 0.5. The contour plots in Figure 3 12 (b) and 13 (b) identify DDS as the dominating cratering mechanism for the near sonic jet Deep shearing in the top of the bed for the near sonic jet could explain why an outer crater does not form under DDS conditions Both the near sonic and the base case simulations predict stagnation pressures at the center of the crater (stagnation region) that are too low for the BCF mechanism. The bearing capacity of sand is on the order of 100,000 P a. The gas static pressure produces a local maximum at the stagnation region, and is approximately 800 Pa and 9,000 Pa for the base case and near sonic simulation s, respectively However, for less permeable soils, like lunar soil, the gas may not permeate through the soil faster than
71 the particles erode so that the static pressure builds up on top of the soil, possibly leading to BCF (Metzger et al., 2009a) In summary, t he predictions of Model A are consistent with the results of Metzger et al.'s (2009a) h igh (near sonic jet case) and low velocity ( base case) regimes Summary The model proposed here (Model A) is based on the models of Johnson and Jackson (1987) and Schaeffer (1987) with frictional viscosity increased relative to frictional pressure of Johnson et al. (1990), is able to correctly predict the cratering behavior of a sand bed impacted by a turbulent j et. Additionally, this work demonstrates that the overall frictional stress magnitude and the relative magnitude of the frictional pressure to the frictional viscosity can greatly affect the model 's prediction, which is further proof that this model, and a ll other empirical models, should be used with caution. The magnitude of the frictional stresses of Mode l A, which additionally employs increased frictional viscosity relative to the frictional pressure, are well described for predicting the cratering beha vior of 100 180 micron beach sand. To 1 second, t his model slightly under predicts the initial cratering behavior reported by the experimental data, but later overshoots and predicts a crater that is too deep. For longer times, the model under predicts th e crater depth. The frictional model is empirically based and therefore a more theoretically based model would produce better quantitative agreement with the measured results as the current state of the model would require further adjustment. T he simulati on predicts a particle bed that behaves much like a liquid when the frictional stress model predict too low stresses, the frictional viscosity is too low relative to the frictional pressure or the frictional stress model is completely neglected. The
72 crate r prediction is also sensitive to the turbulence model implemented. The RSM produces better gas turbulence resolution, which implies choosing the correct turbulence model for a specific application is important. Experimentally, when the jet is turned off a crater remains with the walls at the angle of repose. The two fluid model is able to hold the crater back as long as the frictional stresses mainly frictional viscosity, are high enough. When the simulations are carried out in the axisymmetric domain, a s long as a high enough solids volume fraction is taken to track the surface of the bed, the proposed model predicts the crater to remain after turning off the jet. The effect of the wall on the cratering predictions was investigated and results show the wall initially slightly increases the crater depth growth, and then reduces the growth over time. The model predicts the deepest part of the crater to be slightly away from the wall, which is observed experimentally. Simulations of both the high and low je t velocity regimes are additionally presented Low velocity jet simulations accurately predict the crater growth when the proposed model is employed. Additionally, it is able to identify the cratering mechanisms for both the high and low velocity regimes. The proposed model predicts VE is the dominating cratering mechanism for the low velocity regime. While a simulation of a high velocity, near sonic, jet impi nging on a bed of sand predicts DDS to be the dominating mechanism. The two fluid modeling approach is able to accurately calculate the cratering process when the correct models are applied. A validation of the two fluid model as a method of describing complex flows situations is provided here The two fluid model
73 with proper frictional stress description is able to predict crater formation for a wide range of simulation conditions. Reducing the magnitude of the frictional stress or adjusting the relative magnitude of the frictional viscosity to frictional pressure can result in great changes to the model predictions. This work presents a new form of the frictional stress for sand that lowers the frictional viscosity relative to the frictional pressure.
74 Table 3 1. Nomenclature used in this chapter Term Unit Desc ription C D Drag Coefficient d s [L] Particle Diameter e Coefficient of Restitution Fr [MT 2 L 1 ] Empirical Material Parameter for Frictional Pressure [LT 2 ] Gravity g o Radial Distribution Function I 2D [T 2 ] Second Invariant of the Strain Rate Tensor k g [L 2 T 2 ] Turbulent Kinetic Energy L [L] Length M [M] Mass n, p Exponent in frictional stress model P [MT 2 L 1 ] Gas Pressure P s,col [MT 2 L 1 ] Solids Collisional Pressure P s,fr [MT 2 L 1 ] Solids Frictional Pressure P s,kin [MT 2 L 1 ] Solids Kinetic Pressure Re p Particle Reynolds Number R g [L 2 T 2 ] Gas Phase Reynolds Stress t [T] Time !" [LT 1 ] Drift Velocity Greek Symbols [MT 1 L 3 ] Interphase momentum transfer coefficient ( s [MT 3 L 1 ] Collision dissipation of Granular Temperature # g Gas volume fraction # s Solids volume fraction # s,min Solids volume fraction when frictional stress is activated # s,max Max packing solid volume fraction % g [L 2 T 3 ] Turbulent gas dissipation rate ) [L 2 T 2 ] Granular Temperature s [MT 1 L 1 ] Diffusivity of Granular Temperature + s [MT 1 L 1 ] Bulk Granular Viscosity s [MT 1 L 1 ] Solids Viscosity s,col [MT 1 L 1 ] Solids Collisional Viscosity s,kin [MT 1 L 1 ] Solids Kinetic Viscosity s,fr [MT 1 L 1 ] Solids Frictional Viscosity t,g [MT 1 L 1 ] Turbulent Gas Viscosity [LT 1 ] Mean Velocity Vector [LT 1 ] Fluctuating Solids Velocity s g [ML 3 ] Density [MT 2 L 1 ] Stress Tensor F [Deg] Angle of Internal Friction
75 Table 3 2 Governing Equations Governing Equations Continuity Fluid Phase Continuity Solids Phase Momentum Equation Gas Phase g g # v g !" # t + v g !" $ % v g !" & ( ( ) + + = g % $ g g % P v g !" v s ( ) + g g g g g % $ R g Momentum Equation Solid Phase Granular Fluctuating Energy Balance 3 2 t s # s $ ( ) + % & s # s v s $ ( ) ( ) + = s c ol + s k i n + P s c ol + P s k i n ( ) I ( ) + % & s % $ ( ) / 0 s / 3 1 $ g t + # $ g v s ( ) = 0 s t + # $ s v s ( ) = 0 s s # v s # t + v s $ % v s & ( ) + = s % P + % $ s % P s + v g v s ( ) + s s g
76 Table 3 3. Constitutive Equations Constitutive Relations Gas Phase Stress Tensor g = g v g !" + v g !" ( ) T # $ % & ( ) 2 3 v g !" ( ) I + / 0 Solids Phase Stress Tensor Solid Phase Pressure (Johnson and Jackson, 1987) Solids Collisional Pressure (Lun et al., 1984) Solids Kinetic Pressure (Lun et al., 1984) Solids Frictional Pressure (Johnson and Jackson, 1987) Solids Phase Viscosity (Johnson and Jackson, 1987) Solids Collisional Viscosity (Gidaspow, 1994) Solids Kinetic Viscosity (Gidaspow, 1994) s = s v s + v s ( ) T ( ) + # s $ 2 3 s % & ( ) + v s ( ) I P s = P s c ol + P s k i n + P s f r P s c ol = 2 g 0 s s 2 # 1 + e ( ) P s k i n = s s # P s f r = 0 s s m i n F r s # s m i n ( ) n s m a x # s ( ) p s > s m i n $ % & & & & s = s c ol + s k i n + s f r s c ol = 4 5 s 2 s d s g 0 1 + e ( ) # $ s k i n = 10 s d s # 96 1 + e ( ) g 0 1 + 4 5 g 0 $ s 1 + e ( ) % & ( ) 2
77 Table 3 3. Continued Constitutive Equations Solids Frictional Viscosity equation ( 3 3) (Schaeffer, 1987) s f r = P s f r s i n F ( ) 2 I 2 D Solids Frictional Viscosity equation ( 3 4) (Srivastava and Sundaresan, 2003) s f r = P s f r s i n F ( ) 2 I 2 D + d s 2 Second invariant of the deviator of the strain rate tensor I 2 D = 1 6 u s x v s y # $ % & ( 2 + v s y 2 # $ % & ( 2 + u s x # $ % & ( 2 ) + + + 1 4 u s y + v s x # $ % & ( 2 Solids Bulk Viscosity (Lun et al., 1984) Radial Distribution Function (Sinclair and Jackson, 1989) Collisional Dissipation of (Lun et al., 1984) Drag Coefficient (Wen and Yu, 1966; Rowe, 1961) s = 4 3 s 2 # s d s g 0 1 + e ( ) $ % g 0 = 1 s s m a x # $ % & ( 1 / 3 ) + + 1 s = 12 1 e 2 ( ) g 0 # s $ s 2 d s % & 3 / 2 = 3 4 C D 1 # s ( ) # s $ g | v g v s | d s 1 # s ( ) 2. 65 C D = 24 Re p 1 s ( ) 1 + 0.15 1 s ( ) Re p ( ) 0. 687 # $ % & ( i f 1 s ( ) Re p < 1000 0.44 i f 1 s ( ) Re p ) 1000 + Re p = d s g | v g v s | g
78 Table 3 4. Turbulence Equations (Fluid Phase) Turbulence Equations (fluid phase) Gas phase Reynolds Stress ( v g v g for RSM model) R g = v g v g Boussinesq Hypothesis ( v g v g for k % model, used only for high velocity simulation) v g v g = 2 3 k g I t g g # v g !" + # v g !" ( ) T $ % & ( ) Turbulent Kinetic Energy, k g Equation (High velocity simulation only) (Elghobashi and Abou Arab, 1983) t g # g k g ( ) + $ % g # g v g !" k g ( ) = $ % g t g & k $ k g ( ) + + g t g S 2 g # g g + g k g Mean rate of strain tensor, S S = 1 2 v g !" + v g !" ( ) T # $ % & : v g !" + v g !" ( ) T # $ % & Turbulent Dissipation Rate, % g equation (high velocity simulation only) (Elghobashi and Abou Arab, 1983) t g # g g ( ) + $ % g # g v g !" g ( ) = $ % g t g & $ g ( ) + + g g k g C 1 t g S 2 # g C 2 g ( ) + g g Turbulent viscosity Covariance of gas and particle phase velocit ies (Simonin and Viollet, 1990) with Ratio of characteristic time scales (Simonin and Viollet, 1990) t g = g C k g 2 g k s g k s g = 2 k g b + s g 1 + s g # $ % & b = 1 + C v ( ) s g + C v # $ % & ( 1 s g = t s g F s g
79 Table 3 4. Continued Turbulence Equations (fluid phase) Lagrangian integral time scale calculated along particle trajectories (Simonin and Viollet, 1990) with and is the angle between the mean particle velocity and the mean relative velocity Particle relaxation time due to inertial effects acting on particle phase (Simonin and Viollet, 1990) Characteristic turbulent time scale (Simonin and Viollet, 1990) t g = 3 2 C k g g Drift velocity (Simonin and Viollet, 1990) with D t s g = 1 3 k s g t s g Characteristic turbulent length scale (Simonin and Viollet, 1990) Influence of solids phase on kg (Cokljat et al., 2000; Simonin and Viollet, 1990) Influence of solids phase on % g (Simonin and Viollet, 1990) t s g = t g 1 + C # 2 = | v g v s | # t g L t g C = 1.8 1.35 c os 2 # F s g = s # s $ % 1 # s # g + C v & ( ) + v dr = D t s g s g # s $ # s D t s g s g # g $ # g % & ( ) L t g = 3 2 C k g 3 / 2 g k g = k s g # 2 k g # v g # v s ( ) $ v dr ( ) g = C 3 g k g k g
80 Table 3 4. Continued Turbulence Equations (fluid phase) Modeling constants Gas phase Reynolds stress transport equation (RSM model) (Cokljat et al., 2006) t g # g R g ( ) + $ % g # g v g !" R g ( ) = & g # g R g % $ v g !" + R g % $ v g !" ( ) T ( ) + + g $ % g $ R g ( ) & g # g $ % v g v g v g + g p $ v g + $ v g ( ) T ( ) + & g # g g + R g Interaction of gas and solids phase turbulence (Cokljat et al., 2006) R g = 2 3 k g I Turbulent diffusive transport closure (Lien and Leschziner, 1994) # g $ g v g v g v g % & ( = t g 0.82 R g ) + Linear pressure strain model (Launder, 1975) p v g + v g ( ) T # $ % & = ( g ) T 1 + ) T 2 + ) T W ( ) Slow pressure strain term, T 1 (Gibson and Launder, 1978) T 1 = 1.8 g k g R g + 2 3 k I # $ % & ( Rapid pressure strain term, T 2 (Gibson and Launder, 1978; Launder, 1989) T 2 = 0.6 R g # v g !" + R g # v g !" ( ) T $ % & ( ) + # v g !" R g ( ) + 2 3 P w C w ( ) I + / 0 where w is component normal to wall Production near wall, Pw (Gibson and Launder, 1978; Launder, 1989) P w = v g w v g w v g w x w C 1 = 1.44 C 2 = 1.92 C = 0.09 C 3 = 1.2 C v = 0.5 k = 1.0 = 1.3 s g = 0.75
81 Table 3 4. Continued Turbulence Equations (fluid phase) Convection near wall, C w (Gibson and Launder, 1978; Launder, 1989) C w = x w v g w v g w v g w ( ) Rapid pressure strain term normal to wall, T 2 w (Gibson and Launder, 1978) T 2 w = T 2 n w Rapid pressure strain term, T W (Gibson and Launder, 1978) T W = 0.5 g k g v g w v g n I # 3 2 v g v g w n # $ % & 3 2 v g v g w n ( ) + T / / c l k g 3 / 2 g d w + 0.3 T 2 w n I # 3 2 T 2 w n # 3 2 T 2 w n ( ) T ( ) + c l k g 3 / 2 g d w where dw is the distance to the wall and c l = C 3 / 4 0.4187 Dissipation tensor g = 2 3 g I Turbulent dissipation rate %g equation (RSM model) (Elghobashi and Abou Arab, 1983) t g # g g ( ) + $ % g # g v g !" g ( ) = $ % g t g & $ g ( ) + g # g g k g C 1 R g : $ v g !" + C 2 g ( ) + g # g g
82 Table 3 5 Simulation Parameter s Property Value Particle density 2650 (kg/m 3 ) Particle diameter 140 m Particle particle coefficient of restitution 0.9 Particle wall coefficient of restitution 0.9 Angle of internal friction 28 o Max packing solids volume fraction, s ,max 0.63 Min solids volume fraction, s ,min 0.5 Initial solids volume fraction 0.58
83 Table 3 6 Summary of Frictional Models (n=2 for all models) s,fr P s,fr Fr p Fr p Increased s,fr relative to P s,fr (Model A) 0.1 5 0.05 5 Johnson et al. 1 0.05 5 0.05 5 Ocone et al. 2 0.05 3 0.05 3 Increased Frictional Stress (Model B) 0.2 5.5 0.1 5.5 Increased P s,fr relative to s,fr (Model C) 0.1 5 0.5 5 Srivastava and Sundaresan 3 0.05 5 0.05 5 1. Johnson et al. (1990) 2. Ocone et al. (1993) 3. Srivastava and Sundaresan (2003) I 2 D + d s 2
84 Table 3 7 Predicted Frictional Viscosity and Pressure Associated with Different Frictional Models Frictional Model Max s in the bed at t=0.5 s Magnitude of s,fr at t=0.5 s (Pa*s) Magnitude of P s,fr at t=0.5 s (Pa) Model A 0.591 O(10 0 10 7 ) O(10 10 3 ) Johnson et al. 0.588 O(10 1 10 6 ) O(10 10 3 ) Ocone et al. 0.619 O(10 1 10 2 ) O(10 0 10 2 ) Model B 0.587 O(10 0 10 8 ) O(10 10 4 ) Model C 0.570 O(10 1 10 4 ) O(10 10 3 ) Srivastava and Sundaresan 0.588 O(10 1 10 5 ) O(10 10 3 )
85 Figure 3 1. Schematic of experimental setup
86 Figure 3 2. Axisymmetric formulation computational mesh and simulation domain. Unless otherwise noted, A=15.2 cm, H p =7.62 cm, 1/2D p =0.51 cm. (a) Computational mesh and (b) S imulation domain.
87 Figure 3 3. Predicted crater depth with varying frictional models for a 37 m/s Argon jet
88 Figure 3 4. Crater depth predictions with varying Argon jet velocities (Model A) Figure 3 5. Crater depth prediction with varying gas type (Model A)
89 Figure 3 6. Crater depth prediction with varying nozzle height (Model A) Figure 3 7. Long time crater depth prediction with 37 m/s Argon jet (Model A)
90 Figure 3 8. Experimental crater images and correspon ding simulations at different times
91 Figure 3 9. Simulation domain and computational mesh. T wo dimensional planar (a) S imulation domain and (b) C omputational mesh. T wo dimensional planar with a wall placed at the center (c) S imulation domain and (d) C omputational mesh
92 Figure 3 10. Crater depth prediction with varying simulation domain (Model A)
93 Figure 3 11. Contour plots of solids volume fraction from planar mesh with thin wall simulation for various times. The dotted line indicates the bed h eight at t=0s.
94 Figure 3 12. Solid phase mass flux (m/s) vectors and solids volume fraction iso surfaces predicted by Model A. (a) Argon jet with velocity 37 m/s after 0.5 seconds and (b) A near sonic nitrogen jet after 0.067 seconds.
95 Figure 3 12. Continued
96 Figure 3 13. Gas velocity (m/s) iso surface contours plotted and volume fraction is o surfaces predicted by Model A. (a) Argon jet with velocity 37 m/s after 0.5 seconds and (b) A near sonic nitrogen jet after 0.067 seconds.
97 Figure 3 13. Continued
98 CHAPTER 4 EFFECT OF PARTICLE PROPERTIES ON CRATER DYNAMICS M otivation Software packages are necessary for future (terrestrial) space exploration and for better understanding crater or scour hole formation. Two fluid modeling is the most realistic method for modeling these large scale phenomena. While two fluid models are based on spheres much of the data currently available is for non i deal particles. Cratering and scour hole experiments provide a unique opportunity for validati ng m ultiphase granular flow models. Currently, there is a lack of transient crater growth data for ideal particles, which are beneficial for validating the current constitutive equations. The present state of the literature demonstrates results based on a variety of particles most of which are various types of sand, with scaling parameters built and tested o n either a single particle type or on a few particles with multiple changes made to the properties at once (Kobus et al., 1979 ; Rajaratnam and Beltaos 1977; Haehnel et al., 2008). The results also provide the equilibrium or long term crater growth, which can be difficult to use for validating numerical models as it can take much longer to simulate physical events than they would take in real time. Backg round Rajaratnam and Beltaos (1977) found that the asymptotic depth, D asym scales linearly with the erosion parameter, E c as shown in equation ( 4 1) below. ( 4 1) w here H is the height of the pipe above the bed, m 1 is the slope and m 2 is the intercept. D as y m H = m 1 E c + m 2
99 The erosion parameter is proportional to the densimetric Froude number, Fo ( Rajaratnam and Beltaos, 1977 ). Rajaratnam and Beltaos (1977) did not account for the dependence of particle shape on crater depth. Additionally, as mentioned in Chapter 1, it is not clear whether they report the static or dynamic asymptotic crater depth. ( 4 2) w here v is the jet velocity, p the intrinsic particle density, f the fluid density, and d p is the particle diameter. There are many different ways of measuring the particle diameter. Rajaratnam and Beltaos (1977) likely used the sieve diameter. However, the sieve diameter under report s the size of elongated particles. Sarma (1967) discussed the crater depth growth behavior and functions to desc ribe the growth over time, most of which asymptote. The crater and scour hole data that has been published in the past either reports the asymptotic crater size (Haehnel et al., 2008; Rajaratnam and Beltaos, 1977; Rajaratnam, 1982) or the long term transie nt (on the order of 100 seconds) crater growth (Metzger et al., 2009a; Metzger et al., 2009b). However, the short term crater growth data are useful for validating physical models, as multiphase flow codes can be computationally expensive and take much lon ger in actual time to simulate long periods of time. For example Kuang et al. (2013) performed DEM simulations of cratering of large spherical particles the experiments of which are presented in this chapter carried out to approximately 0.4 seconds. Add itionally, Chapter 3 validates a cratering model for sand by simulating crater growth to 1 second using the two fluid framework. E c = v g f # $ % & ( d p d pi pe H = F o d pi pe H
100 The understanding of scour hole and crater formation is based on work mostly conducted with non ideal particles (i.e. sand and soil), and the literature is currently lacking results with smooth spherical particles. Chapter 4 aims to fill the void by presen ting an investigation that branches from ideal to non ideal particle property cases by first comparing the crater growth rates for smooth spherical particles wi th different densities or sizes Then, the results of spherical parti cles are analyzed against t hose with non spherical particles with varying surface treatments while holding the particle size and density constant. The effects of particle shape are studied using particles with the same density and size bu t different particle shapes (e.g crushed gla ss to spherical glass or stee l spheres and steel cylinders). To date, no study has specifically account ed for the effect of particle shape on the cratering rate or the asymptotic crater depth. Although the effects of particle shape have been ignored, this study shows that they are significant. Here, the particle size distribution effects, which has also largely been disregarded, are briefly explored and show that segregation effects during cratering can affect the transient crater growth rates. The experim ental results provided here offer a plethora of benchmark data for various numerical models. Initial, short term crater growth, to within 1 second, is presented in addition to transient crater growth behavior over long time frames. A dditionally, a modified form of the densimetric Froude number to accounts for particle shape is presented This work is the first to provide a compressive parametric investigation of the effects o f particles properties on scour hole/crater development. Finally, it is the first to explore the effects of, and develop scaling relationships that account for, particle shape.
101 Materials and Methods The imaging system was composed of an 18 mm double Gauss focusable lens mounted on an Edmund Optics EO 0418 high dynamic range camera with a rolling shutter and frame rate of 48.02 frames per second The camera has a pixel size of 10x10 microns, an area of 768x576 pixels, and pixel depth of 8 bits The video was captured and recorded by the ids ueye demo program and the recoded video was then analyzed using a program that was written for M ATLAB which allows the user to record the crater growth information frame by frame. A schematic of the apparatus used to perform cratering experiments is presented in Figure 4 1 and is similar to the one used by Metzger et al. (2009 a ) The particle bed wa s contained in a box with one wall replaced by a clear polycarbonate window. The top of the window was beveled away (with a sharp tip angle of approximately 43 degrees) from the particle bed. The width of the tip of the bevel was found to be less than 0.05 mm. The jet wa s deliver ed into the bed by a pipe that wa s connected to a compressed tank of breathing air The exit of the pipe straddled the beveled edge so that the center of the pipe was in the plane of the bevel tip. T he th ought was that half of the jet impinged on the particl e bed and the crater evolution wa s monitored through the window. T he effect of the windo w on the jet and the cratering was assumed to be minimal, which is difficult to test as wi thout the window it was not possible to monitor the crater formation in situ The gas flows from a compressed air tank through a pressure regulator then a n Alicat M500SLPM D flow meter with a range of 0 500 SLPM and accuracy of 1% F.S. The pressure regulator was used to achieve the desired flow rate. The pipe maintained a length of approximately 1.73 m of straight pipe before the nozzle exit. The g as exited
102 the pipe (0.774 7 cm inner diameter) to form a jet. The height of the pipe exit above the tip of the beveled window was always 2 cm and 5 cm above the top of the particle bed The particles were poured int o the box and were not compacted The top of the particle bed was leveled by carefully dragging a thin rectangular plate acro ss the top of the particles. Sensitivities to the method of filling the box and leveling the surface were tested and results showed negligible dependence on how the surface of the particle bed was prepared for non cohesive materials For cohesive materials the strength of the material increases with consolidation, so care was taken to similarly level the surface for cohesive materials. All experiments were performed with jet s of gas produced from compressed breathing air in Earth's atmosphere and gravity and in standard temperature and pressure. Various gas velocities were imple mented for each particle type. T he range of velocities employed for a particle bed overlapped with velocities used on particles beds for which results were compared. The jet was steady and the velocities implemented never exceeded 0.3 times the speed of sound for air (340 m/s) so that density changes in the jet could be neglected ( Pri tchard 2011). All experiments were performed with turbu lent jets with Reynolds number ranging from approximately 5000 to 44000 (jet velocities 10.71 to 93.74 m/s). Figure 4 2 is a map of the particles instituted here and allows for understanding the overlap in particle properties. In Figure 4 2 particles are arranged by shape, size and intrinsic density. There are two main standard particle size ranges, 0.2 5 0.3 mm and 1 mm, which were used to study the effect of particle density and shape. Particle shape effect s were investig ated in two ways. First, uniform, well defined, smooth
103 cylinder and cubic particles were studied Next, irregularly shaped (frit) particles were used which exhibit more realistic shapes to those found in nature and industry (i.e. crushed glass, olivine and aluminum oxide ). Plastic partic les were composed of either polystyrene or Poly Amid Nylon (PAN). All soda lime (S.L.) glass was used for all glass particles. Unless otherwise stated, experiments with steel particles were performed with stainless steel (abbreviated S. Steel in Table 4 1) The stainless steel shot was produced from cut wire conditioned into to spherical shapes. The results of 1 mm diameter stainless steel shot were compared to true spherical 1 mm carbon steel shot and showed no difference in cratering The relevant particle and bulk material properties were measured and are provided in Table 4 1 The intrinsic particle density, p was measured using the water displacement method. The angle of repose, -, was measured using the angle of piles that remain after a hoppe r discharge type experiment. A rectangular cuboid container was filled with material and particles were slowly poured out through a small slit in the bottom. A shelf was fixed in the center of the container to provide an additional pile for angle measureme nt. The angle of repose was recorded for the two piles at each bottom corner of the container as well as on each side of the pile on the shelf. The shelf was 12 cm long and centered 13.5 cm above the bottom opening. The contain er was approximately 10 cm wi de, 30 cm tall and 30 cm long The width of the slit opening was adjustable and typically was a few centimeters wide, but could be opened to a maximum of 11 cm. The slit was set to a size that allows a slow pour of particles All particle characterization measurements and cratering experiments were performed multiple times and the average and standard deviation are reported.
104 In addition to the angle of repose and intrinsic density Table 4 1 reports the mean sieve, d si mean equivalent sphere volume, d v and mean erosion d E diameters in addition to the sphericity. The erosion diameter a new particle diameter defined here. It is the average of the distance the longest dimension a particle will align in shear flow with the orthogonal For example, the eros ion diameter for a cylinder is the average of the cylinder's diameter and length. The sphericity, volume and erosion diameters for the regularly shaped non spherical (cylindrical and cubic) particles were calculated based on the geometries of the particles For the irregularly shaped particles the RapidVue particle shape and size analyzer measured the sphericity and the Feret's width and length while the Coulter LS13320 technique measured the equivalent volume light scatter diameters. For irregularly shaped particles, the average of the Feret's width and length are approximately equal to the erosion diameter. For the plastic cylindrical particles with 0.38 mm diameter (D) and 0.38 mm length (L) the sphericity was measured with the RapidVue technique and the results compared well to the calculated results, 0.878 and 0.874, respectively. The spherical, cylindrical and cubic particles have well defined shapes with smooth surfaces. Crushed glass, aluminum oxide and olivine are irregularly shaped, angular, parti cles with rough surfaces. Figure 4 3 provides microscope images for 0.25 mm diameter glass spheres, aluminum oxide, olivine, and crushed glass. One set of experiments was performed on large glass spheres with mean particle diameter of 2.3 mm Three jet ve locities were studied 35.36 m/s, 67.18 m/s, and 93.74 m/s. The particles are translucent and light shine d through a few layers of particles making it difficult to for the surface tracking program to detect the surface of the crater.
105 The refore, videos were analyzed by hand. The first 10 consecutive frames were analyzed, and then ever fifth frame was analyzed until 2.5 seconds of cratering w as recorded. Results and Discussions Cratering Large Glass Spheres For 2.3 mm diameter glass spheres the crater quickly reach e s a plateaued size under all jet velocities The depth and width of the plateau that the crater reaches increased with increased jet velocity, as described in Figure 4 4 (a) and (b) respectively. The crater depth is more sensitive to increased jet velocity than the crater width. Figure 4 4 expresses the sensitivity of 2.3 mm diameter glass spheres to increased jet velocity. Initially, the crater depth and width grow the largest which quickly reduces to a steady crat er size. The crater width has a much more dramatic change in initial crat er size to steady crater size ( Figure 4 4 (b)) than the depth ( Figure 4 4 (a) ) Immediately after the splash, particles from the top of the bed near the crater relax downwards and sus pended particles fall reducing the width and depth of the bed. The difference between the early over growth and the ste ady size of the crater increases with increased jet velocities. Possibly due to an initially high er stagnation pressure with increased v elocity, impacting the bed acts as a higher initial force pushing particles away from the impingement region. After the jet impacts the bed, the particles are displaced from their original positions in forming a crater. Initially the particles move in b oth radial and lateral directions, with particles moving no more than a certain number of degrees above the initial bed height. Some particles are eroded at angles allow ing them to exist in the air above the bed For jet velocity 93.74 m/s the highest ejec tion angle is about 30 degrees,
106 for jet velocity 67.18 m/s and 35.36 m/s the ejection angle is about 25 degrees As these lofted particles fall back into the bed they contribute to a reduction in the crater depth and width in Figure 4 4 When the jet initially impacts the bed there is movement of particles away from the stagnation region of the jet. Depending on the jet velocity, the interaction of the jet with the bed can either cause particle movement several particle layers from the stagnation regio n (high jet velocity) or only translate the surface particles (low jet velocity). The increasing volume of particles felt by the initial impact of the jet with increased jet velocity is expressed in Figure 4 5 Figure 4 5 shows the crater evolution in the first 0.0417 seconds under jets with increasing velocity Figure 4 5 (a) correlates to jet velocity 35.36 m/s and only the very top particles in the first layer are displaced from their original positions throughout the 0.0417 seconds However, w ith increa sed jet velocities a larger number of particles in deeper layers are affected by the initial impingement of the jet. Figure 4 5 (b) and (c) are the initial crater development for jet velocities 67.18 m/s and 93.74 m/s, respectively. In Figure 4 5 (b), the displacement of particles 4 to 5 particle layers vertically, which increases to approximately 10 particles deep by the next frame. The highest jet velocity affects the largest volume of pa rticles when it initially impacts the bed. The 93.74 m/s jet initial ly displaced particles around 8 to 10 particle layers down and 20 particles wide, increasing to roughly 13 particles deep and by t=0.0417 s Therefore, with increased jet velocity the depth and width of translated particle in the effected region increases. T he long chains of particles affected by the higher velocity jets are not sustained for long times. As the crater reaches a steady size, the thickness of the layer reduces.
107 After reaching a steady size, shearing of particles along the crater wall for the two higher jet velocities occurs which increases in thickness as the jet velocity increases. The lowest velocity jet does not show a shearing region of particles as the crater reached a steady size. For the 67.18 m/s jet, the shearing layer is about 2 pa rticles thick and for the 93.74 m/s jet the (steady) shearing layer is about 3 to 5 particles thick. Red arrows in Figure 4 6 identify the general trajectories of particles in various regions of the bed, impacted by a 93.74 m/s jet after reaching a steady size crater The lengths of the lines are indicative of the relative velocity of the particles. The particles shear along the surface of the crater with particles moving slower deeper into the bed. The general movement of particles is up and along the cra ter edge, towards the top of the crater. Particles move inwards around the stationary heap of particles collecting at the top edge of the crater as they near the exit. Some particles do not have enough energy to move around the heap, and subsequently they collide and temporarily become part of the heap. Other particles are able to move through or around the heap, and after reaching the peak of their journey fall back into the crater. The trajectories and concentration of recirculating particles result in nu merous particle particle collisions Particles are also seen exiting the edge of the crater at various lateral positions up the crater wall. Particles moving up the wall are knocked from the edge of the crater by colliding with particles falling near the c rater edge. P articles near the top edges of the crater demonstrate little movement The movement of subsurface particles (>2 particle layers deep) occur s when a particle collides or rolls past an adjacent particle, indicating the movement is due to partic le particle forces such as friction Hence, it is theorized here that t he shearing of the glass
108 spheres along the outer e dge of the crater is so great that it induces subsurface shearing of the particles due to particle particle forces This is a newly ide ntified cratering mechanism, as fluid particle forces do not propagate the movement within the bed While the trajectories of the particles around the sides of the crater are consistent with that of DDS, the subsurface particle movements are not consistent with fluid driven flow The particles move whe n contacted by adjacent particles, as opposed to fluid driven shearing in which the shear force of the drag on the particle results in steady flow of particles along the edge of the crater. The progression of red and orange dashed arrows in Figure 4 6 demonstrates the subsurface particle based shearing mechanism. The red dashed arrow represents the trajectory of a particle moving through the bed. After contacting the stationary particle, centered below the orig in of the orange dashed arrow, the stationary particle traverses the trajectory of the orange dashed arrow indicating that the particle particle contact induced the second particle's movement. Also, Metzger et al. (2009a) showed for sand that when DDS domi nates, particles continue shearing to the to p of the crater edge and up towards the bottom of the crater for particles directly below the crater This new cratering mechanism is only e xpected to dominate for very coa rse particles. It is not clear under wha t conditions this mechanism will dominate, but the results indicate that the particles would need to have high inertia in order to displace particles inside of the bed. Particles that erode out of the crater can either deposit on top of the bed or recircu late inside of the bed. A pile of particles forms as p article s escape the crater and deposit on the edge of the crater surface. For the 35.36 m/s jet the height of the mound near the crater edge increases by about two particles, and approximately four particles
109 for both the 67.18 m/s and 93.74 m/s. A fter the crater reaches a steady size, numerous particles recirculate inside of the crater, which experience a significant amount of collisions. In Figure 4 6 the top layer of the recirculating particl es can be identified as a band of slightly blurred particles, while the particles moving very fast inside of the crater are identifiable as light gray streaks. The particles that seem to hang at the top of the crater are easier to see in Figure 4 6 since t heir velocit ies are lower, which is likely due to more frequent collisions dissipating momentum of the particles. A qualitative comparison of the density of gray streaks moving inside of the crater shows that i ncreased jet velocity correlates to an increas ed density of colliding particles The particles colliding inside of the crater are addressed and identified here, for the first time, as a new particle particle based mechanism as they can enhance or attenuate erosion by ejecting particles from the surfac e as well as dissipating momentum of particles from inelastic collisions. These collision based results are consistent with the observations of Berger et al. (2013). To ensure that the wall does not affect the entrained particle induced collision behavior, a set of experiments was performed in the center of the bed far from the walls The jet was directed straight down into the center of the bed and particles were observed jumping arou nd inside of the crater and a layer of slower moving particles was hangi ng on top confirming the collisions are not a result of the wall The angles of both the left and right wall of the steady sized crater were measured for each velocity and found to be between approximately 50 and 70 degrees The angle of the crater does not correlate to the angle of repose, which was measured to be approximately 20 degrees and given in Table 4 1 indicating that the gas traction
110 and layers of shearing particles dominate in the entire range of the crater size, as the walls are supported at such high angles An increase in the angle of the crater walls from approximately 50 to 65 degrees with jet velocity increase from 35.36 m/s to 67.18 m/s, respectively, is found. However, there is not a significant change in the crat er angle between the t wo higher jet velocities. O nce the jet is turned off particles move to s table position s somewhat filling in the crater. Particles shearing along the side of the crater roll to the bottom after the jet is turned off. Additionally, particles that are suspen ded above the crater as well as those entrained inside of the crater settle. The crater depth decreased from the steady jet on crater size for all velocities. There is a small change in the crater depth after extinguishing the jet for the 35.36 m/s jet as the crater reduced to approximately 0.75 cm. Conversely, a larger change is found for 67.18 m/s and 93.74 m/s jets in which the final crater depth is approximately 1 mm and 1.1 mm, respectively. There is a marked increase in the width of the crater after t he jet is terminated compared to the steady jet on width. The jet off width is similar to the peak width experienced initially by the crater. As the jet is extinguished the sidewalls of the cr ater relax, filling the crater, resulting in an increase of the crater width. Effect of Particle Size and Density on Crater Growth Figure 4 7 compares the long term (a) and short term (b) crater depth growth for spherical glass particles of various particle diameters with jet velocity 20.90 m/s. For 1 mm and 0.5 mm gl ass spheres the crater reaches a steady state size quickly skipping the logarithmic growth phase. The particles with diameters 0.181 mm and 0.25 mm have similar c rater depth growths over time for jet velocity 20.90 m/s, but with increased
111 velocity the cra ters grow consistent to what are shown for other particles; the smaller particle initially grows at a slower rate, but overcomes the growth of the larger particle Generally, the crater grows deeper over time, or produces a deeper asymptotic crater, as th e particle size is decreased. Glass spheres with diameter 0.046 mm (46 microns) produces a crater that grows slightly deeper over time than the glass spheres wit h diameter 0.181 mm, even though the ratio of the particle diameters is greater than any other subsequent size tested. However, i n Figure 4 7 (b) the larger particle diameters exhibit larger initial crater growth than the smaller particles The length of time that the initial crate r growth rate for a larger particle is gr e ater than the next smaller size increases as the particle size decreases. Specifically, t he depth of the 0.5 mm diameter spheres overcame that of the 1 mm spheres at approximately 0.25 seconds, while the 0.046 mm particles overca me the depth of the 0.18 particle bed after a pproximately 46 seconds. At higher velocities the time at which the smaller particle crater depth growth rate overcame that of the larger particles increased. For example, with a jet velocity of 31.05 m/s the 0.25 mm diameter particles initially the crater grows slower and deeper than the 0.5 mm diameter particles after around 1.5 seconds, as opposed to 0.85 seconds for a 20.90 m/s jet. Plastic cylinders with the same aspect ratio and different equivalent volume diameters produces crater growth similar to t he results displayed for spheres in Figure 4 7 The increased number of particles that can occupy the same volume as the particle size is decreased explains the increased initial crater growth rate, in Figure 4 7 (b) with increased particle size At 0.1 second the depth of the crater for a specific particle size is measured in number of particles. For 1 mm diameter glass spheres the
112 depth is approximately 0. 8 cm or 8 particles f or 0.5 mm glass spheres, the crater depth i s about 0.6 cm or 12 par ticles, and the crater depth for particle size 0.181 mm is 0.5 cm or 28 particles. In summary, while the crater is initially deeper for the larger particles, the number of particles that are eroded is much less than for smaller particle beds. The crater de pth is generally a good indicator of general crater growth. H owever sometimes the crater depth can grow inconsistently to the overall crater In Figure 4 7 (a) the crater depth for the 0.181 mm and 0.25 mm diameter particles are similar, while the crater depth for 0.046 mm diameter spheres grows slightly faster. In Figure 4 8 the crater width (a) and volume (b) over time is presented for glass spheres with diameter s of 0.25 mm, 0.181 mm and 0.046 mm. Comparing the crater width and volume for 0.181 mm and 0.25 mm diameter spheres reveals the crater growth trend that is typically found; the crater depth and width increases simultaneously For 0.25 mm and 0.181 mm diameter particles the crat er depth is only slightly different for the time mea sured but the difference in the crater width and therefore volume, over time is significant The crater width and volume for the 0.046 mm diameter glass spheres in Figure 4 8 (a) is smaller than for the larger particles. The decreased crater width and vo lume correlates to an increased angle of repose for the 0.046 mm beads which is discussed in below For the smaller size d spheres, the crater depth growth is less affected by size than the larger particle diameters. For 0.046 mm glass spheres, the effects of surface forces, such as cohesion resulting from van der Waals forces and liquid bridges are expected to be significant. Cohesive effects are likely the reason for 0.046 mm spheres crater depth to grow only slightly faster than the 0.181 mm spheres and for the reduced crater width and volume as compared to the larger diameter spheres. Figures 4 7 and 4 8
113 show that, for cohesive materials, the long term cratering behavior of the crater depth cannot be scaled in the same manner as the width and volume. T he sidewalls of the experimental box need to be far enough away from the impingement region to not a ffect the crater growth and t ests were performed to determine the critical sidewall position By performing experiments with reduced box sizes, it was determined that the walls affect the crater growth when the outer crater edge reaches the sidewall. The crater size measurements are made with respect to the original bed height, so the edge of outer crater extends beyond the crater width; the difference between the two positions is determined by the angle of repose. The outer crater width growth increases with decreasing particle size, for non cohesive particles, and increased jet velocity. For examp le, in Figure 4 8 (a) the 0.181 mm glass sphere experiment was stopped before the other particles in order to avoid effects from the experimental sidewall. Figure 4 9 describes the crater depth over time for three different beds composed of particles with different intrinsic densities and mean diameters of 0.2 4 0.3 mm for both long (a) and short term (b) crater growth. Particle size of polystyrene and steel are slightly higher than glass but from Figure 4 7 (a) it is clear that the crater depth growth for particle s in this size range is marginally affected by the particle size. I ncreased particle density, Figure 4 9 decreases the crate r depth for both long and short term growth The effect of particle density on crater growth for 1 mm diameter particles is shown in Figure 4 10 and presents a similar effect to that exhibited in Figure 4 9 (a).
114 The short term crater depth grows similarly to that of 0.25 to 0.3 mm beads, Figure 4 9 (b), in which the increased density particles cratered to a smaller size. Co mparing Figures 4 9 and 4 10 exemplifies a clear difference in the growth functions for the crater depth. The crater depth growth over time for glass spheres with diameter 0.25 mm in Figure 4 9 (a) continued growing in a logarithmic fashion, but in Figure 4 10 the 1 mm glass spheres quickly formed a deep steady state sized crater. The experiments are characterized into two groups. The first is referred to as type A, which quickly reach their limiting, or asymptotic depth. Examples of type A experiments are the glass or steel particles in Figure 4 10 Experiments that do not reach their asymptotic crater depth within the duration of the experiment, comparable to glass and polystyrene spheres in Figure 4 9 are classified as type B. Investigating the trajector ies of 1 mm diameter polystyrene and steel particles during cratering show initial subsurface movement similar to 2.3 mm diameter glass spheres, Figure 4 5 indicating initially DDS or BCF occur. Near when the polystyrene crater reaches its asymptotic dept h, only the VE cratering mechanism occurs as individual particles are identified eroding surface of the crater, but no subsurface shearing is observed. Similar observations of steel 1 mm spheres at later times are of surface layer particles eroding over ea ch other into the flow. The lack of subsurface movement of particles indicates that the cratering mechanism is VE. All 1 mm spheres experience significant collisions inside of the crater after the initial crater growth phase and the number of collisions in creases with increased velocity. In Figure 4 10 even after 100 seconds, the crater depth of 1 mm polystyrene spheres is marginally increasing, while 1 mm steel and glass
115 quickly reached steady crater sizes. The asymptotic depth is reached faster with incr eased particle size, and to a lesser extent with increased intrinsic density. Effect of Particle Shape on Crater Growth In Figure 4 11 the effect of particle shape is expressed for steel particles with equivalent sphere volume diameter diameter of a sp here with the same volume as the particle of approximately 1 mm. T he crater depth over time for steel spheres with diameter 1 mm is compared to steel cylinders with diameter 1 mm and length 1 mm, aspect ratio of 1 (AR= L/D= 1), and steel cylinders with dia meter of 0.5 mm and length 4 mm, aspect ratio of 8 (AR=8). In Figure 4 11 (a) t he steel sph eres steady state, or equilibrium, crater size is smaller than that of steel cylinders with AR =1, but is greater than the steady crater size for cylinders with AR=8 Initially, Figure 4 11 (b), the steel spheres and cylinders with AR=1 have the same crater growth rate to about 0.5 seconds. The steel cylinders with AR=8 have the same init ial growth, but quickly diverge after approximately 0.1 seconds. The crater depth over time for plastic spheres, cylinders and cubes with equivalent volume sphere diameter of approximately 1 mm are compared in Figure 4 12 T he cubic particles have a larger steady state crater size than the cylindrical particles, followe d by the spherical particles. Under lower velocity jets, non spherical particles exhibit more similar steady state crater depths. The effect of particle shape becomes more pronounced with increased jet velocity. Fluctuations in the crater depth occur as pa rticles avalanche from the outer crater into the inner crater. These avalanches can carry a lot of material causing very large fluctuations, as is seen for 0.3 mm diameter polystyrene and 0.046 mm glass spheres. In other cases, these fluctuations are small and could not be differentiated from the
116 noise in the crater depth. For short times the initial crater growth increases with decreased particle density, as in Figure 4 9 (b). The size of these fluctuations directly correlates to the mean and standard devi ation of the angle of repose the material. A large standard deviation of the measured angle of repose indicates a large disparity between the angle of failure and angle of repose of the material. The macroscopic fluctuations in the crater depth are average d out for the previous plots. Figure 4 13 shows the crater depth over time for a single experiment of 0.3 mm polystyrene spheres under a 20.90 m/s jet. Inlayed in Figure 4 13 are four images that represent a single avalanche from the experiment and correl ate to the numbered stars on the crater depth over time data points The crater depth peaks at star 1 and from the inlayed fra me it is clear that the crater is deep, and that the left outer crater wall is steep (which indicates it is about to fail). Moments later, star 2, illustrates the left crater wall fails resulting in material avalanching into the bottom of the crater. Next, star 3, the left sidewall continues to avalanche, as the left wall relaxes to the angle of repose, into the crater further reducing the crater depth. The bottom of the crater depth peak, corresponding to star 4, expresses the shallowest the crater became for the fluctuation. T he crater growth approximately between time 20 and 35 seconds does not show macroscale fluctuations, and grows slower than the period directly before and after an avalanche. After an avalanc he directly after star 4, the crater depth gr ows at an accelerated rate It is theorized here that the crater grows at an increased rate after, as the avalanched mate rial is less compact. Around 35 seconds, there is a drastic increase in the cra ter growth rate leading to star 1. The marked acceleration in growth rate that occurs before material avalanches off of the sidewall is theorized here to be due to
117 increased out er crater angle resulting in increased momentum at the tip of the jet. As the crater sidewall angle increases, the area for the radially expanding jet decreases, which in turn increases the momentum directed to the bottom of the jet. For low intrinsic dens ity materials, the increased momentum to the jet tip is enough to cause accelerated crater growth. Metzger et al. (2009a) showed with a single phase CFD simulation that the jet does not interact with the outer crater when the angle is below 40 degrees. Her e, the outer crater angle before failing is approximately 55 degrees, star 1 in Figure 4 13 and therefore the interaction of the jet with the outer crater can no longer be neglected. For all particles, the angle of the outer crater is the same as the meas ured angle of repose, which is consistent with the findings of Metzger et al. (2009a) for sand. Figure 4 14 illustrates the long (a) and short (b) term crater depth growth for glass spheres, crushed glass, olivine, and aluminum oxide with mean sieve diamet er s between 0.25 and 0.3 mm. The crushed glass grows to a deeper crater than glass spheres, for the same particle size over long times. The particle densities of a luminum oxide and olivine are higher than glass, so it is consistent that the crater depth gr owth for a luminum oxide and olivine are less than that of the crushed glass Interestingly, the olivine and aluminum oxide produce slightly deeper craters than glass spheres at long times, indicating that the effects of particle shape dominate intrinsic density over the studied for long term crater growth Crushed glass has higher friction than glass spheres, as expressed by the increased angle of repose, and the increased resistance to movement of particles under the jet was expected to slow crater depth growth. However, quite the opposite is found.
118 The initial crater growth rate s Figure 4 14 (b), for crushed glass and glass spheres are similar For short time frames, the crater growth of olivine is slightly less than that of crushed glass, which is cons istent with previous results presented here. However, the larger intrinsic density of aluminum oxide results in reduced initial cratering rate compared to crushed glass and olivine. In Figure 4 14 (b) the short term growth rate of glass spheres and crushed glass deviate after approximately 0.8 seconds, cratered under at 20.90 m/s jet. The crater depth growth rate of the glass spheres slows, relative to the rate of crushed glass, when the outer crater for 0.25 mm glass spheres formed, after about 0.8 to 0.9 seconds. Conversely, the crushed glass forms an outer crater after approximately 1.5 to 1.9 seconds. The crater depth growth rate reduces when the outer crater forms as particles at the top of the crater fall into the crater and recirculation begins along the outer crater. The crushed glass depth growth rate at long times is slightly higher than for glass spheres, expressing that even long after the outer crater forms crushed glass craters at a faster rate. This indicates that the difference in crater depth at long times is not due solely to the difference in the time it takes for the outer crater to form. The outer crater angles for non spherical particles, with the exception of steel cylinders with AR=8, are below 40 degrees, the angle that Metzger et al. (2009a) showed not to interact with the jet. Therefore, the increased crater depth growth for non spherical particles does not occur from increased momentum at the jet tip. Similarly, the jet velocity reduces with distance away from the nozzle. The crater depth of spherical particles is less than that of non spherical particles, with the exception of steel cylinders AR=8, resulting in higher jet velocities at the bottom of the crater for spherical particles.
119 The higher local jet velocity for the spherical particles indicate increased erosion of spherical particles would be expected resulting in an increased growth rate at long times. However, this is not the case. Non spherical particles often have slightly lower permeabilities caused by decreased pore size which would result in higher surface pressure and reduced diffused gas velocity in the pore spaces. If differences in permeability, resulting from non spherical particles, account for the disparities in growth rates, then permeability effects should domi nate in the early crater growth, when velocity and pressure gradients are highest. However, the results show the same early crater growth rates when only the particle shape is changed. The crater depth results presented here seem to oppose intuitive effect s of friction, as it was expected that increased friction would reduce crater depth along with overall crater size. However, increased friction of the particles is experienced beyond directly under the jet nozzle. Friction also must be considered in the re circulation as particles must roll or slide down the outer crater to fall back into the inner crater. Therefore, it is argued here that the reduced recirculation rate, caused by increased friction, allows the non spherical particles to crater deeper over t ime than the spherical particles. The growth of the crater width and volume growth over time for glass spheres, crushed glass, olivine and aluminum oxide is much more in line with what is expected intuitively. The olivine and aluminum oxide produces crate rs with similar outer width and crater volumes over long times. Glass spheres eroded the largest volume and widest crater over time, and the crushed glass eroded a slightly lower crater volume and outer width, followed by the olivine and aluminum oxide. Th e outer crater width is determined
120 by the depth of the crater and the outer crater angle and hence, by the angle of repose and the angle of failure of the granular material. The theory of the recirculation of particles along the outer crater explaining the crater depth growth effect is verified for 0.25 mm diameter glass spheres and crushed glass. Particles were tracked at various horizontal and vertical positions as they move d along the outer crater. Figure 4 15 illustrates the tracked particle trajectories in the outer crater for (a) glass spheres and (b) crushed glass over a period of 0.54 seconds. The positions of the particles are provided with lines, and the markers repre sent the particles' position every approximately 0.15 seconds. The trajectories are overlaid on top of the image of the crater correlating to the first frame. The crater images for crushed glass and glass spheres are to scale and with respect to the same c amera position. Figure 4 15 expresses two results: the glass spheres shear down the outer crater in bulk, and crushed glass have moderately stationary particles on the surface and in the bulk of the outer crater. The glass spheres on the top layer of parti cles in the outer crater in Figure 4 15 (a) all move with similar rates, but slightly faster closer to the inner crater. Particles below the top surface are found to move at lower velocities than the particles at the top surface, as they are shearing down the crater wall. The group of tracked particles near the top of the outer crater show that the shearing layer is approximately 3 particles thick, and the group of tracked particles near the bottom shows the shearing layer is more than 6 particles thick. Co nversely, for crushed glass the particle trajectories generally show that there is very little movement of particles, even at the surface of the outer crater. The dark gray line in Figure 4 15 (a) and (b)
121 designates the region of crater that is shearing in bulk. The group of particles near the edge of the outer crater displays more continuous particle movement in the small shearing region between the inner and outer crater. The difference in distance moved by the crushed glass and glass sphere particles in the region near the inner crater demonstrates that glass spheres are moving faster than the crushed glass particles. For the case of crushed glass, a shearing layer a few particles thick does occur when the outer crater avalanches. However, the avalanches occur periodically and the shearing layer is thinner than that of glass spheres. It is not clear why, but there seems to be more instances of particle collisions in the crushed glass case that lead to individual particles quickly moving down the outer cra ter than for glass spheres. While these particles move with velocities much higher than the top shearing layer in the glass spheres instance, they are sporadic and occur at low frequency. The glass spheres have a much larger volume of particles moving with generally higher velocities into the inner crater than the crushed glass case. The theory of reduced recirculation rate is also expected for increased cohesion. Figure 4 15 demonstrates the theory proposed here that non spherical particles crater to a lar ger depth due to increased friction leading to decreased recirculation rate in the outer crater than for spheres. From the results presented here on high aspect ratio cylinders, Figure 4 11 the proposed theory must be expanded to account for much higher friction. It is expected that friction has varied effects on crater depth, as increased friction will cause a deeper crater until the friction (or cohesion) inhibits the erosion of particles under the jet, which could explain the crater growth results show n for glass spheres with diameter 0.046 mm
122 and steel cylinders with AR=8. Note should be made that the decreased crater depth growth found with AR=8 steel cylinders could be due to much increased friction (larger contact area), particle packing effects, or by changes in the fluid particle induced erosion (i.e. drag) with such elongated particles. Crater depth and width increase with decreased particle size, except for 0.046 mm glass spheres, which show a slight increase in crater depth and a narrower crate r than for larger particles. The friction between the smalle r particles may not increase but cohesive forces (i.e. van der Waals) relative to the gravity force are much higher and likely inhibit recirculation similarly to the increased friction of non sph erical particles. The angle of repose, Table 4 1 of 0.046 mm diameter glass spheres is found to be steeper than for 0.181 and 0.25 mm diameter glass spheres. The narrower crater is due to the increased angle of repose of the 0.046 mm glass spheres, Figure 4 8 (a), which would indicate reduced recirculation and therefore should result in a deeper crater. Together, the effect of particle size and reduced recirculation, due to cohesion, should result in a much larger increase in crater depth. However, the 0.0 46 mm glass spheres have only a slightly deeper crater than 0.181 mm glass spheres (Figure 4 7 ). It is argued that only a slight increase in crater depth is found due to cohesive forces inhibiting the erosion of particles under the jet balanced by reduced recirculation along the outer crated and decreased particle size effects. Figure 4 16 compares the long (a) and short term (b) crater depth growth over time for 1 mm diameter glass spheres, 0.5 mm diameter glass spheres and a 50 50 mix (mix 50) by mass of 0.5 mm and 1 mm diameter glass spheres. T he long term crate r depth growth for mix 50 is more similar to the larger diameter spheres. Conversely, at
123 early times the crater depth growth ( Figure 4 16 (b)) is more similar for mix 50 and the smaller diameter spheres An image of the crater edge of mix 50 taken after the jet was extinguished in Figure 4 17 illustrates the segregation that occurs during cratering Along the edge of the crater the top layers are dominated by the larger particles. Nei ther particle size seems to dominate the subsurface space a few layers down, indicating that at la ter times the smaller particles segregate out leaving the larger particles to dominate the crater growth behavior. Scaling Asymptotic Crater Depth The erosio n parameter proposed by Rajaratnam and Beltaos (1977), equation ( 4 2) does not account for the particle shape. The results presented here show that the particle shape significantly impacts the asymptotic crater depth. Some methods of measuring the particl e diameter take into account the particles' shape more than others. For example, sieve diameter is a measure of the smallest dimension that can fit through a sieve hole, while the equivalent surface area diameter can account for the difference in surface a rea with a non spherical particle. An adjusted form of the densimetric Froude number, Fo to account for the particle shape is proposed here and given in equation ( 4 3) E c = v g f # $ % & ( ) d p d pi pe H = F o ) d pi pe H ( 4 3) Nonetheless, there are multiple methods for measuring or reporting the particle diameter, d p Various types of diameters are tested here in both equation s ( 4 2) and ( 4
124 3) and the goodness of fit, which is reported by the correlation coefficient, r, compared in terms of scaling the asymptotic cra ter depths. As the pipe height and diameter are kept constant, all scaled results are provided in terms of the asymptotic depth and the densimetric Froude number. Figure 4 18 (a) provides the measured asymptotic crater depth for type A experiments compare d to the original densimetric Froude number of Rajaratnam and Beltaos (1977) in which the diameter was the mean sieve diameter. Further, Figure 4 18 (a) compares the original from to the modified form to take into account the particle shape by multiplying the mean sieve particle diameter by the sphericity. The original densimetric Froude number, equation ( 4 2) with the mean sieve diameter does a poor job of scaling the asymptotic crater depth, and by simply accounting for particle shape using the sphericit y does not improve the relationship. The sieve diameter under reports the size of elongated particles, which also have reduced and, therefore, when accounting for particle shape along with the sieve diameter only reduced the accuracy of the scaling met hod. Figure 4 18 (b) illustrates the scaling of asymptotic crater depth with modified densimetric Froude number proposed here Fo in equation ( 4 3) to describe the particle shape effect with t he sphericity of the particles along with the erosion diameter. Comparing Figure 4 18 (a) to (b), the circular points represent measurements of spherical particles. The pos ition of the spherical points do not change from Figure 4 18 (a) to (b) since the diameter of a sphere is not affected by the method of m easurement and the sphericity is one The erosion diameter proved to have the highest correlation coefficient of all diameters studied, with and without the sphericity term: projected area diameter, projected perimeter diameter, equivalent surface volume d iameter (the
125 diameter of a sphere with the same ratio of surface area to volume as the particle), and equivalent surface area diameter (the diameter of a sphere with the same surface area as the particle). The optimal description of particle size in the de nsimetric Froude number was found by c omparing the correlation coefficient between the experimentally determined asymptotic depths and the densimetric Froude number for each diameter with and without accounting for the sphericity. The Froude number with th e erosion diameter including the sphericity, equation ( 4 3) produced the best scaling with the highest correlation coefficient, r=0.98. When the densimetric Froude number is described using the original method of the mean sieve diameter, equation ( 4 2) p roduces a scaling with r of 0.88, but accounting for the particle shape, equation ( 4 3) reduced r to 0.83. The sieve diameter produces the worst correlation coefficient for both equation s ( 4 2) and ( 4 3 ) than any other diameter tested. Using the median sieve diameter in equation s ( 4 1) and ( 4 3) resulted in a slight decrease in the correlation coefficient. The densimetric Froude number calculated with the mean equivalent sphere volume diameter, both with and without accounting for the pa rticle sphericity, resulted in r of 0.93. In summary, the final modified form of the densimetric Froude number to account for particle shape is found in equation ( 4 4) below. E c = v g f # $ % & ( ) d E d pi pe H = F o ) d pi pe H ( 4 4) Sarma (1967) discusses multiple possibilities of crater growth fun ctions, one of which is an arctangent (arctan) fit and is provided in equation ( 4 5) Sarma found that equation ( 4 5) fit crater depth data well for two dimensional submerged jets, and the function is extended for use with circular vertically impinging jet s and is used to fit all
126 experiments performed here. Both type A (quickly produced asymptotic depth) and type B experiments (asymptotic depth was not reached within experiment) are well fit by equation ( 4 5) ( 4 5) Where D c is the crater depth, t is the time, and a 1 b 1 and c 1 are fitting parameters. In the limit of t goes to infinity, D c goes to 0.5* *a 1 The asymptotic crater depth for type A experiments, for example glass spheres with 1 mm diameter in Figure 4 7 compared well with the asymptotic depth predicted from 0.5* *a 1 in equation ( 4 5) Sarma (1967) found that the fitting constants a 1 b 1 and c 1 could be related to experimental parameters (e.g. jet velocity, Froude number, particle terminal velocity, noz zle height) for two dimensional submerged jets. The analysis performed here on a 1 b 1 and c 1 does not show the same correlation to experimental parameters. Metzger et al. ( 2009a) provides a logarithmic fitting function that fits the crater growth for shor t times well, which is given in equation ( 4 6) ( 4 6) w here a 2 and b 2 are fitting parameters. Figure 4 19 (a) and (b) compare Metzger et al.'s (2009a) logarithmic fit, equation ( 4 6) to the arctan fit, equation ( 4 5) for both type A and type B experiments, respectively. Crater depth growth of 1.0 mm glass spheres under a 31.05 m/s, in Figure 4 19 (a), illustrates that for type A experiments, equation ( 4 5) fits the crater depth growth well. The logarithmic fit is a p oor fit for type A experiments, and is not able to predict the asymptotic depth, as logarithmic functions do not asymptote. Both equation s ( 4 5) and ( 4 6) are extrapolated to the crater depth growth of aluminum oxide with jet D c = a 1 t a n 1 b 1 t c 1 ( ) ( ) D c = a 2 l n b 2 t + 1 ( )
127 velocity 20.90 m/s in Figure 4 19 (b). Equation s ( 4 5) and ( 4 6) are fitted to approximately the first 90 seconds of experiment, since most experiments are carried out to 90 seconds. The fits are then used to predict the crater depth along the entire length of the experiment. As Figure 4 19 (b) shows, the arctan fit, equation ( 4 5) does a much better job than the logarithmic fit, equation ( 4 6) at predicting the long term crater growth. Equation ( 4 5) predicts the long term growth of type B experiments even though it was only fit to t he first 90 seconds of crater growth. Indicating enough crater development is needed after the elbow in the depth growth over time for equation ( 4 5) to predict the asymptotic crater depth for type B experiments. Metzger et al. (2009a) and Metzger et al ( 2010) report that the logarithmic fit is only appropriate for the initial crater growth period, so it is expected that the logarithmic fit would not be able to predict the crater growth long term. The crater growth for short times is inlayed in Figures 4 1 9 (a) and (b) and expresses that very early in the experiment, equation ( 4 6) does a marginally better job fitting the crater growth than the arctan fit for type B experiments, but not type A experiments. Figure 4 20 compares measured and predicted asympt otic depth taken from the arctan fitting function for type A experiments and the predicted asymptotic depth for the type B experiments scaled by the densimetric Froude number modified to account for particle shape, equation ( 4 4) using the erosion diamete r. The arctan fitting function, equation ( 4 5) correlates well to the measured crater depth. The predicted asymptotic crater depths for type B experiments agree with the measured asymptotic crater depth scaling of the densimetric Froude number from equati on ( 4 4)
128 There are two sets of outlier data points in Figure 4 20 The first is the asymptotic crater depth of 2.3 mm glass spheres, and are indicated with a plus, +, in the figure. The cohesive 0.046 mm glass spheres contribute to the second and the asymptotic depth is indicated by x on the predicted asymptotic crater depth in Figure 4 20 For the case of 2.3 mm glass spheres, with increasing velocity the divergence from the scaling increases. The cratering mechanism for 2.3 mm glass spheres under higher velocity jets results in subsurface shearing and is not dominated by VE. Glass spheres 2.3 mm are the only particle studied here to experience the deep subsurface shearing after the crater reached its asymptotic size, and hence should not be expected to scale with experiments dominated by VE. The outlier data points on the top right section of Figure 4 20 are the predicted asymptotic crater depths, by equation ( 4 5) for glass spheres with diameter of 0.046 mm. The experiments for 0.046 mm glass extrapolated with equation ( 4 5) contain enough crater depth growth to assum e that the predicted asymptotic depth is good. The outlier data points of the 0.046 mm glass spheres indicate that the scaling of the asymptotic crater depth with densimetric Froude number must be adjusted to account for cohesion. Based on the results pres ented here for cohesive particles a correction to equation ( 4 4) should be in the form of equation ( 4 7) ( 4 7) w here / 1 and / 2 are constants that relate to the cohesiveness of the material. Decent correction is made to the scaling for glass sp heres with diameter of 0.046 mm with / 1 =4.88 and / 2 =8.1. However, more cohesive materials need to be investigated in order F o = 1 1 v g # $ $ % & ( ) d f + 2
129 to verify the form of equation ( 4 7) and to gain understanding of the dependence of / 1 and / 2 on the cohesive material properties. S ummary This work presents a rigorous study of the effects of various particle properties on both asymptotic and transient crater growth. The long term transient and initial crater growth do not always coincide, as is demonstrated for increased particle dia meter. The long term crater growth is dominated by recirculation rate and particle properties such as density, size and shape. The long term t ransient crater growth increases with decreased particle density and size. All experiments can be extrapolated with an arctangent crater growth equation, which can then be used to estimate the asymptotic crater depth However, this function must be used with enough crater development in order to avoid under predicting the crater depth at later times. This work is the first to discover that particle shape has a significant effect on crater growth. Non spherical particles tend to decrease the crater width and volume but increase the crater depth over time. Initially the effects of particle shape on crater growth are minimal, but when the outer crater forms the shape effects dominate the crater development over small differences in particle size and density. Additionally, the method of describing the particle diameter is important fo r scaling the asymptotic crater depth in order to account for the particle shape. This work defines a modified version of the densimetric Froude number that accounts for particle shape. Further, as particle size reduces to sizes below 100 microns, cohesion is an important material property and this work shows that the effects must be included for scaling predictions to be accurate. Increased friction, to a point, actually can increase the crater depth for long term crater growth due to decreased flow of par ticles down the outer crater. For the case of high friction or
130 cohesion, it is argued that reduced erosion under the jet causes decreased crater depth growth. Also, a rigorous study of various particle types shows that the outer crater angle is consistentl y the same as the angle of repose of the material. VE is identified to dominate under most of the operating conditions presented here. Two new particle based cratering mechanisms are also identified. Deep subsurface shearing caused by particle particle in teractions are found for large, 2.3 mm, glass spheres. Particle particle force, such as friction, induced shearing mechanism is also found for smaller particles, which results in reduced cratering rates for spherical particles as increased shearing along t he outer crater increases recirculation inside of the crater. Furthermore, collisions in the crater are identified to occur at significant rates. Collisions can, for example, enhance or attenuate erosion rates by scattering particles from the surface or di ssipation of momentum due to inelastic collisions, respectively.
131 Table 4 1 Bulk and Intrinsic Properties of Particles Particle Type d si (m) d v (m) d E (m) p (g/cm 3 ) b (g/cm 3 ) Aluminum oxide ( Al 2 O 3 ) 279 1 361 5 330 0.782 35 .2 0.8 3.9 0.06 1.784 0.0009 Olivine 278 2 378 15 240 0.817 35 .5 0.7 3.4 0.09 1.604 0.004 Polystyrene sphere 0. 150 0. 360 mm 303 4 303 4 303 4 1 32.2 5.5 1.06 0.04 0.600 0.0008 Polystyrene sphere 0.65 mm 646 3 646 3 646 3 1 23.3 0.9 1.05 0.03 0.615 0.0003 Polystyrene sphere 1 mm 1052 1 1052 1 1052 1 1 25.4 1.2 1.05 0.02 0.604 0.001 PAN cube 1 mm x 1 mm x 1 mm 1090 3 1240 1000 0.806 37.7 1.3 1.1 0.04 0.653 0.003 PAN cylinder 1 mm D x 1mm L 1091 1 1140 1000 0.874 34. 8 1. 7 1.1 0.04 0.630 0.001 PAN cyl inder 0.38 mm D x 0.38 mm L 503 6 430 380 0.874 33.6 1. 6 1.1 0.03 0.588 0.001 Carbon Steel Sphere 0.25 mm 295 1 295 1 295 1 1 25.5 1. 9 7.8 0.04 4.430 0.002 S Steel Shot (spherical) 1 mm 1090 1 1090 1 1090 1 1 28 0. 9 7.8 0.04 4.698 0.006 S. Steel Cylinder 0.5 mm D x 4 mm L 612 1 1140 2250 0.617 38 10 7.7 0.03 3.376 0.04 S. Steel Cylinder 1 mm D x 1 mm L 1 093 5 1140 1000 0.874 32 2. 8 7.7 0.03 4.489 0.005 S.L. Glass Spheres 2.3 mm 2270 6 2270 6 2270 6 1 19.7 1.0 2.5 0.03 1.613 0.001 S.L. Glass spheres 1 mm 10 22 1 10 22 1 10 22 1 1 19.9 1. 3 2.5 0.06 1.489 0.003 S.L. Glass spheres 0. 4 0.6 mm 533 1 533 1 533 1 1 21.5 1.3 2.4 0.08 1.500 0.006 S.L. Glass spheres 0. 212 0.300 m m 250 250 270 1 24.6 0.8 2.5 0.02 1.454 0.0007 S.L Glass spheres 0.150 0.212 m m 181 181 181 1 24.1 0.4 2.6 0.03 1.442 0.001 S.L. Glass spheres 0.0 38 0.054 mm 46 46 46 1 35.5 6.0 2.5 0.04 1.385 0.005 S.L. Crushed glass 0. 212 0. 300 mm 252 1 366 9 298 0.7 7 35.3 0.9 2.5 0.0 9 1.135 0.003
132 Figure 4 1 Schematic of Experimental Apparatus
133 Figure 4 2 Map of particles arranged by particle shape, size and intrinsic density. Figure 4 3 Microscope images of 0.25 0.3 mm particles. (a) Glass spheres (b) Crushed glass (c) Aluminum oxide and (d) olivine
134 Figure 4 4 Tra nsient crater depth and wid th for 2.3 mm glass spheres by jets with velocities 35.36, 67.18, and 93.74 m/s. (a) Depth and (b) Width
135 Figure 4 5 Crater evolution for 2.3 mm glass spheres under jets with varied velocity (a) 35.36, (b) 67.18, and (c) 93.74 m/s.
136 Figure 4 6 General trajectories of 2.3 mm glass spheres cratered by a 93.74 m/s jet.
137 Figure 4 7 C rater depth evolution for glass spheres with various particle diameters and jet velocity 20.90 m/s. (a) Long time frames and (b) Short time frames
138 Figure 4 8 Crater width and volume over time for glass spheres with various diameters and jet velocity 20.90 m/s. (a) Width and (b) Volume
139 Figure 4 9 C rater depth evolution for small spherical particles with varied intrinsic density and jet velocity 20.90 m/s. (a) Long time frames and (b) Short time frames
140 Figure 4 10 C rater depth evolution for large spherical particles with varied intrinsic density and jet velocity 31.05 m/s. (a) Long time frames and (b) Short time frames
141 Figure 4 11 C rater depth evolution for stainless steel particles with equivalent sphere volume diameter of approximately 1 mm with spherical and cylindrical particles and jet velocity 67.18 m/s. The aspect ratio (AR) is defined as the length/diameter. (a) Long time frames and (b) Short time frames
142 Figure 4 12 C rater depth over long time frames for plastic particles with equivalent volume diameter of approximately 1 mm with spherical, cylindrical a nd cubic particles and jet velocities of 31.05 m/s.
143 Figure 4 13 C rater evolution of 0.3 mm diameter polystyrene spheres by a 20.90 m/s jet.
144 Figure 4 14 C rater depth evolution of non spherical particles and spherical glass beads with mean sieve diameter s between 0.25 0.3 mm and intrinsic densities between 2.5 to 3.9 g/cm 3 by a 20.90 m/s jet (a) L ong time frames and (b) S hort time frames
145 Figure 4 15 Trajectory of particles m oving along the outer crater for spherical and non sp herical beds. (a) Glass spheres and (b) C rushed glass.
146 Figure 4 16 C rater depth evolution for glass spheres with diameters of 1.0 mm, 0.5 mm and a 50% by wt. mixture of 1 mm and 0.5 mm with jet velocity 20.90 m/s. (a) Long time frames and (b) Short time frames
147 Figure 4 17 Image of crater for 50% wt. mix of 1 mm and 0.5 mm glass spheres after the jet was turned off.
148 Figure 4 18 The asymptotic crater depth scaled to the original and modified densimetric Froude number s. (a) O f the original form (d p =d si ) and revised to incorporate particle shape (d p = d si ) and (b) M odified to incorporate particle shape with new definition of particle diameter ( d p = d E ).
149 Figure 4 19 Fitting of crater depth data (a) U nder asymptotic conditions and (b) P rior to asymptotic conditions.
150 Figure 4 20 Comparison of measured and predicted asymptotic depth from the arctangent fit scaled by the modified densimetric Froude number that accounts for particle shape.
151 CHAPTER 5 CRATERING UNDER LUNAR CONDITIONS Motivation It is essential that multiphase modeling tools be validated against erosion and cratering experiments under low gravity, vacuum and terrestrial conditions if they are to be used for predicting the rocket exhaust interaction on terrain with various possibilities of gravity and atmospheric pressures Recently, Metzger et al. (2009a) performed terrestrial cratering experiments with s ubsonic jets impacting a sand bed Chapter 3 used Metzger et al.'s (2009a) findings to validate a continuum based cratering model, in which t he effects of jet velocity, pipe diameter and height and gas typ e were investigated. Kuang et al. (2013) modeled t he experiments in C hapter 4 performed with large spherical particles in a CFD DEM framework. Investigating the reaction of lunar soil simulants, like JSC 1A and Chenobi, to an impinging gas jet provides important data that can be used for validating physi cal models with relevant materials and for information useful for estimating the reaction of lunar soil to a rocket exhaust plume. Background The crater depth and width over time are useful measurements for validating multiphase flow models. Environment a nd jet conditions as well as particle and bulk material properties such as permeability, particle size, density and shape affect the response of the granular material when impacted by a jet ( Haehnel et al., 2008; LaMarche et al., 2012; Metzger et al., 2009 a; Metzger et al., 2009b; Rajaratnam and Beltaos, 1977 ). I ncreased friction of non spherical particles in the bed actually in creases the crater depth growth and decrea ses the width growth due to reduced recirculation of
152 particles along the outer crater. Metzger et al. (2009b) demonstrated an increase in the crater depth with decreasing particle size. Metzger et al. (2009b) and Immer and Metzger (2010) showed that the crater growth increases with decreased gravity. Metzger et al. (2009a) performed a parametric study investigating the effect of jet properties on crater growth of sand beds and de scribed the dependence in detail. Over time the crater approaches, or reaches, an equilibrium or steady state size, which is often referred to as the asymptotic size (Rajaratnam a nd Beltaos, 1977; Sarma, 1967). Rajaratnam and Beltaos (1977) found that the asymptotic crater depth, D asym scales linearly with the e rosion parameter, E c as shown in equation (5 1) (5 1) w here H is the height of the nozzle above the bed and m 1 and m 2 are the slope and intercept, respectively. The erosion parameter proposed by Rajaratnam and Beltaos (1977) and used by many others ( Aderibigbe and Rajaratnam, 1996; Haehnel et al., 2008; Rajar atnam, 1982), is based on the densimetric Froude number, Fo, and is provided in equation (5 2) below. (5 2) where v is the jet velocity, g is the acceleration of gravity, d pipe is the pipe, or nozzle, inner diameter, d p is the particle diameter, and "! is the difference in the fluid density f and the particle intrinsic density, p Rajaratnam and Beltaos used the mean sieve diameter, d p = d si for particle diameter. Chapter 4 illustrates that the effect of particle shap e must be considered in the densimetric Froude number in order to account for D as y m H = m 1 E c + m 2 E c = v g f # $ % & ( d p d pi pe H = F o d pi pe H
153 particle asphericity Additionally, Chapter 4 shows that multiplying the diameter by the sphericity of the particle and using the erosion diameter, d E as a measure of the partic le size, equation (5 3) produces the highest correlation between the asymptotic crater depth and adjusted erosion number. E c = v g f # $ % & ( ) d E d pi pe H = F o ) d pi pe H (5 3) where is the spheric ity. For the irregularly shaped particles described here the erosion diameter is taken as the average of the Feret's width and length. Metzger et al. (2009a) identified that when VE is the dominating mechanism the crater forms a dual crater shape composed of an inner parabolic crater and an outer conical crater. They found that the inner crater size is relatively constant with time, but the outer crater continues to grow until a steady size is reached. When VE is the dominating mechanism, p articles are ejected from the crater by eroding along and off the lip of t he inner crater. Particles then deposit on the outer crater, on the top of the bed, or outside of the cratering box. As particles land on the outer crater the angle increases until failure, causing an avalanche of particles flowing from the outer crater to the inner crater. The outer crater stabilizes when it relaxes to the angle of repose of the granular material. The particles that are not ejected completely from the crater recirculate between the outer and inner crater as the outer crater walls oscillate between the angle of repose and angle of failure of the granular material (Haehnel et al., 2008; Metzger et al., 2009a). So far it has only been shown under certain circumstances that particles do not exhibit the dual crater shape behavior: when the cohes ive forces are high (LaMarche et al., 2012) or when the crater growth is
154 dominated by a cratering mechanism other than VE (Metzger et al., 2009a; Metzger et al., 2009b). Metzger et al. (2009a) provides a detailed description of previous work done on verti cal subsonic jet impingement on particle beds. Metzger et al. (2009a) identified that the avalanching of the outer crater into the inner crater to cause macroscopic fluctuations on the crater depth over time. In Chapter 4 it is shown that the sizes of the se fluctuations va ry with the particle properties. F or experiments with small polystyrene spheres, there is a sharp increase in the crater depth growth directly before a fluctuation in crater depth. The increased crater growth rate is attributed to the inc reased angle of the outer crater reducing the radial expansion of the jet resulting in increased momentum to the tip of the jet. This work presents two types of cratering experiments: s pherical particles in simulated variable gravity environments and lunar soil simulants in an Earth based setting F or ease in comparison with numerical models variable gravity experiments were performed on s pherical particles with relatively narrow particle size distributions To date, most granular flow theories are derived based on an assumption of spherical particles. The variable gravity experiments described in this work were designed to probe the various cratering mechanisms. The first set of experiments was performed on a low specific gravity material as it was antici pated with this lower specific gravity material, at higher jet velocities and reduced gravity conditions, deep shearing would be induced within the material and the BCF or DDS mechanisms would dominate. As the particle density is increased, the jet velocit y is decreased or the gravitational acceleration is increased, the e ffect of the BCF and DDS mechanisms are expected to decrease and
155 the VE mechanism will begin to dominate. Hence, the second set of experiments was performed on a higher specific gravity m aterial Here, a rigorous examination the effects of particle size distribution content of fine particles, and particle shape of the lunar soil simulant s JSC 1A and Chenobi on cratering is provided The cratering behavior of JSC 1A is compared to a sie ved fraction of JSC 1A and spherical glass particles that approximately match the particle size distribution of the sieved fraction of JSC 1A, as well as a narrow size distribution of glass spheres with mean diameter that coincides with the mean diameter o f sieved JSC 1A. Due to the cost of performing experiments with realistic lunar conditions and increased modeling complexities (i.e. vacuum, supersonic jet, lunar gravity), lunar regolith cratering experiments were performed in Earth's gravity, in standard atmosphere and with nitrogen and air jets. By reducing the complexity of the experimental conditions, the results presented here can be used to validate physical models. This work provid es the first variable gravity, jet induced cratering experiments to operate u nder conditions that attempted to span the VE, BCF and DDS mechanisms. Additionally, t hese experiments are the first to measure jet velocity, gas pressure and temperature during the cratering process and use spherical particles with narrow particle size distributions. This is also the first work to compare the cratering growth rate of multiple lunar soil simulants and to investigate the effects of lunar simulant particle size dist ribution and shape. Finally, it is the first to comprehensively test and report qualitative and quantitative results of the cratering of particle beds in variable gravity and beds of lunar soil simulants.
156 Experimental Methods HOOSH Based E xperiments The H andheld Observation Of Scour Holes (HOOSH) is a fully enclosed subsonic cratering apparatus, adjusted from the form used in past to measure ambient temperature, pressure and jet velocity for variable gravity cratering experiments ( Immer and Metzger, 2010; Metzger et al., 2009b). The apparatus consisted of a polycarbonate box that contained the experiment including a well, which was partially filled with particles. A centrifugal fan re circulated the air inside of the HOOSH T he fan pulled the air from the main experimental section and pushed it through a flow meter into a pipe containing a flow straightener The pipe was situated directly above a sharp, outwardly beveled edge. Thus, the air wa s split in half by the edge, and the jet caused a crater to form in the particle bed. The front of the apparatus was transparent so that the half crater development could be observed and recorded. The inner diameter of the pipe was 0.9525 cm. The crater observation area that contained the particles is 12.4 cm deep, 13.5 cm long and 26.7 cm wide. Figure 5 1 is a diagram of the HOOSH experimental apparatus. Experiments were performed at lunar gravity (1/6 gee) and 2 times Earth's gravity (2 gee) in the NASA's Facilitated Access to the Space environment for Technology ( FAST ) flight in which variable gravity is simulated. The pressure of the air on the plane was about 12 psia and the temperature varied between 20 and 27 degrees C. Experiments were performed in the HOOSH at sea level to test the cratering of the particle beds in Earth's gravity (1 gee). T he pressure of the air for the E arth gravity experiments was about 14.76 psia and the temperature was between 25 and 29 degrees C. The density and viscosity of air at conditions for the experiments were used
157 to calculate the Re ynolds numbers (Re) of the jet. The ground based experiments were compared to experiments done on the flight by matching the jets' Reynolds number. The particle size ranges of the two sets of variable gravity experiments were chosen so that the arithmetic mean diameters would approximately coincide with the arithmetic mean volume diameter of JSC 1A. Unfortunately, the surface forces of the two particles types are different, and may have had marginal effects on the cratering of lower density material in low gravity. A bed of spherical polystyrene particles was used for the first flight and a bed of spherical glass particles for the second. Both particle beds had a particle size range of 150 to 360 microns. The system was equipped with an Alicat M 500SLPM D f low meter with accuracy within 1% F.S. and range of 0 500 Standard Liters Per Minute (SLPM), and could also measure the temperature and pressure of the flow. A centrifugal fan pushed the air through the flow meter to a pipe that released the gas as a jet directed vertically downward towards the granular material. Changing the voltage delivered to the fan controlled the flow rate of the jet, which had a range of 12.7 to 68.6 SLPM (jet velocities 2.87 m/s to 16 m/s). The atmosphere inside of the flight was s lightly reduced compared to sea level, so the jet Reynolds number was used to compare the crater growth between experiments performed in the HOOSH at variable gravities. The Allied Vision Technologies (ATV) Prosilica CV 1280 monochrome camera with resolut ion 1280x1024 pixels and 8 frames per second was pointed directly at the viewing window and recorded the crater development over time. Flashlights were used to illuminate the top surface of the particle bed. The crater surface was identified by the contras t of the illuminated top of the bed with the dark sidewalls. The eye could
158 generally see the surface of the crater and it was manually tracked. For the sake of time, only 1 of every 5 frames was analyzed and the entire crater was mapped. The video camera c aptures 8 frames per second thus the analysis contains data for the crater every 0.625 sec. Each constant gravity section of the NASA FAST flight was taken as a separate period of cratering, therefore each video of a cratering event lasted about 25 sec. L unar Regolith Experiments Cratering experiments were performed in an experimental setup described by Metzger et al. (2009a) and Chapter 4 The experiments were performed in a setup similar to that of the HOOSH. A schematic of the experimental apparatus used is presented in Figure 5 2 Gas was released from a pressurized bottle of breathing air and the pressure of the released gas was regulated to control the flo w rate. After exiting the pressure regulator, the gas flowed through the Alicat 0 500 SLPM flow meter, then through a straight section of pipe approximately 1.73 m long, which was directed downward at the bed of material an d had an inner diameter of 0.774 7 cm. The pipe exit was situated so that a clear window, with an outwardly beveled edge, would bisect the developed jet. The idea was that jet would be split and that half of the jet would expand on either side of the window, allowing for subsurface crater development to be monitored in real time and recorded by a video camera. A MATLAB surface tracking program was used to analyze the crater growth. The camera used for this study was an Edmond Optics high dynamic range camera (EO 0418) with a frame rate of 4 8.02 frames per second and a pixel size of 10x10 microns, pixel area of 768x576 and pixel depth of 8 bits. A plate was placed on top of the bed to block the flow of the jet while the
159 pressure was regulated to achieve the desired flow rate, then the plate w as quickly removed. Ex periments were performed on l unar soil simulants Chenobi and JSC 1A. The lunar regolith and lunar regolith simulants have a large number of fine particles, which have a direct effect on the mechanical properties of the material (LaMa rche et al., 2012). By removing the fine particles, the effect of particle shape can be investigated. Experiments were performed on JSC 1A in its naturally sieved state as well as a sieved fraction from 150 425 microns, referred to as sieved JSC 1A. Glass spheres wide distribution is a mixture of sph eres to approximate the particle size distribution of sieved JSC 1A. Standard sieve fractions of soda lime glass spheres from Mo Sci Corporation between 150 and 425 microns were mixed together in order to attain the desired distribution. A narrow particle size distribution of 212 300 micron glass spheres (glass spheres narrow distribution) is compared to the sieved JSC 1A, as the mean diameter s are approximately equal. Table 5 1 provides the arithmetic mean parti cle size, angle of repose and intrinsic density for the various materials used in this manuscript. The angle of repose was measured using the same method described in the previous chapter A rectangular container was filled with material and then a small adjustable sized slit in the bottom of the co ntainer was opened to allow a slow pour of material After the container emptied, the angle of repose was measured from of the piles that remained in the container. The particle intrinsic density was measured us ing the water displacement method or with the Quantachrome Ultrapyc 1000 Gas Pycnometer Chenobi has a wide particle size distribution. A set of experiments was performed on Chenobi with the full particle size distribution, but the large particles
160 caused local bulk density variations and made it difficult to perform experiments with repeatable results. Particles larger than 2.36 mm were removed from the Chenobi using a standard sieve. The distribution containing particles less than 2.36 mm is referr ed to as Chenobi, and the full particle size distribution as ChenobiC. Figure 5 3 compares the equivalent volume scatter sphere diameter particle size distributions of Chenobi and JSC 1A as measured by the Coulter LS 13320 with (a) the frequency distributi on, f v and (b) cumulative undersize fraction, F v Microscope pictures of JSC 1 A, Chenobi, and glass spheres wide distribution are provided in Figure 5 4 The images in Figure 5 4 provide a picture of the angularity and roughness of various particles. JSC 1A Bulk Density Cratering E xperiments Experiments with JSC 1A with variable bulk density beds were performed in a setup similar to that in Figure 5 2 Only a different flow meter, camera and slightly larger box were used. The flow rate was measured with a sierra TopTrak 820S H 0 500 SLPM flow meter with 1.5% F.S. accuracy. The ATV Prosilica CV 1280 with 8 frames per second was used to record the crater growth over time. A plate was placed on top of the bed to block the flow of the jet while the pressure w as regulated to achieve the desired flow rate T hen a solenoid valve after the regulator was closed, stop ping the flow. Next, the plate was removed and the solenoid valve rapidly opened. This method of turning the jet on is slightly different than was des cribed in the previous section, as the effect of the moving plate is removed. However, the flow rate was problematic to control with the use of the solenoid valve making it difficult to repeat experiments with the same flow rate. The JSC 1A was poured int o the cratering box with a funnel. For compacted beds of JSC 1A a large overburden weight was placed on top of the bed, which was much heavier than the bed of material, and the bed was vibrated The JSC 1A self compacts
161 so some variability in the bulk den sity was attained just from pouring the material. The bulk density of the JSC 1A was found inside of the cratering experiment by measuring the volume and mass. Results and Discussion HOOSH Variable Gravity Experiments The results found here for spherical particles are similar to those found by Metzger et al. (2009b) for JSC Mars 1A crater growth in variable gravity. Figure 5 5 provides the crater depth growth over time for glass beads in Earth gravity, 2 gee and lunar gravity with jet Reynolds number in th e range of 3200 to 3800. T he crater growth rate decreases with increased gravit y. Additionally, t he time t o reach asymptotic depth reduces with increased gravity. In Earth's gravity both polystyrene and glass exhibit the dual crater signifying VE dominat es the crater growth (Metzger et al., 2009a). However, in lunar gravity, polystyrene spheres do not typically form the dual crater shape. Instead, one long crater forms, which is similar to the crater shape described by and Metzger et al. (2009b) indicatin g DDS or BCF dominates. Figure 5 6 compares images of craters for polystyrene and glass at various gravities. In Figure 5 6 the crater shape for polystyrene spheres in (a) lunar gravity with jet Re 6600 is compared to (b) Earth gravity with jet Re 6500. It was difficult to match flow rates exactly with the HOOSH setup but small differences in Re do not affect the crater shape. T he crater shape for glass is similar in lunar gravity (c) and Earth gravity (d), expressing the dual self similar crater shape. Ho wever, glass in 2 gee (e) exhibit particles eroding over the surface and only a shallow depression of a crater forms.
162 VE is likely the mechanism for cratering of glass part icles in 2 gee (Figure 5 6 (e)) as only a thin layer of particles were observed ero ding and the shape of the crater does not indicate any of the other cratering mechanisms. Similarly, f or the velocity range studied polystyrene and glass in Earth gravity and glass in lunar gravity demonstrate cratering behavior of the dual crater shape that Metzger et al. (2009a) identified VE t o dominate It is possible that another cratering mechanism may dominate for glass beds initially struck by higher velocity jets in lunar gravity, since a long narrow crater forms that abruptly splits into an inne r and outer crater region. The cratering behavior of polystyrene in lunar gravity, Figure 5 6 (b), is more similar to behavior Metzger et al. (2009a) identified to occur when BCF or DDS dominate. More investigation is needed to understand why polystyrene only forms a single crater in lunar gravity. The increase in the ratio of particle surface to volume forces, such as van der Waals to gravity, in reduced gravity can lead to cohesion in the particle bed. Increased cohesion of the polystyrene bed in lunar g ravity could change the cratering behavior. However, the behavior of cohesive particle beds, described in detail below, differs from that of polystyrene in lunar gravity Indicating that cohesion is not the dominating reason for the atypical behavior of po lystyrene in lunar gravity. The movement of individual particles cannot be seen on the videos so it is difficult to identify whether VE, BCF or DDS cratering mechanisms dominate The top edges of the polystyrene crater walls appear to be fluidized by the jet. The bed of polystyrene in lunar gravity seems loose, a deep crater forms and the ballooning effe ct of the crater suggests that DDS could be the dominating cratering mechanism as opposed to BCF. In BCF, the jet mechanically pushes the bed downward, f orming a narrow cup (Metzger
163 et al., 2009a). The mechanical force responsible for BCF is expected to consolidate the bed in the region of the jet, or at the very least not dilate, which is contrary to observations for polystyrene in lunar gravity. N umerica l simulations can provide detailed analysis on the flow field, ratio of forces, and subsurface particle trajectories in order to confirm which mechanism dominate in various gravities Comparing the crater growth for polystyrene at two different Reynolds n umber s at lunar and Earth gravity, in Figure 5 7 it is clear that the crater growth is more sensitive to jet Reynolds number at reduced gravity. Additionally, t he crater depth growth rate and the asymptotic depth are much greater in lunar gravity than in Earth gravity. Figure 5 8 exhibits the effect of jet Reynolds number and particle density on cratering in lunar gravity. The asymptotic depth and crater depth growth rates for the lower intrinsic density material are much greater than for the increased den sity material. Figure 5 9 shows that t he widt h of the inner and outer crater, when a dual crater shape exists is dependent on gravity Both the inner and outer crater width for glass spheres decreas es with increased gravity. As the gravity force reduces, there is less resistance against the jet and therefore, less shear stress from the jet is needed to support the particles at an angle above the angle of repose. The inner crater width of glass is inde pendent of jet velocity in lunar gravity, but this could be du e to the relatively small range of jet velocities tested Cratering of JSC 1A: Effect of Bulk Density and Jet Velocity Similar to lunar soil, the JSC 1A is very compressible. Cratering experime nts were performed on JSC 1A with beds consisting of a range of bulk densities, which correlate to the bulk densities measured in the top 10 cm of regolith on the moon by Mitchell et al. (1972) and vary between 1190 kg/m 3 and 1330 kg/m 3 Figure 5 10
164 expres ses the complicated dependence of crater depth growth for JSC 1A on both bulk density and jet velocity. The fine particles in JSC 1A made analyzing the experiments very difficult. The fine particles blew out of the soil and reflected the light in front of the camera obscuring the view of the crater. The crater growth of JSC 1A for various bulk densities and jet velocities in Figure 5 10 is provided for the first second of cratering due to the fine particles masking the view of the crater after long period s. Figure 5 10 is consistent with the findings of Metzger et al. (2009a) and Rajaratnam and Beltaos (1977) that for beds of JSC 1A with the same bulk density, the crater growth rate increases with increased jet velocity However, in the case of a compress ible bed the early crater depth growth is not valid for compari ng jet velocities when there is a significant change in the bulk density of the bed. For the same jet velocity, t he higher bulk density (more compressed) bed exhibits a lower cratering rate th an the bed with a lower bulk density (less compressed) Surprisingly, the more compressed bed cratered with the higher velocity jet presents a smaller cratering rate than the less compressed bed under the reduced jet velocity even with a 10 m/s difference in jet velocities With increased bulk density Alshibli and Hasan (2009) found the friction angle and cohesion to increase for JSC 1A, which indicates increased resistance to flow. Metzger et al. (2009b) establish ed that the asymptotic crater depth was in dependent of the bulk density of the bed. However, the results of Figure 5 10 express that for short times the bulk density effects dominate over the jet velocity effects.
165 Segregation of JSC 1A During Cratering JSC 1A is a brownish dark gray color. However after sieving into various size fractions, the finer portions of JSC 1A are brownish in color and courser divisions are more of a dark gray color. The distinction in relative size of JSC 1A by color allowed the observation of JSC 1A segregation during cr atering. A top down view of a crater formed in a bed of a JSC 1A after the jet was extinguished is presented in Figure 5 11 and illustrates the segregation of JSC 1A particles by size on the top of the bed and at the bottom of the crater. In Figure 5 11 the dark gray/black areas indicate regions dominated by the larger diameter fractions of JSC 1A. The particles that are eroded from the crater are entrained in the flow and either deposit on the top of the bed, are ejected from the box, or recirculate ins ide of the crater. Figure 5 11 shows dark region s in the bottom of the crater and in a ring shaped area surrounding the top edge of the crater The dark region at the bottom of the crater is the pile of course particles that settle after the jet is turned off, and indicates that some larger particles in JSC 1A beco me trapped inside of the crater. L arger particles were observed popping around inside of the crater after it reached a certain size, as they could not escape However, finer particles are able to escape the crater and are carried by the gas further away from the impingement region Therefore, the difference in distance traversed by the various sized particles away from the crater results in the dark gray ring around the crater edge Chenobi and Che nobiC experience similar segregation. Cratering Rates of Sieved JSC 1A versus Glass Spheres Cratering experiments with JSC 1A, sieved JSC 1A, glass spheres wide distribution, and glass spheres narrow distribution were performed with various velocity
166 air je ts and allowed for investigation of fine particles, particle shape, and particle size distribution. Removing the fine particles and cratering sieved JSC 1A made it possible to analyze the experiments for long periods of time and isolate the effects of part icle shape by comparing to glass spheres wide distribution. Comparing the crater growth of glass spheres wide distribution to glass spheres narrow distribution investigates of the effects of particle size distribution. Figure 5 12 shows the crater growth f or JSC 1A, sieved JSC 1A, glass spheres wide di stribution, and glass spheres narrow distribution for an air jet with velocity 25.46 m/s for (a) long and (b) short times. Both sieved JSC 1A and glass spheres crater logarithmically, or arctangent, with time which is typical for most granular materials with this particle size, density and jet velocity range exhibited in Figure 5 12 (Metzger et al., 2009a; Rajaratnam and Beltaos, 1977 ; Sarma, 1967 ). The crater growth for all materials in Figure 5 12 follow the general trend mentioned earlier, that the crater growth increases with increasing jet velocity. For all materials except the lunar soil simulants, when the jet is extinguished, the crater depth slightly reduces as the inner crater filled in. Al though intrinsic density affects the cratering rate, in this case, the densities are relatively similar. The differences in the particle surface roughness and angularity dominates the cratering behavior between the sieved JSC 1A and glass spheres as Chapte r 4 shows the crater growth is dominated by particle shape effects even with a larger difference in intrinsic densities Initially, the crater growth rate of sieved JSC 1A is very similar to both glass spheres distribution s, Figure 5 12 (b), which is consi stent with previous results T he effects of particle shape are not experienced in the crater depth growth until after the formation of the outer crater, which occurs for sieved JSC 1A after approximately 2 seconds.
167 T he JSC 1A craters to a larger depth than the sieved JSC 1A. However, t he sieved JSC 1A initially crater s faster than JSC 1A (Figure 5 12 (b)), due to the cohesiveness of the JSC 1A. A balance exists between decreasing particle size, which increases the crater growth rate, and increasing inter pa rticle forces, like cohesion, which deters erosion JSC 1A has a large percentage of particles that are smaller than sieved JSC 1A. The tradeoff between particle size and cohesion effects between JSC 1A and sieved JSC 1A is expected to cause the crater dep th of JSC 1A over long times to be only slightly greater than that of sieved JSC 1A. The crater width growth for glass spheres narrow distribution is the greatest with time, followed by glass spheres wide distribution, sieved JSC 1A and finally JSC 1A, Fig ure 5 12 (c). The effect of particle size distribution between glass spheres wide distribution and glass spheres narrow distribution has negligible effects in the crater depth for the jet velocity in Figure 5 12 H owever the effects are measured in the cr ater width. The crater width is directly related to the angle of repose of the granular material, as found in Table 5 1 It is expected that for a wider particle size distribution, there would be a measureable effect on the crater depth as shown in Chapter 4 Crater growth for Lunar Soil Simulants Chenobi and JSC 1A Figure 5 13 provides the crater depth growth over time of Chenobi and JSC 1A under air jets with velocities of 25.46 and 20.90 m/s. The Chenobi craters to a smaller size over ti me than the JSC 1A. Additionally, Chenobi has a larger percentage of fine particles than JSC 1A, indicating Chenobi is more cohesive The decreased cratering rate of Chenobi compared to JSC 1A could be due to increased cohesive forces or friction in the Chenobi causing more resistance to the jet. Also, the Chenobi even with
168 removing particles larger than 2.36 mm contains more course particles ( Figure 5 3 ) than JSC 1A, which reduc e the long term crater size. T he crater growth of Chenobi is more sensiti ve to increases i n jet velocity than JSC 1A. The large particles in ChenobiC, and to a lesser extent in Chenobi, may cause bulk density variations in the bed resulting in inconsistencies in the measured crater depth over time. Care was taken for all experiments to prepare the bed in the same method in an attempt to keep the bulk density of the bed constant. Figure 5 14 illustrates the crater growth of a single experiment with Chenobi under a jet of air with velocity of 25.46 m/s. Oscillati ons in the crater depth occur as ch unks of material erode out of the side of the crater and fall to the bottom reducing the crater depth. The numbered stars in Figure 5 14 on the crater depth plot correlate to the inlayed frames of the crater growth Star 1 shows the crater at its, locally deepest just before the crater depth drops in size. A large chunk of material falls from the left crater wall to the bottom of the crater in the transition from frame s 1 to 3 causing a decrease in the crater depth. Finally, t he material at the bottom of the crater erodes away frame 3, and the crater depth begins to increase again frame 4 Contrary to what Chapter 4 shows for polystyrene, the crater depth growth does not accelerate before material avalanches into the crater JSC 1A and Chenobi do not fo rm outer craters that oscillated between the angle of failure and angle of repose. I mplying that for Chenobi and similarly for JSC 1A, there is no change in the outer crater that results in increased momentum to the jet tip, which produces the accelerated growth for polystyrene Similar behavior is experienced by JSC 1A under all jet velocities studied.
169 When cohesive beds are cratered they initially exhibit the formation of a long skinny crater, similar to the observations of polystyrene at lunar gravity, Figure 5 6 (a). However, for cohesive beds, such as JSC 1A and Chenobi, the crater slowly grows wider as chunks of material erode off of the crater sidewall and in extreme instances making an upside down mushroom type shape. Polystyrene in lunar gravity forms a crater that looks more like an upside down balloon below the surface. Instead of eating away at the crater sidewalls, the polystyrene bed deforms to the shape of the jet, as can be seen by the curvature of the right side of the crater in Figure 5 6 (a). The difference in the cratering, erosion in chunks and subsurface mushroom type behavior indicates that cohesion is not dominating the cratering behavior of polystyrene in lunar gravity. Crater Shape Variations Figure 5 15 compares the crater shape for Chenobi (a) JSC 1A (b) glass spheres wide distribution (c), and sieved JSC 1A (d) after 50 seconds of cratering by a 25.46 m/s air jet. Chenobi and JSC 1A form deep scour hole shaped craters and are quite different than th e crater of sieved JSC 1A and glass spheres. T he sieved JSC 1A c rater shape is similar to that of glass spheres both of which exhibit the dual crater shape. T he crater growth of sieved JSC 1A is more similar to that of glass spheres than to JSC 1A. The sh ape of the crater for ChenobiC is similar to what is found for Chenobi in Figure 5 15 T he difference in the crater shape between sieved JSC 1A and JSC 1A further indicates that the cohesion and increased friction caused by the fine particles present in Ch enobi and JSC 1A result in differences in the dynamics of the crater shape The small glass spheres in Chapter 4 experience cohesion due to van der Waals, capillary and possibly electrostatic forces. However, Chenobi and JSC 1A additionally experience cohesion due to mechanical interlocking.
170 Some a dditional observations of cratering behaviors are not reflected in Figure 5 15 As particles of sieved JSC 1A and glass spheres depos it on the outer crater, the angle increases until failure causing an avalanche to occur Next, the outer crater relaxes to the angle of repose of the soil. Conversely, as JSC 1A and Chenobi erode in chunks they do not have an eroding layer of particles t hat can be visualized. Figure 5 16 displays images of the crater that remained after terminating the jet for (a) Chenobi, (b) JSC 1A, (c) glass spheres, an d (d) sieved JSC 1A. Chenobi and JSC 1A maintain the same overall crater shape after the jet is exti nguished ; t he larger particles recirculat ing inside of the crater settle but the steep sidewalls remain Yet, the parabolic inner crater of glass spheres and sieved JSC 1A, in Figure s 5 15 (c) and (d), is a result of gas traction supporting the sides of t he inner crater (Metzger et al., 2009a). I n the absence of this traction, the inner parabolic crater disappears, Figures 5 16 (c) and (d), leaving a conical shaped crater with sidewall s approximately equal to the material's angle of repose. The differences in the behavior of JSC 1A and sieved JSC 1A after the jet is turned off indicate that cohesion of Chenobi and JSC 1A beds, caused by the large number of fine particles, allows the crater to maintain steep sidewalls without the presence of the gas shear st ress from the jet. Scaling the Asymptotic Crater Depth Until now, the effect of gravity on scaling the asymptotic crater depth equations (5 2) and (5 3) was not confirmed. The experiments performed in lunar gravity did not reach the asy mptotic crater de pth. However, that the crater depth growth over tim e is well extrapolated by an arctangent fit, provided in equation (5 4) (5 4) D c = a 1 t a n 1 b 1 t c 1 ( ) ( )
171 where D c is the crater depth, t is the time, and a 1 b 1 and c 1 are fitting constants. After fitting the crater depth over time, the asymptotic crater depth can be estimated as 0.5* *a 1 Previous results from Chapter 4 show that there is good agreement between the measured and the predicted asymptotic crater depth by equation (5 4) The asymptotic crater depths for the experiments performed in lunar gravity were approximated using equation (5 4) and measured for lunar soil simulant materials. The asymptotic crater depths are compared to the measured type A depths, previously shown in Chapter 4, and expressed in Figu re 5 17 (a) and (b). Figure 5 17 (a) shows that the asymptotic crater depth scales with equation (5 2) which uses the mean sieve d iameter for the particle size and has been used widely (Aderibigbe and Rajaratnam, 1996 ; Haehnel et al., 2008; Rajaratnam and Be ltaos, 1977; Rajaratnam, 1982). The modified relationship proposed in Chapter 4 repeated here as equation (5 3) utilize s the sphericity and erosion diameter in scaling the asymptotic crater depth and is provided in Figure 5 17 (b). The form of the erosio n parameter, found by comparing Figures 5 17 (a) and (b) does not make a difference for the lunar gravity data, as the particles are spherical. However Figure 5 17 (b) expresses the modified form of the erosion parameter and, therefore, densimetric Froud e number, in equation (5 3) should be used for non spherical particles as better scaling is produced. The mean sieve diameter was not foun d for the lunar soil simulants, hence t he asymptotic depth s for these beds are only included in Figure 5 17 (b). The asymptotic crater depth for lunar soil simulants is offset from the asymptotic crater depths of non cohesive beds The mean erosion diameters of JSC 1A and Chenobi are small
172 compared to the other particles studied, Table 5 1 which cause high values for th e erosion parameter as calculated by equation (5 3) However, inter particle forces start to dominate as the particle size decreases, reducing the asymptotic crater depth The results presented in Figure 5 17 (b) indicate that equation (5 3) does a good jo b scaling the effects of asymp totic crater depth with gravity and particle shape, but that cohesive forces must be accounted for to avoid overestimating the asymptotic crater depth A nother adjustment to the densimetric Froude number proposed in Chapter 4, which accounts for cohesion in the particle bed, Fo and is provided in equation (5 5) (5 5) where / 1 and / 2 are material specific constants related to the cohesiveness. For all JSC 1A experiments, regardless of jet species, / 1 =2.38 and / 2 =2.2 provided suitable correction to the densimetric Froude number, so the constants in equation (5 1) (m 1 and m 2 ) are consistent with the non cohesive materials. Summary The first small scale v ariable gravity cratering experiments with a full specification of operating conditions jet velocity, pressure and temperature were successfully performed with two different particle beds and provide val idation data for software packages on a quantita tive level and at conditions more closely approaching those in lunar environments. The results of the lower specific gravity material in lunar gravity seem to indicate a mechanism other than VE to dominate. The effects of gravity and granular material re presentative to what is experienced on the moon have been investigated with respect to the interaction of an F o = 1 1 v g # $ $ % & ( ) d f + 2
173 impinging turbulent jet. The experiments performed here indicate that crater depth decreases with increased gravity, crater growth is more sensiti ve to jet velocity in increased gravity, lower density materials are more sensitive to increases in jet velocity in reduced gravity, and the inner and outer crater width increases in reduced gravity. The bulk density of the material can make a large impact on the transient crater growths and can even dominate the jet velocity effects. As the crater grows deeper and wider for Chenobi and JSC 1A, large chunks of material erode off of the walls which indicates that cohesive forces, caused by the fine particle s, significantly affect the crater growth behavior. Neither Chenobi nor JSC 1A shows bulk recirculation of particles along the outer crater due to these cohesive forces. Wide particle size distribution materials (lik e JSC 1A and Chenobi) segregate during t he cratering process, leaving large particles inside of the crater and around the outer edge of the crater while smaller particles deposit further away from the crater area. JSC 1A crater grows deep er over time than sieved JSC 1A, as the particles in sieve d JSC 1A re circulat e between the inner and outer craters and JSC 1A contains smaller particles. S ieved JSC 1A grows a deeper crater than the glass spheres with matching particle size distribution as the increased friction of the non spherical JSC 1A part icles reduces recirculation. The effects of particle size distribution prov e to be minimal in the size range of sieved JSC 1A The crater growth results for experiments performed in Earth's gravity indicate VE to be the dominating cratering mechanism. The crater formation of Chenobi is less repeatable than JSC 1A seemingly due to bulk density inconsistencies and self compaction that occurs with Chenobi Differences in the long and short term crater growth behavior between JSC 1A and
174 Chenobi beds are due t o differences in the particle size distributions Craters grow in JSC 1A and Chenobi much differently than what is found for sieved JSC 1A and glass spheres. The results of this work indicate the large number of fine particles and the non spherical shape o f the particles determines the qualitatively and quantitatively different behavior s
175 Table 5 1 Properties of P articles Particle Type Mean diameter ( m) d 10 ( m) d 50 ( m) d 90 ( m) (deg) # p (g/cm 3 ) d E ( m) HOOSH glass mix 257 159 239 301 24.5 2 2.50 0.02 257 1 HOOSH polystyrene mix 303 212 284 325 32.2 5.5 1.06 0.04 303 1 Chenobi (d v ) 305 15.1 147 895 63.2 8.8 2.71 0.02 62 0.71 JSC 1A (d v ) 268 32.3 187 632 53.8 5.4 2.85 0.04 66 0.75 Sieved JSC 1A ( d si ) 223 107 198 303 34.3 1.1 2.83 0.02 223 0.8 3 Glass Spheres wide distribution 230 121 192 335 25.6 0.6 2.50 0.02 230 1 Glass Spheres narrow distribution 25 4 211 256 297 24.6 0.8 2.50 0.02 25 4 1
176 Figure 5 1 Schematic of the HOOSH Figure 5 2 Schematic of Experimental Apparatus
177 Figure 5 3 Equivalent sphere volume diameter of Chenobi and JSC 1A. Inlayed is the distribution from 0 to 50 microns. (a) Frequency distribution and (b) C umulative undersize fraction.
178 Figure 5 4 Microscope images of particles. (a) JSC 1A, (b) Chenobi, and (c) glass spheres. Figure 5 5 Crater growth for glass spheres in various gravities.
179 Figure 5 6 Crater images of polystyrene and glass spheres in varied gravity. Polystyrene spheres in ( a) lunar gravity and (b) Earth gravity. G lass sph eres in (c) lunar gravity, (d) Earth gravity and (e) 2x E arth gravity. All images are to scale and taken from the same position. Figure 5 7 Influence of jet Reynolds number and gravity on crater evoluti on for polystyrene spheres.
180 Figure 5 8 The effect of jet velocity and particle density on crater growth in lunar gravity.
181 Figure 5 9 Effect of gravity on crater width for glass spheres
182 Figure 5 10 Crater growth for JSC 1A at various bulk densities and nitrogen jet velocities. Reprinted from LaMarche et al. (2012) with permission from ASCE.
183 Figure 5 11 Segregation of a bed of JSC 1A by particle size during cratering. Reprinted from LaMarche et al. (2012) with permission from ASC E.
184 Figure 5 12 Crater depth and width evolution of JSC 1A, sieved JSC 1A, and glass spheres under air jet s with velocity of 25.46 m/s. (a) Depth over long time frames, (b) Depth over s hort times and (c) Width over long time frames.
185 Figure 5 12 Continued Figure 5 13 Crater growth for Chenobi and JSC 1A under various velocity air jets
186 Figure 5 14 Crater evolution for Chenobi. Inlayed are images corresponding to the star r ed points on the crater depth All inlayed images are to scale and taken from the same position.
187 Figure 5 15 Crater shape after 50 seconds by a 25 .46 m/s air jet Images are to scale and taken from the same position. (a) Chenobi, (b) JSC 1A, ( c) G lass spheres wide distribution and (d) S ieved JSC 1A.
188 Figure 5 16 Crater sha pe after the jet was turned off. Images are to scale and taken from the same position. (a) Chenobi, (b) JSC 1A, ( c) G lass spheres wide distribution and (d) S ieved JSC 1A
189 Figure 5 17 The asymptotic crater d epth scaled to the original and modified densimetric Froude number s. (a) T he original form (d p =d si ) and (b) M odified to incorporate particle shape with new definition of particle diameter ( d p = d E ).
190 CHAPTER 6 CONCLUSIONS AND RECOMMENDATIONS This dissertation characterizes the cratering of a particle bed by a subsonic turbulent jet, proves the two fluid framework is able to simulate the cratering process, identifies key variables for future model development, and provides benchmark data that c an be used to validate those models. Permeability of JSC 1A The permeability of JSC 1A is measure d and reported in C hapter 2. JSC 1A is a granular material with a wide particle size distribution, highly non spherical particles, and a high content of fine particles. The Carmen Kozeny relationship, which is used to estimate the permeability of porous media, is shown to correctly predict the permeability of JSC 1A. Two Fluid Modeling This is the first work to use the two fluid model to simulate a jet vertica lly impinging downward onto a particle laden bed. It is discovered that the two fluid framework, with appropriate frictional stress constitutive modeling, is capable of predicting the qualitative and quantitative features of crater growth. Simulations show that the proposed frictional stress constitutive model is capable of predicting the salient features for cratering sand with various types of jets. The proposed frictional stress constitutive model increases the frictional viscosity relative to the fricti onal pressure. The predictive capabilities of various frictional stress constitutive models, including those from the literature, are compared. The two fluid model accurately predicts the effects of jet velocity, gas species, and nozzle height above the be d on crater growth.
191 For both low and high velocity turbulent, subsonic jets, the two fluid model correctly predicts the crater growth and cratering mechanisms. The model is not able to well predict the outer (secondary) crater which forms during the crate ring process, or, at least, the prediction of the outer crater angle is lower than the measured angle of repose. Several reasons are provided to explain this discrepancy. Further adjustment of the overall frictional stress magnitude and the ratio of the fr ictional viscosity to the frictional pressure would provide better agreement between crater depth predictions and measurements. Additionally, the frictional viscosity is inversely proportional to the second invariant of the strain rate tensor, indicating t hat the frictional viscosity is high in stationary regions of the particle phase. Along the outer crater, momentum transfer from the gas phase no longer supports particles as it does in the inner crater. Gravity is not able to draw the material down the an gle of repose, which would result in the formation of the outer crater, as the predicted viscosity is too high. A better prediction for the angle of repose associated with the outer crater would require decreasing the frictional viscosity in the empirical frictional stress model, or developing an improved frictional stress model. Further, the ability of a frictional model to predict the angle of repose can be studied under a simplified case, such as pile formation, which would reduce the computation expense It is recommended that a new frictional stress model be developed from first principles. One recommended route using DEM simulations is discussed later. The Reynolds stress model is superior due to its ability to predict high strain experienced in the s tagnation region. It is possible that the Algebraic Reynolds stress
192 model would reduce computational intensity while maintaining accurate prediction of complex high strain flows; this remains to be explored. Parametric Study of Particle Properties on Crater Growth Chapter 4 provides comprehensive benchmark data for cratering, which are useful for validating both DEM and two fluid models. The effects of particle density, size and shape are investigated individually. Increased particle size and density r educe the asymptotic crater depth and the crater growth rate. The increased frictional stress associated with non spherical particles enhances the asymptotic and transient crater depth growth due to reduced recirculation of particles along the outer crater while reducing the crater width. Investigations of particle size distribution demonstrate that segregation by particle size occurs during the cratering process. Smaller particles are ejected further from the crater, leaving the larger particles to recirc ulate. A new form of the densimetric Froude number, a scaling factor relating the asymptotic crater depth to particle and jet properties, is developed that accounts for particle shape. Additionally, a new particle diameter, the erosion diameter, is incorpo rated within the new densimetric Froude number and is a measure of the dimension in which a particle would align in shear flow. Cohesive inter particle forces reduce the asymptotic crater depth and are not captured by the densimetric Froude number, for whi ch a correction is proposed. Additionally, two new cratering mechanisms are discovered. The first is when deep subsurface particle induced shearing occurs, while the second is associated with the case of particle particle collisions occurring inside the in ner crater. These particle particle collisions affect erosion rates by interfering with particles escaping the crater, whereas particle surface collisions cause particle ejection.
193 The effects of particle size distribution are not well understood, and expe riments should be performed mixing particles of several different sizes in various ratios. Further, cohesion effects must be accounted for in the densimetric Froude number. Future work with well described cohesive materials would confirm the proposed corre ction. Additional experiments with cylinders with varied aspect ratio are also recommended. Crater depth growth over time was extrapolated using a fitting function to predict the asymptotic crater depth. Investigating the relationship of the crater depth f itting function's parameters to jet and particle properties would provide a method of estimating the asymptotic and transient crater depth without any experimentation. DEM simulations of spherical and non spherical particles in a simple 3 D periodic shear flow (dry granular flow particles only), varying the solids volume fraction and friction coefficient, is a recommended method of improved frictional stress constitutive models, as well as particle phase stress models that account for particle asphericity Two fluid and DEM modeling, under the cratering conditions in the experiments, can be used to validate constitutive models and fluid particle interaction models. Experimental Study of Crater Growth in Lunar Conditions Chapter 5 presents results of the f irst cratering experiments performed under relevant lunar operating conditions and materials, specifically of lunar soil simulants and in variable gravity. Cratering of two lunar soil simulants, JSC 1A and Chenobi, reveal that wide particle size distributi on materials experience segregation by size during cratering, even with cohesive effects present due to the large percentage of fine particles. The fine (cohesive) particles in lunar soil simulants cause the material to erode in chunks and cause a single d eep crater with steep sidewalls to form. In addition, the large percentage of fine particles results in decreased crater growth rate. The crater
194 growth increases in decreased gravity. The asymptotic depth and the time for the crater to reach the limiting c rater depth decreases with increased gravity. A change in cratering mechanism occurs with decreased gravity as the intrinsic density is decreased. Experiments performed in vacuum conditions are recommended. Additional lunar soil simulants should be studied both computationally as well as experimentally, to provide more robust predictions of the jet interactions with the various lunar regions that are present on the Moon. Simulations are recommended to investigate the dominant cratering mechanism for reduce d gravity with low intrinsic density materials. Additionally, the spray should be analyzed both experimentally, with techniques such as Laser Doppler Velocimetry (LDV) or Particle Image Velocimetry (PIV), and numerically using DEM and two fluid model ing
195 LIST OF REFERENCES Acosta Iborra, A., Sobrino, C., Hern ndez Jimn ez, F., deVega M. 2011 Experimental and computational study on the bubble behavior in a 3 D fluidized bed. Chem. Eng. Sci., 66, 3499 3512. Aderibigbe, O. O., Rajaratnam N. 1996 Erosion of loose beds by submerged circular im pinging vertical turbulent jets. J. of Hydr Res 34, 19 33. Adzumi, H. 1937. On the flow of Gases through a Porous Wall. Bull. of the Chem. Soc. of Jap. 12, 304 312. Albert, M.R., Shultz, E.F., Perron, F.E., Jr., 2000. Snow and firn permeability at Siple Dome, Antarctica. Ann. of Glaciol 31, 353 356. Albert, M.R., Shultz, E.F., 2002. Snow and rn properties and air snow transport processes at Summit, Gre enland. Atmos. Environ. 36, 2789 2797. Alexander, J.D., Roberds W.M., Scott R.F., 1966. Soil Erosion by Landing Rockets. Hay es Int. Corp., Birmingham, AL. Alsh ibli, K.A., Hasan, A., 2009 Properties of JSC 1A Lunar Regolith Simulant. J. Geotech and Geoenviron. Eng. 135, 673 679. Anderson, T.B., Jackson R. 1967. A Fluid Mechanica l Description of Fluidized Beds. In d. and Eng Chem Fun 6 527 539. Andreas, E.L. 2007. New estimates for the sublimation rate for ice on the Moon. Icarus 186, 24 30. Ant hony, J.L. Maro ne C. 2005 Influence of particle characteristics on granula r friction. J. of Geophys. Res. 110, B08409. Ball, B.C., Smith, K.A., Klemedtsson, L., Brumme, R., Sitaula, B.K., Hansen, S., Priem, A., MacDonald, J., Horgan, G.W., 1997. The influence of soil gas transport properties on methane oxidation in a selectio n of northern European soils. J. of Geophys. Res 102, 23309 23317. Battler, M. M., Spray, J. G. 2009 The Shawmere anorthosite and OB 1 as lunar highland regolith simula nts. Planetary and Space Sc i. 57(14 15), 2128 2131. Bear, J. 1988. Dyn amics of Fluids in Porous Media. Dover Publications, Inc ., New York. Benyahia, S., 2008. Validation Study of Two Continuum Gr anular Frictional Flow Theories. Ind. Eng. C hem. Res. 47, 8926 8932. Berger, K.J., A. Anand, P.T. Metzger, C.M. Hrenya, 2013. Role of collisions in erosion of regolith during a lunar landing. Phys. Rev. E 82 022205.
196 Bttega, R., Felkl de Almeida, A.R., Corra, R.G., Freire, J.T., 2009 CFD modeling of a semi cylindrical spouted bed: Numer ical simulation and exp erimental verification. The Can. J. of Chem. Eng. 87, 177 184. Bird, B.R., Steward, W.E., Lightfoot, E.N., 2008. Transport Phenomena, Second Edition. John Wiley & Sons, Inc. New York Boemer, A., Qu, H Renz, U., 1997. Eularian simu lation of bubble formation at a jet in a two dimensional fluidiz ed bed. Int. J. Multiphase Flow 23, 927 944. Carman, P.C. 1937. Fluid flow through granular beds. Chem. Eng. Res. and Des 15a, 150 166. Carman, P.C. 1950. Diffusion and flow of gases and va pours through micropores. I. Slip flo w and molecular streaming. Proc. of the Royal Society of London. Se r. A, Ma th. and Phys. Sci 203, 55 74. Carman, P.C. 1956. Flo w of Gases through Porous Media. Academic Press Inc., New York. Carruba, V., Coradini, A., 1999. Lunar cold traps: Effe cts of Double Shielding. Icarus 142, 402 413. Chapman, S., Cowling T.J. 1970. The Mathematical Theory of Non Uniform Gases Cambridge University Press, 3 rd Cambridge, UK. Chapuis, R.P., Aubertine, M. 2003. On the use of the Kozeny Carman equation to predict the hydraulic conductivity of soils. Can Geotech J. 40, 616 628. Cho, G.C., Dod ds J., Santamarina J.C. 2006 Particle Shape Effects on Particle Density, Stiffness, and Stre ngth: Natural and C rushed Sands. J. o f Geotech. and Geoenviron. Eng. 132, 591 602. Choate, R., Batterson, S.A., Christensen, E.M., Hutton, R.E., Jaffe, L.D., Jones, R.H., Ko, H.Y., Spencer, R.L., Sperling, F.B., Sutton, G.H., 1968. Lunar Surface Mechanical Properties. In: Su rveyor Project. Part 1: Project Description and Performance. JPL Technical Report 32 1265. Jet Propulsion Laboratory, Pasadena, pp 181. Clearl y, P.W., Sawley M.L. 2002 DEM modeling of industrial granular flows: 3D case studies and the effect of particle shape on h opper discharge. App. Math. Mod. 26, 89 111. Cocks, F.H., Klenk, P.A., Watkins, S.A., Simmons, W.N., Cocks, J.C., Cocks, E.E., Sussingham, J.C., 2002. Lunar Ice: Adsorbed water o n subsurface polar dust. Icarus 160, 386 397.
197 Cokljat, D., Ivanov, V.A., Sarasola, F.J., Vasquez, S.A., 2000. "Multiphase k epsilon models for unstructured grids," Proceedings of the 2000 ASME Fluids Engineering Division Summer Meeting, June 11 15, 2000, Boston, MA. Cokljat, D., M. Slack, Vasquez, S.A., Bakker, A., Monta nte G. 2006. Reynolds Stres s Model for Eulerian multiphase. Pro g. in Comp. Fluid Dynamics 6, 168 178. Colaprete, A., et al., 2010 Detection of water in t he LCROSS ejecta plume. Science 330, 463 468. Cooper, D., Jackson, D.C., Launder, B.E., Liao G.X. 1993. Impinging jet studies for turbulence model assessment I. Flow field experiments, Int J. Heat Mass Trans. 36 10, 2675 2684. Craft, T.J., Graham, J.W, Launder B.E. 1993. Impinging jet studies for turbulence model assessment II. An examination of the perfor mance of four turbulence models. Int. J. Heat Mass Trans. 36 10, 2685 2697. Dan, S., W. Shuyan, L. Gougdong, W. Shuai, L. Yongjian, W. Lixin, 2010. "Simulations of ow behavior of gas and particles in a spouted bed using a second order moment me thod frictional stresses model," Chem. Eng. Sci. 65 2635 2648. Duarte, C.R., Olazar, M., Murata, V. V., Barrozo M.A.S 2009. Numerical simulation and experimental study of uid particle ows in a spouted bed. Powder Tech. 188, 195 205. Elghobashi, S.E., Abou Arab T.W. 1983. A two equation turbulence mo del for two phase flows. Phys. of Fluids, 26, 931 938. Gibson, M.M., Launder, B.E., 1978. Ground effects on pressure fluctuations in the atmospheric boundary layer. J of Fluid Mech. 86, 491 511. Gidaspow, D., 1994. Multiphase Flow and Fluidization Academic Press, San Diego Gryczka, O., Heinrich, S., Deen, N.G., van Sint Annala nd, M., Kuipers, J.A.M, Jacob, M., Mrl, L., 2009. Characterization and CFD modeling of the hydrodynamics of a prismatic spouted be d apparatus. Chem. Eng. Sci. 64, 3352 3375. Gunal M., Guven A. 2006. Prediction of flow patterns in scoured beds caus ed by submerged horizontal jets. Can. J. Civ. Eng. 33, 41 48. Haehnel, R.B., 2011. Particle Entrainment Under Non Uniform Flow Created by an Impinging Jet. (PhD Dissertation). Dartmouth College, ProQuest Dissertations & Theses (PQDT). ISBN 9781267169129. Haehnel, R.B., Dade, W.B., Cushman Roisin, B., 2008. Crater evolution due to a jet imping ing on a bed of loose particles. 11th ASCE Earth & Space Conference, March 3 6, 2008, Long Beach, CA.
198 He, C. 2010. Geotechnical Characterization of Lunar Regolith Simulants (PhD Dissertation) Case Western Reserve University. ProQuest Dissertations & Thesis (PQDT). ISBN 9781124014388. Hilton J.E., Mason, L.R., Cleary P.W. 2010 Dynamics of gas solid fluidized beds with non spherical particle geometry. Chem Eng. Sci. 65, 1584 1596. Hindmarsh, R.C.A, van der Wateren, F.M., Verbers, A.L.L.M., 1998. Sublimation of ice through sediment in Beac on Valley, Antart ica. Geografiska Annaler. Ser. A, Phys. Geog. 80, 209 219. Hodges Jr., R.R., 2002. Ice in the lunar polar regions revisited. J. of Geophys Res. 107, E2, 5011. Hong, R., Li, H., Ch eng, M., Zhang J. 1996. Numerical simulation and verification of a gas solid jet fluidized bed. Powder Tech. 87, 73 81. Hosseini, S.H., Ahmadi, G., Razavi, B.S., Zhong W. 2010. Computational Fluid Dynamic Simulation of Hydrodynamic Behavior in a Two Dimensional Conical Spouted Bed. Energy Fuels 24, 6086 6098 Immer, C.M. and Metzger, P.T. 20 10 Rocket cratering in simulated lunar and Martian environments. 12 th ASCE Earth & Space Conference, March 14 17, 2010, Ho nolulu, HI Jackson R., 1997. Locally averaged equations of motion for a mixture of identical spherical particles and a Newtonian fluid. Chem. Eng. Sci. 52 2457 2469. Jackson R., 1998. Erratum. Chem. Eng. Sci. 53 1955. Johnson, P.C., Jackson R. 1987. Frictional Collisional Constitutive Relations for Granular M aterials, wit h Application to Plane Shearing. J. Fluid Mech. 176 67 93. Johnson, P. C., Nott, P., Jackson, R., 1990. Frictional Collisional Equations of Motion for Particle Flows and Their Application to Chutes. J. Fluid Mech. 210 501 535. Karim, O., Ali K.H.M. 2000. Predictions of flow patterns in local scour holes caused by t urbulent water jets. J. Hydr. Res. 38, 279 287. Klinkenberg, L.J. 1941. The permeability of porous media to liquids and gases. Drilling and Production Practice. American Petroleu m Inst., pp. 200 213. Kobus, H., Leister, P., Westrich B. 1979 Flow field and scouring effects of steady and pulsating jets impinging on a movable bed. J. of Hydr. Res. 17, 175 192. Kok, J. F., Renno N.O. 2009. A comprehensive numerical model of s tead y state saltation (COMSALT). J. Geophys. Res. 114, D17204.
199 Koponen, A., Kataja, M., Timonen, J., 1997. Permeability and effective porosity of porous media. Phys. Rev. E 56, 3319 3325. Kuang, S.B., LaMarche, C.Q., Curt is, J.S., Yu, A.B. 2013. Discrete part icle simulation of jet inducted crate ring process on a granular bed. Powder Tech. 239, 319 336. Kunii, D., Levenspiel O. 1997. Cir culating fluidized bed reactors. Chem. Eng. Sci. 52, 2471 2482. LaMarche, C.Q., Curtis, J.S., Metzger, P.T. 2011 Permeability of JSC 1A: A lunar soil simulant. Icarus 212, 383 389. LaMarche, C. Q., Curt is, J. S., Metzger, P. T. 2012 Cratering of a lunar soil simulant, JSC 1A, by a turbulent subsonic jet. 13 th ASCE Earth & Space Conference, April 15 18 2012, Passadena CA Launder, B.E., 1975. On the effects of a gravitational field on the turbulent transport of heat and momentum. J of Fluid Mech. 67, 569 581. Launder, B.E., 1989. Second momen t closure: present0 and future? Intern. J of Heat and Fluid Flow 10, 282 300. Lien, F.S., Leschziner, M.A., 1994. Assessment of turbulence transport models including non linear RNG eddy viscosity formulation and second moment closure for fl ow over a backward facing step. Computers & Fluids 23, 983 1004. Lun C.K.K., Savage, S.B., Jeffery, D.J., Chepurniy S. 1984. Kinetic theories for granular flow: inelastic particles in Couette flow and slightly inelastic pa rticles in a general flow field. J. of Fluid Mech. 140 223 256. Mauran, S., Rigaud, L., Coudevylle, O., 2001. Application of the Carman Kozeny correlation to a high porosity and anisotropic consolidated medium: The compressed expa nded natural graphite. Transp. in Porous Media 43, 355 376. McKay, D.S., Carter, J.L., Boles, W. W., Allen, C.C., Allton, J. H. 1994 JSC 1: A new l unar soil simulant (pp. 857 866). Proc. Eng. Construction, and Operations in Space IV, ASCE, New York. McKeen, T., Pugsley T. 2003 Simulation and experimental validation of freely bubbling bed o f FCC catalyst. Powder Tech. 129, 139 152. Mehta, M., Renn o, N.O., Marshall J., Grover, M.R., Sengupta, A., Rus che, N.A., Kok, J.F., Arvidson, R.E., Markievicz, W.J., Lemmon, M.T., Smith P.H., 2011. Explosive erosion during the Phoenix landing exposes s ubsurface water on Mars. Icarus 211, 172 194. Metzger, P.T ., Lane, J.E., Immer, C.D., Clements, S., 2008. Cratering and blowing soil by rockets and eng ines during lunar landings. Proceedings of the Sixth
200 International Conference on Case Histories in Geotechnical Engineering, Arlington, VA. August 11 16, 2008, Pap er No. 10.11. Metzger, P.T., 2009. Personal Communication. Metzger, P.T., Immer, C.D., Dona hue, C.M., Vu, B.T, Latta III, R.C., Deyo Svendsen M. 2009 a Jet induced cratering of a granular surface with application to Lunar spaceports. J of Aero. Eng. 22, 24 32. M etzger, P.T., Latta, III, R.C., Sch uler, J.M., Immer C.D. 2009b Craters Formed in Granular Beds by Impinging Jets of Gas. Powders & Grains 2009 Conference July 13 17, 2009 Golden, CO Metzger, P .T., Lane, J.E., Immer, C.D., Gamsky, J.N., Hauslein, W., Li, X., Latta, III, R.C., Dona h ue C.M. 2010 Scaling of erosion rate in subsonic jet e xperiments a nd apollo lunar module landings. 12th ASCE Earth & Space Conference, March 14 17, 2010, Honolulu, HI. Millington, R.J., Quark, J.P., 1964. For mation factor and permeability equations. Nature 202, 143 145. Mitchell, J.K., Houston, W.N., Sco tt, R.F., Costes, N.C., Carrier III, W.D., Bromwell, L.G., 1972 Mechanical Properties of lunar soil: Density, porosity, cohesion, and angle of internal frict ion. Proceedings of the Third Lunar Science Conference (Supplement 3, Geochimica et Cosmochimica Acta), The M.I.T press, 3 3235 3253. Mitchell, J.K., Carrier III, W.D., Costes, N.C., Houston, W.N., Scott, R.F., Hovland, H.J., 1973. Soil Mechanics. In: Apo llo 17 Preliminary Science Report, NASA SP 330 Mitchell, J.K., Houston, W.N., Carrier III, W.D., Costes, N.C., 1974. Apollo Soil Mechanics Experiment S 200 Final Report. Space Sciences Laboratory Series 15, Issue 7, University of California, Berkley. Mol drup, P., Olesen, T., Komatsu, T., Schjnning, P., Rolston, D.E., 2001. Tortuosity, diffusivity and permeability in the soil liquid and gaseous phases. Soi l Sci. Soc. Am. J. 65, 613 623. Nemec, D., Levec, J., 2005. Flow through packed bed reactors: 1. Sing le phase flow. Chem. Eng. Sci. 60, 6947 2957. Niven, R.K., 2002 Physical insight into the Ergun and Wen & Yu equations for fluid flow in packed and fluidised beds. Chem. Eng. Sci. 57, 527 534. Ocone, R., Sundaresan, S., Jackson R. 1993. Gas Particle Flow in a Duct with Arbitrary Inclination with Particle Particle Interactions. AIChE J. 39 1261 1271.
201 Passalacqua, A., Marmo L. 2009. A critical comparison of frictional stress models applied to the simulation of bubbling fluidized beds. Chem. Eng. Sci. 160, 2795 2806. Patankar, S.V., 1980. Numerical Heat Transfer and Fluid Flow Hemisphere, New York. Peng, B., Zhu, J., Zhang C. 2010. Numerical Study on the Effect of the Air Jets at the Inlet Distributor in the Gas Solids Circulating Fluidized B ed Rise rs. Ind. Eng. Chem. Res. 49, 5310 5322. Pritchard, P.J., 2011. Fox and McDonald's Introduction to Fluid Mechanics 8th edition. John Wiley & Sons, New York. Rahmatian, L.A., Metzger, P. T. 2012 Soil test apparatus for l unar s urfaces (pp. 239 253). 12th A SCE Biennial International Conference on Engineering Construction, and Operations in Challenging Environments; and Fourth NASA/ARO/ASCE Workshop on Granular Materials in Lunar and Martian Exploration, Reston, VA. Rajaratnam, N., Beltaos S. 1977. Erosio n by Impinging Circular Turbulent Jets J. of the Hydr. Div. 103 1191 1205. Rajaratnam, N., 1982 Erosion by Submerged Circular Jets, J. of the Hydr. Div. 108, 262 267. Reuge, N., Cadoret, L., Coufort Saude jaud, C., Pannala, S., Syamlal, M., Caussat, B., 2008. Multifluid Eulerian modeling of dense gas solids fluidized bed hydrodynamics: Influence of the dissipat ion parameters. Chem. Eng. Sci. 63, 5540 5551. Rhodes, M. ed., 2008. Introduction to Particle Technology, Se cond Edition. John Wiley & Sons Ltd, West Sussex, England. Rice, P.A., Fontugne, D.J., Latini, R.G., Barduhn, A.J., 1970. Anisotropic permeability in porous media. Flow Through Porous Media Symp. 62, 23 31. Rickman, D., Edmunson, J., McLemore, C. 2 013 A Functional Comparison of Lunar Regoli ths and Their Simulants. J. of Aero. Eng. 26(1), 176 182. Rickman, D., Immer, C., Metzger, P., Dixon, E., Pendle ton, M., Edmunson, J. 2012 Particle Shape In Simulants of the Lunar Regolith J. of Sedimentary Re s. 82 823 832. Roberts, L., 1963. Visibil ity and dust erosion during the lunar landing. A compilation of recent research related to the Apollo Mission, National Aeronautics and Space Administration, Hampton, Va., 155 170.
202 Roscoe K.H., Schofield, A.N., Wroth P.C. 1958. On the yielding of soils. Gotechnique 8, 22 53. Roscoe, K.H., 1970. The influence of strains in soil mechanics. Gotechnique 20, 129 170. Rowe, P.N., 1961. Drag Forces in a Hydraulic Model of a Fluidized Bed : II. Trans. Inst. Chem. Eng. 39, 175 180. Royer, J. R., Evans, D. J., Oy arte, L., Guo, Q., Kapit, E., Mbius, M. E., Waitukaitis, S.R., Jaeger, H.M., 2009. High speed tracking of rupture and clustering in freely falling granular streams. Nature 459 1110 1113. Salvail, J.R., Fanale, F.P., 1994. Near surface ice on Mercury and the Moon: A to pographic thermal model. Icarus 111, 441 445. S antos, K.G., Murata, V.V., Barrozo M.A.S 2009. Three dimensional computational fluid dynamics modeling of spouted bed The Can. J. of Chem. Eng. 87, 211 219. Sarma, K.V.N., 1967 Ex istence of limiting scour depth. J. of the Inst. of Eng. (India), Civil Eng. Div. 48, 84 92 Sau, D.C., Mohanty, S., Biswal K.C. 2010 Experimental studies and empirical models for the prediction of bed expansion in g as solid tapered fluidized beds. Chem. Eng. and P rocessing 49, 418 424. Savage, S.B., 1998. Analyses of slow high concentration flows of granular ma terials. J. Fluid Mech. 377 1 26. Schaeffer, D.G., 1987. Instability in the Evolution Equations Describi ng Incompressible Granular Flow. J. Diff. Eq. 66 1 9 50 Scheidegger, A.E. 1974. The Physi cs of Flow Through Porous Media. University of Toronto Press, Toronto, Canada. Schofield, A.N., Wroth C.P. 1968. Critical State Soil Mechanics McGraw Hill, London. Schorghofer, N., Taylor, G.J., 2007. Subsurface migration of H20 at lunar cold traps. J. of Gephys. Res. 112, E02010. Schrader, C. M., Rickman, D. L., McL emore, C. A., and Fikes, J. C. 2010 Lunar Regolith Simulant User's Guide (No. NASA/TM 2010 216446, M 1295). National Aeronautics and Space Administration, Marshall Space Flight Center. Scott, R.F., Ko H.Y. 1968. Transient rocket engine gas ow in soil. AIAA J. 6, 258 264.
203 Sengupta, A., Kulleck, J., S ell, S., Van Norman, J., Mehta, M., Pokora M. 2009 Mars Lander Engine Plume Impingement Environment of the Mars Science Laboratory, IEEE Aerospace Conference. Paper no 1349. Shinohara, K., Oida, M., Golman B. 2000 Effect of particle shape on angle of internal friction by triaxial compression test. Pow der Tech. 107, 131 136. Simonin, O., Viollet P.L. 1990. Predictions of an Oxygen droplet pulverization in a compressible subso nic coflowing hydrogen flow. Numerical Methods of Multiphase Flow, FED Vol. 91, 65 82. Sinclair, J.L., Jackson, R., 1989. "Gas p article flow in a vertical pipe with particle particle interactions," AIChE J., 35, 1473 1486. Sizemore, H.G., Mellon M.T., 2008. Laboratory characterization of the structural properties controlling dynamical gas transpo rt in Mars analog soils. Icarus 197 606 620. Srivastava, A., Sunda resan, S., 2003. Analysis of a Frictional Kine tic Model for Gas Particle Flow. Powder Tech. 129 72 85. Syamlal, M., Rogers, W., Obrien T.J. 1993. Mfix Documentation Theory Guide, U.S. Dept. of Energy, Office of Fossil En ergy, Tech. Note. Utikar, R.P., Ranade V.V. 2007. Single jet uidized beds: Experiments and CFD simulations with glass and polypropy lene particles. Chem. Eng. Sci. 62, 167 183. Van Wachem, B.G.M, Schouten, J.S., van den Bleek, C.M., Krishna, R., Sinclair J.L. 2001. Comparative Analysis of CFD Mod els of Dense Gas Solids Systems. AIChE J. 47, 1035 1051. Vasavada, A.R., Paige, D.A., Wood, S.E., 1999. Surface Temperatures on Mercury and the Moon and the stabilit y of polar ice deposits. Icarus 141, 1 79 193. Wen, C.Y., Yu Y.H. 1966. Mechanics of Fluidization. Chem. Eng. Prog. Symp. Ser. 62 100 111. W estrich, B. Kobus H. 1973 Erosion of a uniform sand be by continuous and pulsating jets, Proceedings of Research and practice in the water environm ent, 15 th Congress of the Internationals Association for Hydraulic Research, Istanbul, Turkey. White, B.R., Schulz J.C. 1 977. Magnus effect in saltation. J. of Fluid Mech. 81 497 512 Zeng, X., He, C., Oravec, H., Wilkinson, A., Agui, J., and Viv ake, A. 2010 Geotechnical properties of JSC 1A lunar soil sim ulant. J. of Aero. Eng. 23 111 116.
204 Zhou, Z.Y., Pinson, D., Zou, R.P Yu A.B. 2011 Discrete particle simulation of gas fluidi zation of ellipsoidal particles. Chem. Eng. Sci. 66, 6128 6145. Zou, X. Y ., Cheng, H., Zhang, C.L., Zhao Y.Z. 2007. Effects of the Magnus and Saffman forces on the salta tion trajectories of sand grain. Geomorphology 90 11 22
205 BIOGRAPHICAL SKETCH Casey was born and raised in Princeton Ne w Jersey. Casey thoroughly enjoys playing sports, specifically cycling, field hockey and lacrosse. Casey went to University of Delawa re, where Casey studied chemical e ngineering, minored in material science and woman studies and as well as par ticipated in study abroad in Milan, Italy for a winter semester. Casey made deans list multiple semesters. Casey joined the PhD program in the University o f Florida Chemical Engineering D epartment after graduating from University for Delaware Casey also p articipated in a NASA internship in th e summer of her first year at the University of Florida While at NASA, Casey had many opportunities to learn about the space program and see the Shuttle up close. Casey also had the opportunity to fly experiments in t he NASA FAST lunar gravity experience. At the University of Florida, Casey has participated in many leadership and group organizations. Casey has held many officer positions in the Graduate Association of Chemical Engineers (GRACE). Casey won the prestigio us Attribute of a Gator Leadership award for the College of Engineering.